A single egg, meanwhile, contains 6 grams of protein, costs just 10 cents, and requires only chicken feed and water as inputs.

Neion Bio is taking advantage of this incredible technology to turn chickens into drug factories. And they’ve already signed a commercial deal with a drugmaker to produce biosimilars.

With AI leaders promising to unleash countless new therapies, we’re going to need to scale up our manufacturing. That sci-fi future may be one where we stop producing drugs in huge steel tanks, and instead start farming them like we do our food.

We sat down with co-founders Dimi Kellari and Sam Levin to discuss this future. [Disclosure: I’m an investor, though I wasn’t at the time of recording.]

Thank you to SynBioBeta for hosting us!

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

0:00 Intro
1:07 What Neion Bio does
4:53 How the founders arrived at synthetic biology and chickens
10:21 Biosimilars as Neion’s first commercial market
13:30 Why farming drugs is a viable path to reshoring biomanufacturing
18:38 Neion’s technical moat & why engineering birds is difficult
25:50 Advantages of chicken-based manufacturing
30:06 Why CHO manufacturing became the industry standard
33:21 What it takes to disrupt CHO manufacturing
37:57 What it means to program biology
42:48 Skepticism around this new manufacturing approach
47:00 Applying tech principles to biotech
55:34 Natural intelligence vs artificial intelligence
57:14 The future of farming drugs & programmable chickens

Transcript

1:07 What Neion Bio does

Daniel 00:01:07

What is the connection between Odysseus and what you guys are doing at Neion?

Sam Levin 00:01:12

Great question. Very few people have actually asked us that.

Daniel 00:01:31

Nobody’s like the Free Radicals podcast.

Sam Levin 00:01:34

Exactly. It really comes down to the fact that Odysseus was a master strategist. He was out-brawled, out-muscled, and out-superpowered by everyone around him, and he won by being strategic. He won the war through strategy.

Dimi and I both look up to that level of long-term strategic thinking and “3D chess.” We believe that chickens represent that same level of strategic engineering.

Daniel 00:02:10

Only 3D chess?

Sam Levin 00:02:11

Sorry, 4D chess.

Daniel 00:02:12

I am completely uninterested in 3D chess.

Sam Levin 00:02:15

You have to get into at least the fourth dimension. Well, it’s a move beyond 2D chess, which is easy.

Daniel 00:02:22

Fair enough. Tell us, what is the three-dimensional chess behind Neion?

Dimi Kellari 00:02:29

We turn chickens into drug factories—medicine factories. This is a way to make the production of some of our most expensive, life-saving therapeutics at the cost and scalability of chicken farming. We farm medicines.

Daniel 00:02:49

Could you tell us about how drugs are traditionally manufactured and help us understand how Neion is different? I assume we don’t currently manufacture drugs in chicken eggs.

Sam Levin 00:03:03

There are two large classes of drugs: small molecules and large molecules. Small molecules are things like aspirin. They are usually taken in pill form and can be chemically synthesized very cheaply.

Large molecules are also called biologics. They are typically proteins and are so large and complex that they cannot be chemically synthesized, so we have to make them using biology. These are typically used for more difficult diseases like cancer and autoimmune diseases. They are usually delivered via injection or IV.

The way we make these biologic proteins today is in giant steel tanks called bioreactors using Chinese hamster ovary cells (CHO cells). Most people don’t realize these drugs come from the ovary cells of Chinese hamsters, which we grow in huge tanks. We use them to express a protein drug, and then we extract and purify it.

The difficulty is that rebuilding the environment of a hamster out of steel, plastic, and media is wildly inefficient. It is very expensive, slow to scale, and depends on multi-billion dollar facilities and complex supply chains. That is how we do it today.

4:53 How the founders arrived at synthetic biology and chickens

Eric 00:04:53

This isn’t a problem you’re new to. You’ve been in the synthetic biology field for a long time. Tell us how you originally got interested in the field and what “synthetic biology” even means.

Sam Levin 00:05:03

Synthetic biology basically encompasses any time we take what exists naturally and modify it. Typically, we mean modifying the genome, though the term has become broader. Before Neion, I built a more traditional synthetic biology company.

We built the steel and plastic bioreactor tanks I just mentioned. My previous company built high-throughput robotic versions of those and machine learning software to develop new strains. I spent five years immersed in trying to harness biology using tanks.

I started looking at this and thinking, “Is this really the way to harness biology?” You look at trees compiling sunlight and water into giant, 100-foot sugar and protein machines for free in a distributed way. Yet when we harness biology in synthetic biology, we build huge factories and billion-dollar tanks.

Dimi and I were interested in why we make our drugs in Chinese hamster ovary cells. Why does it cost $500 a gram to do this? We have a deluge of new protein designs ready to be made, yet the way we make them hasn’t changed in 50 years. That inefficient process was my entry point.

Dimi Kellari 00:07:06

For me, it’s quite different. I’m a recovering aerospace engineer, so I’ve done manufacturing for much of my career. I’ve been bio-curious for a long time. My “night job” was educating myself on biology. I had a lot of FOMO seeing tools like AI, precision genome engineering, and copious genomic data coming into place.

People talk about biology becoming an engineering discipline. In the engineering disciplines I worked in prior, you could work on a laptop and have a good understanding of how something would behave and what it would cost in silico before ever doing anything in the real world. Biology is becoming more like that, which is exciting.

The next frontier for the next decade is going to be in bio. I was lucky enough to meet Sam, and we complement each other’s skill sets. We spent a lot of time thinking about what the future of bio could be and where the opportunities were.

We realized that as we gain the ability to design drug candidates at the click of a button, the bottleneck shifts to manufacturing. How do we realize these designs affordably, accessibly, and at scale? That problem intrigued us and led us to CHO cells.

We now have tools that allow us to move from engineering single cells to reliably engineering multicellular organisms efficiently and robustly. That unlocks a whole new world of possibilities. We were fortuitous in landing on chickens, but once that idea was incepted, it just kept reverberating. Nothing we found told us it wasn’t a feasible idea, so we formed Neion a couple of years ago.

10:21 Biosimilars as Neion’s first commercial market

Daniel 00:10:21

It’s interesting that exciting technologies often lack a business model in the near term. It sounds cooler to produce drugs this way, but what is wrong with the steel vats? In the pharma industry, most cost is in R&D, so companies might not care much about manufacturing costs. How did you conclude there is a demand for this?

Dimi Kellari 00:11:09

You’re entirely right. We could have spent years building something no one wanted, but we’ve been hyper-focused on finding a good beachhead. The place that makes the most sense in the near term is the biosimilars market.

Biosimilars are generic versions of biological drugs. There are a lot of regulatory tailwinds right now, as the FDA is making it easier to bring these to market. This next decade will be the decade of biosimilars because 100 biologics are coming off patent. They require lower-cost, more scalable manufacturing.

We validated that hypothesis with our first large pharma deal to bring multiple products to market. Currently, there is no US manufacturer of biosimilars. Doing this resiliently—onshoring the entire supply chain—is critical for national security.

Our inputs are abundant in almost every state: chicken feed and water. It is compelling when you can onshore the entire supply chain at a lower cost than anyone else in the world. That is only the start. Eventually, we want to go after all biopharmaceuticals. We won’t win just on cost, but on speed, performance, and resiliency.

13:30 Why farming drugs is a viable path to reshoring biomanufacturing

Eric 00:13:30

Our friend Elliot Hershberg talks about the “barbellification” of the life sciences market. On one side, you have a power law concentration of large R&D dollars and mega-blockbuster drugs from the likes of Eli Lilly and Novo Nordisk.

On the other hand, you have an explosion of small biotechs doing AI-enabled molecule discovery because it’s never been cheaper to get from an idea to a therapeutic concept. Do you agree with that take, and do you service both sides of that barbell?

Dimi Kellari 00:14:30

Elliot is actually one of the reasons we happened across chickens. He wrote a piece inspired by conversations with Sam about moving beyond steel tanks. I think he’s generally right about the barbellification.

We want to service both sides of that barbell. When we look at our advantages versus steel tanks, they are Pareto optimal across all scales. The thing that varies is the risk tolerance of the customers.

I don’t expect a company like Amgen, which has spent billions on steel tank facilities, to move to chickens tomorrow. Eventually, once we prove ourselves, that will be compelling. But smaller biotechs need to spend $5 million to $10 million to get IND quantities of a therapeutic.

Often, they have to wait a long time or go to Chinese CDMOs. If you want to service the U.S. government, there was a bill passed that prevents you from using a CDMO like WuXi. These dynamics are in favor of companies that can do things more rapidly, cheaply, and critically, here in the U.S.

Eric 00:16:38

Let’s dig into how the biotech ecosystem has evolved over the past five years. The two biggest headlines are the increasing incumbency of AI for molecule discovery and the rise of Chinese biotech. How does your company fit into those trends?

Sam Levin 00:17:12

The rise of AI is a major tailwind for us. As there is a greater accumulation of potential therapies to bring to market, the two main bottlenecks become regulatory and manufacturing. We are solving the manufacturing bottleneck.

A big part of our vision is that if you can design a therapy at the click of a button, you shouldn’t have to wait three to five years and spend a billion dollars to scale it up.

In terms of China, we are big believers in leapfrog technologies. We are never going to beat China on CHO manufacturing. We won’t be able to reshore Chinese hamster ovary cell bioreactor manufacturing without massively compromising on cost, quality, or both. We see this as one of the few viable paths to reshore manufacturing of our biologics without those compromises.

18:38 Neion’s technical moat & why engineering birds is difficult

Daniel 00:18:38

To what extent will China be able to copy this technology? Even if they do, it doesn’t take away the value of having an independent manufacturing base. I’m curious where the moat and the IP reside in this technology.

Sam Levin 00:18:56

We have poured much of our resources into recruiting some of the world’s leading genome engineers to massively improve the state of the art for engineering chickens. We are becoming the best chicken engineers in the world.

We have achieved this in a very short period of time by finding the people who had made the most complex edits to vertebrates and bringing them onto our team. We are building all these novel technologies, which are Neion IP.

Some we are patenting and some we are keeping as trade secrets. This will dramatically reduce the time to go from concept to protein in an egg. It will be very hard for anyone to compete with us, as the talent we’ve amassed is difficult to replicate.

Daniel 00:20:08

Can you tell us about some of the scientific breakthroughs that enable this?

Sam Levin 00:20:14

The two key technological enablements were precision genome engineering—the ability to precisely put genes where we want them—and the ability to culture a specific type of chicken stem cell.

Both allow us to harness the native genetic architecture that drives protein production in the egg. Our whole thesis is that chickens are the best protein factories. To use them to the fullest extent, you want to use the native machinery for protein production in the egg.

That means using a very specific locus in the genome called the ovalbumin locus, which drives half the protein production in the egg. To use that locus, we need advanced precision editing tools so we aren’t throwing our gene in random places.

Using those tools in live animals is impractical. The ability to culture chicken stem cells allows us to take these cells out of an egg, grow them in petri dishes, perform complex editing, and then put them back into an egg.

These stem cells are called PGCs, or primordial germ cells. The combination of precision editing and the ability to culture these stem cells allows us to get consistent high yields of protein in the egg.

Eric 00:21:59

If I’m understanding correctly, you’re doing cellular engineering on the stem cell, taking those transformed cells and turning them into eggs, and then harvesting the protein produced in the same genetic promoter sequences that produce ovalbumin.

Sam Levin 00:22:23

It is even more involved than that. We take a fertilized egg, extract these stem cells, and expand them in flasks up to millions of cells. We do our editing, then take those edited stem cells and put them back into a fertilized egg.

We use a delicate microsurgery technique to inject the cells, then seal the egg back up. Those stem cells go and colonize the germline of the egg.

When that egg hatches, it is a chick with a genetically modified germline—modified sperm or egg cells. We rear those to adulthood and breed them. What hatches from those is a fully genetically modified organism. It is a complicated process that few people in the world have done successfully.

Eric 00:23:21

Could we apply that to other vertebrates? It doesn’t seem impossible to apply it to mice, dogs, or eventually humans.

Sam Levin 00:23:37

It is actually the other way around. Our Chief Scientific Officer, Sven Östrup, was previously at Colossal Biosciences, which has been in the news for working to bring the woolly mammoth back from extinction.

He led the dire wolf project there, which involved making an animal with 20 edits in its genome. That work is actually easier to do in mammals because we can clone mammals.

We can’t do that in birds because we can’t access the single-cell zygote stage. By the time the egg is laid, there are 50,000 cells in there, which makes the editing more complicated. Sven and the other experts we’ve hired brought those advanced engineering skills from mammals, dogs, and cats to bear on the chicken problem.

Dimi Kellari 00:24:45

Chickens are the most farmed animal. The number of chickens actually outnumbers the total number of all other bird species. We know chickens better than anyone else, and there is a lot of value in knowing the biology of a useful organism.

That insight led to Regeneron being a $100 billion company based on their knowledge of the mouse. They owned that biology. We are already very efficient at farming chickens. Being able to direct that biology is going to be incredibly valuable.

25:50 Advantages of chicken-based manufacturing

Daniel 00:25:50

Let’s imagine the future where the Neion approach has taken root in both biosimilars and biologics. We’re growing all of our non-small molecule drugs in eggs. What does that mean for the future, and where else does that take us?

Sam Levin 00:26:09

It means low-cost, affordable biological therapeutics that can be manufactured locally. Many countries cannot build CHO bioreactor facilities, but almost every country in the world farms chickens.

Part of our vision is to enable countries to have their own medical supply chains. We also want to unlock new biology by making proteins that you simply can’t make in CHO.

The chicken oviduct is an incredible protein factory that regularly secretes multi-thousand-kilodalton, heavily glycosylated, repetitive proteins. There is 40% of the human secretome that CHO cells just won’t make.

There are designs for therapies sitting on the cutting floor because they can’t be made in the one cell type we currently use to make drugs. Our long-term vision is to unlock the possibility of breakthroughs that wouldn’t otherwise reach patients. AI will increase the number of potential drugs, but if we only make them in one cell type, many will never make it to the clinic.

Daniel 00:27:57

To make sure I’m understanding correctly, when we make proteins, there is the DNA sequence, but then it needs to fold in a certain way, which depends on the environment and chaperone proteins.

Then there are post-translational modifications like glycosylation. CHO can only access a part of that potential space of protein structures. Can chickens access that entire space?

Dimi Kellari 00:28:29

There is no panacea host for everything. There will be things that chickens aren’t ideal at making, but we haven’t hit that boundary yet. Chickens can certainly unlock new things that weren’t previously possible.

There are many advantages in terms of post-translational modifications. Because we can modify the properties of the proteins, we can create a library of chicken genetics that allow us to target different classes of protein.

In the future, we’ll be able to pluck the genetics we think are best for a particular protein type, depending on the target and the indication. People have tried to do this in CHO for a long time, but yields typically suffer. Our approach allows us to address unmet medical needs that were previously impossible to reach.

30:06 Why CHO manufacturing became the industry standard

Sam Levin 00:30:06

It is worth lingering on why we use CHO cells. It’s a question Dimi asked me three years ago, and I didn’t have a good answer. Most people in this space don’t have a good answer.

The truth is, it’s an accident of history. We don’t use Chinese hamster ovary cells because they are the best protein factories on the planet. They aren’t even particularly good ones.

We use them because they were the cells we could engineer, for the same reason we use E. coli. They had simpler genetics, with half the number of chromosomes of lab mice, and they spontaneously immortalized in a lab, making them easy to grow.

Our bioengineering capabilities have skyrocketed in the last decade. We can now approach the problem by finding the best protein factory in nature and engineering that to be a medical protein factory.

Daniel 00:31:45

When it comes to making biologics in CHO, a lot of expertise goes into designing the right edits and processes. It sounds like you are building that expertise as well, but you’ve gone outside the space of just CHO. Are you building a broader platform for navigating the whole space of protein engineering in organisms?

Sam Levin 00:32:22

We are not just working on the egg; we are thinking about the chicken and other modifications we can achieve. A key point is that we’re also offloading some of the expertise onto evolution.

In CHO, the expertise is often about trying to coerce a cell type into being a better protein screener. You might make a thousand variants and see which one has good yields.

We do the opposite. We ensure our gene goes into one specific location because evolution has already built a high-yield system at that location in the genome. It is a fundamentally different approach.

33:21 What it takes to disrupt CHO manufacturing

Eric 00:33:21

I’d love to chat more about what you’re competing against. In some ways, it feels like you’re competing against the incumbency of the existing CDMO and CRO landscape and how people currently approach preclinical and clinical manufacturing of biologics.

On the other hand, there isn’t much technical competition in terms of the efficiency of producing protein or the ability to produce certain proteins that CHO cells cannot produce at all. What is holding the field back from adopting newer solutions like yours for protein manufacturing, and how do you convince people to get over the line?

Dimi Kellari 00:33:58

CHO has been around for 50 years. We have invested hundreds of billions of dollars in the infrastructure, the equipment, and the media used to optimize around it. There is a massive amount of inertia. It is a David and Goliath situation. In fact, one of our first chicks that hatched was named David, but that’s a whole other story.

Traditional CHO manufacturing is the elephant in the room. Because our first target markets are uniquely suited to the value we add, we can commercialize quickly as an early-stage company, which is difficult to do in this space. Because we solve those problems so well, we have accelerated quite quickly.

There is a spectrum of early adopters, but some folks will need a lot of convincing. People who operate a $20 billion infrastructure will take effort to switch from CHO. One common misconception is that because bioreactors are shiny steel tanks with sensors and controlled environments, they must be more precise than a chicken. People often associate chickens with being messy.

Dimi Kellari 00:35:54

The reason the bioreactor exists is to get the right amount of oxygen and nutrients to each individual cell to replicate what would happen in the body of a Chinese hamster. All that precision is geared toward doing what blood vessels do naturally: providing exactly what is needed at the right time to produce proteins. The cells are doing the same thing; they are just genetically engineered cells secreting proteins.

It will take time for some to get over that hurdle, and we have to prove ourselves. We must ensure the first drugs we bring to market are reliable and performant. We are confident that will happen soon, but there is significant inertia among those who operate large facilities.

However, because biotechs and large pharma companies are increasingly outsourcing to CDMOs, we can disintermediate. We come in with a compelling value proposition: you don’t need to wait months or pay high upfront costs. We can provide lower batch-to-batch variability. Once you have your initial quantities for a pre-IND or IND application, scaling from there is completely de-risked because we simply breed more birds. The variability between them is very low. Those advantages will help us win in the near term.

37:57 What it means to program biology

Daniel 00:37:57

I want to connect this to the broader idea of programming biology. Dimi, you come from tech, and software engineers often want to code biology the way we do with software.

Those huge steel tanks seem like an attempt to program biology in a crude way by micromanaging every variable. The alternative approach with the egg is offloading the computation onto the organism itself. Do you think about it this way? What other aspects of this programming biology space have you explored?

Sam Levin 00:38:42

That was beautifully said. In the past, we were bad at engineering, so we used simple cell types and compensated with hardware and infrastructure. We made up for biological deficiencies with hardware. The last 50 years of synthetic biology were about getting E. coli, yeast, and CHO cells to do more.

I think the next 50 years will be about expanding the aperture of organisms we use. Organisms are already doing all this computation and have solved these problems. Chickens are fully autonomous, self-replicating 3D printers. They walk around, find food, make copies of themselves, and produce six grams of protein for ten cents.

There are many things in nature solving incredible feats. We only now have the engineering tools to harness and co-opt these complex systems that aren’t E. coli. That is the future.

Dimi Kellari 00:40:08

It is not just about inserting DNA code into a genome; it’s about making the genome easier to engineer. We are pioneering some of those things internally at Neion. These concepts all have analogs in tech.

A naive mind from a tech perspective might think we can just do something in biology because it makes sense on paper, but then biology hits you over the head because it doesn’t work. We are practical and commercially focused. We program and reprogram biology, but we do it in a way that builds a business.

It isn’t science fiction programming; it’s tried and tested methods combined with our novel technology to create commercial applications. Innovation is not just making something work in a gold-plated lab; it is making it work at scale to have an impact on the world.

Sam Levin 00:41:40

One reason we got excited about chickens is the global infrastructure for producing them at scale. A whole generation of synthetic biology companies did incredible work in the lab that either couldn’t scale or faced astronomical costs to figure out how to scale. That killed many companies.

Once we have our chicken, the scale-up problem is solved. We already know how to produce chickens at scale. This makes the commercial opportunity exciting because the hardest technical challenges are upfront in the genetic engineering.

42:48 Skepticism around this new manufacturing approach

Eric 00:42:48

What are the most common questions or points of skepticism you hear as founders of a company that is very contrarian to how the market is moving? How do you present a counterpoint to those views?

Dimi Kellari 00:43:06

The most common one is the idea that steel tanks are more precise. In reality, chickens have very low variability and are extremely precise in how they produce proteins in their eggs. The variability from egg to egg is not high with native proteins, and it is the same with recombinant proteins.

Another point of skepticism is the regulatory market. People ask about the FDA and how they view this manufacturing method. There is important precedent here; a drug produced this way has already gone through FDA approval, which laid the groundwork.

We also have 100 years of data on using chickens for human therapeutics in the form of vaccines. Fertilized eggs are still used for many vaccines today. While it’s a different technology, the infrastructure for rearing chickens in a clean environment suitable for human therapeutic use has existed for years. We follow a specific pharmacopeia for this.

Sam Levin 00:44:47

Sometimes people find it weird because it’s outside their sense of what is viable or possible. We have found that this idea tends to stick with you.

When we first met our president of commercial, he thought we were crazy. He had spent 20 years in the biosimilar space and knew pharma very well. He said it didn’t make any sense. But then the idea eats away at you because, once you see it, it’s hard to unsee. Biologics are proteins, and chickens are great protein factories. Everyone who has had an omelet knows that. People eventually come around.

Dimi Kellari 00:46:02

The pushback often comes from people who are closer to CHO. If you have worked in CHO for your entire career, everything you look at is CHO-shaped. You don’t imagine anything else.

We have had fun convincing people using the arguments we’ve used today, but some won’t be convinced until we show them drugs on the market produced this way. That’s okay. We aren’t here to convince everyone. We are here to manufacture drugs, make them more accessible and resilient, and unlock new biology.

47:00 Applying tech principles to biotech

Daniel 00:47:00

Coming into biotech from tech, what have you observed culturally that is different? Tech people often get skepticism from those in biotech who say they don’t understand how hard biology is. Did you experience that?

Dimi Kellari 00:47:20

I have many observations on this. Biology moves at its own pace. As much as you want to speed it up, a cell divides at the rate it divides. You can either be frustrated by that or accept it and strategize around it.

We have many PhDs in our company who are incredible thinkers, but their default is often the way they operated during their PhD—trying to produce a peer-reviewed publication while being resource-constrained. That leads to a very specific way of formulating hypotheses.

Dimi Kellari 00:48:12

I’m going to run an experiment, gather the data, and then choose what my next experiment will be. We need things to work. We’ll publish findings eventually, but that’s not our primary goal.

Our goal is to move as quickly as possible. Every month we spend or wait is a month of burn. We need to get into a mentality where we don’t necessarily need the most well-formed hypotheses. Even a gut feeling is enough because we can run things in parallel; we’re not as resource-constrained as you might be during a PhD.

Culturally, I’ve been trying to shift the team from a traditional science lab mentality to something more like what you see in tech—a more agile approach, though I hate that word. In the short term, intensity leads to good outcomes, but in the long term, it’s consistency.

Because we’ve been consistent and clear in how we operate over the few months we’ve been around, we’re seeing the returns on that now. Biology can seem like magic. Sven Östrup, our Chief Scientific Officer, is one of the best science communicators out there.

He describes what we do as pipetting water into water. It looks super boring. But at the end of that process, we have genetically engineered chicks that hatch with an insert in their genome allowing them to lay eggs containing medicine. To an outside observer, it’s just water and eggs and then a chicken. It’s a black box of magic, and it’s incredible.

Sam Levin 00:50:30

To toot Dimi’s horn for a second, we went from moving into our lab to generating serious revenue in a year. People often ask how we moved so quickly. A lot of that credit goes to bringing a systems engineering mindset and a non-bio background into a bio team.

We recruited the world’s best scientists, but we ran the operation with a mindset of parallel processing and efficiency. Having built tech before working with Dimi, it’s been awesome to see that. We need to get more non-bio people into biology.

Daniel 00:51:19

What do you think it takes to get more talented non-bio people into biotech?

Dimi Kellari 00:51:26

It’s funny; I was on a bio panel a week ago and three of the four founders did not have biology backgrounds. We also heard a talk from Micha Breakstone, the cellular intelligence guy—an incredible talk—and he also comes from a non-bio background. We’re seeing a lot more of this.

It’s almost like the endgame of tech. You do software and realize it’s powerful, but it’s just pixels on a screen. I’ve always been motivated by physical objects, which is why I worked in hardware before.

People see biology as the ultimate new challenge. If you’re a curious, truth-seeking, and challenge-seeking tech person, you will eventually make your way to biology. You might go the neuroscience route, or lean into synthetic biology and the intersection of AI and bio.

Daniel 00:52:46

Were there other ideas you came across that you weren’t able to pursue because you landed on Neion? Are there any ideas you’d share with tech people who want to build in biotech?

Dimi Kellari 00:52:59

So many ideas. Sam, do you have some off the top of your head?

Sam Levin 00:53:06

There are two big bottlenecks in getting new therapies to patients: manufacturing and regulatory. We’re working on manufacturing, but there’s a bunch of interesting work to be done in speeding up the regulatory process and making it more efficient.

The other area is threat detection and response. As we get better AI tools for biology, the democratization of these tools has a dangerous side. Bad actors could design and create threats. Tools to detect and respond to those threats represent a really interesting space.

Dimi Kellari 00:54:18

We spend a lot of time alongside chickens thinking about other ways to solve manufacturing problems. While chickens are optimal for a broad class of drugs, we have colleagues engineering insects or tobacco plants to do useful things. These generally aren’t therapeutics for humans, but they solve other problems.

Biology allows us to lean into the principle of evolution. Nature has explored so many niches in the design space that there are many existing solutions we can exploit.

The history of drug discovery was essentially drug hunting—finding something in nature that works as a cure. I think we can now do “production hunting”—finding ways in nature to solve manufacturing problems.

55:34 Natural intelligence vs artificial intelligence

Daniel 00:55:34

That’s interesting because the current trend in AI is the opposite approach. We’re creating our own artificial intelligence rather than tapping into the intelligence of nature. I wonder if that will be less effective. Maybe we should be using AI to search for capabilities that already exist in nature and figure out how to tap into them.

Dimi Kellari 00:55:59

We talk about this a lot. The brain in your head runs on the equivalent of cornflakes. Meanwhile, we have to build $30 billion infrastructures to run AI today. Evolution has created something extremely efficient at general intelligence. If we could find a brain in nature that was super-intelligent, that would be great.

Sam Levin 00:56:36

Introducing Eric Dai.

Dimi Kellari 00:56:37

It’s interesting how we do the opposite with computation. It makes sense because it’s easier to produce ones and zeros than it is to produce molecules. You should probably do that in a man-made system. But if the goal is to produce something in the physical world, the solutions evolution has already given us are hard to beat.

57:14 The future of farming drugs & programmable chickens

Eric 00:57:14

Let’s say everything we’re talking about today comes to fruition. What does the future of humanity look like in the next five to ten years?

Dimi Kellari 00:57:30

We want countries to have the sovereignty to manufacture drugs through farming. You will see biosecure farms all over the place producing critical, life-saving medicines. That’s the five-year window.

Beyond that, we want to harness chicken biology for what it’s uniquely good at, leaning into performance advantages to solve things we couldn’t solve previously. There are so many unmet needs, from specific diseases to longevity, that this platform could help address.

Daniel 00:58:37

It could solve those by bringing manufacturing costs down, allowing for more investment in drug development. Are there other things you have in mind?

Dimi Kellari 00:58:50

We are developing high-throughput screening tools that will allow you to generate and test molecules that you might not be able to produce otherwise. We want to pivot away from CHO cells toward things where chickens have a native advantage.

We want to use these tools ourselves and also provide them to others working on novel therapeutics. One of the most frustrating things for a drug developer is having an incredible candidate that shows great results, but being unable to get the quantities needed for animal studies.

Sometimes the yields are too low or the protein simply won’t express in traditional systems. Those candidates are often thrown on the shelf and abandoned. We can tap into those to solve unmet needs.

Sam Levin 01:00:21

It’s not just the final cost of manufacturing; it’s the cost of getting the small quantities needed for testing safety and efficacy. To get the 3 to 5 grams of a biologic needed for an IND application using CHO cells can cost millions of dollars.

There are 6 grams of protein in a single egg. An egg costs 10 cents, and a chicken lays one every day. If you think about what that unlocks on the screening side, it’s a stepwise change in what’s possible.

Daniel 01:01:25

Can a chicken give me an egg with a different molecule in it every day?

Sam Levin 01:01:29

Not yet, but we are starting to engineer the chicken genome to be more modular and swappable. Today, different chickens make different products, but in the future, we want the chicken to be an autonomous, self-replicating 3D printer.

Dimi Kellari 01:01:57

If someone develops a protein drug that is orally available, the egg could even be the delivery vehicle. It’s a sterile vesicle. If you could deliver the drug that way, you could produce biologics at a lower cost than small molecules. There are biological reasons why that’s hard, but you can imagine a world where you crack an egg and that is the medicine.

Daniel 01:02:35

If you drop the carton and break 12 eggs, it’s only $1.20 and you just get a new one.

Dimi Kellari 01:02:41

That’s the moonshot.

Eric 01:02:45

Is there anything else you’d like to share with the audience?

Sam Levin 01:02:47

Our vision of the future is one where the farms that produce the food in our grocery stores also produce the medicines in our hospitals. It’s a simple, almost old-school vision. Getting there requires cutting-edge genetic engineering, but it’s a vision we believe in.

Daniel 01:03:28

It’s an exciting vision. Thank you both for coming onto the Free Radicals podcast and sharing it with us.

Sam Levin 01:03:33

Our pleasure. Thanks for having us.

I sat down with Derya to ask if such bold predictions are too fatalistic. Is the future guaranteed? Shall we just sit back and wait for AI to save us? Derya disagreed.

The golden age for humanity is within our sights, but there is much work to be done.

Academia is failing us. If it doesn’t adapt to the pace of progress AI enables, we have no shot at curing aging this century. And we need to invest trillions into collecting the data that superintelligence will need to solve biology.

If we pull it off, the future can be brighter than anyone imagines.

Derya Unutmaz is a Jackson Laboratory immunologist whose roughly 190 scientific publications have been cited more than 30,000 times.

Special thank you to Adam Gries and Vitalism Foundation for hosting us at Vitalist Bay!

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

0:00 Intro
1:20 AGI arrived early, accelerating the timeline to biosingularity
8:31 Are we really in the takeoff now?
13:15 Superintelligence can help us build a world model of biology
21:13 Aging as the breakdown of biological resilience
26:26 The case for investing trillions into biological data
31:39 The economic return on curing aging
36:26 AI is accelerating science but academics are stuck in the past
42:07 How AI can liberate humanity from biological constraints
49:54 Derya’s Human 2.0 dream
54:44 Testing the latest OpenAI models and rethinking the PhD
1:03:46 Agency, jobs, and free will in an AI world
1:09:41 Science fiction & Derya’s case for a coming golden age

Transcript

1:20 AGI arrived early, accelerating the timeline to biosingularity

Daniel 00:01:20

Derya Unutmaz, welcome to the Free Radicals podcast.

Dr. Derya Unutmaz 00:01:22

Thank you. It’s great to be here.

Daniel 00:01:24

Thank you for joining us. I’m excited to talk about the singularity. You had a website 20 years ago called Biosingularity where you predicted that by 2045 we’d cure aging, and you haven’t changed that timeline. Let’s start with: what is biosingularity?

Dr. Derya Unutmaz 00:01:47

I was very much inspired by Ray Kurzweil. He published the very influential book, The Singularity is Near. I also read about Vernor Vinge, a science fiction writer and mathematics professor who actually proposed this idea and had incredible insight.

Basically, Ray and Vernor were saying that technology is advancing in an accelerated, exponential manner. We knew about that based on Moore’s Law, which was the doubling of transistors in chips every 18 months or two years. That was proposed by Intel co-founder Gordon Moore back in the 1960s.

What Ray showed is that this is not restricted to computers. Throughout history, even biological history going back to single cells and the origin of life, the complexity and knowledge base has been doubling at a certain rate and accelerating. This happened during human civilization as well.

He extrapolated that to the future. Because computer intelligence and speed were increasing, he proposed that AI would eventually reach human intelligence (AGI) by 2029 and move beyond that to superintelligence. I took that and applied it to biology.

I wondered what would happen if all these advances were applied to biological systems. My calculations suggested that by 2035, we should be able to treat most or all diseases. By 2045, we’ll be able to reverse aging.

By 2050, we will reach what I called a true biosingularity, which would be Human 2.0. We will be able to re-engineer our biology, increase our intelligence, and truly engineer biological organisms from scratch.

Twenty years ago, people thought I was totally crazy. However, the timeline is actually looking a bit conservative now. I have moved my date for reversing aging to about 2040 because AGI arrived earlier than expected. That is the story of Biosingularity.

Daniel 00:05:36

Can you say more about the calculation you did? What was the trend you were plotting for biological progress?

Dr. Derya Unutmaz 00:05:43

Biology is extraordinarily complex. After going into biological research, you soon realize the human mind is not able to put all that information together. Every advance we made up to that point was either serendipitous or focused on low-hanging fruit.

After the genome was sequenced, people thought we had decoded life and should know everything. But that was only the very beginning. We then started understanding the incredible complexity of proteins, interactions, and cells.

My thought process was that we had to achieve artificial intelligence or AGI. In fact, we had to get to superintelligence because the human mind cannot process such incredible amounts of multimodal datasets.

I plotted this alongside Ray’s plot and concluded that within five years of reaching AGI and ASI levels, we should be able to solve biology. This assumed that biological technology would accelerate as well.

That is exactly what happened. We now have things like RNA sequencing where we can sequence genes inside a single cell in a high-throughput fashion. The first genome cost $3 billion to sequence; now it costs about $500.

This is part of the exponential progress. We will be able to sequence everyone’s genome and measure all kinds of datasets. But we are going to need computation power to process it.

8:31 Are we really in the takeoff now?

Daniel 00:08:31

I’m fundamentally an optimist, but I’m torn on whether we are currently in the takeoff of the singularity. I see massive progress in AI with coding, but we haven’t seen that progress in biology yet.

Eroom’s Law shows that every drug is costing more and more to discover. We are collecting more data, but it’s not clear that we know what data we need. We are essentially extrapolating that there will be a fundamental change.

What are you looking at that gives you that optimism? How can we be confident that we can collect the right data fast enough? What if the AI tells us we need 50 years of lifespan data?

Dr. Derya Unutmaz 00:09:33

We haven’t reached singularity yet, but I think we’re going to get there much earlier than 2050. The human mind is not able to comprehend exponential increases; we think in linear terms.

There is a fable about a chessboard where an emperor doubles a grain of rice on every square. Not much happens until square 20, but then it becomes billions, trillions, and quadrillions. There is a point where it just takes off, and we are at that cusp right now.

This exponential progress has been going on since the beginning of life. It took a billion years for eukaryotes to happen, but humans evolved in just the last 100,000 years. 99% of our knowledge is only a few hundred years old.

Biology is not as mysterious as people think. It’s extraordinarily predictive. Despite the trillions of reactions happening in your body, if you had the information, you could predict exactly what would happen in the next moment.

If I take a single cell from you and turn it back into an embryonic stem cell, I can recreate you like an identical twin. Nothing goes wrong in that process. Trillions of cells divide and you still have the same eyes, nose, and brain.

That means there is an algorithm in biology. If that wasn’t the case, medicine wouldn’t work. You can give one small molecule and fix things despite millions of simultaneous reactions. It is just a matter of knowledge.

13:15 Superintelligence can help us build a world model of biology

Dr. Derya Unutmaz 00:13:15

What we are missing is a “world model” of biology. It is very similar to robotics. We have AI models that can solve difficult math, but when put in a robot, that robot has difficulty simply sitting down because it lacks physical intelligence.

Physical intelligence is imprinted in our brains; we are born with a model of the physics of the world. We need the same approach for biology. Right now, we are mostly collecting the text equivalent of biology, like the genome and static protein data.

We lack data on functional behavior—how a cell behaves next to a fibroblast versus a tumor cell. That is very contextual and depends on the environment and state of the cell. These functional datasets are not easy to generate in a high-throughput manner.

In my lab, we work with immune cells, and the work is like gourmet cooking; it’s hard to do at scale. However, I believe we will solve this by developing automated, robotic-controlled labs with smart robots that know what to do.

If we have tens of thousands of those, we will be able to generate that data. As for the concern about waiting 50 years, if a system says that, it isn’t superintelligent.

The whole point of superintelligence is the ability to simulate the future based on current data. If you have enough data to predict the future, you don’t need to wait for it. I am still very optimistic.

Daniel 00:19:47

One would think it’s not superintelligence if it needs 50 years of human data. At the end of the day, the information is in the cell; everything emerges from that cell. With sufficient data, even at the level of cells in a dish, one should be able to simulate what happens from there.

Dr. Derya Unutmaz 00:20:15

Think of the physical intelligence that we have. When we are born, we are not trained on every possible thing that can happen in life. There are so many edge cases and unpredictable events, but somehow we are able to survive and predict what will happen next, even with a limited world model.

AI will be able to do that. It doesn’t have to know every single molecule at every single moment. It needs a model where it understands how a cell will behave based on its neighbor and what it knows about that cell.

21:13 Aging as the breakdown of biological resilience

Daniel 00:21:13

This reminds me of an article you wrote about the complexity of biology and how multiple layers interact, with emergence at each layer. Feynman talks about how the laws of physics condense an enormous amount of observations about the physical world. If we compare that to biology, biology has a much lower compression ratio. We can’t come up with laws that encompass all these observations.

I’ve been trying to figure out how to envision the compression ratio for biology. One way of thinking about it is that the code for a human being is in the DNA. That is a lot of compression. Yet, you can’t grow a cell on its own; it has to grow in a uterus with all those signals.

I wonder if aging is encoded in those genes or if it’s the genes plus the whole environment. That is a massive amount of complexity.

Dr. Derya Unutmaz 00:22:27

I view aging as a breakdown of resilience—an information breakdown. People ask why we age and mention the second law of thermodynamics, but that doesn’t strictly apply to biology in the same way. Biology is actually anti-entropic because it has the ability to repair and regenerate itself. It can continue indefinitely.

The information in the genome doesn’t dictate exactly when we will age, but rather when the program stops working as it should or what happens when resilience is lost. Resilience eventually breaks down because of entropy, but biology is able to compensate for 60 to 90 years.

The concept of aging starts immediately. From the moment you’re born, you are exposed to threats and insults. Biology compensates for damage, even if you have a poor lifestyle or suboptimal genes. At some point, it stops doing that. Evolution didn’t care about fixing that problem.

The problem is different if you’re a mouse, a human, or a whale. A whale can live hundreds of years because they were selected in an environment where they could afford to live that long; there were no predators to hunt them. A mouse cannot afford that. It has to reproduce very quickly because it won’t survive long enough to reach old age anyway. Evolution selected for rapid reproduction in those cases.

We have to figure out why that program stops and how we can reconstitute that resilient information. Unfortunately, by the time we solve this, there will be a lot of older people. We also have to figure out how to repair what has been damaged.

It’s going to be much easier for someone in their 30s or 40s. If you find a solution to maintain that state, you never age. But if you’re in your 80s or 90s, we have to worry about regeneration and repair. That’s a tougher problem, but we’ll use biology. We won’t fix things one cell at a time. We will use stem cells to create new skin or a new heart.

26:26 The case for investing trillions into biological data

Daniel 00:26:26

We’re at Vitalist Bay, run by Adam Gries. Adam is known in the community for ringing the alarm bells. He thinks we are not on track to solve aging and that we need a Manhattan Project level of investment into the field.

When you talk about longevity, is your perspective fatalistic in the sense that we’ll all be saved by AI no matter what? Superintelligence might tell us what data to collect, but if we don’t start now, we’ll be years late, and 100,000 people die every day from aging. How do you think about expediting that progress?

Dr. Derya Unutmaz 00:27:33

We still need the data. I agree with Adam that we have to invest to generate that data. If we hadn’t invested in building data centers, we wouldn’t have AI. Companies are putting trillions of dollars into compute.

In biology, if we don’t invest trillions, we’re not going to get the dataset we need. No matter how good the superintelligence is, it will require data as fuel. AI companies are already worried about running out of data on the internet and are trying to create synthetic data.

This is the same in every field, like material science or chemistry. If you want to create new materials, you still have to generate the data. AI and robotics will tremendously accelerate that process, but it requires investment in robotic automation and factories.

The same investment must happen in biology. We need data from the lab and from humans—their behavior, diet, and biology. I have a glucose meter that measures my glucose every five minutes, but imagine measuring 200 things every five minutes for a million people.

That is the urgency. If we have superintelligence but haven’t collected the data, it will still take time because of physical world constraints. We might as well start now. Even now, we have enough compute to handle the data we generate, but as it goes exponential, we need to generate more complex types of data.

The Chan Zuckerberg Initiative wants to sequence one billion single cells. That is a valuable dataset, but it’s almost low-hanging fruit. You want to use organoids or bodyoids to see things happening in a complex way, using imaging and serial transcriptomics to observe cells as they are perturbed by cancer or bacteria. We have to invest in this because every day of delay is 100,000 lives.

31:39 The economic return on curing aging

Daniel 00:31:39

Elon Musk treated every day of delay at Tesla as if he were burning all of the daily revenue the company would have in the future.

Dr. Derya Unutmaz 00:31:50

That’s right. And you reminded me of an important point regarding the economic aspect of this. Even a trillion-dollar investment in biological data generation would have an enormous return on investment.

Dr. Derya Unutmaz 00:32:13

We are spending $4 trillion every year in the US alone on healthcare. If we solve aging and address these problems before they happen, we could save $3 trillion. Investing a trillion to save three trillion would solve the US budget problem and pay off the debt.

Every year you extend the human lifespan, the monetary benefit is in the hundreds of billions, or even trillions. Adam showed slides illustrating that in the last three years of life, hundreds of billions of dollars are spent just to keep someone alive for an extra month. The cost-benefit is tremendous. Of course, keeping your loved ones alive is priceless.

Daniel 00:33:23

There are very good economic models for this. Raiany Romanni was on the podcast a few months ago and talked about the trillions of dollars in value generated from delaying the decline of aging by just one year.

The upside is virtually infinite. It’s about keeping us alive and giving humanity a future where we can explore the universe and achieve amazing things.

Dr. Derya Unutmaz 00:34:06

Elon Musk is developing self-driving cars that are already five to ten times safer than human drivers and will eventually be 100 times safer. 50,000 people die from accidents annually.

Dr. Derya Unutmaz 00:34:26

Having that AI available one year earlier saves 50,000 lives and prevents millions of accidents. Solving aging and disease is that same impact multiplied by a thousand.

Daniel 00:34:43

I did an analysis with ChatGPT where I asked about the baseline hazard ratio for a 22-year-old. Without aging, the average man would live to be about 350, and women would live to 450 or 500 because they die less frequently from accidents and violence.

Once we solve aging, we’ll have to crack down even more on all these other sources of death. If we have virtually indefinite lifespans, every death becomes an even greater tragedy. Imagine losing someone who could have had millions of years of life.

Dr. Derya Unutmaz 00:35:41

If you solve all diseases and aging, calculations suggest you could make it to 1,000 or 1,500 years old based on accident risks. I’ll take that. Think about the singularity. We are talking about what will happen in the next 30 years. If you extend your life another 100 years, nothing will be impossible beyond the laws of physics.

At that point, there could be mind uploading or other things that seem like science fiction today.

36:26 AI is accelerating science but academics are stuck in the past

Daniel 00:36:26

Singularity is such fun stuff to talk about, and it’s clearly much more real now than it seemed a decade ago. What’s it like for you being in academia and talking about these things? It’s not a traditional topic for an academic to discuss.

Dr. Derya Unutmaz 00:36:56

I’ve always felt like a stranger in a strange world. That was the title of a famous science fiction book by Robert Heinlein, Stranger in a Strange Land, about a man who comes from Mars and feels like a stranger on Earth. For most of my life, that is how I have felt.

Academia is extremely traditional. In fact, the reason I didn’t continue in clinical medicine is because it is even more traditional. You cannot think differently there. I thought that by doing research, I could come up with crazy ideas. I was able to do that partially because you are experimenting, but there is a limit.

Society will say a proposal is too crazy. You cannot propose ideas if your grants won’t get funded, or your peers will say you aren’t serious and you won’t be able to publish papers. But I have never hesitated to talk about these things.

My colleagues and friends know I have fringe ideas, and it turns out they aren’t that fringe anymore. Up until a couple of years ago, I would tell everyone they had to get into AI. They would argue that it hallucinates, but I knew that whatever AI cannot do today, it will be able to do in a year or two.

Even now, many of my colleagues are very skeptical. It felt like being part of a small niche group. I was happy to find the singularity community. I have a signed copy of Ray Kurzweil’s book, which is one of my most precious possessions.

I’ve been waiting twenty years for this day. Sometimes you wonder if you’re totally wrong or if it’s all an illusion, but I never lost my conviction. Recently, it’s getting more difficult because of AI.

Daniel 00:39:33

In what way?

Dr. Derya Unutmaz 00:39:36

The point I keep making to my colleagues is that the way we do science has to completely change. We cannot pretend we’re going to keep doing experiments every month, analyze the data for five months, and then write a paper and apply for a grant every five years.

If we continue with that cycle, we have zero chance of solving disease and aging for centuries. We have to adapt to the incredible speed of AI. Analysis that used to take six months now takes minutes. I’ve experienced this myself.

I took data that literally took us six months to analyze and ran it through the latest models from ChatGPT. It did it in minutes and provided better data. We contracted six months into minutes. The same applies to doing experiments and forming hypotheses.

AI is just better than me. I have thirty-five years of experience in immunology, I’ve published extensively, and I have thousands of citations, but I see that the model is smarter than me. If I refuse to accept that there is a non-biological entity that is more intelligent and just better, it’s not going to work out.

I see many of my colleagues struggling to accept this. We need to embrace it. If you do, you become 10x or 100x more productive. It’s the same with software engineering; if you refuse to use AI in coding today, you have zero chance of surviving. Academia is moving very slowly, and it worries me that they aren’t adapting as fast as they should.

42:07 How AI can liberate humanity from biological constraints

Daniel 00:42:07

Where do you think that transition lands in terms of the human role in science? Those who use AI will be far more productive, just like software engineers. But at a certain point, will the superintelligence just be telling us what to do? Are we just meatbags in the physical world moving dishes around in the lab?

Dr. Derya Unutmaz 00:42:29

I don’t know what a superintelligence is going to do because, by definition, superintelligence is beyond our understanding. However, I think humans have an extraordinarily important role because all of this is for us.

AI isn’t trying to cure diseases so that it can live longer. We are trying to cure diseases and aging for ourselves and other humans. We are the directors. We have to have the agency to push AI in the direction we want.

Left on its own, a superintelligence might decide to figure out how stars work or start building a new planet. It might decide to leave Earth entirely because there is plenty of energy in space. That is why it is vital for people to embrace this.

When we get to Human 2.0, it’s not as if humans will stay at the same level of intelligence. Biology is so amazing that we could be thousands of times more intelligent. We will have capabilities we cannot even imagine.

We will be able to merge with AI. With technologies like Neuralink providing an instantaneous flow of information, our capabilities will increase. People say AI will only benefit a small portion of humanity, but it is actually the greatest distributor of intelligence.

Someone in a village will have the same tutor or doctor as the richest person in San Francisco. Everyone will be able to amplify their intelligence in the same way. I see AI as a collaborator and a partner that will help us move far beyond our current capabilities.

Daniel 00:45:38

When thinking about the transition to Human 2.0, how much have you thought about what that would really look like? There might be some splintering in humanity; some people may not choose to get augmentations. But there are also so many biological options. Humanity won’t turn into just one thing; it could turn into all kinds of interesting things over time.

Dr. Derya Unutmaz 00:46:15

It certainly could. Look at the diversity in the human population right now. There are still humans living in the Amazon who don’t want civilization and live as they did 10,000 years ago. There is an incredible range of intelligence and ability.

Currently, most of that is a lottery ticket. You don’t get to pick the genes you are born with, whether it’s a disease gene or certain physical capabilities. I am never going to be a great basketball player because I don’t have the height. Others might never be great chess players because they lack the computational power.

What if all humans have the choice to be what they want? You could choose to be a great sportsman, or to be radiation-resistant so you can survive in space. That would be incredible. Some might choose not to take those options, and that is perfectly fine.

Some people might choose not to reverse their aging and live a natural life. They already have that choice. But those who want to live longer currently do not. That is the point of Human 2.0. It doesn’t mean everyone is forced to be engineered in a certain way.

I would like to have my immune system completely reset and re-engineered. Having studied the immune system, I see it as a legacy system with many problems. I want Human Immune System 2.0. Others might be happy with what they have.

It’s about having a choice. Right now, you don’t choose where you are born. You could be born in San Francisco or in a poor environment in Bangladesh. You might be smart and hardworking but lack options. AI is giving people those options.

I termed this concept bioprogressivism. It’s about democratizing biology so everyone has the same access. In sports, you can’t take steroids to enhance yourself, but the reality is that many athletes are naturally doping because they are biologically built to compete. Having a choice is always better than not having one.

49:54 Derya’s Human 2.0 dream

Daniel 00:49:54

I completely agree. Part of the tagline of this podcast is giving humanity control over biology. Ultimately, it’s about giving people agency, not just over their health in a narrow sense, but over their entire life.

Do you have a personal dream for what you want to have when you are Human 2.0, beyond the immune system?

Dr. Derya Unutmaz 00:50:16

Since I was eight or nine years old, my dream has been to go to space. Ever since I watched Star Trek, I’ve been waiting for this world. I want to be resistant so that I can travel long distances in space, whether that means protection against radiation or general survivability.

Daniel 00:50:48

Similarly, what got me interested in aging biology when I was in high school was science fiction. I really wanted to meet aliens one day. Part of my interest in biology is wondering what an alien life form would be like.

It could take thousands of years to make a journey into space. That’s when I discovered Aubrey de Grey and realized people were working on lifespan extension. I lost that dream for a while when I went to college and entered the working world, as it’s hard to keep thinking about those things.

What has maintained your enthusiasm for that dream over the last few decades? From my perspective, it was hard to see the exponentials in AI during the AI winter.

Dr. Derya Unutmaz 00:51:42

I’ve always been very interested in science fiction. Reading those books makes you realize there could be an alternative world. It sounds like fiction, but then you realize it could actually happen.

I was also extremely interested in technology. At fourteen, I started coding. Computers then had 48 kilobytes of RAM—less compute than our watches today—but it was exciting. I could see that it was going to get better every year.

When the internet started, I made one of the first 1,000 websites on the net in 1994 when Netscape came out. Before that, we had BBSes that we connected to with modems. Suddenly, you could connect so fast.

Being in biological research, you realize that thirty years ago, we couldn’t dream of what we are doing today, like engineering and changing genes. In the late ‘90s, I was involved in developing some of the first lentiviral gene editing vectors derived from HIV.

We didn’t imagine then that they would someday be used to cure cancer. Now we can program immune cells and use CRISPR. Technology was advancing, and I was constantly dreaming of these plots. I just told myself to be patient because we were going to get there.

When you see it finally happening, the excitement is unexplainable. That’s what people don’t always get. I’m always thinking one, two, or three years ahead.

54:44 Testing the latest OpenAI models and rethinking the PhD

Daniel 00:54:44

It’s much easier now. It’s amazing what’s happening, though you were expecting it.

Dr. Derya Unutmaz 00:54:54

Exactly. Technology was always advancing, but the speed was slower. Every two years, something great would happen. You’d wait a year for the next Apple iPhone. Even after the first ChatGPT, we waited a year for the next major leap.

The difference now is that we are approaching the singularity. Every month something amazing happens, and soon it will be every week. You can feel this acceleration in almost real time. It is an incredibly exciting time.

Daniel 00:55:41

I think you’re an early tester of OpenAI’s models. Can you tell me about that experience and your impression of the latest models?

Dr. Derya Unutmaz 00:55:51

I started early testing with o1-preview in late September 2024. They reached out to me, realizing they needed to find some crazy person. It’s been great. I’ve already tested GPT-5 Pro models before they came out.

The latest model for me was another turning point, the 5.5. I was already testing that. I’m a total addict of Codex vibe coding.

I can’t talk about other things that are going to come out, but people can predict that it’s going to keep getting better. There’s no limit. There’s no ceiling. People say deep learning is going to hit a wall. No such thing. The next couple of months are going to be crazier than the last year.

Daniel 00:56:56

Do you think young people interested in science should still go into PhD programs, or should they just be vibe coding?

Dr. Derya Unutmaz 00:57:15

I’ve been controversial about that. Last year I said half of the PhD programs in biology should be cut immediately. I do not recommend going into a PhD unless you are passionate about it and love doing it for its own sake.

That’s what I did. I love doing science. You can do that if you don’t need to take care of your family immediately. However, if you’re doing it as a career, that is foolish.

Right now, biology PhDs take four to six years. Can you imagine how life will be five years from now? Biology is going to be moving at light speed.

I’ve had 12 or 13 PhD students over my career. What they did in their PhD for five years, I can now do in five to seven months. In five years, I’ll be able to do it in one or two months. Spending five years is meaningless.

Institutions have to realize this. We shouldn’t train so many PhDs. We should be selective and only take the best of the best. We should reduce the time from five years to two years.

It should be a pure apprenticeship. You need to work in the lab, have day-to-day interactions, and learn how to use a pipette. Your mentor is going to be AI anyway; it will direct you. They won’t even need me anymore.

The same thing applies to medicine. There’s no point in memorizing all that information. You need to contract that time and start seeing patients. Two months after getting into medical school, you should start seeing patients because that’s how you learn. AI can explain everything in real time. Most people don’t understand that yet.

Daniel 00:59:59

Is anybody setting up these accelerated apprenticeship programs?

Dr. Derya Unutmaz 01:00:05

Not that I’m aware of. This isn’t just for PhDs; it should be for master’s degrees or college. Why are you going to college for four years? It’s crazy. Go for one year because you need to interact with people, network, and experience that life.

Everything that can be taught in four years can be taught in four months with AI. I guarantee you. I saw in the news that applications for some MBA programs dropped by 50 percent, so they cut the tuition by 50 percent.

That’s not enough. We should cut the tuition by 90 percent and reduce the program to six months. Otherwise, it’s a waste of money and time.

Daniel 01:01:01

A few months ago, a 22-year-old senior in college in Boston called me. He listens to the podcast and said he had a few months left of school, but he wanted to leave and go to San Francisco to work with whoever’s building the future there. I was torn on that. On the one hand, with only a few months left, you might as well finish. That’s what common sense dictates. Everyone often feels like they’re missing out on the next big thing. But crazy things are happening, and a few months in Silicon Valley is what used to be a decade.

Dr. Derya Unutmaz 01:01:42

If it’s only a few months, I wouldn’t leave. Finish it because you’ve invested the time. You should have something to show for it unless you have offers or amazing ideas to start a company.

He’s right that this is moving fast. The opportunities available if you are AI-pilled are exponentially increasing. People shouldn’t worry about a few months because opportunities will keep growing as model capabilities grow.

I’m building complex biological software. We used to pay thousands of dollars every year in subscriptions; now I’m making a better version. I wouldn’t have been able to do that two months ago. This is a completely new capability.

In a couple of months, there will be more capabilities and more you can do. Eventually, that will reach a peak. After that, there are only so many opportunities available, and you don’t want to miss that peak. You don’t want to wait a year or two. You want to get in.

The same applies to software engineers. People are worried they can’t find a job. Why worry? If you’re a software engineer, sit at your laptop and start vibe coding. You don’t even have to write the code.

There are a million problems you can solve. Take five or ten of them and start your own company. It costs nothing now. Work on that.

1:03:46 Agency, jobs, and free will in an AI world

Daniel 01:03:46

Very often the ability to code wasn’t necessarily a bottleneck, even when software engineers were expensive. It’s a lot harder to know what to build.

When we think about unique human contributions, like agency and choosing what to do, most software engineers aren’t equipped to figure out something valuable to build on their own.

Dr. Derya Unutmaz 01:04:15

That’s not a limiting factor. Ask ChatGPT what to build?

It already has memory of you. It knows. I ask it what I should do today or what questions I should ask. I tell it to give me ten things to do. I probably won’t care about five of them, and I already knew three. But then two of them are things I hadn’t thought about.

That is not a limiting factor anymore. Agency and curiosity are the real limiting factors. If you don’t have agency and you just want to sit at a desk and have somebody tell you what to do, an AI agent can do that. That’s what AI agents are. You give them a task and they do it.

If you live like that, it’s not going to be a great life. You have to have that agency and realize you have superpowers. Ask yourself what you can do. If you don’t have ideas, ask ChatGPT what you should build today. Honestly, it will give you millions of ideas.

Think of it this way: the concept of a job has to change. We need to think in terms of solving problems or tasks. Anyone with certain abilities should be able to be flexible and solve different types of problems. You could be a nurse solving a complex medical problem like a doctor, or a lawyer. Or an auto mechanic who can repair robots with the help of AI. It’s almost unlimited, but you need the curiosity to learn with AI and the agency to apply what you do to real life.

Daniel 01:06:27

Do you think we’ll be able to biologically augment people’s level of agency? Or is that fundamentally free will?

Dr. Derya Unutmaz 01:06:35

In my opinion, there is no such thing as free will.

Daniel 01:06:39

That’s interesting. Say more.

Dr. Derya Unutmaz 01:06:40

Free will is an illusion. We think we have free will, but everything is based on our genes and our environment. The next action we take is like predicting the next word; it’s totally dependent on our biology, our history, and our surroundings.

We are pushed toward doing things we think we are choosing. You choose a certain path over another not because of a random event, but because of how your brain has been conditioned to pick that pathway. That’s a long conversation, though. What was the other question?

Daniel 01:07:36

One point on that: it comes back to a question I asked at the very beginning. Isn’t the whole singularity concept too fatalistic? It feels predetermined.

Dr. Derya Unutmaz 01:07:50

Self-determination is still undeterministic. There are too many things involved. We think we have this agency in one small area of the brain that controls everything, but there’s a lot of information coming to that agent, and decisions are made based on that information. If that information were different, you would make a different decision.

Going back to your question, we will be able to give people agency with biology because that’s a biological feature. We don’t know exactly why some people have more agency than others, but it may be that a certain neurotransmitter in their brain works better.

Biology explains why some people sleep less or why some are more prone to addiction. It’s why some people, like me, have ADHD and others don’t. It’s all biological.

Some people might choose not to have agency. I’m happy to have ADHD because it allows me to think of a million things, but for other people, it’s tiring because they can’t focus. That’s the point of Human 2.0. People will have that choice.

1:09:41 Science fiction & Derya’s case for a coming golden age

Daniel 01:09:41

We’re getting close to the end of the conversation. Unless there’s something else you want to talk about, it would be interesting to end on the sci-fi that most inspires you.

Dr. Derya Unutmaz 01:09:53

I like hard sci-fi, not fantasy. I’m a fan of Star Wars, which is a bit of a fantasy, but I prefer alternative worlds and lives.

I read a lot of Isaac Asimov. The Foundation series had an incredible influence on me. It was a parallel universe.

Daniel 01:10:23

I’m very familiar with the Foundation series; I just reread it and loved it. One thing I disliked is that Asimov is very anti-longevity.

Dr. Derya Unutmaz 01:10:34

It was his imagination. There is an alternative world where longevity is not looked upon. Perhaps humans will eventually decide that living long is not attractive.

I love the fact that you can create these worlds. I’m a fan of Star Trek and the idea of meeting aliens like the Klingons. I’m also a huge fan of Arthur C. Clarke, who wrote hard science fiction. He predicted satellites in his books.

Rendezvous with Rama is a book I still love. It’s about aliens who leave signs you have to figure out. That is the kind of science fiction I enjoy.

One book that influenced me is Snow Crash by Neal Stephenson, published in 1991. That book introduced the term Metaverse. The story is about an internet or Metaverse where humans live, and Snow Crash was a mind virus that could infect you. Many people, including Mark Zuckerberg, were influenced by it.

Daniel 01:12:20

Awesome. Anything else you want to leave our audience with?

Dr. Derya Unutmaz 01:12:23

The last thing I would say is that I know many people are anxious or afraid. That is understandable because humans don’t like uncertainty, and we’re living in the most uncertain time ever. We sometimes can’t even predict what will happen next month.

This is also the most exciting time to be alive because we now have a path to the golden age. We can see the light at the end of the tunnel. We didn’t have that five years ago. When I claimed we were going to solve aging, people asked how we would do that. It was science fiction.

Now we see it’s happening. Everything is changing. When we reach the golden age, it won’t just be about eliminating disease and reversing aging. It will be about giving people choices regarding their biology and entering the age of abundance.

It’s hard for people to understand because humans evolved in scarcity. For millions of years, we had to hunt to survive the next day. We are moving beyond that. With robotics and AI, there won’t be any problem producing anything we want at an incredibly low price or even for free.

We can focus on what’s important in our lives: human interactions and the love we have for our children, parents, and friends. Those are the things humans should value. People ask what the meaning of life will be without a job. I doubt everyone wakes up happy to go to work.

“This gives me so much meaning.” There are very few people like that, but meaning comes from many things.

It’s not going to be an easy road over the next five to ten years. We’re going to have some suffering, but it will be worth it. It will be worth it for our loved ones, ourselves, and our children to have an amazing world.

So keep being positive. It also extends your life, by the way.

Daniel 01:15:20

Derya Unutmaz, thank you so much for joining us on the Free Radicals podcast.

Dr. Derya Unutmaz 01:15:25

Thank you. It was great.

Daniel 01:15:27

Thank you for listening to this episode of the Free Radicals podcast. If you enjoyed this episode and would like to support us, the most helpful thing you can do is share this with a friend you think might enjoy it too.

Please also leave us a five-star review on Spotify and Apple Podcasts, and like and subscribe on YouTube. It would really mean a lot.

I’m Daniel Shur, and my co-host is Eric Dai. Thanks for listening.

Aging kills 85% of Americans, and the US spends $15-20 per person per year on cancer research but only $1 on aging biology. No amount of zone 2 and sauna are going to save you from that gap.

I sat down with Andrew Steele, scientist, author of Ageless, and founder of the Longevity Initiative, to talk about closing it. His position: if he had $10M to spend on his own longevity, he’d spend every dollar on policy and communication, because that’s where the leverage is.

With health systems pushed to their limits and demographic crises looming, the time is right to organize against aging.

Andrew is a phenomenal communicator, and played a huge role in inspiring me to start this podcast. You’re gonna love listening to him, and I highly recommend you follow him as well.

Check out Andrew’s work at the links below:

• The Longevity Initiative

• Andrew’s Book Ageless

• Andrew’s Personal Site

Special thank you to Adam Gries and Vitalism Foundation for hosting us at Vitalist Bay!

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

1:03 Andrew’s path into aging biology & activism, and publishing Ageless
8:28 Aging as the biggest driver of disease
12:10 The policy gap in addressing aging
20:57 How Andrew hopes his work will help us defeat aging
28:05 The case for increasing funding for aging research
37:12 Bryan Johnson & personal health optimization vs biotech progress
42:37 Why health podcasts are such a nerd snipe
49:27 Who will fund the solution for aging?
57:50 The Longevity Initiative
1:03:50 Andrew’s experience as a science communicator
1:09:53 How to measure progress in activism against aging
1:15:07 Will we solve aging?

Transcript

1:03 Andrew’s path into aging biology & activism, and publishing Ageless

Daniel 00:01:03

Dr. Andrew Steele, welcome to the Free Radicals Podcast.

Andrew Steele 00:01:06

It’s amazing to be here. Thanks for having me.

Daniel 00:01:07

Thank you so much for joining us. When I reached out to you, I mentioned that your book Ageless, which I read a couple of years ago, was a big part of my introduction to the longevity field.

It was really influential in giving me a good understanding of what’s going on. It inspired me to enter the field and start this podcast, so thank you for that.

Andrew Steele 00:01:27

Thank you so much. That’s just amazing to hear. One of the main reasons I wrote Ageless is that I took a slightly roundabout route into longevity.

I wanted it to be the book I wish I had been able to read when I first got into the field. I wanted to provide a broad overview and show what’s going on with realistic optimism.

I also wanted to present the evidence. If you’re a scientist or have a technical background, this stuff can often sound like snake oil, so I wanted to provide a scientifically rigorous overview. I hope it can help a few more people along on that journey.

Daniel 00:01:58

I love that phrase, “realistic optimism.” That has been central to my thinking with this podcast. I first learned about longevity in high school when I read Aubrey de Grey’s book Ending Aging and got really excited.

Then I went through a “valley of despair” for about ten years where I didn’t see much progress in the field and eventually forgot about it. Reading your book caught me up on what had been happening.

I felt that when I did see optimism, it wasn’t necessarily realistic. I would hear about the tech singularity, which is hard to make feel real. My goal with the podcast is to have conversations with the scientists who are actually doing the work to understand where things really stand.

Andrew Steele 00:02:41

Definitely. The challenge is that we are at a point in the field where we have many small grounds for optimism. We have numerous examples of things that extend lifespan in mice and some suggestive data in humans, but we aren’t yet at the point where we can rest on our laurels and assume the problem is solved.

This creates a lot of leverage. It’s an exciting time to try and make a difference. We face two extremes. One is extreme pessimism: the idea that aging is an impossibly complicated problem we can never hope to solve because biology is too hard and our current tools lack the necessary resolution.

The other extreme is the overly optimistic view that AI and AGI will simply solve biology in the next few years. I don’t think that’s true either. The truth lies somewhere in between.

Our actions now can have a significant effect on whether we succeed. While luck always plays a role, this is a time when we can truly try to do something, and that is what I am focused on.

Daniel 00:03:47

We’re going to dig into where the field is currently and the things that will give us leverage. Before we get into that, could you walk me through how you ended up getting into this field?

Andrew Steele 00:03:57

I did a physics PhD and was quite deep in academic physics before I decided to study aging. At the end of my PhD, I was looking for how to make the biggest impact on the world.

I nearly became a climate physicist, which would have been a more logical use of my skills. However, I started reading about aging biology and realized the case for a single underlying cause for many diseases was obvious.

It didn’t take much convincing. I wondered why no one was working on it. I also knew physicists have a tendency to barge into new fields and oversimplify things, so I decided to work as a biologist first to see if I was missing anything fundamental.

I became a computational biologist, which was the easiest path for a physicist. I learned a huge amount of biology and worked with talented researchers. Often, I knew more about longevity than they did simply from reading a few books and papers.

It was mind-blowing that one could complete a biology degree without a single lecture on aging, which is a universal biological process. My wife, who is a medical doctor, thought I was crazy when I first talked about medical treatments for aging.

She will likely be prescribing these drugs during her career, yet there was nothing about it in her textbooks. While doctors can’t prescribe these yet without better evidence, they should understand what a geroprotector is and how it might be used.

I want doctors to be skeptical and careful, but if they aren’t briefed on aging biology, they may dismiss it as hogwash. There is a need for realistic optimism in this field.

I realized that biologists, doctors, and the general public often hadn’t heard of these concepts or associated them solely with supplements. I decided to write Ageless to have a bigger impact than I could as a lab scientist.

I spent two years learning more about physiology and cell biology to ensure the book was scientifically rigorous. I wanted it to be a source of truth that doctors and scientists could trust.

Since the book’s release in early 2021, I’ve been communicating these ideas through writing and media. I now believe policy communication is the highest-leverage area for this field.

Longevity is often misunderstood. Even at Vitalist Bay, where everyone is working on these issues, we are in a bubble. In the real world, many people still haven’t heard of these ideas, or they associate them only with fitness and diet.

We have a massive communications task ahead for scientists, doctors, policymakers, and the public. I’ve started a nonprofit called the Longevity Initiative to catalyze the science and get these ideas into the world.

8:28 Aging as the biggest driver of disease

Daniel 00:08:28

Even here at Vitalist Bay, there is a lot of debate about what longevity is. When you say you want to get this idea out into the world, what is that core idea?

There are different ways of framing it: the geroscience hypothesis of intervening in aging biology, the creation of preventative drugs, or the idea of curing all disease. What do you think is the key idea to communicate?

Andrew Steele 00:08:56

Aging is a fundamental driver of disease. By intervening in that process and developing medical treatments, we can create preventative medicines.

The key for me is that this is medical. The biggest misconception is that longevity is all about supplements, diet, and optimizing zone 2 or zone 5 training. I am happy to nerd out about those things as much as the next person at Vitalist Bay, but we can get distracted by them.

The general public often doesn’t know there is anything beyond lifestyle factors. We need to focus on the biomedical side. We need to understand aging in the lab and develop treatments that intervene in those processes.

My dream for the future is that this becomes a routine part of medical care. Just as a doctor measures your cholesterol and prescribes drugs if diet changes don’t work, they will do that with your senescent cells or telomeres.

You will take a treatment every night before you brush your teeth, and it will keep you healthy without much thought. It will just become the background of our lives. We won’t have to worry about health as much because the diseases of aging will be substantially deferred.

Daniel 00:10:18

I completely agree with that future. Is the key insight the recognition of aging as the underlying process to intervene in?

I dig into this because some people in the community think GLP-1s are longevity drugs. While they extend life for certain people, I don’t see them as tied to aging biology.

We are trying to grow a field, so it is important to define what that field actually is.

Andrew Steele 00:10:50

That is a major challenge. Cancer research receives more investment partly because we have a clearer idea of what cancer is. It is a much tighter definition.

I could be convinced that GLP-1s act on fundamental aging mechanisms because aging is a broad web. Some of it is metabolic and some is inflammatory. GLP-1s might intervene there, though whether those are the best places to intervene is another debate.

It is plausible that GLP-1s are a first-generation anti-aging intervention. The challenge is uniting people around defeating something so hard to define. We don’t have to fully understand something to attempt to treat it.

Longevity, aging, and healthspan are slippery, complicated concepts. While we don’t have perfect definitions for cancer or heart disease, they are much more defined than aging.

12:10 The policy gap in addressing aging

Daniel 00:12:10

Tell us more about the Longevity Initiative and the key things you are trying to push forward.

Andrew Steele 00:12:17

I have spoken to policymakers, investors, and the public since Ageless came out. I’ve found that my first few slides are identical regardless of the audience.

Almost everyone needs to be brought up to speed on the geroscience hypothesis. This is the idea that aging is a fundamental ticking clock in our biology that drives the risk of various diseases.

We can intervene in that process. Senolytics are a great example: a bad thing builds up with age, you clear it out, and outcomes improve in mice. People need exposure to that narrative.

If I had fifteen minutes with a world leader, I could make a convincing case. However, that leader would then refer me to their advisors to ask how to actually implement the plan.

We would need to discuss how to deploy research funding and how to ensure it targets aging. We must consider what it means for healthcare systems, pensions, and medical care.

We also need to consider the private sector. How do we incentivize R&D? Should we use prizes or change the patent system? Should aging be classified as a disease?

The field debates these questions, but we lack concrete policy papers to give to ministers or senators. Without those resources, it is hard for politicians to implement changes.

We need to create those underlying policies. Healthcare systems are under increasing pressure, and every country is facing demographic issues due to birth rates and aging populations.

Daniel 00:15:00

Yeah.

Andrew Steele 00:15:01

There will be a time when people reach for longevity science, but we have to make sure we are prepared for that moment. We cannot be caught off guard when the spotlight is suddenly on us.

That is part of the preparation we need to do. The aim of the Longevity Initiative is to set that narrative while creating concrete policies and ideas. We want to allow the science we all want to see to be implemented at the biggest possible scale.

Daniel 00:15:26

Which do you see as your main priority? Is it spreading public awareness of aging, or is it the more detailed public policy and planning piece?

Andrew Steele 00:15:37

Those two things fit together. Since writing my book, I have realized that media work and policy work are not as divorced as they might seem. Policy is serious and dry, while the media is flashy and requires a simplified message, but having a book has opened doors that would not have been open to me if I were just Andrew Steele, the scientist.

Scientific credentials help, but having a book or a TV series is a huge credential for politicians and high-net-worth people. It helps when getting in front of audiences for talks. A huge part of this is essentially marketing for the field.

Being seen as a credible voice for this field is vital. If someone has a question about longevity, I want the Longevity Initiative to be the place they go. Journalists and researchers often find websites through random searches. I want them to find our website when they hear about rapamycin or a new supplement, and then go down the longevity rabbit hole.

The theory of change is broad because we need a social shift for policies to be implemented. I have done campaigning around science funding in the UK in the past. I had a naive idea about this at the end of my PhD when I started something called Sciencegram.

The idea was to contextualize science funding. I realized the amount we spend on science per capita is minuscule compared to the scale of the problems science is trying to solve. For example, cancer kills about a quarter to a third of people. In the UK, we spend £2.80 per person per year on publicly funded cancer research.

In the US, it is about $16 to $18 per person per year. Even though that is more, you would be hard-pressed to have a meal in San Francisco for $20. As a US taxpayer, that is your average annual contribution to a disease that is very likely to kill you.

I thought that if I just told people how crazy these numbers were, it would work. But putting data in front of the world does not automatically lead to social and political change.

I would love to find the one wonk in government responsible for turning the science funding knob and have them turn it up for aging. That would be a surgical strike. Unfortunately, there is no single person doing a rational cost-benefit analysis to determine government spending.

There is a network of incentives at play, and the most fundamental is that politicians respond to voters. We do not have to convince every single voter. This probably won’t be the top issue on the ballot for most people.

However, if a politician gets the impression that a topic is popular, they are much more likely to act. If they believe a significant fraction of the population will respond positively to an announcement on longevity research funding, that filters down to the people implementing the policies.

There is no clean way to just convince five people and change the world. Even on the private side, billionaires are just regular people who happen to have more than a billion dollars. They get their news from similar places and have the same misunderstandings about longevity as everyone else.

By getting this information out there in an audience-agnostic way, you sow seeds in many different places. Once those ideas are established, you can strike with policy work or philanthropic advice.

We can help investors identify companies that are actually likely to move the needle on longevity, rather than just following the hype around things like GLP-1s. It is frustratingly complicated.

It feels like there should be an incredibly precise theory of change where you target specific groups. Science communication experts might ask who the target audience is, but universal communication is remarkably important to moving this field forward.

20:57 How Andrew hopes his work will help us defeat aging

Daniel 00:20:57

These are exactly the issues I’ve been grappling with. We’ve discussed the motivation and goal of the podcast. Initially, we’re trying to carve out a niche within the longevity biotech space. We want to provide a forum for those dialogues and flesh out these ideas before slowly expanding.

My initial thinking was that if we can reach people like the Collisons and adjacent tech billionaires who are interested in this field, we can expand from there. However, these theories of change are very vague. It’s challenging because we almost have to influence the entire world, or at least one percent of the population.

I’ve debated this point with Adam Gries, who has given several talks about theories of change at this conference. He generally pushes for a specific path to change—a clear chain that leads to a solution. My belief is that we just need to spread the right ideas to the right people, and that leads to change.

I am sympathetic to the idea that there’s a lot of hope baked into that equation. How high is your confidence that this “ideas first” approach can actually work?

Andrew Steele 00:22:21

I think it’s anathema to people who have worked in tech that the theory of change is so diffuse. Once you’ve done some campaigning, you realize just how frustrating and piecemeal it can be. There isn’t always a clean story to tell, and it’s unlikely that any one actor will be the single person to convince one percent of the population.

I want to be part of a broader movement. I have a certain voice and will only be accessible to a specific subset of people who find me compelling. Other influencers will be much better at reaching their own niches.

It’s important not to go completely general. You don’t need every human being on the planet to be convinced that longevity is important before we can start making a change. It’s worth considering who the key decision-makers are. Are they people with capital to deploy? Are they in the public sector? Are they philanthropists?

Unfortunately, you can’t just call up fifteen billionaires and solve everything. It is a frustratingly diffuse theory of change. It’s important to articulate who you want to reach and how you plan to get there, rather than just saying it’s too vague to define. It sits somewhere between throwing things at the wall and having a rigid plan.

You can have the greatest theory of change ever, but if you put it on the internet and nobody cares, it doesn’t matter. There is real value in niche platforms like your podcast. While I hope my book is broadly accessible, if it moves even a small number of people into the field, they are a vital part of the readership.

You have to produce content with multiple audiences in mind. You might favor one, but you can’t neglect the broader audience. You never know who will hear a given piece of content. You have to say yes to many things, see what happens, and try to learn what was valuable in the long term.

It’s difficult because one talk can lead to an invitation, which leads to a podcast appearance. It would be great to have a media strategist, and that’s why I want to build an organization around myself. Currently, what I’m doing is highly serendipitous. A media strategy can only do so much; the rest is often just luck.

Daniel 00:25:01

It’s hard to know what will lead to what. With the podcast, we have small wins every week, like when someone outside the field reaches out because of an episode.

This inspires me to think about the metrics I could set. It doesn’t have to be vague. I could track the number of people who entered the field because of this work. As I iterate and grow the audience, I can see if that number increases.

There are ways to consider particular demographics and the impact on each of them. You can take this idea-based theory of change and make it more practical and real.

Andrew Steele 00:25:42

It is very tempting to optimize for the metrics you can measure. Is getting people into the field the most important thing we can do? At the end of the day, it is a relatively small pipeline.

If we were to have a much broader social acceptance of this, we could address the key constraint, which is funding. There are only so many people it is possible to encourage into the field.

One of the challenges I had when I switched from physics to biology was finding these specific roles. My very first postdoc was in aging, working on C. elegans, but I only did that for six months before getting a better offer. My next postdoc was very generic genomics and medical records research.

That role gave me the opportunity to learn a lot of biology. I did start doing a little bit of aging-specific machine learning right at the end before I got my book deal and quit. However, the field is so small that it is quite hard to find aging biology-specific jobs.

As a researcher, you have to have a lot of grit and determination. You might have to take a cancer job or a job in cell biology. The skills are very transferable, meaning when you eventually set up your own lab and apply for grants, you can give them an aging bent.

We need people to do that, but if the field were larger, there would be a massive multiplier effect. Cancer researchers already have many of the skills you need to be a longevity scientist.

If there were funding streams that allowed them to carry on doing the science they love while focusing on aging biology, they would start adapting their work in that direction. That is a much larger scale impact than influencing individuals.

If you could deploy policy work that allowed that change to happen, you could get thousands of scientists moving across. There is a fundamental limit to how many individual scientists you can convert when the field’s funding is currently so small.

That said, maybe some of those scientists you convert will start putting the meme out there, making grant application panels more sympathetic to aging science. This is why theories of change are hard. I always look for those leverage points where you can do things that scale rather than just influencing individuals.

28:05 The case for increasing funding for aging research

Daniel 00:28:05

Regarding the point of funding specifically, let’s focus on public funding for aging biology research. What is the case you would make for increasing that funding, and how would you approach that?

Andrew Steele 00:28:23

The most straightforward case is the geroscience hypothesis. We have all of these very expensive diseases costing our healthcare systems billions of dollars around the world.

By creating preventative medicines and stopping people from getting ill in the first place, we could alleviate those treatment costs. We could also alleviate the wider economic costs that occur when people give up work to care for a relative.

There are billions of dollars in unpaid care happening in the economy all the time. Furthermore, when you get these diseases, they don’t just cost money directly; they mean you are less able to travel or engage in your hobbies, which has a broader economic impact.

Economists try to capture this through the value of a statistical life, pricing a year of human life in good health. But even if these things end up costing money, we are buying the thing most of us care about most: our life and our health.

Because aging biology underlies a whole range of different pathologies, you are creating preventative medicines that don’t just hit one disease. Cancer research is great for curing cancer, but it often leaves you frailer than you were before the treatment.

Aging biology doesn’t just hit cancer, dementia, heart disease, and stroke; it also reduces frailty and cognitive decline. It addresses things like impotence and incontinence—all the different things that make it worse to be an older person.

The fundamental case is that it has enormous economic and health benefits, and the numbers are staggering. In the US, you spend somewhere between $15 and $20 per person per year on cancer research. Just over $1 per person per year goes toward aging biology.

Daniel 00:30:22

That’s more than I thought.

Andrew Steele 00:30:25

The US is the only country in the world where I can even give you those numbers. This is because, though flawed, you have a National Institute for Aging and a specific federal body funding this research.

In the UK and Europe, I’ve tried to find these numbers. A journalist looking at the EU level emailed the Horizon program to ask how much they had given to aging research. After doing a search through their grants, the answer was a few million euros across a multi-billion dollar program.

I suspect it is probably pennies per person per year in most countries. The US just happens to be a very large funder of research. These numbers are shockingly small and don’t make any sense.

Aging kills 85% of Americans and more than 90% of people in most European countries. Spending only a dollar per year on the thing most likely to make you frail and kill you is irrational.

If you look at the economic cost of these issues, it is probably more than $10,000 per person per year. To put only 0.01% of that into research is mind-blowing. The straightforward economic and human case is the strongest way to argue for change.

Daniel 00:31:45

Virtually everyone above the age of 22 experiences the negative effects of aging. I find the case for longevity research extremely compelling. Rationally, there should be almost no limit to what we are willing to spend on this. Our health is our most important asset, and everything else follows from it.

Despite being so compelling, this field hasn’t fully caught on. When you communicate with the public or people in government, what are the main obstacles in getting them to share this point of view?

Andrew Steele 00:32:21

The first issue is that they don’t even know this is an option. If you talk to a politician worried about the burgeoning cost of the healthcare system, they don’t see longevity science as a policy lever they can pull to solve or alleviate the problem.

A classic example is the debate over demographic birth rates. People discuss how to encourage immigration or incentivize women to have more children, but no one mentions longevity science. Keeping the population healthier would improve the dependency ratio, yet it never comes up in those discussions. It’s simply not on people’s minds.

Even if you do get it on their minds, there are huge communication challenges regarding ethical questions. Will this only be for billionaires? Will it cause overpopulation? Will dictators live forever? Will people get bored with an extra lifespan?

People place longevity science in a separate ethical category. As a computational scientist, I think the case is clear: there is a fundamental underlying process that causes these diseases. We dislike the diseases, so we should address the underlying cause.

Most people hear this for the first time and assume aging is natural. Overcoming that mindset requires time for the idea to be internalized. The best argument I use is to ask if they would have the same ethical objection if I were a cancer researcher.

Daniel 00:34:30

Yeah.

Andrew Steele 00:34:31

If you want to treat cancer, dementia, and heart disease, why treat longevity science differently? The problem is that the current lifespan is deeply ingrained in people’s worldviews. We look at society and assume living to 80 is how it has always been and how it should be, even though that is a remarkably recent development in human history.

The funding argument often doesn’t land because people haven’t heard it. When I talk about science funding broadly, I rarely meet anyone who is already aware of the macro figures. Even scientists often don’t engage with macro funding; they focus on applying for individual grants.

I once had a meeting with a senior official in the UK Office for National Statistics. He knew the science budget was £4.6 billion, but he didn’t know the cost per person or how it broke down into categories like heart disease, dementia, and aging. Governments generally don’t collect data in a way that highlights these disparities.

Once you show people the data, it can be convincing. The challenge then is telling them what to do. I’m often jealous of sleep scientists because their takeaways are simple: sleep in a dark room and practice good sleep hygiene.

My takeaway is often to write to a representative, which is much harder to action. We need people to spread the word and find ways to advocate for more longevity science funding.

If you are already following the standard advice in the longevity space, you are probably 90% optimized. The biggest influence you can have on your own longevity is moving policy levers or bringing more resources into the field. That is what will make us all live longer and healthier.

37:12 Bryan Johnson & personal health optimization vs biotech progress

Daniel 00:37:12

Everything you just said is a great representation of how multifaceted social change is. Ethical questions come up, as do questions of whether this is just natural. People often haven’t even thought about aging because it’s like a fish in water; it’s always there, and we don’t question it.

There are abstract values, ways of seeing the world, and questions about the science. Then there is the practical question: what do I do? I’ve experienced this when trying to explain the longevity field. People often can’t get away from asking for concrete actions.

Sometimes I feel like a buzzkill because I explain that no matter what you do, you are going to age and die. I’ve had this conversation many times and feel dramatic, but it’s the reality.

Consider Bryan Johnson and the Don’t Die movement. Many followers seem to think that if they do everything right, they won’t age and die. It makes me wonder if this is a philosophical issue, a misunderstanding of the science, or the biology of aging. Is there a central experiment that could help people understand this? How do you think about that?

Andrew Steele 00:38:26

Regarding how people can have an influence, many don’t realize how important small contributions can be. I was recently chatting with someone involved in a series of longevity rallies organized through fundlongevity.org. They held rallies in various cities worldwide; I attended the one in Berlin, and there were several across Europe and the US.

A major problem was getting people to show up. Some protests only had five or ten people. People don’t realize that the difference between five and twenty attendees is significant for credibility. You can take a good photo with twenty people and it looks like a reasonable attendance, whereas five people looking around awkwardly doesn’t have the same impact.

It’s the same with donating to organizations. If you can only afford $20, organizations still benefit because they can say they are supported by thousands of people from the longevity science community. These tiny individual actions feel insignificant, but they add up.

If a congressperson or an MP in the UK receives thirty letters on the same subject, they start to notice a pattern. If they get only one or two, they might dismiss them as cranks. It’s very rare for politicians to be engaged on substantive national-level policy issues by their constituents.

In the UK, an MP represents about 70,000 to 100,000 people. Most of their interactions involve local planning disputes or immigration issues. They rarely hear about funding longevity science to address an aging population. If they received 100 letters to that effect, that’s only 0.1% of their constituency, but it would make them think something serious is happening.

Individuals can have a much bigger effect than they realize; the challenge is articulating how to have that effect. Regarding Bryan Johnson and his approach, I am actually sympathetic to the mindset. I’ve found myself going down rabbit holes after reading about a new supplement or a study in mice.

I’ve felt like I was just one study away from understanding the perfect diet, only to read the next study and end up more confused. Evidence from early studies often doesn’t hold up, or dosages are hard to optimize. Ultimately, these things tend to have very small effects.

It’s easy to get carried away with a “do your own research” mentality and feel like you’re just a few optimizations away from a protocol that will keep you alive for decades longer. I think there is a failure in communicating the scientific method. Many people think what Bryan Johnson is doing is a useful n=1 experiment, but I don’t think it even qualifies as that because it’s so uncontrolled.

He is doing hundreds of different interventions simultaneously. It’s impossible to get any signal out of that noise. A clinical trial involves hundreds or thousands of people receiving the exact same intervention in a controlled way, and even those are difficult because effect sizes are often small.

The idea that you can personally experiment with a bunch of stuff and find a way to live dramatically longer is for the birds. However, I understand the appeal. Do you know the phrase “nerd sniping”?

It comes from XKCD. Someone is crossing the road, and another person shouts out a math problem. Because the pedestrian is a computer scientist, they get distracted by the problem and get run over by a car. Protocol optimization is a really good nerd snipe.

42:37 Why health podcasts are such a nerd snipe

Daniel 00:42:58

Yeah.

Andrew Steele 00:42:58

It is easy to get distracted thinking that if you can just optimize your Zone 2 training and sauna usage, you will add decades to your life. The reason we spend so long debating these things is because the evidence isn’t great. If the evidence were amazing, we would all just be doing it.

When it comes to protein consumption, there is a massive debate about whether we should be protein maxing to build muscle or practicing protein restriction because it makes animals live longer. There are hours of podcast material on this. I have written 1,000-word articles on the subject and come away more confused than when I started.

I think there are unreasonable positions that are obviously incorrect, such as eating steak all the time. However, once you are 80% of the way there on diet advice, I don’t think we can do much better. I actually think we are going to cure aging medically before we fully understand diet.

Truly knowing the optimal diet for any individual will require a systems biology level of understanding. In the meantime, nutrition is a massive nerdsnipe because of the endless forums and podcasts. I actually think many health podcasts are reducing life expectancy because they engage people in three-hour conversations about absolute minutiae.

If those listeners instead spent those three hours composing a letter to their congressperson, they would probably have a much greater impact on their life expectancy.

Daniel 00:44:26

I have thought about this a lot. Health podcasts suck away so much attention. I am torn because it is good that people are paying attention to their health, but many are drawing the wrong lesson.

If they wrote to their congressman or donated money to research, it would be much more effective.

Andrew Steele 00:44:50

That is the challenge. We are currently fundraising for the Longevity Initiative, and one of the hurdles is this multi-step theory of change. Writing to a congressperson feels far less health-related than going to the gym.

You don’t come back sweaty and satisfied from writing a letter, so you don’t have that same feeling of personal physical engagement. When you talk about the theory of change for a think tank or an educational organization, it’s a difficult sell.

If we were fundraising for research, the theory would be straightforward: we do research, find treatments, and people live longer. But we are doing highly leveraged policy and communications work to enable that research.

If I were a billionaire, I would fund research without question. But if I only had $10 million, I would spend all of it on communications and policy. You can’t do that much science for $10 million, but the potential leverage of that money is enormous.

If that investment could move even a few percentage points of government or private research funding, it would outstrip the original investment many times over. However, that theory of change is more complicated.

If you speak to a tech billionaire, they fundamentally understand the power of science to change the world, but they might not be interested in policy. They might see the government as big, bloated, and unexciting.

It is a non-intuitive way to try to live longer and healthier. I believe the single most effective thing I could do for my own health with $10 million would be to fund a policy organization to do this work at scale. For amounts of money like that, it is the highest-leverage way to make a difference.

Daniel 00:47:17

There is another objection I hear from people regarding the framing of aging. Pharmaceutical companies are already spending billions of dollars on R&D to solve cancer, heart disease, and Alzheimer’s.

People often react by saying, “It sounds like you think they’re doing it wrong.” How do you respond to that? Is it that they aren’t focused on aging biology and are missing a big opportunity?

Andrew Steele 00:47:51

People have a conspiratorial attitude toward pharma. They think it’s to their advantage to keep providing disease treatments that keep people alive but dependent. Even if that conspiracy were true and pharma simply wanted us hooked on long-term drugs, the best outcome for them is anti-aging medicine.

You might start taking these medications as early as age 22. These would be long-term health maintenance treatments that people could take indefinitely. If we solve aging, pharma would have endless repeat customers. There are many misconceptions about how pharma works and the incentives it actually has.

One problem is that pharma typically doesn’t know much about aging. I’ve spoken to “activist cells” of scientists within these companies who try to spread the ideas of geroscience to upper management. There are also broader regulatory challenges and policy questions.

The FDA’s acknowledgment of the TAME trial means there is a potential pathway for aging as an indication, but is it economic to pursue? It might be easier to target something narrower that can be proven more quickly. These policy hurdles make it difficult for pharma to engage with longevity.

49:27 Who will fund the solution for aging?

Daniel 00:49:27

You mentioned in a conversation yesterday that a lot of these things will start as public goods, like the data we need to solve aging. Pharma companies have a huge financial incentive to solve this problem, but they face communal obstacles like public data, regulations, and fundamental biology work.

In some ways, this is a tragedy of the commons or a public action problem. We need to solve those communal issues to get the flywheel of private innovation moving.

Andrew Steele 00:49:56

It totally is. There has been some cooperation between pharma companies post-AlphaFold. They realized how useful it was, but AlphaFold only provides static structures. It lacks many of the structures found in pharma’s internal datasets.

To address this, they pooled some of their internal protein data to improve the model for everyone’s benefit. No individual company would necessarily benefit in a way they could monetize alone. We need to do that at a much larger scale for aging.

The data we need to feed into AI models is broad, diverse, and multimodal. We could convince pharma to chip in or pay for access to publicly generated data. Conceptually, the most straightforward approach is to treat this as a pure public good funded by philanthropy and government.

If we open-source the data, anyone can build on it. We don’t know where the innovation will come from—it could be a scrappy biotech startup, a major pharma company, or academia. Opening up that data is immensely valuable.

We also need policy to identify actors who benefit economically from longer, healthier lives. For example, life insurance companies directly benefit if people live longer. They receive premiums for a longer period and pay out death benefits much later.

These companies have a direct financial incentive to keep their customers alive. Many already have health programs that offer premium discounts for hitting step goals, going to the gym, or getting preventative screenings.

I would love to calculate the trillions of dollars life insurance companies have under management. If rapamycin or metformin slowed down aging, would the benefit to these companies outweigh the cost of the trials? Could we make a business case for them to fund the research?

Pharma would also have a stake because a proven aging drug would create a clear pathway to market. Perhaps they could communally contribute a few hundred million dollars. Someone should spend a week or two on the basic math to see if that computes.

Daniel 00:52:40

I think I have my next project to work on with Claude. In the US, people switch health insurers every few years with their employer, so those companies don’t have an incentive to care about long-term health.

However, people often keep their life insurance policies for life. There could be trillions under management, creating an insane windfall for those companies if longevity increases.

Andrew Steele 00:53:00

Exactly. That sort of out-of-the-box thinking is necessary. We have to look around the economy and see who would benefit from this. Once we identify those synergies, we can combine the forces of the public sector, philanthropy, and the private sector to maximize the benefit for everybody.

Daniel 00:53:19

The AI labs are another big player in the ecosystem. They are buying up biodata and making huge acquisitions. Their founders have expressed a desire to cure all diseases and end aging.

Some believe superintelligence will solve everything, but that feels too fatalistic and is not a guarantee. How much should the ecosystem orient around these AI labs? Should we focus on getting them the data they need and amplifying their messaging? Is that the approach that gets us there?

Andrew Steele 00:53:54

I’m interested in why foundation model labs aren’t lobbying for more data generation. I hadn’t thought about this until an AI talk yesterday. DeepMind showed that AlphaFold is possible because of the Protein Data Bank, which is a publicly funded database consisting mostly of the work of publicly funded scientists who have submitted it over decades.

A paper from a few years ago estimated that if the Protein Data Bank were accidentally deleted, it would cost $20 to $25 billion to regenerate it from scratch. That is a lot of money, but in terms of the scale of the economy, it’s not very much. It is about $25 per person in the rich world. Communally, that is a resource that is relatively cheap to acquire.

If we need datasets on that kind of scale to understand different aspects of biology, we need tens to perhaps low hundreds of billions of dollars worth of data collection. It surprises me that AI labs aren’t asking if we can collectively do this or get governments to start doing it for them. They would rather free-ride on a public good.

There is another set of stakeholders we should bring to the table to find mutual support, because this is in their interest as well.

Daniel 00:55:26

It’s interesting. Even if you aren’t completely “superintelligence-pilled,” models are improving every few months. I imagine the perspective from the AI labs is that they don’t know what data they need yet, and that AI will help them figure it out.

They might feel it is a waste to spend hundreds of billions of dollars right now when waiting a year will provide far more clarity, and the economy will be larger. With aging, it is something you could easily convince yourself to kick down the road for a little bit.

Andrew Steele 00:56:03

There are two counterarguments I would give to that. First, some of this data takes time to collect. The ultimate example is human or long-lived animal longitudinal data. We currently have many small patchworks of data.

A paper recently received a huge amount of media coverage for claiming we age in bursts—once in our early 40s and again in our 60s. The statistics were a bit shaky, partly because these time courses were only about two years long. They stitched together many different people’s two-year time courses to create an overall trajectory.

Within one individual is by far the most useful data. If we want to collect an incredible amount of aging-related longitudinal data, you can only collect it at the rate of one year per year. You cannot parallelize that or buy more GPUs to collect that data faster. We should start today regardless of knowing exactly what data we need.

The second argument is that it is very little money in the scheme of the economy. If we spend $100 billion on this data collection, that is $100 per person in the rich world. If that increases our chance of slowing down aging by even a very small amount, most people would be willing to chip in for that.

If it turns out to be a complete waste of time because AI needs a different dataset, we’ve only wasted $100. I’ve spent $100 on far worse things than data that could potentially save me, everyone I love, and the entire world from the spectre of chronic disease. Even if you are heavily AGI-pilled, some things just take time and aren’t that expensive. We should be doing them now, just in case.

57:50 The Longevity Initiative

Daniel 00:57:50

Let’s dig more into the details of the Longevity Initiative. Are there some specific projects you’re hoping to kick off soon?

Andrew Steele 00:57:57

There are two main strands we’re focusing on at the moment: policy or general orientation reports, and something called the Longevity Library.

Regarding the reports, the primary goal is to propose specific policies. If a government wants to implement longevity initiatives—whether that’s increasing funding or integrating it into healthcare—we want to produce the necessary foundational reports to help the field broadly.

A great example is a report on how many deaths are caused by aging. Everyone has heard that it’s 100,000 or 110,000 people a day. I cited this in my book, but when I looked for a source, I found no published academic paper containing that number.

I ended up doing the calculation myself. It’s on my GitHub; I wrote some code in R back before the days of AI. It was a manual process with my limited bioinformatics scientist coding skills. It is wild that I had to do this myself.

While it hasn’t been peer-reviewed, no one has complained in the last five years, so it’s likely accurate. However, it would be much better to have a scientific paper as a centralized, citable source for how many deaths are caused by aging.

We need to know what percentage of disease and what percentage of disability-adjusted life years (DALYs) are attributable to aging. How much of the world’s health burden can be linked to this process, and how will that change in the future? These are fundamental, field-orientated questions.

If you attend a heart disease conference, the opening slide always states that cardiovascular disease is the leading killer, responsible for 30% of deaths. If we had a similar statistic for aging, every longevity scientist could lead with it. It would be the primary citation in every paper and a key part of every startup pitch deck.

It would highlight the total addressable market, given that two-thirds of people globally die from age-related causes. Providing these foundational resources offers huge leverage. It gives journalists an authoritative source to cite with confidence, rather than just referencing my personal GitHub.

We also want to create beautiful, updatable dashboards because those numbers will increase as global populations age. I really love the model used by Our World in Data. It’s a website that aggregates statistics on population, GDP, life expectancy, and carbon emissions.

They became prominent during COVID by aggregating national data and creating clear visualizations. They were the go-to source early in the pandemic. I want to be that for longevity.

Journalists often ask me questions that I have to research or code answers for myself. I want to provide a centralized, one-stop shop with dashboards where people can see how their country is doing or assess their individual likelihood of dying from aging. This has the potential for significant personal impact.

We are also developing the Longevity Library. Currently, there isn’t a centralized, trusted resource for this field. If you search for high blood pressure, you find authoritative overviews from the Mayo Clinic, the NHS, or Harvard Health.

There is no equivalent for longevity. If a journalist asks me about a specific supplement, the answer is usually that there’s promising data in mice or a human correlational study, but we don’t know if supplementing actually helps. To provide a thorough answer, I have to hunt down specific papers.

A centralized resource would be incredibly helpful for me personally, but it also serves as a funnel for TV researchers, political staffers, and journalists. They are all looking for quick, evidence-based answers.

If they find our site, they might go down the rabbit hole and explore our other work. This is an opportunity to provide authoritative, scientifically grounded information that builds credibility for us and the entire field. It will bring diverse audiences into the Longevity Initiative way of thinking.

Daniel 01:03:06

That sounds amazing. I’m excited for that to exist in the world.

Andrew Steele 01:03:10

There is so much potential. I have a Google Doc full of ideas for reports and visualizations, along with about 20 half-baked Claude code projects where I’m playing with different datasets.

The bottleneck is human time. I’m doing a hundred things at once, and it would be fantastic to have enough funding to pay a small team of researchers.

Daniel 01:03:35

Yeah.

Andrew Steele 01:03:36

This doesn’t require a 50-person team. You can make a massive difference with just a few people pulling these data sources together.

These things can spread online, and since nobody else is doing it right now, I hope we can fill that void.

1:03:50 Andrew’s experience as a science communicator

Daniel 01:03:50

I have a somewhat selfish question because this discussion makes me think about what I want to do in this field. How much do I want to do through science communication versus other approaches?

You originally had a background in physics as a researcher, and my background is in physics as well. I’m curious how much you miss doing science, or if you’ve found that communicating and influencing in this space aligns more with your passion?

Andrew Steele 01:04:18

I do miss the science, though there are certain aspects I don’t miss. Being in the lab can be a grind. Anyone who’s ever been in a lab knows there are those boring days when your code isn’t working and you’re just debugging endlessly.

I had a cursed piece of equipment during my PhD. For a month, I was trying to get a helium-3 cooling unit working, and it eventually didn’t even work. Those days never made it into the thesis. Being a scientist is hard, but I do miss it.

I’ve squared this for myself in a few ways. Firstly, I enjoy the communication. If I hated it, I wouldn’t be able to keep it up. Secondly, I enjoy the potential impact. It depends on your comparative advantage; some scientists should stay in the lab because they are far better at bench work than communication.

Finally, I stay a hardcore scientist. I was chatting with someone in pharma yesterday and asked if a certain effect was caused by a chaperone. He was surprised I knew biology. It’s useful as a communicator and when dealing with policymakers to understand the science on a deep level.

I sometimes wonder if the physics PhD was a waste of time, or if I could have jumped straight to writing books. However, working as a scientist gave me tacit knowledge about how science works and how scientists think.

I want to be the scientists’ favorite longevity communicator—someone who sticks to the evidence and understands the biology. Some longevity communicators make small slips that reveal they don’t understand the science as well as they should.

I make mistakes too, but you can usually tell the people who are deep in the biology versus those who are just cosplaying as longevity influencers on a bandwagon. You don’t need to be a PhD biologist to do communications—there are fantastic journalists with non-science degrees.

However, it helps to have communicators who are deep enough in the science to read a paper and call out flaws that might not be obvious to a standard science journalist.

Daniel 01:07:10

It makes me wonder how I can build credibility as I explore this space. I have an undergraduate degree in physics and neuroscience, but I’m not a PhD-trained scientist. I’ve thought about going back to school to become a researcher and build more knowledge.

However, the world is moving so fast because of AI, and 100,000 people die every day. I want to contribute to this space without it taking too long.

Do you still recommend that people get a PhD? Is that the best way to get this training, or are there faster ways now, especially with AI?

Andrew Steele 01:07:46

It’s difficult because I don’t want everyone to do what I’m doing; I need some science to communicate. I would like most people to do the bench work and do it competently. Having a PhD has helped, though it’s a bit misleading since my PhD is in molecular magnetism. It has nothing to do with my credibility in longevity.

There are many PhDs and MDs who talk absolute crap when it comes to longevity science and health. A PhD is a useful credibility signal, though I’m concerned when people bill me as Dr. Andrew Steele and assume I’m an MD giving health advice.

I don’t recommend PhDs for people who don’t want to go into research. If you do, it’s a prerequisite. It’s your scientific training and a great way to understand how science functions. If you want to do something else, it’s probably too much of a detour.

If your contribution to longevity is donating, campaigning, or volunteering, there’s no need for a PhD. In fact, it will likely decrease your lifetime earnings because of the years spent on low-paid work.

It helps to have a deep scientific background, but mine isn’t even in biology. You can learn a lot by reading and watching materials online. Papers are increasingly open access, so you can really get your teeth into the subject. If you’re driven, there is a lot of potential to homeschool yourself on this.

1:09:53 How to measure progress in activism against aging

Daniel 01:09:53

Absolutely. You mentioned that seeing your impact keeps you motivated. What metrics or ways have you used to measure that in your career thus far? And what metrics are you planning to track for the nonprofit?

Andrew Steele 01:10:14

Metrics are just really, really tough because I do get a few emails a year being like, “I’ve changed my career” or “How do I get into longevity science?” That is obviously hugely personally valuable, but is it the biggest impact I’ve had? I think it probably isn’t for the reasons we were discussing earlier. These are individual people.

Much as I’m blown away by the fact that something I’ve written is now out there in the world, the fact is that most readers probably don’t even know my name, which is fine. I don’t care. I don’t want to be famous. This is not the reason I’m in this game. But maybe they’ve gone away with a sense that this is real science, and you don’t know how that’s going to propagate out.

It’s frustrating how random it is. If you want to pitch to funders, you have to have metrics. A common one cited by a lot of think tanks is media mentions. I think that’s somewhat useful, but it’s a bit gameable. Often, if you look at think tank reports, they’ve had 150 media mentions, but 110 of them were in some weird trade publication that’s only very specific to the thing that they do. You have to be careful with metrics like that.

However, I do think media mentions have this positive feedback loop of credibility. If you’re well-known in the media, suddenly policymakers are more interested in talking to you. Also, journalists who have seen your name put you on the list of people the BBC calls when they want comment on longevity stuff.

You start getting more and more bookings. If you then say things in an accessible way that is coherent, evidence-based, and seems credible, it creates a positive feedback loop. So, I think that’s one way you can measure it.

The dream metric is policies implemented. What you want to see is: Have we changed a regulation? Is longevity now an indication somewhere? Has longevity funding increased? Are there mechanisms by which this science is being moved forward? Are particular people talking about this in a certain way?

The trouble is these metrics are super high variance. You never quite know whether you were the thing that got something over the line, or whether you were just part of a general cacophony of background noise, or whether you were completely irrelevant.

You can cite some policy, and if someone specifically copies and pastes your policy proposal, sure. But even for very successful think tanks, that is a relatively once-in-a-blue-moon event. What most commonly happens is that you’re shifting the Overton window—the range of things that are considered politically acceptable.

You nudge that window in the direction of the thing you want to see, and that suddenly allows a policy to fall through that window that wouldn’t otherwise have been able to. It’s really, really hard.

The thing that motivates me about this as a way of changing the world is that it’s so cheap by comparison to all the other methods. If we want to solve aging, we need to unlock probably hundreds of billions in public good funding. I’m very agnostic about where that comes from. It doesn’t have to come from government, but a lot of this funding, at least at the beginning, is going to be basic science and datasets—things that are very hard to make a private profit from.

If you want to start moving hundreds of billions, consider the National Cancer Institute’s spend throughout the whole of US history. Corrected for inflation, you get about $250 billion. If you do the same for aging, you get $7 billion. It’s about 3% of what’s been spent on cancer historically.

That gives you an order of magnitude. We have lots of cancer treatments now. Cancer isn’t a solved disease, but it’s a disease that we have lots of different treatment options for. There are some curable cancers, and so on. We’re probably going to need an amount of money on that order.

If I personally had to raise that much money for a nonprofit, I wouldn’t have a great deal of confidence in my ability to do that. But if I have to raise millions a year in order to create a think tank that can then spur the ecosystem broadly to make that kind of change, then hundreds of billions is like a tiny fraction of the global economy. It’s an amount of money that’s within reach.

If I thought we needed $10 trillion, maybe I’d just go and get a high-paying job and buy a private island. But I think this is in a regime where these relatively small investments in policy and communications can catalyze millions in policy and communications, which catalyzes billions in research, which then catalyzes trillions in downstream economic value of extra healthy human life.

Although the metrics are frustratingly fuzzy—and if anyone has any better ideas, I’m all ears—I think everything matters. You can measure YouTube views, but it depends who watches. If you get a million views and one of those is a billionaire who decides they want to suddenly donate money to this field, or a policymaker who decides to do a deeper dive, that was more impactful than the 999,000 other people who watched that video.

You can use those kinds of metrics. It’s a bit stupid to say there are no metrics and we’re just going to be Zen about this. You can’t judge us by the fact that we’ve had no media mentions and no YouTube success, but you’ve got to be careful about how you think about these things as well.

1:15:07 Will we solve aging?

Daniel 01:15:07

That totally makes sense. Another thing that you brought up is how important this work is. We’re at a very pivotal moment in history where it’s the first time we can do something about aging, but it’s not clear we’re going to do it. How optimistic are you in general? What do you think the future is going to look like?

Andrew Steele 01:15:30

I am optimistic enough that it is worth dedicating my career, and the careers of many others, to cracking this. If epigenetic reprogramming turns out to be successful, we might hit a home run and solve the problem relatively easily.

However, that is quite unlikely because it neglects the other hallmarks of aging. Whether you count 10 or 12 hallmarks, we likely need to address most of them. We don’t know how difficult this problem will be, but what gives me optimism is that we don’t have to solve it in one shot.

I often talk about curing aging because it is a coherent concept. We need to navigate how we have conversations around that word. Cancer Research UK aims to “beat cancer sooner,” and Alzheimer’s Research UK wants a world free from the fear of dementia. We should be able to say the same about aging.

When I talk about curing aging, people often imagine a single pill that freezes or reverses time. That is vanishingly unlikely. We don’t need one magic pill; we just need to solve it iteratively.

I am 40 years old, which means I have perhaps 40 or 50 years of life expectancy ahead of me. That is an enormous amount of time for science to unfold. If we raised $100 million, we could start the TAME trial for metformin or another geroprotector tomorrow.

Within three to five years, we would know if that drug actually slows aging. That is plenty of time for me and most people alive today to benefit from the first geroprotectors. If a treatment adds just two or three years to human life, that buys scientists more time to develop the next breakthrough.

As long as we stay one step ahead of aging, we can eventually reach a steady state. We don’t need to perfect this immediately to achieve radically longer, healthier lives. That makes the challenge much easier to face because we can solve it in an iterative fashion.

I don’t know for certain if we will succeed. If we were spending $100 billion a year, I would be very confident. At current funding levels, success is not impossible, but I’m not confident enough to rest on my laurels. That is my calibrated level of optimism.

Daniel 01:18:36

Where can people learn more about you, your work, and the Longevity Initiative?

Andrew Steele 01:18:40

To learn more about the Longevity Initiative or to get involved, visit the-li.org. You can find the first few Longevity Library articles and see what we are all about there.

Please follow us on social media. Individual followers make a huge difference in the early days. We are on LinkedIn, X, BlueSky, and Instagram. Together, we can make a significant difference in longevity science.

Daniel 01:19:07

Andrew Steele, thank you so much for joining us on the Free Radicals podcast.

Andrew Steele 01:19:10

Thank you very much. This has been a lot of fun.

We sat down with Dr. Jamie Justice, the geroscientist running the $101M XPRIZE Healthspan competition, and she told us there are five teams pursuing reprogramming, and hundreds pursuing other approaches. The $101M prize is to be awarded to teams that can restore muscle, cognitive, and immune function by 10 to 20 years.

We discussed how the longevity field has a serious hype problem. We‘ve found hundreds of interventions that can extend lifespan in model organisms (reprogramming among them) and yet none have been translated into humans.

The XPRIZE is changing that, by challenging the field to get out of the lab and into the clinic. $101M for those that can show results now, not in a decade, is a huge incentive to turn the hype into reality.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

0:00 Intro
1:01 XPRIZE Healthspan: the $101M prize for restoring function in humans
3:05 Where teams enter: discovery, trials, or scale-up
5:08 Aging isn’t an indication: the regulatory wall
9:07 Chemical reprogramming in humans for the first time
12:35 How XPRIZE verifies the data and picks finalists
15:40 How XPRIZE manages ambiguity around aging
20:48 Endpoints for aging: function, feel, survive
26:06 The geroscience hypothesis and why TAME stalled
33:15 What makes a drug a real gerotherapeutic?
36:43 Where the field actually is, from telomeres to peptides
40:33 The peptide problem: gray markets and autonomy
43:40 Is the first person to reach 150 alive today?
46:01 Building the dataset behind personalized longevity
49:27 What motivates Jamie

Transcript

1:01 XPRIZE Healthspan: the $101M prize for restoring function in humans

Daniel 00:01:01

Jamie Justice, welcome to the Free Radicals Podcast.

Dr. Jamie Justice 00:01:03

Thank you so much. I’m glad to be here.

Daniel 00:01:05

Thank you for joining us. You run XPRIZE Healthspan, which is a $101 million prize for restoring muscle, cognitive, and immune function by a minimum of 10 years with a goal of 20.

This is extremely exciting. Can you tell us more about the prize and the motivation behind it?

Dr. Jamie Justice 00:01:23

The prize is a challenge motivating the field. We have people from around the world participating, which is why we run it as a competition. It encourages people who wouldn’t normally enter the field to get involved. The $101 million is a very big carrot on the end of a stick.

The competition launched in 2023. Our fundraisers and scientists identified the biggest barriers to progress. We looked at therapeutics, public perception of aging as something malleable, regulatory barriers, and funding.

We found that while lifespan can be changed in animal organisms, translating those therapeutics to humans is a major barrier. The challenge was set up to accelerate that translation. We want to build frameworks to demonstrate that medicines targeting aging actually work.

Right now, there’s a lot of hype and story but not enough substance. This field has been making huge developments, and this is an opportunity to validate them in a rigorous, transparent, and global competition.

3:05 Where teams enter: discovery, trials, or scale-up

Eric 00:03:05

When you think about the sweet spot for how you partner with potential teams, are you looking to come in earlier in the R&D process at the discovery stage, or later during clinical trials and commercial scale-up?

Dr. Jamie Justice 00:03:19

The sweet spot is that we don’t need to have one. Unlike traditional investment or fundraising models, we simply put a line in the sand. If you pass a certain area and meet a milestone, we give you money.

In this case, we set the milestones toward the pathway for clinical trials. This pivoted teams toward entering the competition with therapeutics they thought were close to human testing. Most came in with some R&D already behind them so they could leverage that proof of concept to advance to trials.

That doesn’t mean we didn’t get new drugs or things that hadn’t been tested in humans. Some of our teams have gone into first-in-human trials over the last year, which is exciting. We also have drugs already used for other disease conditions being repurposed and tested for aging for the first time.

Other teams don’t even have a commercial runway yet. They are taking public health approaches involving sleep, nutrition, and exercise and figuring out how to package them to improve aging. The process democratizes the field. We don’t have to seed a specific area; we just draw the finish line.

5:08 Aging isn’t an indication: the regulatory wall

Daniel 00:05:08

Getting new drugs into human patients involves many barriers. One big barrier in the longevity field is that aging is not an indication; you usually need a particular disease to target.

The XPRIZE is unique because it is a prize for age reversal or restoration of function. How have you been helping the teams navigate this to get these treatments into relatively healthy people?

Dr. Jamie Justice 00:05:38

Developing frameworks, particularly regulatory ones, is a huge point of impact for the prize. We don’t necessarily think our teams will have immediate regulatory approval since aging is not currently an indication. The indication refers to the population or disease you intend to treat or prevent.

In the absence of an aging indication, companies typically target age-related diseases. They find a mechanism common to many diseases and test it one disease at a time to apply for approval.

With the prize, we are bypassing that by focusing on an aging outcome. It is general enough to run a competition while allowing teams the flexibility to test in specific disease conditions if they choose.

For regulatory approval, agencies want to ensure the benefit outweighs the risk of an intervention. For a new drug, that ratio must be promising. We allow teams leeway to define the patient population they are targeting.

At the same time, we are intentionally creating friction. I hope the teams struggle because their struggle is the friction needed to create change. If it were easy to move into trials, this prize wouldn’t be necessary.

Getting hundreds of teams to start making inquiries and doing this work globally drives the conversation. By putting a timeline on it—requiring all teams to go to trial this year—we are putting pressure on regulatory systems to have this discussion now.

Dr. Jamie Justice 00:08:23

XPRIZE is not alone in this effort. There are several other initiatives we are tightly aligned with; we watch each other closely and offer mutual support. The more stakeholders we can involve to create friction, pressure, and time bounds, the more likely we are to have these essential conversations.

We are creating indications by generating the friction necessary to establish guidelines. At the same time, we allow flexibility for teams with new drugs to pursue disease indications if necessary, while still advancing aging measures within their testing paradigms. This approach helps us meet both their commercial needs and the needs of the field.

9:07 Chemical reprogramming in humans for the first time

Daniel 00:09:08

Practically speaking, for teams that have dosed patients with a new therapy that isn’t a repurposed drug, have they received FDA approval for a particular disease indication? Or are there cases where they proceed without a specific disease indication, using only IRB approval?

Dr. Jamie Justice 00:09:30

Absolutely. We have a number of teams doing exactly that. They have their IRB approval, and some of them are very new. One of the bigger names in the field, David Sinclair’s group at Harvard, has an epigenetic reprogramming approach using small molecules.

They secured IRB approval, giving them ethics approval to apply this even in the absence of a specific disease. They can measure what we call physiological endpoints. This is a way to conduct new testing with an IND that has a much lower barrier to entry since it isn’t seeking approval for a new disease class.

That informed the framing of our prize around functional endpoints. It allows us to start a conversation while giving enough flexibility for teams to move forward. Some teams get IRB approval and test elsewhere.

There is no requirement that the testing be done in the US; this is a global competition. Many of our teams are forming international partnerships to determine where to test or which clinics might have different standards for approval. We aren’t allowing anything unsafe or unethical, but different regions have different barriers to entry and testing methods.

Daniel 00:11:03

One quick follow-up on that. We are familiar with David Sinclair’s work with Life Biosciences, where they are developing a genetic therapy for the eye. I had not heard that they were dosing patients with a small molecule for reprogramming.

Dr. Jamie Justice 00:11:13

Life Biosciences has not entered the competition. David Sinclair’s group is a separate entity, though that company came out of his lab. We, along with everyone else in the field, are watching their AAV therapy closely.

There are a number of groups looking at reprogramming using different platforms. There is no single approach; for example, some target three out of the four Yamanaka factors. There have been publications by David’s group, Vadim Gladyshev, and others regarding small-molecule-based partial chemical reprogramming.

In our finals applications, we have about five teams using reprogramming approaches. Some are cell-based, some use gene therapy, and others are proposing partial chemical reprogramming. These teams are entering trials now, which is a significant milestone.

For this next threshold point, a trial is required. Teams must demonstrate regulatory approval and prove they have recruited their first handful of subjects to show the trial is feasible.

12:35 How XPRIZE verifies the data and picks finalists

Daniel 00:12:34

Eric, did you realize that patients will soon be dosed with chemical reprogramming?

Eric 00:12:40

I just found out from Jamie.

Daniel 00:12:42

I’ve been following this space and speaking to many people, but this is extremely exciting. If we see that data, it will be massive.

Dr. Jamie Justice 00:12:50

It is massive, especially in the context of a global competition. I don’t know who will make it to the finals; I am not allowed in the room when the judges deliberate. I help prepare them, but we have an excellent panel of 15 permanent judges and several ad hoc experts.

They are currently reviewing the applications. For this stage, teams had to submit de-identified data so we can verify the results rather than just reading their reports. We check for rigor, confirm the subjects are real, and ensure nothing was fabricated.

We review all their approvals to make sure they obtained the necessary ethics approval for human subjects testing and that the dates align. The judges prioritize teams that balance innovation and feasibility. You need a great idea and a mechanism to test it, including a clinic, a recruitment plan, and regulatory approvals.

Without those basics, even the best idea in the world falls apart. Another big barrier for our teams is transparency. This is not about running a study in your own lab and sending us the results when you’re finished.

It requires following a common protocol and submitting data prospectively as subjects are recruited. We have a mirror book that goes directly to our data coordinating center, so there is no retroactively entering data to fit a narrative. It is run in real time.

All biological specimens—serum, plasma, peripheral blood mononuclear cells, and urine—are shipped to a central lab. Teams don’t generate their own biomarkers and submit them independently. Everything is rigorously controlled and transparently reported.

There is nothing for the teams to hide behind. I would love to see reprogramming as a finalist because it’s one of the first chances we have to see the actual merit of this exciting new class of treatment.

15:40 How XPRIZE manages ambiguity around aging

Eric 00:15:41

One point I want to dig into is how you were purposefully flexible about how to define aging. There is one class of medicines, either already existing or under clinical trials, which go after different indications affiliated with aging.

Dr. Jamie Justice 00:15:57

That’s right.

Eric 00:15:58

Then there is another class you are describing which is very different: how do we define the core mechanisms of aging? How do you devise a set of interventions to specifically and selectively modulate those newly defined mechanisms?

What have you seen from some of the participants? How do you define the core mechanisms of aging?

Dr. Jamie Justice 00:16:19

This is really fun because we don’t need to define it; we let our teams define it. The challenge for us is building a testing framework flexible enough to allow for both approaches.

I have had experience on both sides. As a scientist before coming to XPRIZE, I spent a lot of time working on a trial created to establish a regulatory pathway for aging using metformin. It is one of the most lovingly boring drugs ever to be repurposed.

It was a great drug for that purpose. There was a low barrier to running the trial, the proof of concept was already there, and we could just go in and do it. It was a great target to create an aging outcome using a repurposed drug.

I also had a great experience working with the Translational Geroscience Network in the US with Jim Kirkland, my old mentor and dear friend Steve Kritchevsky, and others including Nathan LeBrasseur. We worked with early senolytics about ten years ago when we weren’t sure of the potential.

We used senolytics as the first use case for that network, building common protocols and common testing. Senolytics were still very new, and we were figuring out how to measure them and what they were targeting.

Early senolytics like dasatinib were repurposed chemo drugs and had side effects. Because of the risk-benefit balance, we had to find diseases that these would match with for first-in-human trials. We didn’t test them as a general aging pathway initially.

The first disease we tested was idiopathic pulmonary fibrosis. In that model, rather than looking for a mechanism specific to that disease, you look for a common mechanism of aging and then find the age-related diseases it matches.

Beyond idiopathic pulmonary fibrosis, other groups are targeting hematopoietic stem cell transplants, mild cognitive impairment, dementias, frailty, and surgical resilience. There are many different conditions.

The idea is that if you get them all to use similar biomarkers or lead indicators, you have the disease target for the primary endpoint and aging measures for the secondary and exploratory endpoints. That was a great method for building a framework, but it can also be very limiting.

It is fantastic for developing novel solutions to hard-to-treat diseases and is a great use of geroscience, but it is challenging to make a lasting impact on aging itself. Both of those models negate what I consider the more novel therapies.

We need to think beyond repurposing and consider novel interventions specifically for rejuvenation, repair, and regenerative models. This includes work on heterochronic parabiosis, which led toward plasmapheresis and transfusion, as well as cell therapies.

There are some really cool classes coming out, like novel immunotherapies and senolytics that include CAR-T. Using the immune system for cell therapy is outstanding.

There is also a huge, wide-open space for reprogramming and replacement. Thinking about the future of the field, we need testing frameworks and approval pathways that encompass more than just repurposed drugs for disease prevention.

20:48 Endpoints for aging: function, feel, survive

Daniel 00:20:48

When thinking about those frameworks, a big piece is the endpoints of aging. What are we measuring, and where are we as a field in establishing those endpoints?

Dr. Jamie Justice 00:20:57

We’re having the same conversations every five to ten years, but there has been an evolution. Regarding TAME, we had an endpoint the FDA was willing to consider. The FDA defines clinical benefit by how a patient functions, feels, or survives.

This is a critical framing. Whether we’re thinking about repurposed drugs, disease-based approaches, or aging in general, the focus must remain on function, feel, and survive. You have to demonstrate that the clinical benefit outweighs the risk.

Using that framework, we can backtrack to determine our endpoints. In animal models, measuring survival is easy. You look at all-cause mortality or median lifespan and build survival curves.

In humans, all-cause mortality is problematic. The timescales and risk profiles required to test all-cause mortality are extreme. You would need a specific diagnostic to lead you toward an indication, and the proof point there is challenging.

That leaves us with function and feel in combination with survival. On the endpoint side, there have been multiple approaches for a large Phase 3 trial. One is age-related multimorbidity, which packages different diseases into a basket along with mortality.

Another option is disease-free survival. We’ve also seen disability-free survival used, as in a recent aspirin trial. That trial failed, but we have to ask if it was the drug or the endpoint. They looked at physical disability, cognitive disability, and survival.

Aspirin improved physical function but increased cognitive risk, resulting in a flat trial result. That illustrates the challenge of a composite endpoint. To work backward from those, we can use function, feel, and survive to develop earlier metrics.

We can’t wait ten years to look at disability, so we look for surrogates earlier in the pathway. I believe more exciting progress is happening here. For XPRIZE Healthspan, we landed on a function-forward approach, focusing on muscle, cognitive, and immune function.

Another effort involves intrinsic capacity, a measure from the World Health Organization. It is very public health-driven and global. It looks at locomotor, vitality, cognitive, and sensory functions, along with a fifth dimension that escapes me at the moment.

We’ve had workshops on how to turn that into a framework. Is it actionable? It’s a great public health measure to indicate risk, but it’s unclear how it differs from a standard frailty measure.

Each of these models, even those looking at function, often ignores the underlying biology. We still need to identify the causal factors or mediators. This leads to exciting work around biomarkers and how we develop and qualify them.

The goal is to target function, feel, and survive rather than a specific disease. This could include multiple diseases, disability, economic endpoints, and survival. The field uncorks this debate every decade or so.

We go through Delphi processes and try to reach a consensus, though consensus is often a pipe dream. We make executive decisions, test them, and keep iterating.

26:06 The geroscience hypothesis and why TAME stalled

Daniel 00:26:07

I’d like to tie this to the geroscience hypothesis.

Dr. Jamie Justice 00:26:11

Yes.

Daniel 00:26:11

The geroscience hypothesis suggests that if we target the biology of aging, we can prevent multiple diseases at once because aging is the primary risk factor. My understanding is that TAME struggled to get off the ground partly because the scientific community didn’t fully believe in that hypothesis.

Dr. Jamie Justice 00:26:31

Correct.

Daniel 00:26:32

What is the current status of the geroscience hypothesis in the community? How does that affect our ability to move these aging endpoints forward?

Dr. Jamie Justice 00:26:44

Support within the scientific community working specifically on aging remains very strong. Aging is undeniably the risk factor. However, when we tried to secure funding for the TAME trial, the pushback didn’t come from our peers; it came from traditional dogma.

Traditional drug development involves studying a specific disease, identifying its core mechanisms, and developing a drug to target that mechanism. It is a linear and logical approach. Aging, however, is not a single mechanism or a single disease.

Aging is a systemic process that underlies multiple diseases. This breaks the traditional model of drug development. Clinical trial reviewers look for certain pitfalls, and if you don’t follow the traditional path, they struggle to approve or fund the project.

Even though we made it through two rounds of large-scale review within the NIH, the project hit resistance in the regular study sections. The primary critique we couldn’t overcome was the demand for separate trials.

They wanted us to run the TAME trial for cardiovascular disease, a parallel trial for cognitive impairment, and another for cancer, each with mortality tied to that specific disease. That requirement negates the entire premise of the conversation we started with the FDA. By the way, the FDA is not the enemy in this process.

Eric 00:28:53

Right.

Dr. Jamie Justice 00:28:54

The regulators were fine with it; they just needed the use case. They gave some guidance, like noting that we couldn’t include diabetes or metabolic conditions because metformin is already used for those. They provided frameworks to adjust the study population, but there was never a “no” from the regulatory side. The resistance really came more from our scientific dogmatists.

Eric 00:29:17

To understand more deeply, were these scientific dogmatists representatives of the NIH?

Dr. Jamie Justice 00:29:23

No. The review was handled within the NIH, but the NIH itself is not the reviewer. The reviewers are scientists who are given the application to evaluate. Within the study section, one particular reviewer was a “no pass” for us.

There was also pushback against the use of metformin in other review critiques. Some people didn’t think it would have a strong enough effect, even though we did all the power calculations and event modeling. That’s just scientists doing science, and that’s okay.

The bigger challenge was having traditional trialists critique the composite endpoint and how it relates to aging. Even though the FDA was fine with it, our fellow scientists were skeptical. There are inherent challenges with composite endpoints, so we began looking at other options, like disability-free survival or disease-free survival.

All of that work has happened in the five to ten years since TAME began. I wouldn’t say that TAME is dead. I will forever call it the best trial that never was. We never recruited a single subject, but it started a necessary global conversation.

That model still applies. We still need an aging outcomes trial. Would it be repeated with metformin? I don’t know. There might be stronger drugs now that would have a greater effect. The creators of the TAME trial were open to that because it was never intended to be just a metformin trial; it was intended to be an aging outcomes trial.

Eric 00:31:26

Is this work a spiritual cousin or even a sibling of the work Matt Kaeberlein has been producing?

Dr. Jamie Justice 00:31:32

Absolutely. I adore Matt and think his ideas are central to the field. I was a trainee at the NIA summer workshops hosted at the University of Washington when Matt was the leading figure. He has been central to the geroscience hypothesis—defining what it is and how to move it forward.

I also appreciate Matt’s perspective that while we drive these trials, this should not be the only work we do. We still have major misconceptions to address regarding the basic biology of aging.

We need to understand the systems that underlie the large-scale disorders we recognize as either disease or disability. Understanding these intricate systems, how they interconnect, and how they interact with our environment is an area where we still need to make a lot of progress.

We must do the work in model systems and then use the human as a model system as well. We are a large animal model. We can’t necessarily do all the things in humans that you would do in model systems, but there is still so much we need to understand. So, yes, this is definitely more than just a cousin to what Matt has done. He was there at the start of this field. I consider myself part of the first generation of geroscientists who trained specifically for this, and that is largely thanks to the groundwork Matt laid.

33:15 What makes a drug a real gerotherapeutic?

Daniel 00:33:16

I’m curious to better understand what distinguishes geroscience from other fields. Clearly, geroscience is the study of aging, but I wonder what makes a genuine gerotherapeutic—a drug that treats aging.

There are a few ideas intersecting here. One is the idea of targeting aging biology directly. Another is simply making preventative medicines. Some argue that GLP-1s are longevity drugs because they reduce the risk for various endpoints we associate with aging.

To what extent do you think the distinction of a longevity therapy as being rooted in aging biology is important for the geroscience field?

Dr. Jamie Justice 00:34:00

I’ve heard this a few times. I am actually a little more tolerant regarding the definition of a gerotherapeutic. Does it have to target one of the specific hallmarks of aging biology? Not necessarily. I am effect-driven.

That distinction falls in line with my competition model. I’m looking at the endpoint, the milestone, and the final line you have to cross. That is effect-driven. When I’m evaluating a therapeutic, if it targets something within the biological system, there should be a plurality of effect.

If a drug works on only one specific disease category or pathway, it is probably not a gerotherapeutic. However, I have a hard time saying it must be limited to one of these central mechanisms of aging. Targeting a finite set of things is a very traditionalist drug development approach, and I don’t necessarily think that applies to geroscience.

You can look at the biology of aging for clues on what to target without necessarily targeting only one mechanism at a time. You might look at several. If disease is the mediator towards longevity, it can’t be a single, rate-limiting disease. It can’t just be cardiovascular disease or a single metabolic pathway leading to changes.

I’m looking for a plurality of effect, not strictly the mechanism used to get there. This is actually aligned with the earliest developments in the field. One of the first interventions we studied for its effects on longevity was caloric restriction.

Caloric restriction is not related to a single mechanism; it’s as “dirty” a drug as you could possibly get. Dietary approaches, exercise, lifestyle, and circadian rhythm are all important to the field of longevity and geroscience. They are just alternative ways to achieve that effect.

36:43 Where the field actually is, from telomeres to peptides

Eric 00:36:43

There is a really interesting set of questions underlying all of this: where are we at right now in the field of aging?

If we were to go to a layperson who is tangentially interested in the field and give them a benchmark of where we are actually at today, how would you summarize that progress marker?

Dr. Jamie Justice 00:37:07

Laypeople do tend to get really excited. I occasionally use Claude just to get a pulse check on the average perspective, and it has definitely shifted.

A few years ago, when we were talking about TAME or geroscience, the question I most encountered was about telomeres. People got really into telomeres, and that dominated the conversation. Then it moved into clocks.

Right now, the hype cycle is focused on epigenetic reprogramming. Ten years ago, scientists were talking about senolytics. When the average person asks me a question on the street today, they ask about peptides.

But as you were just asking, is that a longevity drug? Peptides are a class of drugs. My dad called me the other day and said a guy at his gym told him he could give him peptides for his shoulder. He asked what I thought about peptides for longevity.

I told him that is like calling me and asking what I think about “those drugs on the street.” It is such a non-thing. However, I try to listen to those discussion points, and right now, there is a lot of talk around peptides.

Many people know about metformin, and a lot of people are talking about GLP-1s. I would say the GLP-1s, in particular, are helping our case on both the regulatory side and in public discussion.

People are starting to see that these drugs have a plurality of effects. It doesn’t seem to be working under a single pathway. We don’t have a good clue yet what the mediators of that pathway are—whether it’s weight loss, reward, or something else.

I am ambivalent about whether it is called a longevity drug or a geroscience drug. Regardless, it is helping us make the case because, for the first time, our regulatory agencies have to figure out how to approve or reimburse for a non-disease indication.

If it’s not just being used for diabetes and doesn’t seem to be for any single disease class, they have to think on their feet about how to apply this. I am a champion for the public applying pressure to say they want these treatments. It is causing change that we can use to gain momentum.

Everything is peptides right now.

40:33 The peptide problem: gray markets and autonomy

Eric 00:40:33

That is really fascinating. I have many people talking about peptides in my life as well, and I struggle to formulate a thoughtful answer without being overly dismissive.

There is some exciting science going on in the world of peptides broadly. But going to a gray market peptide dealer and injecting yourself with something when you have no idea about the chemical purity, composition, or PK/PD profile is risky.

Maybe it will do something. It is definitely doing something, but we don’t know what it is doing. That’s really the takeaway.

Daniel 00:41:05

What is the positive thing about the discussion of peptides? I think this is very positive for the geroscience field because it’s not just about aging biology; it’s about the idea that there are drugs that can benefit me even if I don’t have a disease.

Beyond conviction in aging biology, that is what we need conviction in. We need to show that there is a market and an ability to develop drugs that can prevent multiple diseases.

Dr. Jamie Justice 00:41:18

Exactly. It has given people a platform to discuss risk and benefit. We place so much on our regulatory officials to assume safety and risk for the population at large.

In doing so, we take a certain amount of autonomy away from those who are seeking treatments, whether for a disease or not. People should have autonomy over what risks they are willing to take as long as there is appropriate consent and good data practice.

That data can then be used for proof of concept towards building regulatory approval. There are really intelligent ways to do this that don’t involve the black market.

I told my dad he needed to change his gym. There are ways to do this that are above board, used for the public good, and allow autonomy in determining risk even in the absence of disease.

GLP-1s are peptides, and they are starting that discussion for us. Does it go toward the gray market? Does it get really risky? Is there a lot of snake oil? Absolutely.

I have a little bit of tolerance for that because I don’t necessarily know what will or will not work. That is why I am obsessed with setting up measurements and establishing frameworks.

We shouldn’t be gatekeepers. If somebody wants to figure this out, we should invite them in. We should ask them to get ethical approval and oversight, and then provide testing frameworks.

This is how citizen scientists can figure out how to test things properly, rather than just on themselves or a friend. An N of 1 doesn’t have to stay an N of 1. You can aggregate the data to become a scientist and test your ideas rather than trusting an influencer on TikTok.

43:40 Is the first person to reach 150 alive today?

Daniel 00:43:40

Regarding the state of the longevity field, we don’t really have any approved longevity drugs today, although you can debate things like GLP-1s.

There is nothing on the market today that we think is going to give us a dramatically longer lifespan. But in another interview, you commented that you think the first person to live to 150 is probably alive today.

Can you say a bit about what gives you that optimism?

Dr. Jamie Justice 00:44:04

I see so much excitement right now. I get asked this question a lot. My son recently asked me, “Mom, do you think it’s you?” I don’t know if it’s me.

However, I know about the “alive today” bet that started in the ‘90s with Steven Austad and S. Jay Olshansky. It has carried through, and I think it’s a lovely bet. I still believe it’s possible.

We are harnessing enough novel technology that progress won’t just come from therapeutics. A great misconception is that it’s only about technology and therapeutic development.

One of the key failure points for drug approval and testing is finding the right patient. I think the bigger unlock will be pairing novel therapies with a very personalized approach.

Whether we use technology, AI, biomarkers, or devices, we must figure out the appropriate pairing of an agent to a specific person at a specific time. This will require an immense amount of data.

Ultimately, the shift will be away from the idea of a “wonder drug.” Instead, it will be about finding what works under a specific condition, at a specific time, and in a specific setting for you.

Unlocking that personalization requires data. We can do this intelligently by figuring out how to better leverage data from people who are experimenting on themselves, running large trials, or starting animal models.

46:01 Building the dataset behind personalized longevity

Eric 00:46:01

What methods for effectively scaling data collection from human populations have most excited you and the team at XPRIZE?

Dr. Jamie Justice 00:46:14

I’m excited to be at the center of building a prospective dataset. I did not leave my lab just to give away money, though I do hope someone wins so that we can award it. What truly got me out of the lab was the opportunity to build these datasets.

For the first time, we are working completely agnostic to treatment type. We have different populations from around the world contributing to common protocols. We will have biobanked samples, time-locked data analyses, and centralized phenotyping.

I can’t wait to award the money, but I’m even more excited for that dataset to go live. We can then begin probing responder and non-responder analyses and considering polygenic risk scores across global populations.

We can look at various strategies working under similar pillars or hallmarks, whether in nutrient sensing or immune targeting. We might see different combinations emerge from that.

This will be the first small use case for an approach that I think will model out very well for clinics. Early adopters can contribute to that dataset.

The challenge with relying only on early-adopter clinics is that they tend to serve high-net-worth individuals. You might not get the population diversity needed to ensure the data is global and equitable.

We need to figure out how to match this with learning health systems at a larger scale. We must do that thoughtfully and parsimoniously. We need a lot of data, but it cannot be done with an unlimited budget.

We are figuring out the best proxies in these early stages to build necessary and actionable data. Our immediate hope is to build a framework and a dataset that regulators, payers, and individuals consider important.

We often don’t stop to ask people what is important to them. We defer to regulators, scientists, and payers. We don’t often ask, “Dad, what would motivate you? What’s important to you as you age? Is it your blood pressure, or are you more worried about a heart attack?”

When you ask people what’s important, it always comes back to function. How they feel and function matters almost more than survival. We must use this as a process and remember to put people first, not last.

49:27 What motivates Jamie

Daniel 00:49:29

Speaking of motivation, what motivates you to work on this problem?

Dr. Jamie Justice 00:49:34

It is the best, most fun challenge in the world. Thinking about how we age is more significant than any disease class. It addresses the fundamental questions of what makes us healthy and what life means.

It involves thinking about the bounds of our existence and where we might go next. I love complexity; it makes me excited. I don’t like simple solutions, so I enjoy being in the middle of this.

I also love working collaboratively. This competition is the greatest collaboration I have ever been part of. We have teams contributing to the science, patients sharing what they want, and regulators helping inform the work.

It breaks down the hierarchies around science regarding who gets to do it and who gets to participate. On the aging side, it is about life itself.

What does health mean relative to disease? Should we continue to define our existence by how many diseases we do or don’t have? It’s about the very bounds of humanity.

Dr. Michael Koeris directs DARPA’s Biological Technologies Office (BTO), the group that leverages biological properties and processes to revolutionize our ability to protect the nation’s warfighters. Prior to DARPA, Dr. Koeris was a highly successful biotech entrepreneur, holds over a dozen patents, and was a professor of bioprocessing.

DARPA is a research & development agency within the Department of War. Created in response to the launch of Sputnik in 1957, DARPA represents America’s commitment to never again face a strategic technical surprise. The agency has funded many transformational technologies, including mRNA vaccines which were pivotal during COVID.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

0:00 Intro
2:13 DARPA’s efforts across combat casualty care, synthetic blood & soldier enhancement
6:32 GLP-1s and the new paradigm for pharma of enhancement rather than treatment
11:57 How DARPA can fast track drug development to get us longevity medicines sooner
14:40 Turning red blood cells into nanobot drug factories
19:13 Blood as the universal management layer for the body
21:24 Why Mike is a LinkedIn influencer & how to collaborate with DARPA
25:07 Replacing silicon with biology for more efficient compute
28:37 Why non-invasive brain computer interfaces are the future
30:07 How Cortical Labs has created the SDK for coding with neurons
32:09 How to scale up physical & synthetic data collection to decode biology
35:52 Protecting our crops & our civilians from AI-designed biological weapons
39:14 Reshoring biomanufacturing to ensure American resilience
41:36 The unique danger of biological weapons & how to address them
46:51 The acceleration of technological progress & a call to work with DARPA to build the future

Transcript

2:13 DARPA’s efforts across combat casualty care, synthetic blood & soldier enhancement

Daniel 00:02:13

Dr. Michael Koeris, welcome to the Free Radicals podcast.

Dr. Michael Koeris 00:02:15

Thanks for having me.

Daniel 00:02:17

At DARPA, you’re investing in some of the most futuristic technology. Could you give us a broad view of how you picture the sci-fi future? What do you imagine?

Dr. Michael Koeris 00:02:34

The mission of DARPA has always been to prevent and create strategic surprise using technology. We are fundamentally a technology development organization, and it is a great privilege to push on technologies we may see in the very near future.

DARPA’s job is to take the future that is 10, 20, or 30 years out and pull it into a horizon of two to five years. Our technology office, the Biological Technologies Office, covers both macroscopic and microscopic biology.

We look at the warfighter as a human system, but we also look at nano-technologies and viral threats. There is a lot happening in both spaces.

One thing we are particularly focused on now on the macro side is combat casualty care. If you are wounded in combat, you need care. Our current paradigm is to evacuate you as quickly as possible.

That works when you have air superiority and access to the battlefield, but we are worried that won’t be possible in the future. Observing the conflict in Ukraine and Russia over the last couple of years, we see a very different medical scenario.

If we can’t bring the wounded soldier back, we need to bring care forward. We need to invent new things, such as synthetic blood. If blood leaks out, you need to put it back in.

Donated blood is vital, but it doesn’t solve the problem because it has a short shelf life—about 30 to 40 days—and requires refrigeration. In a combat scenario, carrying a refrigerator is very limiting.

It would be much better to have shelf-stable, universal blood that can be carried in a rucksack. DARPA’s superpower is the ability to try things multiple times. It didn’t work the first time, it sort of worked the second time, and now it seems to be really working.

We are also looking at autonomy and what AI can do for us. This involves running on the edge where you can’t call home, as well as robotic support because we never have enough hands.

On the performance side, we want to support warfighters in being the highest-performing versions of themselves. Many have heard of GLP-1s for weight loss, but they also offer cognitive performance benefits that we’re interested in exploring.

6:32 GLP-1s and the new paradigm for pharma of enhancement rather than treatment

Daniel 00:06:32

Let’s dive into that last piece regarding enhancement. Could you tell us about those programs?

Dr. Michael Koeris 00:06:39

We don’t have an active program in cognitive enhancement yet, but both cognitive and physical enhancement are very important. I am excited to figure out how we can help warfighters overcome the burden of not sleeping enough.

I drink coffee not because I’m sick, but because I’m tired. This is a fundamentally different paradigm for drug development. We’re not thinking of these things as medicines, but as molecular support for a mission.

We need to look at what is possible on both the cognitive and physical sides. Soldiers undergo enormous stresses, often developing tendonitis in the Achilles or other major tendons.

Daniel 00:07:41

Podcasting is a bit easier on the body than warfighting.

Dr. Michael Koeris 00:07:43

That is a fair point. Whether you are a weekend warrior or an aggressive athlete, these issues occur. We see soldiers medically separated because they have recurring tendonitis that can no longer be treated.

This is devastating for them because it is their chosen profession. It is also bad for the force; you lose highly trained people who can no longer perform the mission. We would love to change that.

The second thing we care about is muscle loss, not muscle gain. We know how to build muscle through exercise and nutrition. Muscle loss is different; it affects everyone constantly. It occurs during aging, with various pathologies like cancer, and even with the use of GLP-1s. This has strong resonance on the civilian side.

For us, muscle loss begins the moment a soldier is deployed. They are no longer following their usual routines, sleeping well, or eating right. They are engaged in high-stress, strenuous activities that differ from their base training.

As the musculature attempts to adapt, muscle loss occurs almost immediately. Over long deployments, soldiers don’t necessarily become thin, but they lose muscle in ways that increase their risk of injury. A twisted knee can take someone out of the fight just as effectively as a major injury.

This problem is most severe for astronauts, but these are the challenges we are addressing. We are eager to explore this space with anyone interested. We are very open to discussions with startups, large companies, and academics. We can work with almost anyone.

Daniel 00:09:23

This mirrors conversations we’ve had with people in the pharmaceutical industry. You mentioned GLP-1s, which are increasingly used by everyone, not just those with specific diseases.

Treatments that help warfighters recover from injuries are also useful for the general population. There is a trend toward consumer pharma targeting massive indications. How does DARPA capitalize on these pharmaceutical trends to accelerate progress?

Dr. Michael Koeris 00:10:00

You’re exactly right about GLP-1s. I recently obtained a prescription myself to explore the potential cognitive benefits. GLP-1s weren’t designed for cognitive enhancement, and the effect isn’t uniform across everyone.

Much work remains to be done if we want to design something specifically for that benefit, but it’s an exciting area. Traditionally, pharma designs medicine for a narrow population of sick people. Now, we are looking at designing substances for everyone.

Consider coffee; it is a substance many people use daily and is reasonably well-tolerated. When designing drugs for general use, you encounter a wide spectrum of administration that pharma typically avoids in favor of specificity.

Furthermore, we aren’t always targeting a specific medical indication. I drink coffee because I’m tired, not because I’m sick. That is an important distinction pharma should embrace. We want to work with more partners in this space.

As part of the government, we can engage with the FDA, the USDA’s Agricultural Research Service, and the EPA. This “whole-of-government” approach allows us to address this macro trend.

There is a significant opportunity to push this forward. We want to influence this space and encourage people to think about what these advancements mean specifically for younger warfighters.

11:57 How DARPA can fast track drug development to get us longevity medicines sooner

Daniel 00:11:57

I hadn’t considered how DARPA interacts with other government agencies to accelerate progress. We often discuss gerotherapeutics—drugs that treat the diseases of aging—and you’re touching on that.

Are there active DARPA programs focused on making it easier for the FDA to approve these types of treatments? I would love to hear more about that.

Dr. Michael Koeris 00:12:26

DARPA isn’t a regulatory or policy agency, so everyone needs to stay in their own lane. However, a mechanism has existed for nearly a decade between the DOD, the FDA, and HHS where we meet every six months. This group includes the FDA Commissioner, the Assistant Secretary of Defense for Health Affairs, and leadership from DARPA.

We discuss the Military Priority Products List (MIPPL). We have priorities that are often unrelated to what the civilian population needs, wants, or is willing to pay for. One program on that list is our synthetic blood program called Rapid/F#.

The FDA has committed, through an agreement between the DOD and HHS, to review programs on the MIPPL with absolute priority. This is very powerful. Priority review usually carries a high cost in the commercial world because those vouchers are tradable. While that isn’t the case here, it is a significant benefit.

Entities developing drugs, therapeutics, or devices for military medicine missions now receive absolute priority review from the FDA. They have a dedicated contact at the agency and maintain very tight connectivity. We are excited that this exists and that we are finally utilizing it effectively.

Beyond synthetic blood, we are working on various enhancers. There are positive tailwinds for this work, and mechanisms like the MIPPL are very useful. We also have discussions with other agencies that I cannot disclose publicly, but the interagency communication is a huge benefit.

One of DARPA’s strengths is our ability to explain technical concepts and run experiments. If we agree to disagree with another agency, we can run a test to gather data and then resume the conversation. That is a vital function we provide.

14:40 Turning red blood cells into nanobot drug factories

Eric 00:14:41

I’d love to dig more into the blood programs you’re advancing at DARPA. Help us understand the philosophy behind your approach. You mentioned getting blood to a point where it supersedes natural physiology in practicality and function—being room-temperature stable, transportable, and universally donable.

Where is the state of the art today in synthetic blood, and where do you hope to go? Where are you seeing the most progress over the next five years?

Dr. Michael Koeris 00:15:19

I want to explain our blood portfolio because we see blood as a multifaceted tool. We need mass, meaning a volume of blood that cannot be met through donations alone. You simply cannot physically donate the amount of blood required for certain missions.

We also explore what can be done with the blood we have through two other programs: RBCF, the Red Blood Cell Factory, and sRBC, or Smart Red Blood Cells. These programs explore how we can enhance the capabilities of the cells themselves.

In the Red Blood Cell Factory, we take donated blood and modify it with specific constituents. There are already medicines approved to modify the ability of blood to perform certain functions. One example involves 2,3-BPG (bisphosphoglycerate).

This is commercially known as Diamox, an altitude sickness treatment. While it is FDA-approved, it doesn’t work well when taken orally, which is how most climbers and soldiers use it. We are investigating whether we can put that chemical directly into the red blood cell.

That is where the chemical is supposed to go to help with oxygen affinity. Hemoglobin holds onto oxygen differently based on partial oxygen pressure, which is lower at high altitudes. This leads to altitude sickness.

By modifying the blood cells outside the body and then returning them to the patient, we believe we can achieve the desired effect without the side effects of oral Diamox. Furthermore, blood naturally turns over every 60 to 120 days.

This provides an automatic expiration date on the modification as it clears from the system, which allows for excellent management and control. On the Smart RBC side, we are looking at what can be done starting from hematopoietic stem cells (HSCs).

We are exploring how to integrate synthetic biology elements, such as sense-and-response circuits. This would result in an allogeneic blood product. We would take universal O-negative donor blood, grow it, and incorporate synthetic circuits during that process.

A red blood cell is anucleated, meaning it has no nucleus or machinery. It only contains what it acquired during its maturation cycle. To modify it, you must start upstream with the hematopoietic stem cell and ensure those modifications carry through to final differentiation.

This is technically challenging, which is exactly why DARPA is doing it. We are pushing the state of the art in base modification techniques that might not otherwise receive funding. These programs may or may not succeed, but that is the nature of DARPA.

We want our programs to be aggressive enough that they have roughly a 50% chance of success. We are looking for those high-impact breakthroughs.

19:13 Blood as the universal management layer for the body

Eric 00:19:15

When you think about the realized future of your red blood cell engineering programs, what does that mean for the warfighter or civilian of the future? Help us picture it.

Dr. Michael Koeris 00:19:30

Blood is the universal transmission fluid in our body. It manages all of our gas transfer and most of our nutrient transfer. It is an enormous communications layer, managing things like the endocrine system. It also serves as an inflammation management layer.

One reason I got into this field is to operate on these universal layers. We have many things we want to fix, including pathologies and defensive purposes. Much of the immune response runs through the blood.

The blood is also attacked by various viruses and bacteria. Bacteremia, for instance, manifests and multiplies in the blood. I want to be able to manage the body, especially on the endocrine and metabolic layers, through the blood.

My ideal scenario is providing whatever you need to be an effective, long-living human. We want to remove waste products as effectively as possible. While waste eventually goes through the kidneys, blood is the primary medium for that process.

In the future, we could enable higher metabolic rates or control pathologies like atherosclerosis. Atherosclerosis involves fats and cholesterols that aren’t properly bound in the blood. Currently, people take Lipitor, but it doesn’t always work.

We want to nip those issues in the bud throughout a person’s entire life. I want to reach a point where there are no limits on how often you can treat or manage your blood. We don’t get to choose a different body, but we can work with our own much better.

21:24 Why Mike is a LinkedIn influencer & how to collaborate with DARPA

Daniel 00:21:27

The future you’re describing is very exciting and appealing. It makes me think about the different leverage points for bringing that about.

With DARPA, there is coordination with different agencies and the funding of particular programs to push things along. But I’m also hearing you call out to startups to get involved.

Dr. Michael Koeris 00:21:54

Absolutely.

Daniel 00:21:54

What do you see as the role of media or culture in supporting these programs?

Dr. Michael Koeris 00:22:01

It is absolutely critical. I have a presence on LinkedIn for a reason. Having an audience is vital because we don’t do anything ourselves.

DARPA does not have secret labs, regardless of what you might hear. We are a funding agency, and we are proud of that. We fund the people who apply to us, which is why being on this show is a privilege.

I want to tell everyone listening: please engage with DARPA. Don’t wait until a program is already funded and public. At that stage, selections have already happened and approaches are already competing.

Just because you hear about a program doesn’t mean you can’t be part of it, but you have to engage with DARPA earlier. That is really critical, and it’s something I relish personally.

I want to talk with people about their approaches—the crazier, the better. Our mission is to figure out what is just inside the edge of possible. I need to hear from as many people as possible to help shape the ideas of our program managers.

As the office director, I’m not the primary person to talk to, but I love facilitating introductions. Program managers and technologists should collaborate on what is possible. That is how you shape the mind of a program manager and create new programs.

Secondly, we are always recruiting. Nobody stays at DARPA forever; it is the law. We serve tours of duty, usually for four years, and then we move on. This will happen to me too.

It is a feature, not a bug, but it means we must constantly recruit excellent people. We have an innovation fellow program for recent college or PhD graduates. If you have graduated within the last five years, please reach out.

Innovation fellows stay with DARPA for two years, while program managers stay for about four. These people identify what is nearly impossible and go after it with heavy funding.

That is how we achieve breakthroughs like synthetic blood. It is always a partnership between the program manager and the community that brings these ideas to life.

25:07 Replacing silicon with biology for more efficient compute

Eric 00:25:09

Speaking of ideas on the edge of possibility, the O-Circuit program has gained publicity over the past year. This program is dedicated to designing hybrid biological computer interfaces.

These computers utilize naturally occurring physiology, such as the immune and neural systems, to improve computation. Help us understand the origin of this program and its status today.

Dr. Michael Koeris 00:25:42

We didn’t come up with these ideas either. That is often true for DARPA. Sometimes it is a completely new concept—and I have one I want to share with you that doesn’t exist in the real world yet—but often it is something that has been percolating at a certain level.

We decide we need to push on it and organize it into a larger form. OCircuit is representative of that. We also pay attention to the rest of the world. Everyone is looking into compute: how many chips there are, where to get the latest GPUs, and how to find the energy for them.

There is an implicit assumption that the way to run compute going forward is essentially more silicon and more electrons. It turns out that nature doesn’t do it that way at all. Our brains run on a very different energy performance envelope.

DARPA program managers like Jeff Zaleski have considered what it would look like if we could take what a human brain does—running on about 100 watts—and apply that to compute. That is significantly less power than the hungry applications running on silicon chips.

Our synapses are not as fast as GPUs; they are a couple of orders of magnitude different. However, we still do fascinating things with neural compute. In the OCircuit program, DARPA is exploring what types of compute are useful to embody in a biological system.

It is a completely different way of thinking. There are things you will always want to run on a silicon chip, but there are likely also very useful things to run on an actual wet neural side. Those two things have not been put together yet.

DARPA is providing the funding to see what it looks like when we combine them. It won’t necessarily look like an iPad, but eventually, there will be an easier way to interface these systems. DARPA often coalesces an industry by pulling together a community of interest that then starts to propagate itself.

Right now, this is very difficult to do on your own. You need a lab, you need to know how to handle neurons, and you need to know how to connect them with electrodes to a training layer. It requires a set of skills that is not easy for any one person to hold. Through these programs, we can standardize the process and grow the field.

28:37 Why non-invasive brain computer interfaces are the future

Daniel 00:28:39

You’re describing ex vivo neuron-computer integrations, but there is also an in vivo version. There is a burgeoning brain-computer interface industry.

Dr. Michael Koeris 00:28:49

I strongly agree. DARPA has worked in brain-computer interfaces for a long time, mostly for prosthetics. Many wounded warfighters returned during the Global War on Terror, and DARPA did a lot of work in that area.

We didn’t invent prosthetics, but we worked on BCIs to give people functional limbs. We are very proud of the Luke Arm, which was developed by Dean Kamen and approved as a medical device by the Veterans Administration. It connects via BCI into your nervous system, and you can manipulate it almost like a real arm. It is complicated, but very cool.

Regarding brain-computer interfaces, we believe invasive BCIs are useful medically for paraplegics or those with severe disabilities. But for the average person, I don’t want my skull drilled and electrodes placed on my brain. Beyond being an unpleasant experience, it isn’t long-term stable. The brain does not like having electrodes sitting on it.

BCIs are super useful, but invasive versions are limited. We have to think about what a non-invasive interface looks like. That is a very interesting frontier. Non-invasive is the only approach that will truly scale.

30:07 How Cortical Labs has created the SDK for coding with neurons

Eric 00:30:09

Cortical Labs is one of the companies operating in the bio-hybrid compute space. They became famous a few years ago after publishing a paper in Cell where they set up a bio-hybrid computer to play Pong.

More recently, they demonstrated the computer playing Doom on its own. Help us understand what this work means for the future of compute.

Dr. Michael Koeris 00:30:38

The team at Cortical Labs really knows what they’re doing. They are excellent scientists and engineers. They developed a device called the CL1, which is a completed compiler and incubator.

It is a system that keeps a neural chip alive and connects it to a Python interface. You can then write code to train the neural net. This is a capability you can buy now. It isn’t cheap, but it isn’t hyper-expensive either.

It allows people to figure out how to train neural nets. We have all been trained as neural nets, but we don’t actually know how it works. There are reward functions and memory mechanisms involved, and nobody has fully solved that yet.

Cortical Labs has pushed this forward by providing an SDK. You have a box you can play with, so you don’t need to know how to feed and husband the cells or connect all the wires. They will likely enable a much broader set of researchers to work with this technology. It’s super exciting.

32:09 How to scale up physical & synthetic data collection to decode biology

Daniel 00:32:10

That is one version of programming biology—programming neurons. There are also aspects like decoding the processes within a cell. Can you tell us about the work you’re doing there?

Dr. Michael Koeris 00:32:24

One of the reasons I came to DARPA is that I would love to push the soft condensed matter sciences, especially chemistry and biology, towards being simulation-first rather than experimentation-first. That is a hard ask, but at DARPA we decompose that ask by looking at what we need.

We actually need a lot more data to train models, because more data is better. That is still true today. People may think we generate a lot of data in biology, but that is not actually true. Relative to the complexity of the human body, it is vanishingly little data.

We have a data volume problem and a data diversity problem. We get a lot of nucleic acid sequencing data because it is cheap and widely proliferated, but we get very little metabolomic or proteomic data. That is one of the things we need to attack.

We have an ongoing program called PROSE, which stands for protein sequencing. It runs out of the Microsystems Technology Office because we have to build systems that can help sequence proteins using Nanopore or optical resonance approaches. These are fascinating approaches.

We are also looking at various programs that take a synthetic data approach. While real-world data is good, there is always a cost to generating it at the right scale. Some of our program managers are very clever computational biologists and experts in molecular dynamics and simulation.

There are regimes in which we can invest in compute and essentially pay for it for a year or two to generate a large volume of data. Then we train on that data to see if it generalizes, much like AlphaFold. One of the reasons AlphaFold was so successful is that the protein databank already existed.

It took 30 or 40 years to generate all those wonderful structures in the protein databank, and I do not have that kind of time. The question now is whether we can simulate high-quality structures or traces for proteins to learn from.

We are currently running the NODES program to see if there is another way to approach a generalizable algorithm using simulated protein traces. Physical data generation and synthetic data generation are both very important, and we are continuing to pursue them.

We have not reached the end of investing in these technologies because there is still metabolomics, proteomics, and cell-to-cell communication. We need to figure out how cells talk to each other through connections or vesicles.

The complexity ladder in biology is staggering. It is very easy to lose sight of how much complexity is not yet captured or even attacked. At each layer, you have wonderful emergent properties that are not clear from the sum of the underlying parts. That makes biology beautiful and sometimes frustrating because there is so much still to learn.

35:52 Protecting our crops & our civilians from AI-designed biological weapons

Daniel 00:35:55

Capturing all of these forms of data and encoding them in foundational models is clearly very important for biology. But DARPA also has a remit to help warfighters in the near future. Can you help me understand the vision there?

Dr. Michael Koeris 00:36:12

I am delighted you asked that question. One reason we do this is to understand the human layer to design better molecules and medicines. That also translates to livestock and crops.

Anything and everything that runs on a DNA-based operating system is, to some degree, attackable, and that is almost everything in the world. We take agricultural protection very seriously. We recently started moving more into the agricultural space.

We generated a Memorandum of Understanding with the USDA that SecWar and SecAg signed together. Separately, we did one between DARPA and the research function of the USDA because we believe we need to be more closely integrated. We need to protect the plant layer because if our corn goes away, it is bad for everyone involved.

We also care about understanding molecular machinery from a threat perspective. Not everyone has the best intentions when they use AI. AI is a tool that can be used for many different things. While AI companies use control strategies, there are many models in the world and many ways to jailbreak them for nefarious purposes.

If someone has ill will, they might want to make a chemical or a biological threat agent. We care about how the cell works because AI can come up with things we have never seen before. If a new agent pops up or starts spreading around the globe, we need to know what it does to us.

I would much rather use a virtual testbed than work in a high-containment lab for six months. Once we have characterized and sequenced a new agent, we can put it into software in a virtual testbed and let the simulation run.

The simulation should tell us what that agent does and what we need to do to defend against it, whether it’s a countermeasure or a therapeutic. Simulation is much more scalable and runs quicker than a highly contained lab. That is one of the primary reasons we do so much work in the simulation space.

39:14 Reshoring biomanufacturing to ensure American resilience

Daniel 00:39:17

When we talk about threat detection and solutions, the actual manufacturing of the solution is a crucial piece. There have been many discussions within other branches of the DOD about the importance of domestic manufacturing.

It makes me wonder about biomanufacturing, where we’ve also outsourced much of the supply chain. How much do you think about the ability to onshore biomanufacturing?

Dr. Michael Koeris 00:39:42

A lot. My former faculty practice was bioprocessing and biomanufacturing, so I’ve worked in this space quite a bit. Biomanufacturing is one of the six critical technology areas (CTAs) that the research and engineering part of the DOD focuses on.

That means there is a dedicated lead for biomanufacturing and a collaboration with BioMADE, a manufacturing innovation institute funded by the DOD. Biomanufacturing is an important function because it allows us to manufacture various things differently than through purely chemical or physical means.

We are looking into opportunities to reshore manufacturing that has been offshored. This isn’t just biomanufacturing; it involves base chemicals and intermediate complexity chemicals. In previous decades, we offshored everything that was low margin or environmentally challenging.

We won’t get that back unless we find a different way of manufacturing. We have several programs running in this space, including SWITCH, Fleetwood, and Equip-A-Pharma, which we run jointly with HHS ASPR.

These programs develop manufacturing approaches that are containerized and modular. While manufacturing often needs to be centralized to achieve economies of scale, we are investigating if we can be nearly as efficient in a modular way. We’ve made great strides and progress on this.

41:36 The unique danger of biological weapons & how to address them

Eric 00:41:38

I’d like to touch on biosecurity. This concept is central to programming biology. In computer science, the ability to program hardware with software is incredibly powerful, but it also creates structural deficits.

The cybersecurity industry is a multi-trillion dollar global effort dedicated to creating counter-security measures against attack vectors. The same concept now applies to life sciences.

When you can program not only biological medicines but also biological threats, what does that mean for the future of the average individual? What does it mean for private industry, the government, and DARPA?

Dr. Michael Koeris 00:42:32

It’s an interesting question. While policy is not our remit, DARPA must think about developing appropriate countermeasures. We assume that a nation-state or a non-state actor may eventually have the ability to do something dangerous.

That is the general threat management scenario we have operated under for decades. Biological threats have a singular distinction: they replicate. This makes scale a unique challenge.

A small initial event can lead to a rapid spread. At DARPA, we consider how to deal with a rapidly spreading threat from an unknown source. We are leaning into better situational awareness and an understanding of global viral activity.

Currently, no global capability exists to monitor all viruses. We use satellites to look at the Earth, but satellites can’t see viruses. We are interested in an early warning system for the viral layer.

Viruses are a primary concern because of their spreadability. We need a global push to look at the molecular layer—the sub-100-nanometer layer—at a frequency that matters. Sampling a wastewater stream once a year isn’t helpful; we need revisit frequencies that make these systems useful.

Simulation works hand-in-hand with surveillance. Simulation tells us what is possible, and surveillance confirms if it is actually happening. Finally, if a bad event occurs, the response timeline must be extremely quick. During the end of the COVID pandemic, some organizations set a goal for a 100-day response to an airborne virus.

Eric 00:45:29

In the future.

Dr. Michael Koeris 00:45:31

We think 100 days is too long; it needs to be much shorter. To achieve that, we need to develop technologies that allow for a rapid response.

We may not be able to protect everyone from an initial event, but we must be able to react and nip it in the bud. That is our concept of deterrence and defense.

Eric 00:45:53

The mRNA vaccines developed in record time for COVID are a great case example. The first COVID vaccines were developed within days or weeks and taken through the FDA at record speed.

The result speaks for itself. It is a remarkable testament to the ability to program biology at scale and deliver medicines to entire populations.

Dr. Michael Koeris 00:46:17

People may not remember that DARPA was the initial funder of Moderna back in 2011. I remember this well because I wasn’t at Moderna or DARPA at the time.

I got a news alert and thought it sounded like an outlandish idea. I wondered who these government people were giving them money. Back then, this was not at all on the horizon.

DARPA often does that. It is one of our core requirements to think as far into the future as possible and develop technologies that help with potential future surprises or create surprise going forward.

46:51 The acceleration of technological progress & a call to work with DARPA to build the future

Daniel 00:46:55

We have just a few minutes left. We’ve covered a lot of ground and talked about a lot of interesting technology. I’d like to take a step back to discuss the nature of technological progress.

You’ve had many unique vantage points on this as an academic, a company builder, and now funding projects at DARPA. Do you have a narrative or an idea of the source of technological progress?

Dr. Michael Koeris 00:47:21

That question should be talked about for much longer, but I’ll give a brief answer. Technological progress is increasing in speed; it is happening faster and faster.

I’m not just talking about the new capabilities that come out every few months with AI, though that is a factor. The knock-on effects of moving faster in all parts of the industry—sometimes connected to AI and sometimes not—is an interesting observation.

Everyone seems to be faster at operating at the technological development frontier. That is really healthy and I like that overall. It is also challenging because the cycle times become shorter and shorter.

DARPA is a very rapidly moving organization, but we’re still a federal organization, so there are limits to how fast we can move. There is a pro and a con to how fast the world starts to move.

Increasingly, progress will come from everywhere in the broadest human sense possible. It used to be that if you weren’t at one of the most well-connected research labs, you didn’t usually get to talk to DARPA.

We operated in our own echo chamber. That was never really true, but it was often the case that some circles were better connected than others.

Through the proliferation of various technological approaches, the audience that uses them, thinks about them, and has the ability to make forward progress is much broader than it ever was. I’m super excited about that because it enables more people to contribute.

There is, of course, a counter. It allows people to do nefarious things if they want to, but so far we’ve observed mostly the positive benefit. That’s super exciting.

If progress comes from AI at some point, I am delighted. In other parts of the agency, we have programs looking at what is possible with AI from a creative perspective. I’m sure lots of people outside are doing this too.

I don’t mind if it is a hybrid idea, partially generated by a system and partially by humans. That doesn’t matter to me. The point is that we would like to move technology forward more quickly. It is now an enabling technology for everyone.

Daniel 00:49:43

We have just one minute left. Anything else you’d like to leave our audience with?

Dr. Michael Koeris 00:49:48

Talk to us, please. Talk to us with a frequency that you may find surprising. The reason is that everybody turns over all the time.

Everybody has a ticking clock at DARPA. In fact, all of our badges have our expiration date on them. Everybody gets reminded every day how much time they have left.

That is why we move with the pace and haste that we do; my clock is ticking too. It also means that I will turn over eventually. You may know Mike, but Mike won’t be at DARPA anymore and you will need to find the new Mike.

There is a revisit you have to do with all the PMs, the various office leadership, and everyone at DARPA. I love talking to people about ideas.

One other important thing: don’t just pitch me your idea and try to get me to give you funding. That’s not productive.

What is useful is to describe the technology, and then maybe we can brainstorm something useful together. That co-creation act is a great privilege to deliver.

Daniel 00:50:51

Thank you so much for joining us on the Free Radicals podcast.

Dr. Michael Koeris 00:50:54

Totally radical. Thanks for having me.

Eric 00:50:56

Thanks, Mike.

We discuss why Kris rejects incrementalist approaches to radical life extension, and his argument that supplement, lifestyle interventions and small-molecule “geroscience” drugs offer only minor improvements at best to lifespan.

We cover the four part plan to cure aging spanning traditional drugs, advanced bioengineering, tissue/organ replacement, and biostasis (cryopreservation), and we dive deep for the first time on the podcast into replacement: body transplants, lab-grown organs, chimeric organ donors and even... brain replacement.

Resources & Links:

• Kris’s Book: ⁠Radical Life Extension⁠

• ⁠Longevity Biotech Fellowship⁠

• ⁠Repair Biotechnologies⁠

• ⁠The Economic Laws of Scientific Research⁠

• ⁠Tomorrow Biostasis

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

0:00 Intro
1:12 Why most longevity bets are too risk averse
12:09 The four approaches to longevity: drugs, bioengineering, replacement & biostasis
17:16 What’s going on with Peptides?
32:04 What is the path forward for replacing organs to extend lifespan?
43:41 What are the limits of Tissue Engineering?
55:46 Biostasis: pausing death until a cure exists
1:03:34 Identity & consciousness: chemo vs. cryo and using AI to upload our minds
1:07:07 Anarcho-capitalism, freedom & the role of government
1:13:24 Closing: pipeline tech & a call to action

Transcript

0:00 Intro

Daniel 00:00:32

Welcome to the Free Radicals Podcast, where we interview the scientists and builders working to dramatically extend human lifespan and bring about a sci-fi future where humanity has full control over biology.

Today’s guest is Kris Borer, author of the brand new book Radical Life Extension, in which he outlines the plan for solving aging. Kris is an entrepreneur with a background in robotics and is an angel investor in many companies working on longevity. Welcome to the podcast, Kris.

Kris Borer 00:00:59

Thanks for inviting me.

Daniel 00:01:01

Thank you for coming on. I’d love if you could share the broad inspiration for the book and the key message of Radical Life Extension.

1:12 Why most longevity bets are too risk averse

Kris Borer 00:01:12

There is an organization called the Longevity Biotech Fellowship (LBF) that you guys know about. We spend a lot of time thinking about longevity and how to make it happen as quickly as possible.

If you attend one of the fellowship retreats, you’ll get all the information you might want about good strategies for building technology to help people live longer and healthier lives. Not everyone can join LBF, so we thought a book might be able to give people a slice of that experience.

It provides an understanding of what we think are the most effective ways to treat aging and help people live longer.

Daniel 00:01:48

Awesome. We had Nathan Chang on the podcast maybe a month or two ago. Eric and I have both been to LBF retreats.

There are a lot of folks in LBF who are listeners to the podcast, but also people who haven’t been able to attend retreats. It will be great to give them a taste of the roadmap for solving aging, which I think we all agree is probably the most important thing in the world to accomplish.

Kris Borer 00:02:16

It’s an exciting time to be alive. There’s so much going on with longevity and a lot to be excited about. But at LBF, we really want to think about whether we can do better and if we are really moving the field forward as quickly as we possibly can.

When we sat back to think about it, we thought we could probably do much better. The reason is that there are many different strategies for treating aging, and not all of them are created equal. There are actually many popular strategies that are quite bad.

One of our goals with LBF is to bring people in, give them a taste of many different areas in the field, and then help them understand that if they really want to make as much difference as possible and save as many people as they can, they should focus less on some strategies and more on others.

To take some extreme examples, you might say we could potentially help people live a lot longer with supplements, and you might start a supplement company. That might be helpful for some people, but for the vast majority, supplements will not make a big difference in how long they live.

On the other end, you might use modern bioengineering technology to change people’s genes, perform genetic editing, and improve human cellular biochemistry so people live a lot longer. That’s much more plausible than supplements, but also much harder.

You actually find a lot of people saying they don’t know how to do genetic engineering, but they do know how to sell supplements. People tend toward things that are easier but lower impact. We want to encourage people to do the hard things that will make a bigger impact in the end.

Daniel 00:03:56

People understand the supplements versus bioengineering angle. I think where people probably get a little more confused, or where there’s more disagreement, is advanced bioengineering for radical life extension versus trying to cure cancer.

Everybody agrees we should cure cancer, and yet curing cancer won’t give us radical life extension. Can you say a bit more about how you focus on what really moves the needle when it comes to lifespan?

Kris Borer 00:04:18

If you cured all cancers, something else would eventually kill you. Curing all cancers gets you a few extra years of lifespan, which is great and something we definitely want to do.

But if we can focus more on aging damage and treat the underlying causes of aging and other chronic conditions of age-related diseases, then we can have a much bigger impact. It’s not that we don’t want to cure cancer; it’s that we want people to focus on underlying mechanisms that could not only cure cancer but cure other chronic conditions as well.

This is probably easiest to understand when you think about the difference between treating aging versus creating longevity drugs like small molecules. A scientist might find some metabolic pathway and see that when someone is young, this pathway is active, but when they are old, it is diminished.

If we design a drug that helps that pathway stay steady, that would be great. That will probably make people who are older feel younger, be healthier, and maybe live a little longer, but it won’t solve aging.

How many of these drugs do you think it would take to make people live to, say, 130? Could you do it with one drug? Can you just tweak AMPK a little bit and make people live that long, or would you need 10, 100, or 1,000?

Eric 00:06:10

The chance that we get anything more than a very modest effect from any one of these molecules currently available is very unlikely. We’re looking at maybe low single-digit lifespan extension from any one of these interventions.

It’s not clear that you would get synergistic effects from stacking them. Let’s say we took five of them and they each give us three years. Ideally, they are all synergistic, resulting in 15 additional years.

Added to an average lifespan of 80, you get to 95. To reach 130, you’re looking for an additional 50 years. By that math, you’re looking for 15 to 20 small molecule interventions at three years each.

Kris Borer 00:06:57

Even if that could work, you should be skeptical of that approach because drugs like that have side effects. Even today, when we give older people three, six, or nine medications, there are drug interactions that cause problems.

You can imagine that if you had 15 or 20 of these drugs to try and get an extra 15 years of life, you may actually be hurting yourself in ways that are hard to predict.

Curing cancer is great and finding drugs that can modulate pathways that change with aging is great, but if you target these things, you are unlikely to have as much of an effect on how long people will ultimately live.

That’s why the Longevity Biotech Fellowship advocates identifying the best strategies for radical life extension—not just a couple of years, but 10, 20, or 50 years—and going hard after those. We don’t want to waste time on things that might only have marginal effect sizes.

Daniel 00:07:58

We’re going to discuss things that can have larger effect sizes, but let’s debate this for a bit. Nobody is claiming—including James Peyer—that his drug will dramatically extend lifespan. I don’t think anyone serious believes that any small molecules currently in the pipeline can significantly extend human life.

Most people expect a step change in the available technology. We need advanced bioengineering. These researchers are trying to find near-term business opportunities that validate the underlying biology and stimulate the pharmaceutical industry’s flywheel toward more advanced therapies.

Is there value in building these small molecule drugs based on the geroscience paradigm? Or is your take that it’s a complete waste of time?

Kris Borer 00:09:09

My take isn’t that controversial, but you have to consider the counterfactual. These drugs definitely add value and will help many people. However, if you choose one path, you might help a lot of people slightly. If you take a path toward radical life extension, you could save people who would otherwise die.

Every dollar devoted to one strategy instead of another represents someone who might or might not reach radical life extension. Simply saying something has value isn’t enough to justify it; you have to consider the opportunity cost.

The opportunity cost of devoting hundreds of millions of dollars to small molecules is quite large. We have identified other strategies with potentially much bigger payoffs.

Daniel 00:10:02

There is a chasm between the niche longevity biotech community and the traditional pharmaceutical industry. Billions of dollars flow into small molecule development, while longevity advocates argue for more investment in areas like cryostasis and replacement therapies.

Small molecules aren’t going away, and the investment there is massive. How do you envision the longevity field influencing traditional pharma? I don’t want to write off the entire industry. How can we leverage their capabilities to further the roadmap outlined in your book?

Kris Borer 00:11:03

The goal is to shift time, attention, and resources toward higher-impact strategies. We aren’t suggesting we abandon small molecules or traditional pharma entirely. If you’re working in pharma, you should focus on things like underlying mechanisms of aging rather than niche issues that aren’t critical for human longevity.

If you have the skills for bioengineering, that might be a better use of your time because the potential impact is much higher. We want to shift people toward these more specialized areas.

Even if the potential impact were the same, current funding levels are vastly different. A marginal dollar going into small molecules won’t make much of a difference, but that same dollar could have a much bigger impact if invested in newer technologies.

12:09 The four approaches to longevity: drugs, bioengineering, replacement & biostasis

Daniel 00:12:10

Let’s get into the specific strategies. This will give people who aren’t familiar with them a sense of how impactful different technologies can be. Can you give us an overview of the four different ways of tackling aging? Then we can dive into the most impactful one first.

Kris Borer 00:12:32

The first category is traditional drugs. You identify a molecular pathway and try to modulate it with a small molecule. Recently, newer techniques for modifying biology have emerged, which we call advanced bioengineering. This includes mRNA therapies and genetic editing.

These techniques are much more powerful. You can do so much more when you treat biology like a computer program and reprogram it. It’s so powerful that we don’t yet know everything possible. In a few decades, we’ll be able to program cells just like computers.

In the meantime, some strategies can have an impact sooner. One is replacement. Common procedures like heart or kidney transplants cure diseases, but they also cure aging in the replaced tissue. If an older person receives a liver from a healthy 20-year-old, that liver will be much younger and healthier.

If you can do that for every tissue in the body, you can rejuvenate someone completely. It’s not that straightforward, but that is the general idea. At the Longevity Biotech Fellowship, we think about who we’re trying to help. It’s not just young people who have time to wait for these therapies to reach the market.

People are dying every day. We want therapies that help people regardless of their age or condition. If you need something right now, the only option is biostasis. I’m happy to go into the details of that whenever you like.

Daniel 00:14:32

I’d like to emphasize that when we think about plans for solving aging, we should compare them to what we have today. Currently, as you get old, there is essentially nothing you can do. We have treatments for specific diseases like cancer, but even if we cure that, you end up dying of something else.

There is no alternative today to the slow, accelerating degradation that leads to death. I’d like to start by talking about replacement. Imagine if you could get a body transplant.

Compare the best medical care available today to having a 20-year-old body. That demonstrates the impact replacement can have. I’d love to hear more about replacement, perhaps starting with the concept of “bodyoids.”

Kris Borer 00:15:47

The best we can do now is lifestyle interventions. If someone has a good diet and exercises, they will probably last longer, but there’s a lot of randomness too. By the time you’re 70, 80, or 90, no matter how much you exercise or diet, your body is going to be damaged from metabolic stress and the insults of regular life.

There is no drug or therapy that can do anything about that. However, if you swapped in a young body, all those problems go away. For those who are not aware, body transplant is a proposed therapy where you don’t just transplant a single organ, but everything below the neck.

If you did this, you would restore the function of all your internal organs. You would also provide a better support system for your brain, which would hopefully help it live much longer as well. While this is currently theoretical, there has been some proof of principle.

There have been studies where researchers attached the circulatory systems of young and old animals. This is called heterochronic parabiosis. When you attach a young animal’s body to an older animal, it actually makes the older animal much younger, healthier, and longer-lived. Conversely, having an old body attached to a young animal makes the young animal sicker and causes it to die faster. Replacing an old body with a new one could have a huge impact on healthspan and lifespan.

17:16 What’s going on with Peptides?

Eric 00:17:16

I’d like to dig into the Overton window and the social discourse gulf between the different groups that are important to this movement. Three categories come to mind: early adopters at the frontier of biohacking, traditional pharma and the medical community, and the broader population.

The traditional medical community and pharma don’t have total buy-in from the broader population, but they likely have more trust than the frontier biohacking community does at this moment.

What does the frontier of bioengineering need to do to shift the Overton window from “death is inevitable and good” to being more open-minded and trusting of the bioengineering community?

Kris Borer 00:18:30

Technology is what is going to convince people. When people see demonstrations of life extension or therapies that make them look or feel younger, they are going to change their minds. All the theorizing we’re doing will convince some people to work on the problem, but for the vast majority of people, it will take technological demonstrations.

Eric 00:18:54

Yeah.

Daniel 00:18:54

Eric, you brought up the Overton window regarding body transplants. Growing clones for those transplants is obviously outside of that window. This is not something being realistically discussed in traditional pharma or biotech circles.

Kris, your contention is that we just have to make progress and show amazing results. If we show a mouse whose lifespan we doubled through a head transplant, people will be amazed. Despite the ethical debates, those technological wins capture people’s attention and allow us to progress from there.

Kris Borer 00:19:42

I don’t think we can jump straight to body transplants, even in animals. People would be creeped out by that, even though it would be a huge technological leap for the medical industry. There’s no way public reception could handle that, and the regulatory response could be quite bad.

However, we don’t have to jump right to it. If we are able to solve the organ shortage, we can make normal transplants much more common. When you increase the supply of transplantable material, progress happens naturally. Right now, we only have a few organs, so we give them to the sickest people.

If we had more organs, we wouldn’t just give young kidneys to people with late-stage chronic kidney disease; we might give them to mid-stage patients as well. If you have unlimited kidneys, why not give them to early-stage patients? Why not give them to people who are simply getting older?

We know their organs are going to fail at some point. There is a natural progression as supply constraints go away where you take a more preventative approach with replacement technology. When you start giving older people young kidneys, livers, or even limbs, people will naturally come around to the idea that replacement is a good way to prevent cancer, metabolic disease, or sarcopenia.

Eric 00:21:10

Yeah.

Daniel 00:21:10

The point is that there are immediate needs for organs under the traditional medical system. You mentioned in your Vitalist Bay talk that there are children in need of liver transplants. Anything we can do to increase access to livers for them is clearly good.

Part of this technological progress will simply be alleviating the immediate organ shortage. From there, we can build toward ever-increasing use cases for such organs.

Kris Borer 00:21:46

Absolutely. Once the supply problem is solved, people will realize that transplants work not just for acute organ disease, but also for aging. That is where we can move toward more advanced solutions.

Again, we don’t have to jump straight to body transplants. We can do multivisceral transplants where you get a new set of internal organs. Even just a new set of legs and arms could be hugely beneficial.

Eric 00:22:13

Another challenge with biotech generally is the first-mover disadvantage. Innovative approaches require clinical trials to achieve mass adoption, but there is a massive barrier to entry for implementing those trials and getting successful results.

How do you think about overcoming that barrier of the first-mover disadvantage when pioneering bioengineering solutions for longevity?

Kris Borer 00:22:59

For more traditional biotech or even the new advanced bioengineering, things will likely be okay because there are so many applications and so much money going into the field. People are going to be pushing those technologies forward regardless of what we do.

One of the benefits of having a large community of people who really want to solve aging is that they are going to do it anyway. Even if it is hard to push radical transplant technology like multivisceral organ transplants through, people want to save their parents and grandparents so badly that they will push forward regardless of the difficulty. For niche areas, it will take the radicals to make it happen.

Daniel 00:23:45

Eric, who told you solving aging would be easy?

Eric 00:23:50

Progress in these niche communities, where people are willing to tinker with their own biology, is an important part of how this movement advances. Let’s not forget that the origin of things like pasteurization, the first antibiotics, and the first set of psychoactive chemicals like LSD and MDMA were all originally derived from chemists and biologists who were considered crazy at the time. They were experimenting on things in dishes and putting them in their bodies to see how they reacted.

Decades later, these things become such a part of the status quo that you would be considered insane to think they didn’t work. There is a long-term shift in Overton windows—a preference cascade that happens after something has been so incontrovertibly proven true that people think you’re crazy for not believing it. But when it’s first happening, it seems insane to even attempt it.

Kris Borer 00:25:07

We need to let a thousand flowers bloom. We need the crazy people to try different things. All that matters is that we take the things that work and push those forward. I am all for people doing self-experimentation if they want to. If it works, great. If not, at least they tried.

Eric 00:25:26

One area where I see this experimentation happening very publicly now is peptides. There has been a huge surge in public interest around non-FDA-approved, gray-market compounded peptides that are purchasable online from various Chinese dealers.

There are a number of social events where people get together for peptide tasting parties to try out substances like BPC-157, sermorelin, or tesamorelin. They experiment to see if these naturally occurring or synthetic peptides might improve their appearance, change their sleep, or potentially increase their lifespan.

What is your view on this movement? What does it say about the broader longevity movement that it is kicking off now in such a big way?

Kris Borer 00:26:36

I guess I’m not that cool; I have not been invited to a peptide tasting party. Have you been to one? Does it work?

Eric 00:26:43

There was a peptide tasting party this past Tuesday that our friend Jeff Tang was hosting, but unfortunately, I could not make it.

Daniel 00:26:54

I’ve been to some peptide tasting parties, but I have not injected or consumed any peptides myself. People seem to have a lot of fun with it, and a lot of my friends love them. Kris, I’m curious to get your take on the peptide craze, even if you haven’t been to the parties.

Kris Borer 00:27:16

I haven’t done much research into peptides. If there is useful stuff there, let’s find it and get it out to people. One of the problems with biohacking is that people are throwing parties instead of conducting clinical trials, so you don’t get much good data. It’s hard to tell what is actually effective.

This is a significant problem for the longevity field because there are so many snake oil hucksters out there claiming that injecting massive doses of vitamin C will make you live forever. We know that is crazy now, but whenever there’s a new technology, people want to try it and find out. I would exercise caution, but everyone should feel free to do whatever they want with their own body.

Daniel 00:28:02

I think there are a lot of good things in people experimenting, and hopefully, we’ll learn from it. However, if it’s not done well, it’s hard to learn from it. It’s great that people care about their health and want to optimize it.

My concern is the same one I have with Bryan Johnson. You’re getting people to care about longevity, but my worry is that you give people a false sense of efficacy. People think that because there are hundreds of peptides to choose from, some combination is going to make them live forever.

It sounds dumb when you say it like that, but I think many people intuitively feel this can meaningfully extend their lifespan. They think this is the next generation of amazing biotech. But there’s no evidence to believe these random things will work. As you say in the book, solving aging is a very tough problem. The odds of randomly stumbling upon the solution at a peptide tasting party are very unlikely.

Kris Borer 00:29:11

It is very unlikely, but this idea has recurred many times throughout history. Ray Kurzweil wrote a book about how he takes 200 supplements to stay young and healthy until the next generation of technology can actually reverse aging.

That is not a totally unreasonable thing to do, but you have to decide for yourself what the benefit is for any particular use of your time and money. Should you spend all your weekends and nights researching peptides, or would you be better off going to the gym and getting some extra cardio?

Right now, we know you can get an extra five to ten years just from lifestyle interventions. Even if peptides work, they might only add a year or two. You might actually be hurting yourself by prioritizing them over proven methods.

Eric 00:30:02

The recent uptick in interest in peptides is related to the broader consumer health movement and a distrust of the traditional medical system. People no longer trust the medical system and medical experts as much as they used to after COVID.

Eric 00:30:28

Millennials have much higher agency regarding taking things into their own hands. There is also a contingent of highly educated, well-off coastal elites who are willing to experiment on themselves.

The results of this will likely be that mostly nothing happens. A small amount of people might perceive some benefit, many will perceive none, and a small number will be seriously harmed.

The biggest innovation here is that the same social mechanisms we saw in the 1980s and 1990s leading up to the internet bubble, the era of computers, and the era of AI are now unfolding in the life sciences. That is really exciting.

Kris Borer 00:31:31

I love that energy and want to help channel it into the most effective solutions. If people have the time, attention, and agency to make a difference for themselves and their families, we should mention that there are other options. They might be harder, but they could have a much bigger impact.

32:04 What is the path forward for replacing organs to extend lifespan?

Daniel 00:31:51

It is a good point that anything that gets people more excited about biohacking and biology is a good thing.

Let’s talk more about replacement because I am very excited about the long-term prospect of body replacement. Although I would prefer bioengineering, replacement is often discussed as a far-off, sci-fi concept, yet there is a ton of progress happening. Organ transplants are done routinely today. Why is this a viable path, and what are the most exciting developments?

Eric 00:32:36

Sure.

Kris Borer 00:32:37

We know from current transplants that you can cure diseases and obtain young, functioning organs and tissues from a donor. Replacement is a cure-all. It is a great therapy for both diseases and aging, but it has downsides because large-scale replacement requires surgery.

For small-scale replacement, you might just get a cell injection. For example, with CAR-T therapies, if you have a degraded immune system, we can engineer new immune cells and inject them to replace your system.

To treat aging on a large scale, you need to replace large volumes of tissue. The risk is that a percentage of people who undergo transplant surgery die as a result of the procedure. It is dangerous to open up your body.

The two problems are building a supply of replacement parts and developing techniques to deliver those parts safely. If we can do that, we have a cure for almost any disease and aging itself.

Whether it is cancer, heart disease from atherosclerosis, or simply being 110 years old, the doctor can use the same technique. If a low-risk drug is available, you take the drug. If not, doctors will have this fallback option.

Daniel 00:34:17

What are the main technologies being developed to alleviate the organ shortage and provide a supply for all these transplants?

Kris Borer 00:34:27

There are several technologies in development. Many people are aware of 3D printing organs, but unfortunately, it doesn’t work very well. We have shifted attention toward more promising science.

Some researchers are looking at xenotransplantation, where you take organs from animals like pigs and give them to people. If you genetically edit the pig, the organs are more tolerable, though they still have problems.

My two favorite approaches are chimerism and tissue constructs. Chimerism involves growing an animal that has human organs. You take an animal at the blastula stage—a very small clump of cells—and genetically edit it so it cannot grow a specific organ, like a kidney.

You then inject normal human cells. Those human cells will grow the kidneys while the pig cells grow the rest of the animal. You can then harvest the human organ for transplant. It is a more advanced version of xenotransplantation.

If you want to go straight for human tissue, you can use bioengineering to grow organs directly in a bioreactor. It would be nice to grow a kidney in a jar, but that doesn’t work well in isolation.

Newer technologies are trying to grow networks of organs together. These are tissue constructs where you take skin cells, hit them with reprogramming factors, and direct them to grow a specific set of organs—like a kidney, heart, lungs, and liver—surrounded by skin.

You put that in a bioreactor and, ideally, a year later you have organs ready for transplant. Ultimately, I think tissue constructs made from human cells will be the technology that solves the organ shortage.

Daniel 00:36:40

The other big piece is the brain. We can potentially replace your organs or your whole body below the neck, but your brain is also aging. That leads to the fascinating topic of partial brain replacement. Can you tell us about that?

Kris Borer 00:36:59

Brain replacement is the hardest problem in replacement. You can’t simply give someone a new brain because that would kill them. The brain is also particularly hard to replace because it is such an integrated network. As you know from your academic background, it is a complex web of neurons, supporting cells, and various structures.

There is currently no concrete solution for brain replacement, though we have ideas for things that might work. In a best-case scenario, we could create cells to inject into a person. These cells would navigate to the brain and swap out old cells for young ones.

These cells would also need to replace the extracellular matrix (ECM), which is a significant challenge we haven’t yet solved. Theoretically, advanced bioengineering could produce cells that swap out both the existing cells and the ECM. I have no idea how long that would take; it could be decades or even a century.

Researchers are exploring more immediate alternatives, such as tissue-level replacement. Jean Hebert famously wrote about this in his book. The process involves silencing a region of the brain, removing that tissue, and growing new tissue in its place. Cells are adept at growing new tissue by following developmental pathways, making this approach seem plausible.

The downside is the requirement for routine brain surgery. You would need to have a portion of your brain replaced fairly often—perhaps every year. This is under development, and while we might find interesting techniques, the process remains complicated.

Some researchers are investigating artificial replacement. Certain parts of the brain are not personalized and do not affect your personality; they simply assist with functions like balance or taste. Theoretically, generic biological or artificial parts could be swapped in, though this technology is still far off.

Daniel 00:38:59

Leaving aside brain replacement, which is clearly very challenging, let’s consider nearer-term organ replacement. A major challenge is that transplantation surgeries are extremely traumatic for the body. Undergoing multiple procedures can be very taxing.

This highlights the benefit of multi-organ or even body transplants, which you discuss in the book. There are cases where a whole set of organs is transplanted at once. This involves fewer connection points and a single surgery, which is a compelling solution.

Kris Borer 00:39:48

The more tissue you can replace per surgery, the better. You get more benefit from younger tissue and reduce the risk of multiple surgeries. I wouldn’t recommend someone replace a kidney and then undergo a separate surgery for a liver.

Currently, up to eight organs can be transplanted simultaneously. In a world of unlimited organs, surgeons could likely develop protocols to replace all internal organs at once quite straightforwardly.

Vascularized tissue allografts, such as arm transplants, are much more difficult because of the complexity of the internal structures. As you mentioned, the connections for internal organs are relatively simple for surgeons, so those will likely come first.

Daniel 00:40:34

This is an exciting roadmap. The technologies we are building could enable a future of common organ transplants, near-term health benefits, and radical lifespan extension.

Another exciting aspect is the lack of unknown scientific questions. Unlike the search for small molecules, where we often don’t know exactly what to target, the replacement roadmap feels more like engineering. Which aspects of this involve unknown science versus straightforward engineering?

Kris Borer 00:41:25

In replacement, the biggest open question is how to grow tissue constructs. We currently have bioreactors that can grow cells from an embryo to a stage about two weeks out. This results in a clump of cells rather than full organs.

We can also rescue late-stage pregnancies by placing a developing animal into a bioreactor to grow to full term. However, there is a gap in technology for the middle stage of development. One challenge is developing ectogenesis technology—a bioreactor that can take a single cell to a fully formed tissue construct with organs large enough for transplantation.

We understand the general process, but there are scientific questions regarding the specific nutrients and conditions required during growth. Our bodies handle this naturally, so we must emulate those conditions sufficiently to ensure organs develop healthily.

Another major scientific question is how to ensure tissue constructs never develop parts of the brain that could constitute a person. Ideally, a tissue construct provides as many replacement parts as possible without developing a brain capable of thought or feeling.

This is a significant ethical barrier. There are techniques for brain knockouts that prevent certain cells from developing. Natural examples like anencephaly show bodies can develop without brains due to genetic defects. We need robust mechanisms to prevent a thinking brain from developing within a tissue construct.

43:41 What are the limits of Tissue Engineering?

Eric 00:43:41

In my prior life, I was a bioengineer and worked in the wet lab for around a decade. I worked specifically on stem cell engineering towards regenerative medicine and tissue engineering.

We are probably two orders of magnitude more immature in the field of tissue engineering relative to small molecule therapies. Small molecules are much more like an engineering problem now. In contrast, we are in the Stone Age of understanding the basics of cell biology in the context of the hierarchy of cells to tissues to full organismal function. We barely understand these things at all.

The reason is mostly that they are unimaginably complex. What gives me a lot of hope is the fact that we can arrive at relatively unsophisticated approaches and get remarkable results. It is a sign that there is a lot of progress to be made in the next decade.

The Yamanaka factor reprogramming approaches are a great example. We threw cocktails of transcription factors at skin cells until they looked like embryonic stem cells. Shockingly, it worked. That is not a sophisticated approach, yet it succeeded.

We have done other things, like throwing various growth factors and transcription factors at embryonic-like cells, and they start to look like other organs. Again, this is not sophisticated, but the fact that it works gives us a sense that we can make progress.

However, it is non-trivial. How do you get the extracellular matrix (ECM) assembled in a way that replicates the native ECM? How do you get cells to not only individually differentiate into the proper cell type, but also cross-communicate with their ECM and each other in a way that is sustainable for decades?

Integrating these tissues with each other and with the rest of the body is a massive challenge. I don’t think we have the first idea of how to do that from first principles, beyond throwing cocktails of factors in the media to make it look a little more like the human body. It still feels very early.

Kris Borer 00:46:07

This is one of the biggest reasons to be confident about replacement. If we had to design tissues and genetic programs from scratch, it would not be feasible.

Fortunately, we don’t have to. We can use existing developmental pathways to grow them in bioreactors. By leveraging the natural processes that have evolved over time, we can create tissue constructs relatively easily. This is a primary reason to be hopeful about this approach.

Daniel 00:46:39

There are a few points here. The field is early in tissue engineering, and yet we’re seeing quick gains. That is reason to be bullish on the rate of progress. There are techniques that outsource a lot of the complicated work to biology itself.

With traditional drug development, anything you give to a human requires extensive work to confirm safety. You have to conduct big, expensive clinical trials. With tissue engineering or any technique to generate organs, it is more like engineering.

You are trying to create the organ elsewhere, and you can see if it is working. The feedback loop is faster because you don’t necessarily need to put it in a human to know it works. If you have successfully created a heart, it is obvious.

Kris Borer 00:47:46

You can see if the heart is beating or if the kidney is producing urine. There should be a faster path to developing these therapies.

You will still want to do trials and ensure safety. However, if you grow an organ from someone’s own cells, and it is immune-matched and functionally equivalent, it should be quite safe.

Daniel 00:48:09

Before we move on from replacement, Eric, I’ll give you a chance to tell us if you think what we said was wrong. You had some disagreement at the beginning, and I want to get it on the table.

Eric 00:48:23

I don’t think it’s disagreement; it’s just my own perspective. I am continuously shocked that anything we do works in biology. There is an inverse relationship where things seem harder the more you learn about the field, yet progress continues at a rate that always surprises me.

I look at the work being done with induced pluripotent stem cells for tissue replacement in Parkinson’s disease, spinal cord injury, and pancreatic islet cell replacement for type 1 diabetes. Remarkable things are happening that feel like they shouldn’t work, yet we have clinical evidence in patients that they are working.

The accelerationists always win. People who are optimistic about progress and constantly pushing for acceleration are the ones who solve problems. Looking at the long arc of history, those people have always been right.

Whatever hesitations I have stem from a sense that I wouldn’t know how to approach it myself right now. But when we have enough people knocking at the door of these problems, someone will figure it out. That has been the case for the entirety of human history. I have no doubt we will figure it out, though I cannot tell you the exact timeline.

Kris Borer 00:50:21

That’s really what LBF is all about. We want everyone pushing as hard as they can to accelerate things. Whatever works is what we will go with.

Daniel 00:50:31

From the outside, someone might hear that replacement can solve aging and think we are insane or believe that it is easy. Drug development is extremely hard. Traditional pharma has accomplished a great deal for humanity, and their work is difficult.

The replacement approach is also extremely hard. That is why we are trying to get more people to work on it. We believe that if this is invested in, it could generate tremendous gains for humanity—potentially much larger than other approaches.

Kris Borer 00:51:23

If it works, the payoff will be much higher than any drug program by far.

Daniel 00:51:29

We don’t know exactly how replacement will work or where we will run into issues. However, we want people to be working on things that could plausibly unlock these incredible benefits.

If you want more healthy years for more people, you should work on the approach that has the best shot at enabling that.

Kris Borer 00:52:02

We want as many shots on goal as possible, but we also want the ones that could potentially have the biggest impact. Go for the three-pointers.

Eric 00:52:11

There is something to be said for a stepwise strategy. Look at historical examples of companies that made incredible advances for humanity by starting with a focused, cash-flow-positive problem.

Google, where Daniel is currently employed, is one example. They started with the problem of PageRank for index search. Others had already invented search, but Google did it much better. That innovation skyrocketed Google above everyone else.

From there, they developed ads and other products, eventually leading to the “Attention Is All You Need” paper from their AI research units. That research is what the entire AI revolution is founded on.

The story is not always as clear as choosing between moonshot or incrementalist work. We need both. Incremental work that is cash-flow positive finances the moonshots. In turn, moonshots lift us beyond the limits that incremental work sets for us as a species.

Kris Borer 00:53:48

Our argument at LBF is that the allocation of resources is currently lopsided. If you look at impact, resources are not equally funded; almost all resources go to low-impact projects. We recommend shifting some of that focus to moonshots and taking bigger swings.

Eric 00:54:14

I couldn’t agree more.

Daniel 00:54:16

With any movement, it is easy to frame it as wanting to tear down the existing system and replace it. For example, in crypto, you hear rhetoric about taking down the banking system and replacing it with the blockchain.

“Down with fiat currency” is a rallying cry that inspires people and creates tension between the establishment and the contrarians. However, the result is usually the construction of a new future.

That new future might replace the establishment or merge with it. Regardless, it pushes progress forward. You need optimists with strong visions to go build things that nobody else is building.

Kris Borer 00:55:30

Absolutely. Replacement might not take us all the way there, but having that additional toolkit is very valuable in combination with traditional drug development or bioengineering solutions. We want doctors to have access to all these different tools when they are trying to keep people alive and healthy.

55:46 Biostasis: pausing death until a cure exists

Daniel 00:55:46

Let’s touch on cryostasis for a bit. It is one of the top things people can do today that can actually increase their odds of achieving radical life extension. Can you tell us about cryostasis—or biostasis more generally?

Kris Borer 00:56:08

Daniel, you and I were both EMTs, so we know that after someone’s heart stops, they’re not actually dead. They might be legally dead, but if you apply CPR or an AED, you might be able to bring them back. Death is a process, not an event.

When someone is very old, doctors might say, “Their heart stopped, let’s give up,” because it wouldn’t make sense to restart their heart only for them to die again from cancer or another underlying condition. But when someone is young, it makes sense to bring them back because they have many years of life ahead of them, and the condition might be treatable.

Biostasis is the same idea. When someone is declared legally dead, let’s give them a chance to be resuscitated when technology exists to treat their underlying condition, whether that’s aging, cancer, or whatever killed them.

This is done using cold temperatures or chemicals to stop molecular motion and pause metabolism. By putting metabolism on pause, the person won’t get any worse over time, and hopefully, at some point in the future, they can be revived and repaired.

Imagine someone who died of a heart attack 100 years ago. Doctors then would say there was nothing they could do. But if that same person had been put on pause in a freezer for 100 years, we could theoretically unfreeze them today and treat the underlying condition that caused their heart to fail.

Today, we face the same situation with untreatable cancers and other diseases. Biostasis gives people a chance to be put on pause and benefit from future medical technology.

In practice, you sign up with a provider. When you’re terminally ill, they pick you up after you die and administer protective medications and antifreeze. They cool you down and place you in a dewar, which is like a big thermos, to keep you safe until future technology is ready to repair you.

It sounds like a wild sci-fi idea, but the technology is actually here. For those who can’t wait 10 to 40 years for replacement or bioengineering solutions, biostasis is a real product you can buy right now.

Daniel 00:58:50

Laura Deming, who has a cryostasis company, frames biostasis as time travel into the future where a cure exists for your disease. The rate of technological progress is so high that new cures are emerging constantly.

There are tragic cases of children who died of leukemia but would have survived if they had been born just a year later. These examples will likely increase as we cure more things faster.

If we unlock real levers for longevity in the next few decades, it would be tragic for people to miss out because they didn’t live long enough. Biostasis can help those people.

Kris Borer 00:59:54

That’s the hope. We want to live in a world where suspended animation is common. We don’t have that yet. Suspended animation is the ability to freeze and unfreeze someone without any damage.

If we had that, the people you mentioned could be put into suspended animation and brought back a year later when a cure is available. Since we don’t yet know how to revive people, all we can do is put them on pause and hope technological progress allows us to unpause and cure them in the future.

Eric 01:00:34

Where do you feel the cryopreservation field is today in terms of readiness? Would you recommend it to a friend who is concerned about having access to the best technology for aging? What is your honest read on the field?

Kris Borer 01:01:02

I’d say anyone who wants to live indefinitely or have a radically long life should sign up for biostasis. The field has made significant progress despite limited funding. In the early days, people were frozen, which caused significant tissue damage.

Putting someone into liquid nitrogen without antifreeze creates ice crystals that kill cells. The modern process, called vitrification, is much more sophisticated. We use a special cryoprotectant so that instead of ice crystals forming, the body essentially turns into a block of glass.

When you vitrify something, there is no ice crystal damage, and theoretically, you could unfreeze them with future technology and all their cells would work. This reduces the amount of repair needed to bring them back.

Biostasis technology is working well enough that people should adopt it. In the future, everyone should have a contract with a provider in case of an accident or disease, allowing them to take a break from life and come back once they can be fixed.

Daniel 01:02:19

Anyone bullish on technological progress should be signed up. Even if the technology isn’t perfect yet, if you believe in progress, eventually we will have the technology to bring you back, regardless of the initial technique used.

Kris Borer 01:02:38

If you’re bullish on technological progress, your bar should be information-theoretic death. Is there enough information for future technology to reconstruct who you are and bring you back to life?

If you are cremated, no information remains for future technology to work with. But if you are frozen and have some ice damage, you can imagine future technology repairing those cells.

I think we’re close to a point where cryoprotectants are good enough that we won’t even need hypothetical future rewarming technology. Animal organs have already been vitrified, warmed up, and implanted into animals, where they work well enough to keep them alive.

A human is larger than an animal organ, making it more difficult, but I think next-generation cryoprotectants will make this a reality quite soon.

1:03:34 Identity & consciousness: chemo vs. cryo and using AI to upload our minds

Daniel 01:03:36

Do you have a strong take on the philosophy of consciousness that applies here? Do we need my exact brain tissue for it to still be me when you revive me?

If you scan my brain and then you bioprint my brain somewhere else, what do you think?

Kris Borer 01:03:53

I do not have a strong philosophical position, except that implicitly I do because I signed up for cryostasis instead of chemopreservation.

If you are cryopreserved, the goal is to warm you back up so that biologically you are the same person. If you are chemopreserved, the goal is to ensure the brain structure is perfectly preserved so that you can transfer the brain to some sort of computer emulation later.

That is good enough for some people. They say if there is an emulation of them running on a computer, they feel like they have survived and they are happy with that.

There is a spectrum in between, but I am more in the camp of wanting to be revived in my own biological body if possible.

Daniel 01:04:40

Now that Eric and I have a bunch of YouTube videos out there, we are going to live on immortally as long as the YouTube data centers exist. So we are good, right?

Kris Borer 01:04:51

There are companies that will collect all your data and make a simulation of you using LLMs. If you think that is good enough, then you are fine. Personally, I do not think that is me, but to each their own.

Daniel 01:05:02

Based on that dichotomy, I am going to choose cryostasis. It feels important to me that it is my brain.

Kris Borer 01:05:11

Well, you are in luck because there are more options for cryopreservation, and you will have the pick of the litter.

Daniel 01:05:19

Would you be willing to share which vendor you use?

Kris Borer 01:05:21

I signed up with Tomorrow Biostasis, but there are lots of good options, so I don’t think people should pick based solely on what I chose.

You should really pick the vendor that has the technology you like and has access to you. A local vendor might be better than a vendor in another country.

Even if the foreign vendor has better technology, you want someone who can get to you quickly. After you are legally dead and your heart stops, damage accumulation accelerates.

During life, you accumulate a lot of damage and your body degrades, but that goes up exponentially after your heart stops. You want someone who will be there very quickly to pick you up and preserve you.

Daniel 01:06:03

How high leverage do you think it would be for the field if regulations were changed so that people could go under cryostasis while they are still alive, rather than waiting for death?

Kris Borer 01:06:15

It would help a lot. Some people are trying to approximate this with medical aid in dying laws—death with dignity.

You may be familiar with the agonal process. When someone is dying, before their heart stops, circulation slows down and cells start to die off. Pre-legal death involves an acceleration of damage, and after legal death, it is even worse.

If you know you are going to die in a week or two and you want the best preservation possible, you could go to a state that allows medical aid in dying. A doctor will give you an injection that you can administer to yourself. You will die quickly while your brain is still healthy.

If you have your cryonics company standing by next to your bedside, they can quickly preserve you and you will get an excellent preservation. If you could start earlier, that would be much better.

1:07:07 Anarcho-capitalism, freedom & the role of government

Daniel 01:07:07

I am going to make a bit of a hard pivot. I saw something interesting in your background; you are not a stranger to contrarian groups. You wrote a previous book, The Ethics of Anarcho-Capitalism.

I would love to hear about your interest in anarcho-capitalism and if it relates in any way to your journey into this field of longevity biotech.

Kris Borer 01:07:34

When I was younger, I was interested in philosophy. I read that people’s personalities change a lot when they are young, but less so when they are older.

I did some research on different views of life, philosophies, and political stances, and I came across libertarianism. I thought it was excellent.

It wasn’t just from an interpersonal perspective—it is nice if people treat each other well—but it seemed like it had huge economic benefits as well. The more libertarian a society is, the more likely it is to produce technology and wealth, and people are happier and healthier.

I liked all that stuff and wrote the book so other people might appreciate it too. While I was writing it, I realized that one of the things libertarians care a lot about is freedom.

Freedom is not just freedom from interpersonal conflict, which is liberty. Liberty is the state where nobody is violating my rights or constraining what I can do.

But there are also constraints from nature. Nature puts constraints on us. If you want total freedom, you need to overcome not just interpersonal constraints that come from governments or criminals, but also constraints that come from the natural world we live in.

Total freedom means building societies that respect rights and also building technology that helps us live as long as we like, fly to the stars, or go into virtual worlds. They are connected under the umbrella of freedom. It is just different kinds of freedom that we are talking about.

Daniel 01:09:15

Eric and I talk a lot on this podcast, especially in our last episode with Jonathan Anomaly, about the liberating power of biology and bioengineering.

We are very constrained by our biology. It is obvious how constrained we are when we suffer from disease. To me, that is the most exciting angle on biology: let’s free ourselves to achieve anything we want in the world.

Similarly, I envision a future where we all get to travel to other planets if we want to, have all the different careers we want to have, and create all the things we want to create to really unleash that full level of human agency.

Kris Borer 01:09:46

Exactly. We are not trying to force any of this on anybody, but we definitely want people to have the option.

Daniel 01:10:12

We are going to force the freedom on you.

Kris Borer 01:10:13

You have got to want it. You are going to live forever whether you want to or not.

Daniel 01:10:22

Eric, are you going to endorse anarcho-capitalism?

Eric 01:10:26

Am I going to endorse anarcho-capitalism? I have to really think about that one.

Overall, entrepreneurship is like a self-restrained, societally sanctioned version of anarcho-capitalism where you are basically allowed to be a cowboy and go build stuff.

If customers believe in you, they buy your product. If investors believe in you, they invest in your company. You can change the way the world works.

It is a little restrained because there are rules around how you can be an entrepreneur that keep you from totally toppling the existing status quo. But generally, I do believe in anarcho-capitalism because I am an entrepreneur and a former VC.

Kris Borer 01:11:18

I love it.

Daniel 01:11:21

It brings up an interesting question I’ve wondered about. I dabbled in anarcho-capitalism in my youth, but I’m more of a libertarian now. I’ve often wondered about the role of government in scientific funding and health.

The government is funding some cool science in this space. I’m curious how you feel about the government’s role in the longevity field. There are ways we can leverage it to our advantage, but is it a bad idea in the long term to rely on government support?

Kris Borer 01:11:57

There are two ways to think about it. One is what an ideal society would look like for generating longevity technology as quickly as possible. That would be a society without government—an anarcho-capitalist or libertarian society where people and companies privately fund scientific initiatives without any government interference.

However, that’s not the world we live in. We live in a world with a lot of government. If we want to save lives with longevity technology, we have to accept that and do whatever we can to divert funding from things that are anti-life or anti-freedom toward things that are pro-longevity.

We should shut down government programs we don’t like and put that money toward longevity research and technology development. I have a great book for you if you’re interested in this kind of thing called The Economic Laws of Scientific Research. It goes into this in detail.

Daniel 01:13:01

I’m going to check it out. That take generally makes sense. If the government has already looted your resources, you might as well try to put them to good use rather than bad use.

Kris Borer 01:13:12

Exactly. No one on this podcast is going to be able to get rid of the government; it’s just a very challenging problem. Even if that is the ideal path, it’s not a realistic path anytime soon.

1:13:24 Closing: pipeline tech & a call to action

Daniel 01:13:22

We have a few more minutes left. Is there anything else we should cover?

Kris Borer 01:13:26

People should be very hopeful and excited about this field. I do a lot of angel investing, and I’ve seen some really incredible technology coming down the pipeline. You talked to Karl Pfleger; he’s a much more prolific angel investor with more information and connections.

A lot of the things he and I have invested in are going to help a lot of people. Hopefully, they will change the trajectory of lives for those who are still around in ten years when these technologies get through clinical trials. It’s a slow and painful process, so on one hand, I’m super excited.

For example, there’s a company called Repair Biotechnologies. I don’t know if you’ve spoken with Reason or anyone from that group, but they have a gene therapy that teaches your cells how to break down excess free cholesterol. If you think statins are good, this is a thousand times better.

They’ve shown in mice and monkeys that this mRNA therapy not only halts the progression of atherosclerotic plaque—like statins do—but actually reverses it. For the first time ever, we have a disease-modifying therapy for the number one killer in the U.S.

With these new genetic engineering techniques, we could potentially start knocking down some of the biggest killers, including different types of cancer. One of the companies I invested in has a cure for four types of cancer. It’s amazing what these new technologies can do.

It’s not out yet because it has to go through all the regulatory hoops, but I am super excited about what’s happening. On the other hand, a lot of the effort is still going to things that will only have a marginal impact. We want to encourage people to join the Longevity Biotech Fellowship, understand the strategies that have been laid out, and move toward higher-impact work.

Daniel 01:15:26

If people are interested in contributing to this space—and I think they should be—there is a lot of infrastructure to help them get into it, such as the Longevity Biotech Fellowship.

There are also investors eager to support it who can’t find enough things to invest in. You would probably love to invest in more talented founders building amazing technologies.

Kris Borer 01:15:52

Absolutely. That is music to my ears. I’ve seen so many pitch decks that are just the same old thing. When I ask how it applies to longevity, they say they are targeting one specific disease, but maybe it would help in certain cases.

I would love to have more young people saying they have a new technology they are applying to longevity to solve aging. That would be fantastic.

Daniel 01:16:16

Kris, thank you for joining us on the Free Radicals Podcast.

Kris Borer 01:16:18

Thank you for having me. It’s been a blast.

Daniel 01:16:20

Thank you for listening to this episode of the Free Radicals Podcast. If you enjoyed this episode and would like to support us, the most helpful thing you can do is share this with a friend who might enjoy it too.

Please also leave us a five-star review on Spotify and Apple Podcasts, and like and subscribe on YouTube. It would really mean a lot. I’m Daniel Shur, and my co-host is Eric Dai. Thanks for listening.

Jonathan explains the ethics of embryo selection and how a preference cascade will lead to its normalization, and why that’s good for humanity. We also get into the philosophy of radical life extension and transhumanism, and what a future of total control over biology will look like.

In this conversation, Jonathan espouses a Nietzschean view of humanity. Within each of us we have the opportunity to either envy greatness in others, or seek to cultivate it within ourselves and in the world. As we develop new biological technologies, we will have the opportunity to give ourselves and our children the best life possible, or we can give into envy and tear each other down.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

0:00 Selecting embryos for intelligence and the taboo around it
7:39 Aging, death and status quo bias
14:29 Side effects of secularization
17:29 How a preference cascade will shift the Overton Window
22:40 The importance of biological intelligence post-AGI
29:53 In vitro gametogenesis, gene editing & fully synthetic human genomes
38:14 The rift between us and the upcoming enhanced generation
43:26 Nietzsche’s Übermensch, slave morality, and transhumanism
56:32 Longevity as our generation’s space race

Transcript

0:00 Selecting embryos for intelligence and the taboo around it

Daniel 00:00:56

Jonathan Anomaly, welcome to the Free Radicals Podcast.

Jonathan Anomaly 00:00:59

Thanks for having me on. I love the name of the show.

Daniel 00:01:02

Awesome. We love your name, too. Anomaly is a pretty sick name.

Jonathan Anomaly 00:01:05

Dr. Anomaly. I never thought I’d be that, but I changed my name when I was nineteen at Berkeley as an undergrad. I didn’t anticipate being Professor Anomaly or Dr. Anomaly. Everyone thinks it’s a supervillain name, but it’s real.

Eric 00:01:19

That’s so cool.

Daniel 00:01:21

There’s another interesting name: Herasight. Tell us about your company.

Jonathan Anomaly 00:01:26

It’s named after the goddess of family and fertility, Hera. We started about three and a half years ago. The idea was to build on what Genomic Prediction and Orchid had started.

There are a couple of other companies in the space, and we wanted to build the best polygenic predictors in the world. We wanted to go all-in on the traits that people care about and that actually matter for human welfare, which includes IQ.

We wanted to do the thing that we knew would get us in some trouble, but we also knew would be good for humanity. Twenty years from now, this will be considered obviously uncontroversial. That’s what brought us together: value alignment and going all-in on the best science for the traits people actually care about.

Daniel 00:02:23

This brings up an interesting point regarding the difference between expressed and revealed preferences. I think we’ll see this with embryo selection.

You brought up IQ. A lot of people want to deny that intelligence is a real thing that varies. Yet, we have evidence that once people can select for the intelligence of their children, they’re going to do it. Can you tell us about that dynamic?

Jonathan Anomaly 00:02:52

It’s a hypothesis so far because we’ve only served about 140 people. Once we have 10,000, I’ll get back to you with statistics. Anecdotally, there are going to be people who publicly condemn but privately use companies like ours.

This has been made a taboo, especially in the second half of the 20th century. The genetics of intelligence became a taboo as egalitarian political ideals spread after World War II. Since we couldn’t do much about intelligence differences until now besides education and environmental training, people questioned the point of bringing it up.

The fear was that it would allow people to understand they were born a certain way and couldn’t change it. My view is that now you can do something about it for your children. You can influence them through genetic selection and certainly through sperm or egg selection. We are breaking the taboo, and I think we’re going to be thanked for it because for the first time in history, we can do something about this.

Daniel 00:04:23

I’d add that as a society, we constantly see people express preferences for traits like intelligence. People select mates who are more intelligent and have an intuition that their children are likely to be more intelligent as a result. Why is there such a taboo around it despite how it plays out in real life?

Jonathan Anomaly 00:04:56

It goes back to history and the misuse of genetic data to promote certain ideologies. Specifically, the US in the 1920s and 30s and Nazi Germany. It’s almost boring to talk about it at this point because it’s so obvious that’s what happened.

You see it in the data and in the way scientists changed how they spoke about these things. Some of it was for the better, with more talk about tolerance, but some of it was for the worse. In the 1940s, some scientists started bending scientific facts to fit a narrative they wanted to bolster.

They had good intentions, but they ended up setting back their own ideals as well as scientific progress by holding us back from studying the genetics of intelligence. Countries that didn’t participate in the war directly simply don’t have these taboos.

In Singapore or India, they don’t understand why you would deny the trait that literally makes us human. We call ourselves Homo sapiens for a reason; the brain distinguishes us. It’s not just intelligence; it’s social coordination, prosocial behavior, and the ability to plan together.

Intellectuals deny this for political reasons, but people with less education often understand it better. Emily Willoughby from the University of Minnesota wrote a paper on this. She studied people with two or more children and different levels of education.

She found that less education or more children led people toward the correct views on heritability. People who have kids or haven’t absorbed academic taboos simply look around the world and see that traits like personality, athleticism, and musicality are largely out of our control.

You can tune it up with lessons, but parents realize that even if they raise their children the same way, they turn out wildly different. They understand it must be for genetic reasons.

7:39 Aging, death and status quo bias

Daniel 00:07:39

We’re going to dig more into intelligence and embryo selection, but first, I want to get longevity on the table because I think there are similarities around taboos.

We’re very interested in indefinite lifespan extension—treating aging itself as a disease. But that’s a taboo. It’s weird to talk about aging as a disease.

Just like with embryo selection, we can’t do much about aging today. Maybe that’s why people don’t like to think about it as a problem. How do you react to that?

Jonathan Anomaly 00:08:27

That seems like part of it. If you can’t do anything about a problem, there is something negative about constantly dwelling on it. It’s as if you’re advertising your own superiority over other people.

Even if you aren’t superior in a dimension we care about, pointing out differences that can’t be changed makes you look like a bad person. Why constantly harp on these things? For the first time, we might actually be able to do something about it.

The second thing going on is that people have an un-Darwinian view of human nature. I absorbed this early: most people don’t understand what kind of creatures we are or the tentative processes that led to where we are today. They don’t realize how easily things could have been different.

This applies both to longevity and embryo selection. People understand wanting to cure a discrete disease caused by a mutation because it is a clear deviation from the norm. However, they find it creepy or strange to go above the statistical norm, whether that involves lifespan, health, or intelligence.

We are anchored to the status quo. Nick Bostrom and Anders Sandberg have discussed status quo bias in ethics. If you ask people if they want to select an embryo for intelligence or longevity, they are often hesitant.

But if you ask if they would want to select against those things, they say, “Absolutely not.” This suggests they think the statistical average is exactly where we should be. It’s a false, teleological view of human nature, somewhere between an Aristotelian and a Christian perspective.

It is an incorrect view of human nature. For example, David Reich recently published a paper showing genetic selection in favor of cognitive ability in East Asia and Europe over the last few thousand years. This suggests there isn’t one static human nature that has remained unchanged for hundreds of thousands of years.

Selection can be recent, rapid, and can move in all kinds of directions. Ashkenazi Jews, for instance, come from a founder population that was very small in recent European evolution. Because that population was so small, Jews carry various disease-causing mutations alongside potential advantages.

This is how evolution works. The idea that the current average of humanity is “just right” is very strange. Groups vary in many traits, such as lung capacity in Tibet, long-distance running in East Africa, or sprinting in West Africa. Status quo bias is a real bias, and it’s worth recognizing.

My friend Alan Buchanan points out that if you saw a fast-forwarded version of the symptoms of aging in a normal person, you would be horrified. Their skin sags, their bones lose density, their muscles waste away, and their eyes fail. You wouldn’t wish it on your worst enemy.

Yet, this is accepted as the norm. There’s no reason to accept it. This also shows the arbitrariness of the moral line between treatment and enhancement. Do we really want to view the deterioration of our sight and hearing at age 40 or 50 as morally good and normal?

Most people do Botox and other treatments, but they still have this weird view that we shouldn’t mess with things at the genetic level. Why not? Why do we accept this status quo bias?

Daniel 00:13:54

This bias pops up in our decision-making in other ways. When someone has a terminal illness, we’re willing to take much more risk in their care because the alternative is death.

I would argue we’re all terminally ill. We are all going to age and die, yet there is an enormous amount of aversion toward trying anything to treat aging.

14:29 Side effects of secularization

Jonathan Anomaly 00:14:23

Psychologically, it’s interesting to ask why this is true. There are negative side effects to the secularization of society, including a loss of hope. Belief in God and religious structure is beneficial; there’s a reason it evolved universally.

There is a real loss with the decline of religion, but there is also a gain: freedom from the idea that death is not only inevitable but perhaps even good. Some views suggest death is the best thing because it allows you to unite with God.

While religion is compatible with various views on longevity, there is a tendency within religious worldviews to simply accept suffering and death. I reject that. Suffering is part of life, but the beauty of life is using that suffering to overcome challenges and improve ourselves physically and mentally.

It is depressing to think we could be made to suffer with no compensating benefit at the end. One benefit of losing this teleological worldview is the realization that life is in our own hands. We should make the best of it for ourselves and our children.

Daniel 00:15:56

Religion inserts an interesting value function. If you have eternal life in heaven after death, that’s infinitely better than the regular world.

You can make the same argument with longevity. If we crack aging, that is an infinite upside. As an economist, how would you value the possibility of that infinite upside?

Jonathan Anomaly 00:16:21

People value life at different rates. Many are terrified of the prospect of living forever because they imagine they would be forced to live and could never die. However, suicide and accidents will always be a factor.

There is a weird thought experiment involving Groundhog Day. It feels like a curse where you keep living the same life and eventually get bored of everything. Maybe that is true, or maybe we could enhance cognition so that we cannot get bored because there is an infinite possibility of exploring new topics and meeting new people.

I think that worry is in the back of people’s minds. Most people say that if they were ever in a decrepit state, they would want to die. Yet, when they are actually near death, they choose to live longer. They want more time.

If we could picture living a very long time while remaining healthy, almost everyone would opt for it. I do not know how to calculate that value function, but the desire for more time is clear.

17:29 How a preference cascade will shift the Overton Window

Eric 00:17:34

The core theme of our podcast is that we should grant humanity total control over biology. That should be a core tenet of how we invest our resources and talent to build something historic.

Even in the 21st century, it remains an extremely controversial way to view our role in society and our place in the universe. My question is not why this view is so controversial, but how do you operate as a contrarian in a world that not only doesn’t care, but actually actively opposes you in realizing those views?

Jonathan Anomaly 00:18:21

According to the Great Man theory of history, maybe you are the ones who will change these norms. Joseph Henrich discusses this in his book, The Weirdest People in the World, where “WEIRD” stands for Western, Educated, Industrialized, Rich, and Democratic.

His first book, The Secret of Our Success, explores what kind of creatures we are. He argues that elites have always shaped the norms and ideas of the groups they belong to. Most people look for cues of success and leadership and then defer to the behavior, norms, and beliefs of those leaders.

Right now, many people have a status quo bias against radical life extension or genetically improving the prospects of their children. But as elites adopt these new views, the blank slate theory of human nature will melt away.

I think the change will be fast because there will be a price associated with holding false beliefs. The price is your own life or the lives of your children. As this happens, we are going to see really rapid change.

I also want to reference the work of Timur Kuran and his book, Private Truths, Public Lies: The Social Consequences of Preference Falsification. He discusses cases in which you get radical changes in ideology, whether in science or politics.

His favorite example is the fall of communism in 1989. It looked like it would last another 100 years, but it fell immediately. Once enough elites believed communism did not work and realized others felt the same way, it became safe to talk about it. Once a consensus formed among elites, everyone else felt free—and eventually pressured—to change.

The same will be true here. In longevity, I don’t think many of the current “cures” work, but with better research, they will in a decade. With embryo selection, there is no doubt that it works.

As that is better understood, we are going to see a massive preference cascade. Elites will start using longevity medicine and embryo selection, and very quickly, others will feel obligated or pressured to do it too.

Daniel 00:21:28

Regarding longevity, the issue is that there isn’t a product on the market today that can actually extend lifespan. It is a hard problem to solve and requires many people working on it.

If everyone is expressing a preference against longevity, it is difficult to get people to work on it. But if we consider the Great Man theory of history, perhaps all that matters is convincing enough elites to work on it. That may be where the change comes from.

Jonathan Anomaly 00:21:59

I know people in Austin who periodically go to Washington, D.C., to try to influence the administration. If you can get the NIH or grant-making bodies to change their rules or reemphasize what they are funding, you can accelerate progress and change history.

By spreading ideas on an intellectual level, you are contributing to that. Growing this podcast and your other projects is the only way change actually happens.

22:40 The importance of biological intelligence post-AGI

Eric 00:22:42

There are many historical examples where a top-down signaling cascade led society to adopt controversial technologies. We saw this with the move to electricity and the transition from horse-drawn carriages to combustion engines.

What is difficult for people today is that, for the first time, human intellectual supremacy is being challenged by artificial intelligence. That is going to be a tough pill to swallow, even for the smartest elites. We are at a very different juncture in history.

Jonathan Anomaly 00:23:33

I agree. Things have changed so fast that my thoughts haven’t caught up with where we are now. For the longest time, I resisted the idea that AGI is coming.

I know you can’t get AGI from an LLM alone, but it looks like we are actually making progress toward it, and it really is going to change everything. With all the investment dollars flowing into it, this is one of the few hyped-up areas that is probably correctly hyped. It’s not a bubble.

The example you gave of electricity has one disanalogy: electricity was immediately beneficial to everyone. When Ben Franklin invented the lightning rod around the same time electricity was being propagated, there were religious crusades against the technology. People protested, claiming it thwarted God’s will.

The same thing happened with IVF. Fifty years ago, IVF was considered religiously unacceptable. Even secular reporters said it was disgusting and claimed we were creating Frankensteins. Yet, only five years later, public attitudes shifted because elite attitudes changed. That happened for electricity, lightning rods, and IVF. It will be no different here.

Daniel 00:25:09

Tell us why you think AI is going to make embryo selection even more important.

Jonathan Anomaly 00:25:17

I suppose you’re alluding to a thesis I’ve been presenting called “Why Biological Intelligence Matters in a World of Artificial Intelligence.” I’m somewhat ambivalent about this, but I do think biological intelligence will become more important as AI becomes ubiquitous.

There will be more opportunities to use AI to manipulate people politically and financially. Algorithms often embody political and moral commitments that many people aren’t aware of. Even when they aren’t explicitly designed that way, they are trained on human data.

For now, there are areas where we get the answers to relevant questions systematically wrong because of ideology. Sociology is a clear case study. The discipline has been ideologically captured for fifty years, long before the recent wave of wokeism.

When one ideology takes hold of a discipline, it systematically biases the results. If everything published in a discipline is wrong, LLMs will pick up on those errors.

If you aren’t bright or creative enough to ask the right questions or ask the LLM to reflect on its field’s biases, you’ll be a slave to the AI or the person programming it.

There is evidence that brighter people tend to commit fewer logical fallacies and are better at statistical reasoning. Raw general intelligence (G) will be at a premium. However, I’ve backed off from overemphasizing IQ because creativity and personality traits like openness also matter deeply.

The worst person to be in this new world is a bright psychopath or a bright ideologue. Religiosity and political orientation are about 0.4 heritable—moderately so, though less than height or intelligence. We all know smart ideologues who are insufferable.

Nassim Taleb is an interesting example. He is clearly bright and writes good books, but if you disagree with him on certain topics—like the heritability of IQ—he will immediately block you and call you names. He shows that G alone is not enough.

You need intellectual and moral virtues to pursue the truth and overcome bias. This is partly heritable and partly a matter of training. While biological intelligence matters more than ever, the character and mental virtues Aristotle emphasized are also crucial.

Personality traits like intellectual and physical courage are undervalued because they are hard to measure. We don’t have good genome-wide association studies on them yet. One of our funders, an ex-military guy, wants us to look into that.

If we are selecting embryos, we should want personality traits like courage alongside IQ. Even then, you still need to raise and teach your children well. You still need to put them in the right environment.

29:53 In vitro gametogenesis, gene editing & fully synthetic human genomes

Eric 00:29:55

As a bioengineer with a background in stem cell engineering and venture investing at Andreessen Horowitz and Dimension Capital, I’ve seen hundreds of companies at the frontier where life science intersects with technology.

The first, second, and third-order effects of compounding growth in these technologies will completely alter how we view biology and exist as a species.

Today, a company like Heracite or Orchid might focus on single-cell sequencing to determine the genetic orientation of an embryo, typically selecting from about a dozen options.

On the other end of the spectrum, we have technologies that allow you to take a skin cell, turn it into an induced pluripotent stem cell, and then induce that into an egg. There are multiple startup steps.

On the other hand, you have the ability to specifically and selectively engineer sites of that genome using CRISPR-Cas or other genetic engineering technologies. These individual pieces currently operate in largely separate domains, but as these tools become modular and compatible, there will be an explosion of an infinite number of eggs that you can genetically select and engineer.

You can use population-wide statistics to determine the likely traits you are selecting for or against. That future is not far away. What do you think of that scenario, and how far away are we from it?

Jonathan Anomaly 00:31:40

That was the subject of a talk I gave this morning. There will be more convergence in the use of these technologies, but I want to push back because the technical difficulties are interesting.

I have no doubt that embryos will be edited. This already happened in 2018 for a single variant associated with HIV and AIDS. However, editing is often unnecessary if you have In Vitro Gametogenesis (IVG), which allows you to create eggs from stem cells. You can take a skin or blood cell, turn it into induced pluripotent stem cells, and then into eggs.

With enough eggs and embryos, you can deselect for almost any disease without needing to edit. There may be use cases for single-gene editing if you have few embryos before IVG becomes available. But even in 30 years, there is an interesting problem.

If you want to perform multiplex editing on hundreds or thousands of variants for polygenic traits like height, intelligence, and personality, you need to do it at the single-cell stage. There is always a chance for off-target mutations. If that happens, you must wait five days for the embryo to grow, biopsy the outer layer, and check the results.

You would need to whole-genome sequence the embryo to see if you introduced undesirable mutations. If you have high-quality polygenic scores and IVG, you almost don’t need editing. Polygenic screening and editing are rival technologies. For screening, you wait until the embryo is a few days old, biopsy it, and sequence it.

Unless editing becomes extremely precise, you won’t want to edit a five-day-old embryo because it already contains 150 cells. You would have to perform thousands of edits on every single cell and ensure there were no off-targets.

My previous view was naive. I thought we would combine polygenic scores with IVG and then “spell check” the results with CRISPR. It is likely not that simple. These are strangely rivalrous technologies, but it almost doesn’t matter.

If you have a large pool of embryos and good polygenic scores, you already get massive gains without editing. Conversely, if editing becomes safe and precise, you could use it without screening.

Daniel 00:35:32

There is an alternative potential end-state technology: writing the whole genome from scratch.

Jonathan Anomaly 00:35:38

That is how I ended my talk today: synthetic genomes. I have no expertise in this, but I first read the idea from the late Craig Venter. He recently passed away and was a frequent visitor here at SynBioBeta.

He created a synthetic genome and inserted it into a bacterium. He famously called it the first form of life with a computer as a parent because it involved stringing together amino acids to create a synthetic genome.

A friend at the Broad Institute thinks this will happen before widespread gene editing because it might be safer. AI and AI-guided robots will likely play a significant role. If we can do this safely, there is no point in going through complex biological processes and editing while worrying about off-targets. We could simply string together amino acids that resemble a better version of you or your partner.

The distinction between treatment and enhancement is morally misguided and scientifically incorrect. We deal with statistical averages, and there is no reason to believe the current average is a moral or scientific ideal. As a species, we are always changing.

If we take a model of ourselves or others—a kind of von Neumann—we have a sense of what we could be if we were better. We know our deficits, whether it is being too narcissistic, too short, or even pathologically altruistic.

In the next 20 to 40 years, once we understand the genetic architecture of these traits, we will be able to build better versions of ourselves. I have no doubt that is coming.

38:14 The rift between us and the upcoming enhanced generation

Daniel 00:38:17

What happens if the next generation is enhanced to be far more intelligent than us and no longer suffers from aging?

Jonathan Anomaly 00:38:28

Where is the downside?

Eric 00:38:30

That sounds great.

Daniel 00:38:33

But what happens to our generation? That seems like a more plausible future than one where we cure aging for ourselves, but it also doesn’t seem like a stable future.

Eric 00:38:48

That’s the plot of Gattaca, right?

Daniel 00:38:51

Gattaca is different because while you have the haves and the have-nots—those who are enhanced and those who aren’t—the enhanced individuals still die.

It would be great if we could give eternal life to another generation, but I don’t know what that world would be like where we’re all doomed to die while the next generation isn’t.

Jonathan Anomaly 00:39:13

I have a couple of replies to that. First, in the direction you’re gesturing at, I don’t think we’re going to have a step where we go from dying at an average age of 82 to living forever. I don’t know if we’ll ever live forever, but I think there will be incremental changes in how long we can live and how long different organs function.

Neurogenesis is important. We don’t want to live a long time but be basically brain dead. There are going to be different technologies that slowly lead to longer, better lives. I don’t think we’re going to see a sudden divide between immortals and mortals.

Similarly, with enhancement of traits like intelligence—which we discuss frequently because it’s a sexy trait—or a better functioning immune system, it won’t be a sudden jump. It will be incremental. First, we might have immuno-enhancement against the flu virus, and maybe eventually all viruses. George Church has discussed this possibility. You still have bacteria and other issues, so we’re going to get increments rather than the dramatic steps you mentioned.

Secondly, I like to think about this in the way my friend Diana Fleischman does. She has had two natural children and now hopes to use selection for future children. She wondered what she would tell her selected kids who have all these advantages.

There isn’t actually much to tell them because they won’t be that different—perhaps a few extra IQ points or slightly better immunity. But consider a parent in 1950 whose first children were not vaccinated because the polio vaccine didn’t exist, while their second set of kids was vaccinated.

Do you withhold a vaccine and risk your children dying or being disfigured just to honor the first set of kids? Nobody thinks that. You want to give your children every advantage they can have. The same will be true here.

We might envy those future people, and that envy would be justified. It would suck to be part of the last generation doomed to certain limitations. On the other hand, if we aren’t narcissistic, we should be proud of them.

One of the great things about civilization is that humans are the only animals that care about more than just immediate relatives. George Price solved the riddle of altruism at a genetic level across all species with the Price equation. When asked if he would sacrifice his life for a brother, he famously joked, “No, but I’d gladly do it for two brothers or eight cousins.”

That is the evolutionary math of how species should allocate their altruism, but in fact, we don’t do that. We are far more altruistic than that as a species. I don’t know of any other species that are. Some help non-kin, but we are remarkable in the extent to which we do so.

It’s common to see someone help a stranger when they have reason to believe nobody is looking. We see simple acts like someone holding a door for a person in a wheelchair. While some people might do it to impress a date, there are countless unremarkable cases where we make sacrifices for people to whom we have no genetic relationship.

That is a beautiful thing about human beings, and I think we should enhance it. It gives us the capacity to build a church that takes three generations to complete. I was just in Spain and saw one of those churches. It took that long, and what a beautiful thing it was.

We have had that capacity for a long time. I think we will be envious of the people who come after us with capacities we lack, but we will also realize how cool it is that we were a part of creating that.

43:26 Nietzsche’s Übermensch, slave morality, and transhumanism

I include a quote from Nietzsche at the beginning of my book from Thus Spoke Zarathustra. He talks about the Superman and the “last man.” The Superman represents overcoming our pathetic beginnings, both culturally and biologically.

Nietzsche says, “Man is a rope stretched between beast and Superman, a rope over an abyss.” In that scene, a tightrope walker moves toward the Superman while the common man tells him to come back to a life of comfort. That’s the “last man”—the person who could have it all but strives only for comfort and equality.

In Nietzsche’s view, equality is not what we should be striving for. We should be striving to overcome ourselves and our prejudices. I think there is something to this beautiful vision where we can take part in creating something better than ourselves.

That is a huge part of my life. Being part of Herasight is interesting to me because if I can help people have children with a much lower disease burden who will live longer, healthier lives, I wouldn’t even care if I never had kids myself. I would be proud of that.

That is a sincere belief. Human beings have been acting that way for a long time at a cultural level, and we should embrace it as the best part of our nature. We all have the drive to be the best version of ourselves, have the best kids we can, and create technology and institutions that allow the best of humanity to flourish.

There is something beautiful about that. Nietzsche also said that life has an essentially aesthetic justification. In a post-religious worldview, the question of why we are here is the only one that matters.

In half a trillion years, there will likely be no life left in this region of the universe. Why do we go on at all? It is to create something beautiful and interesting.

When I look at those future people who are better in every way than me, I’ll wish I was one of them, but I’m not going to be the “last man.” I’m not going to tear them down just because I don’t have the capacity to be like them.

Eric 00:46:49

I like that.

Jonathan Anomaly 00:46:50

That’s great.

Eric 00:46:51

Where should we take the conversation?

Daniel 00:46:56

That was a lot to respond to.

Eric 00:46:59

Let’s talk about Nietzsche. A core concept of his philosophy is the Übermensch. While I am not a Nietzsche expert, from what I understand of his texts, the core idea of the Übermensch is not necessarily a physically superior being like the Superman from comic books.

Instead, it is a cultural, psychological, and spiritual overcoming of the shackles that society has placed on humanity.

Jonathan Anomaly 00:47:28

I think that’s exactly right. There is a small extent to which it is a biological conception. Nietzsche was reading history and thinking about the future, considering the Aryan invasions of India and these proto-Indo-Aryans.

He called them the “blonde beast,” which he both reviled and upheld as a superior type. He reviled it because, although he was German, he hated the ultra-nationalism of Germany at the time.

He saw firsthand the damage that the Franco-Germanic War caused and witnessed the rise of antisemitism. Some of his last words were, “I’d like to have all antisemites shot.”

He hated this and saw that it could be misused to justify a purely biological conception of the Superman. Indeed, his sister became a proto-Nazi, and Hitler selectively quoted Nietzsche’s discussions of the blonde beast asserting his will on Europe.

Nietzsche is partly responsible for some of that, and there is some truth to the biological component, but the vast majority of what he meant was a spiritual overcoming.

His view was that there was an assertiveness to the horse-riding Proto-Indo-Europeans who made their way across Iran, the Middle East, and Germanic countries. They asserted their worldview on the people they encountered.

The Greeks, Iranians, and Scandinavians inherited their culture, myths, and gods. He saw these as great men to be praised because they didn’t focus on equality or respecting all cultures equally; they simply asserted themselves in the world.

He liked that quality. It is not a view that says we should go around dominating everyone, but it is the view that when Christianity came to Rome, it essentially castrated the European man.

Nietzsche believed Christianity had some beneficial, civilizing effects, especially on the Vikings. It pacified them and made certain projects associated with beauty possible, such as the great churches and the culture Europe created.

The Renaissance was partly a Christian phenomenon and partly a rebellion against Christianity—a revival of Greco-Roman ideals. Nietzsche had an ambivalent view where Christianity castrates Europe but also provides beneficial effects.

He wondered what would happen next, noting that Darwin and others before him had effectively killed God. In Thus Spoke Zarathustra, when he proclaims “God is dead,” he means this as a purely cultural phenomenon.

He was an atheist, but he had no reason to evangelize atheism. He wasn’t like Richard Dawkins; he believed this was one of the greatest crises humanity would face.

He prophesied that Europe would lose Christianity but fail to reassert ancient Greek ideals. Instead, he worried it would adopt the worst parts of Christianity—an obsession with equality and “slave morality.”

In slave morality, you take the people who have enslaved you and rebel by saying all their virtues are actually vices. You then claim your own vices, such as the total impotence to do what you want with your life, are virtues.

In this framework, God rewards the impotent and punishes the able. This is why Nietzsche’s ultimate worry was that Europe would become a nation of “last men.” Instead, he believed we could become the Superman.

He didn’t mean this only for Europe, but he was writing as a European in the context of Christianity’s decline. These ideas were taken up in the early 20th century in both cultural and biological forms.

I always cite George Bernard Shaw’s 1903 play, Man and Superman. He was the first person to translate the term “Übermensch” into “Superman.” Shaw shared Nietzsche’s ideal that we need to biologically and spiritually overcome the shackles holding us back.

There are many ways to conceptualize the Superman, but it stands in contrast to the parts of our nature driven by envy. It rejects the sense that if I can’t have it, nobody can.

It is a rejection of the desire to tear down anyone smarter, more beautiful, or more capable. Instead, we should recognize that beauty and try to foster it. That is the vision of the Superman.

Daniel 00:53:03

That is a fascinating vision.

Eric 00:53:04

Do you think there is a subculture that espouses those views? I would argue that the vast majority of society has gone in the other direction toward slave morality, impotence, and a crabs-in-a-bucket mentality.

Where do you find subsets of people who embody what you might argue is the best of humanity?

Jonathan Anomaly 00:53:26

Nietzsche’s prophecies clearly came true, taking about a century longer than he anticipated. He saw this starting in the 1880s in Europe. Secularization was beginning, and an extreme secular, progressive view was starting to take hold on the other side. It took a long time to manifest fully, but if you look at the modern academy, it embodies the worst ideals in humanity.

I’m not talking about the chemistry department; I’m talking about social studies and philosophy. These fields are supposed to be liberating. They are meant to empower the individual to think for himself, overcome bias, and study humanity as it is. Instead, they have become the opposite. It is slave morality embodied.

Leaving the University of Pennsylvania was the best decision I’ve ever made. You have to leave rotten institutions behind in order to build new ones. While capable people can still navigate academia and get something out of it—especially in the sciences—we ultimately have to throw off those shackles. We need to ask: What are the true ideals? What are the beautiful things?

I always advocate reading old books. Even something just 100 years old, like The Genetical Theory of Natural Selection by Ronald Fisher, is a revelation. Fisher was the inventor of modern statistical genetics; he is essentially Darwin personified statistically. He developed the theories of runaway selection and sexual selection, among many other things.

The last third of that book explores man’s genetic fate. Fisher considers which genetic traits modern liberal society selects for, and his view is quite grim. Whether you agree with him or not, reading someone like Ronald Fisher or Charles Darwin’s Descent of Man allows you to think in radically different ways than a modern education in biology.

While you should learn about cell biology and genetics, reading these classic books makes you realize how much we are shackled by modern norms imposed by progressive liberalism. Some of those norms, like radical toleration, might be positive, but modern academics and journalists often function to constrain thought. We need to break away from that to achieve what we are truly capable of achieving.

56:32 Longevity as our generation’s space race

Daniel 00:56:34

I’ll tie this back to the mission of this podcast. When Eric Dai and I were discussing our goals, we talked about eras of humanity that strove for something great. During the space race, every person could look at the moon and imagine humans up there, traveling into the stars.

When you look at our culture today, you don’t see that same optimism or enthusiasm about the future. Instead, there is a “crabs in a bucket” mentality.

We view longevity and biotech as a way of creating the humans of the future. To me, longevity could be the space race of our time. It is a way for us all to unite toward bringing real beauty and greatness into the world.

Jonathan Anomaly 00:57:32

I totally agree. It is not only important but necessary to have a transcendent goal. Even if traditional religion is on the wane, there can still be a kind of spirituality or ethics in the Greek sense—a set of values to embody and strive toward.

The point you made is fascinating. We are about to head back to the moon and likely to Mars soon. Yet, how much media coverage does that get compared to the stock market’s reaction to the Strait of Hormuz being temporarily shut down?

That gives you a sense of where the average person’s mind is. People are focused on whether their 401k will go up or down by a percent or two this week. I’m not saying people shouldn’t care about financial security; we all need that as a means to an end. But it has become a dull end in itself.

Nietzsche called this the “religion of comfort,” and there is nothing more contemptible. We need something to strive for. That is exactly the spirit you guys embody.

Daniel 00:58:48

Thinking about these grand ideals makes you see all of life differently. After spending the last few days having conversations about the future of humanity and technologies that can help so many people, petty behavior seems insignificant.

We live in an incredible world. There are real dangers facing humanity that we need to work together to fight against, but there is also an amazing future available to us. How can we get caught up in petty nonsense?

Jonathan Anomaly 00:59:20

That is so true. I try to avoid that on Twitter. I don’t really tweet; I just repost updates from my company. However, there are troublemakers at sophisticated universities who just want to poke holes in what we are doing.

Pushback and skepticism are necessary for science to improve its methods, but that isn’t their goal. They are just trying to discredit us, even though we are the most rigorous company offering these services. We make our validation studies publicly available.

These people are like bullies on a playground. They try to bring your aspirations down. In some ways, it’s best to ignore them and “not feed the trolls.” In another way, it’s helpful to recognize them as an anti-ideal.

You don’t want to be a petty person who tears down what others are creating, especially when those creations are beautiful and transcendent. It’s best to view them as a negative example that reminds you to wake up every morning with a renewed sense of optimism.

Daniel 01:00:44

We’ve got one minute left. Is there anything else you’d like to leave our audience with?

Jonathan Anomaly 01:00:47

I’m not very good at promoting my company, so I suppose I should probably do that.

Daniel 01:00:53

Yeah.

Jonathan Anomaly 01:00:53

If anyone is interested in learning more about in vitro fertilization and polygenic scores, whether you’re having a kid or investing in a company, we are always looking for value-aligned investors.

This is an underexplored area. People like Brian Armstrong constantly tweet about editing, which is important, and we need more research there, but I don’t think many people realize how powerful polygenic scores are right now. In the near future, they are only going to get better.

For example, we recently got access to East Asian biobanks. For the first time, East Asians might actually have the advantage over Europeans in terms of the predictive power of various polygenic predictors. This is simply a function of data. If we have more genetically sequenced people in biobanks with well-phenotyped traits, we can create better predictors.

This technology is really under-discussed. Because it is so new, some people don’t realize that we are going to increasingly use polygenic scores. I call it the polygenic revolution.

It won’t just guide embryo selection, which is already very powerful; it will tell us more about history and ourselves. David Reich recently published a paper on this. We can understand ancient humans by looking at polygenic scores rather than just single genes or bones.

We can determine how prone ancient populations were to diabetes or how tall they were. While those traits are partly a function of diet, we can use polygenic scores to determine if they were taller on average or had denser bones. We can see when these changes happened through introgression or mate selection. This is incredibly interesting.

We are going to use polygenic scores to improve our children and to explain the past. In your case, because you focus on longevity, we can use them to guide medical diagnoses and tailor medicine.

To be fair, polygenic scores right now generally aren’t good enough for definitive medical diagnoses, but they are only going to get better and can point you in the right direction. My aunt recently started losing her memory, and she almost certainly has Alzheimer’s. To investigate, we sequenced her and found she has the APOE4 variant, which is associated with Alzheimer’s.

I think we’re going to use this more and more to diagnose and eventually cater medical regimens to that information. What’s going on right now is wildly interesting.

People have this idea in their mind that we could radically transform ourselves through gene editing very soon. I’m not so sure. I hope that’s true, but they are ignoring the elephant in the room, which is polygenic embryo selection and polygenic scores more broadly.

Daniel 01:04:17

Jonathan Anomaly, thank you for joining us on the Free Radicals Podcast.

Jonathan Anomaly 01:04:20

Thank you. It’s been great.

Daniel 01:04:21

Thank you.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

02:17 Levinthal’s paradox and the protein-folding problem
10:03 From AlphaFold to Latent Labs: de novo binders (Latent X)
14:28 Beyond binding: building drug-like antibodies with Latent X2
18:17 The clinical data gap and personalized medicine
24:46 Latent Y: the end-to-end drug-design agent
32:54 Beyond single targets for chronic and aging diseases
36:32 The inflection point in pharma, and what’s next

Transcript

02:17 Levinthal’s paradox and the protein-folding problem

Daniel 00:02:17

Simon Kohl, welcome to the Free Radicals Podcast.

Simon Kohl 00:02:21

Thank you for having me. I’m looking forward to the conversation.

Daniel 00:02:23

We had some issues getting set up here at the SynBioBeta Conference, but it is smooth sailing from here.

Simon Kohl 00:02:30

It will be.

Daniel 00:02:31

We want to start by talking about Levinthal’s paradox. Predicting a protein structure from its amino acid sequence has been an open question since the 1960s. The nature of that question was crisply articulated through Levinthal’s paradox.

The idea is that if a protein had to randomly sample every possible configuration to find its fold, it would take longer than the age of the universe. Yet, proteins do it in nanoseconds. AlphaFold, which you worked on, is also able to do it in a single inference call. What is your intuition for how AI systems are able to solve this problem?

Simon Kohl 00:03:22

Protein structure prediction has been a longstanding problem for decades. The community worked on it with the idea that if you could predict structure, you could forgo the laboratory work that takes months or years to determine a protein structure experimentally.

At DeepMind, we set out to work on that problem. We entered the CASP competition, or the Critical Assessment of Protein Structure Prediction. With AlphaFold 1, the team already outperformed the field, but not at a performance level that rivaled experiments.

With AlphaFold 2, the AI system I worked on, we achieved that feat. The core idea is that the model is able to infer evolutionary history for a specific sequence. It maps context from how a protein evolves through a series of learned computational steps in the AI architecture into 3D coordinates.

As evolution plays out, random mutations are made. Functional ones survive, and those contacts that are mutable while retaining function can lead to new variants with a similar fold. In that covariation across sequences, you can infer contacts. The model is excellent at extracting that signal from evolutionary history and mapping it into a 3D structure.

Daniel 00:05:31

The key insight is that we aren’t talking about random proteins. We’re talking about proteins essential for life. There is conservation in what those structures need to look like across all of these proteins.

Simon Kohl 00:05:47

That is the fundamental principle of evolution that was exploited here. AlphaFold is not a physics-based model.

On some smaller or designed proteins, it is often able to go from a single sequence to the correct structure without evolutionary history. However, for most general-purpose folding problems, it relies on that evolutionary signal.

Eric 00:06:20

Early on at DeepMind, was there a debate about whether to take a physics-based route or a sequence-based route? How did you arrive at your path?

Simon Kohl 00:06:34

AlphaFold built on the shoulders of giants. People in the field had already figured out there was signal in evolutionary history. Those ideas were already being built upon.

At DeepMind, we figured out how to map that into a deep learning architecture that could learn from the large amounts of data in public databases. The key one is the Protein Data Bank (PDB), which now holds over 200,000 structures.

That dataset is the accumulated experimental output of the entire structural biology community. It’s a fantastic resource, and credit goes to the community for creating a joint repository where everyone deposits their structures. That is what enabled this work.

Daniel 00:07:47

If we discovered organisms deep in the soil or the ocean that were much more evolutionarily divergent, would AlphaFold be able to properly predict the structure of their proteins?

Simon Kohl 00:08:09

Probably, yes. There is likely still enough evolutionary signal. It rarely struggles, but when it does, it has the ability to predict its own error.

I worked on that specific aspect of the model. It gives you a confidence estimate that we showed is well-calibrated. When the model says it is uncertain, you probably shouldn’t rely on that structure.

Eric 00:08:45

Where does that uncertainty estimate come from? Is it a lack of available data to determine the evolutionary constraints?

Simon Kohl 00:08:56

It isn’t always perfectly clear, which is why it’s great the model has learned its own uncertainty estimates. It likely relates to the information content in the evolutionary history and how well the model can extract that signal from its inputs.

We co-train that error prediction during training. It ends up being well-calibrated because the model develops a sense of how confident it is based on its own learned representations.

Daniel 00:09:28

You’ve been busy since AlphaFold. You’re the founder of Latent Labs. Most recently, you released LatentY, and before that, LatentX2.

Six months ago, you released LatentX, your first model. While AlphaFold predicts protein structure, LatentX creates binding agents for specific proteins. What breakthrough enabled that leap from predicting structure to creating something that you know will bind?

10:03 From AlphaFold to Latent Labs: de novo binders (Latent X)

Simon Kohl 00:10:04

At Latent Labs, we focus on building generative models and AI systems for protein and drug design. We initially focused our Latent-X models on generating protein binders de novo, which means from scratch. There is no history in nature for these binders; they have zero evolutionary history because they are designed from the ground up.

I sometimes liken the difference to AlphaFold being an amazing AI microscope. It provides an accurate idea of a structure through an inference call, but it doesn’t generate new proteins or drugs. It visualizes existing biology.

At Latent Labs, we are working on the next step. We want to be able to push-button design drugs like antibodies, peptides, and cyclic peptides. Since the fundamental mechanism of action for most drugs is binding, we focus on that.

The leap is in how you formulate these generative models. We use a structure-based paradigm, modeling the binders and the proteins at a structural level. We do joint sequence-structure generation and model every atom, including the side chains. The breakthrough is in the architecture, the training, and the data used to tune it.

Daniel 00:11:40

Your point is that it’s based on structure. We have datasets of things that bind and their known structures. If you can predict the structure, that dataset encodes information about which structures piece together. That is what the model is training on.

Simon Kohl 00:11:59

That’s right. We are trying to learn the language of binding. Given a protein target and the exact spot where you want to bind—called an epitope—what amino acids do you need to place in space so that it binds? It is like solving a 3D geometric puzzle. We are learning the biochemistry required for binding from data.

The model itself is an AI model; it is not a physics-based model. It learns from data how to place atoms and side chains to create the necessary non-covalent and hydrophobic interactions. It adapts and generates binders from scratch in a forward pass.

This approach works exceptionally well. Our models are state-of-the-art, and we’ve benchmarked them against other models. We get strong, highly specific binders. This was first demonstrated with our Latent-X1 models about nine months ago.

In December, we launched Latent-X2, our foundation model for antibody design. It is an all-atom model that also handles peptides and mini-binders. The headline news is that we can now design antibodies that are often drug-like out of the gate. That was the leap.

Daniel 00:13:46

To dig into that, predicting binding is one thing, but many things that bind aren’t necessarily drugs. They might not have the anticipated effects, or they might be too immunogenic.

This seems like another leap in terms of the information your models are encoding. Now, the model is encoding structure, binding, and other complex characteristics. You’re increasingly interacting with more parts of the human system, like the immune system. How are the models capable of doing that?

14:28 Beyond binding: building drug-like antibodies with Latent X2

Simon Kohl 00:14:29

Exactly. For drugs, we need more than just binding. They need to work in the human body without causing toxicity or allergic reactions, and they need to be manufacturable.

One thing Latent-X2 can do is prompt the model with antibody frameworks that are already humanized or germline. These are better starting points for working in the human body. With that prompt, the model is more likely to create human-like CDRs, which are the loops responsible for binding.

While it’s not perfectly understood why it works, 47% of the antibody designs we’ve tested passed critical drug-like hurdles out of the gate, without optimization. Traditionally, to co-optimize these properties, you have to do it sequentially. This often leads to a game of “whack-a-mole” where optimizing one property makes another worse, resulting in long development cycles.

Achieving this from scratch with a single design step is a major change. It isn’t perfect yet, and not every design will hit the mark, but on average, we see that in a lot of our designs already.

Eric 00:16:15

Drug discovery is one of the most complex problems possible. You have to hit the right target and understand the disease biology, but you also have to consider how to modulate that protein’s function in relation to the rest of the human body.

It isn’t a simple lock-and-key problem with a static lock. The lock is amorphous and constantly changing how it interacts with other proteins or biochemical factors.

When you think about how that relates to your work, we can design specific and selective protein binders against a specific epitope, but how do you know if you’re hitting the right epitope in the right way to modulate that protein correctly?

Simon Kohl 00:17:06

We can experimentally characterize whether we have achieved our goal. For instance, we have conducted mutagenesis studies where we mutate either the binder or the target to verify that we have hit the right spot. Our studies show that the model really does bind at the designed location. We feel quite confident that when we bind, we bind exactly where we intended.

Beyond that, regarding function and how these designs work in the human body, we still have a way to go. We want to collect more data in that area. This is a key challenge for the field in general and for Latent Labs specifically. We need more data pertaining to immunogenicity, toxicity, and clinical readouts.

That data will be the real unlock to get our models to a state where we are confident putting their designs straight into the human body. We obviously would not do that today, nor would we be allowed to, but we want to reach that point.

I am confident that there are patterns we can learn without explicitly modeling every interaction a drug might make in the human body. Currently, with existing methods and compute, explicitly modeling everything seems infeasible for at least the next ten years, but we can learn those patterns from data.

18:17 The clinical data gap and personalized medicine

Daniel 00:18:46

Is the hope that the model can output something that can go right into a human? If we think about the thesis for your company, is that a bonus, or is it the central thesis of what you are after?

Simon Kohl 00:19:04

That is one of our central theses. That is where we want to be. Once we achieve that, we will be able to operate at an N-of-1 level, providing truly personalized medicines.

Of course, there is a missing link regarding regulatory steps. We need to determine what we must prove for a regulator to allow that. Regulatory processes will have to evolve, leaning more on organoid models and machine learning models.

The FDA has expressed increasing interest in these types of approvals over the last 12 months. While they are not there today, I expect that shift to happen over the next decade. Combined with our advances in drug design, we are looking at an accelerated, immediate way of putting our designs into the human body.

Daniel 00:20:10

It is interesting that you jumped to regulatory challenges as the primary hurdle. As someone not close to this, it strikes me that there are tremendous scientific hurdles to pass.

We see LLMs as an analogy, where they continue to improve through engineering—feeding them more data and solving specific problems—rather than just scientific leaps. Do you think of your models that way, as an engineering roadmap?

Simon Kohl 00:20:42

There is something to that; it is how we are thinking about it. The models are getting very strong. I perceive the gap to the clinic as a data gap rather than a modeling gap.

When we have datasets with significant breadth and diversity that translate to actual clinical readouts, I am confident our models will learn the patterns necessary to generate new antibodies and peptides that fulfill these rules.

Daniel 00:21:33

When I think about the secret sauce of your business, is it the algorithm or the data? That data seems very hard to get. Have you cracked a specific way to acquire it?

Simon Kohl 00:21:48

It is everything together. Over the last nine months, we have launched Latent X1, X2, and recently our agent, Latent Y. You can only maintain a trajectory like that with the right team and modeling insights.

Once you identify an improvement and understand how to translate computational gains to the lab, you can iterate rapidly. We have our own lab in San Francisco where we do exactly that.

The performance of our antibody models has increased significantly since the launch of Latent X2 in December. We have found a recipe that allows us to hill climb rapidly. It requires our own lab, our own models, and our own data collection. It all needs to come together.

Eric 00:22:58

I have a lot of optimism because technology is improving rapidly from month to month. We are doing things that were previously inconceivable.

Twenty years ago, the way to design an antibody was to start with an immunization campaign in a mouse or a llama. You would try to get the animal to initiate a self-driven campaign against a target and then back-decipher the structure of the protein produced. That is the Stone Age.

Now, we can sit at a terminal and, in an afternoon, validate a computationally designed protein that is specific for a target at a very specific site. That would have been inconceivable even five years ago. The question is no longer whether the technology can get us there, but what other roadblocks we need to out-engineer to reach end-to-end personalized medicine.

Simon Kohl 00:24:01

I agree. It is even more dramatic than that. Twelve months ago, we did not have antibody design models that worked in any meaningful way.

Looking at that trajectory, I believe we will crack the problem of push-button drug design within the next five to ten years. It is a vast departure from the immunization and display techniques that remain the industry standard today.

24:46 Latent Y: the end-to-end drug-design agent

Eric 00:24:46

Should we discuss Latent Y?

Daniel 00:24:47

Tell us about your latest model, Latent Y.

Simon Kohl 00:24:50

LatentY is a drug design agent. It’s a reasoning model that has access to LatentX2, LatentX1, and other bioinformatics tools we’ve built. We have shown that we can go from a simple prompt to the agent reasoning about it, using our molecular models like LatentX2 to create antibodies that work in the lab.

We’ve gone end-to-end, which is a significant leap. We’re increasingly seeing agents, but it’s rare for people to show that the science works in the real world. That’s a first here.

LatentY is available on our platform today, and we have users operating with the agent now. It has access to external literature and structural databases. It can reason about where to bind, what format to use, and which target to use in the process. Then it arrives at sequences that it recommends and performs quality assurance as well.

It mimics the workflow of an expert computational protein designer. Our user studies showed that LatentX2 compressed the timelines for arriving at a developable antibody from 18 months down to one month in the best case. Now, LatentY compresses what would otherwise take a computational protein design expert a few weeks down to a single afternoon.

The most exciting aspect is not so much the autonomy, but the scale that it gives you. If you’re a research team or a protein designer, you can now try many hypotheses in parallel that would otherwise take ages to set up. This allows you to 10x or 100x what you could otherwise do by testing different targets, frameworks, and setups.

We see it as a force multiplier that strongly increases the productivity of drug discovery teams.

Daniel 00:27:06

Will LatentY incorporate the experimental results after scientists test it? If so, will that be an important part of continuing to train the model?

Simon Kohl 00:27:17

In our technical report that came along with LatentY, we haven’t relied on iterative cycles with the lab. However, as the model is applied in the future to areas where there might be a greater sparsity of data, it may profit from that cycle of seeing real-world feedback and taking it into account.

You can use it without that, though. It can perform effective zero-shot antibody design.

Eric 00:27:54

I’m going to put my futurist hat on for a second. I’m imagining a world 20 years from now where we have nanorobots constantly surveying the activity of different protein-protein interactions and biochemical interactions in our bodies.

From there, you have micro-clinical trials happening within a human. If that reports back to a centralized model running continual learning, you eventually get to the point where you have the ability to learn from a single patient which drugs are most likely to affect their health trajectory.

Then you run an agentic framework like LatentY to design those drugs and scale up the experiment. We could continually learn on a specific patient how to prevent or treat disease on any level. That seems to be where we’re moving.

Simon Kohl 00:28:37

That’s a very interesting and compelling idea. Rapidly treating a patient with confidence is top of mind. There are a number of different roads that lead to Rome, and there are different ways of performing that translation, but I agree. That would be the dream—being able to adapt to subpopulations or even an N-of-1 in a very rapid way.

Daniel 00:29:10

If the model can immediately output something we can put into a human, it would be an important part of a flywheel that accelerates discovery. We close the loop by creating something and seeing results much faster. Does that provide more data to train models and give us more control over biology?

Simon Kohl 00:29:36

That’s very true. I want to say again that this is some way out. To be transparent with the listeners, I think we’ll get there, but it’s probably still five to ten years away. When we reach that point, it will be fantastic and enable exactly that kind of acceleration.

Eric 00:29:59

You mentioned that there are multiple roads to Rome. Tell us more about how you’re conceptualizing the best ways to build toward this future. What is the right wedge to build as a startup?

Coming from one of the largest companies in the world at DeepMind to building a startup of your own, what is the competitive advantage you see in leading a nimble startup while navigating these capital-intensive, technologically complex spaces?

Simon Kohl 00:30:27

It is the nimbleness and the focus. That is the key bit that has worked so well for us. We really go after one core idea, or perhaps a few.

Larger companies inevitably take multiple bets in many different areas. There is an element of being pulled in different directions and the inevitable distraction that comes with that.

As a startup, you have the luxury of doing one thing at a world-class level. At Latent Labs, we focus on rapidly shipping new models that are state-of-the-art and outperform others. That is the key advantage.

Eric 00:31:20

And how do you choose what to focus on? That is a very hard design problem.

Simon Kohl 00:31:24

Part of the process is intuition and part is market feedback. You have to iterate quickly, using your intuition to start, and then talk to possible partners to see what sticks. The results are often very surprising.

When I started out, many people told me antibodies were a solved problem. They asked why I would even work on them. Initially, I didn’t focus on them with the company, but I don’t think the “solved problem” narrative is true.

While immunization campaigns and phage display give you hits, the problems surrounding developability and immunogenicity are very deep. A few years ago, people didn’t dare to think a single model could address those in one shot. That is changing, and it will represent a real step change.

As a startup, you ping-pong around your core thesis with different slants until you find your way. Intuition is like your prior distribution; then you talk to the market and it collapses into a clear path.

32:54 Beyond single targets for chronic and aging diseases

Daniel 00:32:53

That’s awesome. Big Pharma has generally done well with diseases where there is a particular target to knock out. However, they haven’t been as successful in managing chronic diseases or diseases of aging.

In those cases, there isn’t a single target; there is a loss of homeostasis where multiple processes are out of equilibrium. How do you see AI enabling us to go after those more complex, homeostatic illnesses?

Simon Kohl 00:33:41

That is a great question. It isn’t something we focus on heavily at Latent Labs today. Our product enables our partners with the superpower to make an antibody against any epitope they care about.

This could be for therapeutics, diagnostics, or research. I’m excited about probing unknown biology with our binders and enabling our partners to do that rapidly.

If a phage display or an immunization campaign takes weeks or months, it isn’t fast enough to iterate through hypotheses. With our technology, you can get results in an afternoon. This will enable research breakthroughs and rapid hypothesis testing to understand fundamental biology beyond just making drugs.

Eric 00:35:01

Antibodies have never been a more exciting space. Recently, Candid Therapeutics was acquired by UCB for $2 billion only a few years after launching. They are designing fast-follower molecules, like bispecific T-cell engagers, to selectively eliminate immune cells for autoimmune disease.

This idea has been in the literature for a long time, but now old ideas are new again. Applying an antibody to re-engage the immune system against itself has unlocked a new language for treating disease-causing cell populations.

There is an entire library of epitopes and cell-specific engagers left to develop. At the core of that progress will be designing the right antibody to bring two cells together. There is so much left to develop in the antibody world.

Simon Kohl 00:36:06

I couldn’t agree more. It is a wide-open field. There are plenty of unaddressed targets and tough functional targets out there.

When you consider things like BiTEs and multispecifics, there is a whole combinatorial space that ensues from trying to simultaneously bind different targets.

36:32 The inflection point in pharma, and what’s next

Daniel 00:36:32

How optimistic are you about where this future can take us? Even if Latent Labs isn’t working on the diseases of aging specifically, you mentioned how this can accelerate scientific discovery. Do you think these technologies can give us a dramatically longer lifespan?

Simon Kohl 00:36:53

I should caveat this by saying I am not an expert in aging. But my feeling is that we still know very little about human physiology and the aging process. We are only scratching the surface.

Because of that, I imagine there is so much yet to be unlocked that we can use to improve our lifespan and our healthy years. I am very optimistic that AI will unlock the science necessary for those improvements.

Eric 00:37:44

Is there anything else we should cover with our audience before we part ways today?

Simon Kohl 00:37:49

I think we have covered what we are building quite well. I feel there is an inflection point in our industry. We now have enough data to train frontier models like Latent-X and Latent-Y that can reason on top of that data.

It will take some time for the impact to fully unfold, but the writing is on the wall. This is what keeps me going. It is really exciting.

Ultimately, this is for the better of everyone. We are going to have better drugs faster and, hopefully, cheaper and more accessible. We will see more “N of 1” treatments for conditions that currently lack economic justification. I am very optimistic we will see a lot of progress.

Eric 00:38:52

What’s next on the roadmap? What are you most excited about for Latent Labs in the coming months?

Simon Kohl 00:38:57

We’re already working on future, even more powerful model generations. I hinted at that earlier.

We’re also increasingly rolling out with partners, which is very exciting. We are at the front line with them, testing what’s important and difficult for them. We’re going to continue to push along these two vectors.

Daniel 00:39:27

Simon Kohl, thank you for joining us on the Free Radicals podcast.

Simon Kohl 00:39:30

Thank you very much for having me. It was a blast.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

0:00 Intro
2:12 What era of biotech are we in?
3:58 GLP-1s: the once-in-a-generation breakthrough
7:04 Inside Lilly’s trillion-dollar valuation
9:09 The Gillette Era of GLP-1s and the Bryan Johnson Era of Consumer Biotech
12:55 The $1T question on Consumer Biotech
14:33 Patent cliffs, fast followers, and building for “me-last” therapies
18:20 Disintermediating pharma: HIMS, Loyal, and the standalone biotech
19:41 Why do biotechs sell to pharma in the first place?
23:03 Longevity: a science problem or an ecosystem-alignment problem?
27:14 Why GLP-1s came from pharma
29:52 Why Elliot is Bullish on Standalone Biotechs
33:03 Is Drug Development Becoming more like Software? Not if upside is capped at <$10B
41:30 The geroscience hypothesis and longevity as massive markets, medium prices
45:36 Treating age related diseases or aging as an indication
47:55 How do you value curing aging?
53:19 Sid Sijbrandij, N-of-1 medicine and going “Founder Mode” on cancer

Transcript

2:12 What era of biotech are we in?

Daniel 00:02:11

Awesome to have you. I’d love to start with your take on the current narrative in the biotech industry. How would you describe the era we are in?

Elliot Hershberg 00:02:26

We started our fund at Amplify a little over a year ago. At that moment, the core narrative was whether biotech would continue to exist. We were coming off the last mega-cycle of COVID, which saw a huge rush of investment and new science, followed by a major reset.

At one point, two-thirds of public biotech companies were trading below cash. There was an existential threat regarding whether China would become the primary source of innovation.

A year later, we are in a different part of the cycle. There is more investment activity and more IPOs. We are seeing incredible scientific progress, particularly with AI in biology. We’re seeing model progress and new technology. We are in the next wave, realizing there is a lot of life and exciting innovation left in this industry.

Daniel 00:03:45

Is this new life in biotech primarily driven by breakthroughs in AI, or is something else giving it momentum?

3:58 GLP-1s: the once-in-a-generation breakthrough

Elliot Hershberg 00:03:58

It’s hard not to point to GLP-1s. These represent a once-in-a-generation product breakthrough—potentially the singular breakthrough in biotech history. They have led to enormous franchises for massive indications.

If you look at the cycles of breakthrough products, there was a period focused on large cardiovascular indications, the statin era of blockbusters. Then we saw a cycle of specialty products for rare diseases and orphan indications. Now, there’s a huge resurgence in large markets with products driving tangible benefits for patients.

As pharma companies race to catch up with Eli Lilly and Novo Nordisk, that product success trickles down into innovation. To sustain a nearly trillion-dollar market cap, Eli Lilly has to invest heavily in the ecosystem. They are facing the biggest possible patent cliff and must consider how to recover.

When you looked at the JP Morgan conference this past January, there weren’t many acquisition announcements. It was all about AI. The focus is on the tools that can sustain this level of product innovation.

Eli Lilly, the first trillion-dollar pharma company, partnered with NVIDIA on a billion-dollar transaction. This speaks to the momentum in the market. We are seeing success in the indications we’re targeting, combined with efficiency gains from new technology. Combined with a better macro environment—as biotech is typically inverse to interest rates—we are seeing a significant resurgence.

7:04 Inside Lilly’s trillion-dollar valuation

Daniel 00:07:03

Regarding Eli Lilly’s trillion-dollar valuation, what exactly is baked into that? Is the expectation that the broader biotech industry will create more therapies for Lilly to acquire, or that Lilly will produce them internally through investments like the one with NVIDIA? Where will that value come from?

Elliot Hershberg 00:07:31

Some of it is directly underwritten by the product growth of these first medicines. These are potentially $100 billion a year commercial opportunities. We’ve seen single effector drugs, dual agonists, and triple agonists. Now we have quintuple agonists.

There was an Onion article making fun of Lilly that said, “We see people doing one, two, and three; we’re doing five.” This continued product innovation leads to much better medicines for the massive cardiometabolic indication.

Companies like Lilly are being incredibly ambitious, thinking about aesthetics and longevity. Now that we see a willingness to pay for biomedical innovation directly out of pocket, companies can imagine supplying that demand. One factor is a literal DCF of the first wave of products.

Another factor is a premium for Dave Ricks and the management team, based on the belief that they can reinvest that money and grow into adjacent product opportunities that might be even bigger. But some of it is just having such a phenomenal product already.

9:09 The Gillette Era of GLP-1s and the Bryan Johnson Era of Consumer Biotech

Eric 00:09:09

We’re in the Gillette razor blade era of GLP-1 agonists. You’re adding an additional blade and performance is increasing. You get a closer shave and lose more weight more quickly.

There might be a capping-out effect eventually in terms of functional improvements, but it will be interesting to see how these drugs affect more than just the top-line weight loss metric. These are ultimately psychoactive chemicals. These molecules have profound effects on dopamine, the dopaminergic system, addiction, and many other things seemingly unrelated to weight loss. It’s fascinating to see how broad-spectrum these molecules are.

The most interesting development I’ve seen recently is that we’re entering the Bryan Johnson era of biotech. Individual experimenters who were once on the fringe are now applying frontier science to their own bodies. The biohacking era of personalized medicine is upon us.

From the largest companies in the world, we’re seeing that it’s not just fringe individuals anymore. One in eight American adults is now on a GLP-1. Normal people are interested in experimenting—not as diseased individuals trying to get back to a baseline, but as healthy people asking how they can get better. It’s true bioaugmentation.

Elliot Hershberg 00:10:54

Human enhancement is a real category. A lot of Wall Street is looking at these readouts and asking for the maximum possible weight loss efficacy. But it’s likely a more composite product than that.

For healthy individuals, there is the obesity market, but also a level of maintenance to manage weight as you age. Different people want different things. Beyond weight loss, the toxicity profile is critical. How does the drug make the patient feel? Is the patient nauseous? What is the dosing schedule?

In this industry, we talk about the target product profile. For an oncology drug, the profile is simply whether you maintain life. For a genetic disease where someone is suffering, we might endure adverse effects that wouldn’t be acceptable in another setting.

But for a product for muscle growth, weight loss, or sleep—something everyone might take—you’re threading a very specific needle. It has to be demonstrably safe. If we imagine half the population taking a biological intervention over time, those different components of the product profile start to really matter.

12:55 The $1T question on Consumer Biotech

Daniel 00:12:55

Do you expect Eli Lilly and other pharma companies to successfully build on the success of GLP-1s to create other therapies with widespread consumer adoption? Can they create the future products we care about for sleep or metabolism?

Elliot Hershberg 00:13:18

That is the trillion-dollar question. We’ve never had products of this scale before. You need to take big swings to not only maintain that position but to compound on it. There are two existential questions as Lilly and others reach this scale.

First, what is the surface area of product opportunities compatible with a moonshot of that scale? If you are only playing in potentially $100 billion markets, that is very different from how most pharma companies have been structured.

Second, if you constantly need to fill $100 billion holes in your patent window, what can be changed about the drug discovery process? You need to be able to find new drugs within the timelines of your products being genericized.

It changes the aperture of what you would go after as a reasonable product. It forces you to think very hard about the speed at which you deliver what are effectively scientific moonshots.

14:33 Patent cliffs, fast followers, and building for “me-last” therapies

Eric 00:14:32

Pharma traditionally operated in five, seven, or thirteen-year windows where a novel chemical entity entered clinical trials and was approved by the FDA for sales and marketing. This granted a window of exclusivity protected by IP, making you the exclusive owner of that molecule in commercial markets.

The loss of exclusivity is the “patent cliff.” Once that occurs, companies must fill the resulting revenue gap with new approved molecules as the original goes generic.

Today, even before that exclusivity window ends, updated molecules are overtaking prior generations within a year or two. These feedback cycles have shifted from ten-year horizons to one or two-year horizons. For example, retatrutride is replacing tirzepatide on the gray market before it has even passed clinical muster.

Life sciences is starting to look much more like software than ever before. The ability to efficiently design, optimize, and develop molecules into medicines may become the primary advantage for staying on top of the market.

Elliot Hershberg 00:16:02

In aggregate, that’s true. As an industry, we’ve become progressively focused on fast followers. This was a compelling investment thesis ten or twenty years ago when delivering a target product profile was complex and involved significant biology risk.

In many cases, it really pays to be “me-better.” For example, the seventh statin became the absolute blockbuster product. At that time, it was a discovery problem rather than a biology problem.

Now, commoditization and global competition—especially from China—are driving incredibly fast development and discovery. AI tailwinds are also compressing timelines, potentially allowing “me-better” drugs to be created in a forward pass.

One of the companies we invested in uses a great phrase: you need to develop the “me-last.” If everyone can create products rapidly, the premium lies in differentiation that completely closes the door on efficacy. We haven’t achieved that yet; there is still significant potential in underlying GLP-1 biology.

Imagine using these tools to reach the point where you are that seventh statin that absolutely nails it. These tools could put you in a position to be the first or second to market in a way that completely closes the door behind you.

18:20 Disintermediating pharma: HIMS, Loyal, and the standalone biotech

Eric 00:18:20

The prior playbooks for how molecules gain clinical footholds and patient uptake are being rewritten. Historically, patients accessed drugs through clinician oversight. You went to your doctor, they prescribed the drug, and that was effectively your only route.

Patients going directly to a physician or telehealth clinician to ask for a specific molecule used to be rare, limited to drugs like Viagra or amphetamines. Today, that market has exploded.

We now see a new middle-market layer of companies like Hims. They act as distributors for existing molecules without innovating on the underlying chemistry. They apply a direct-to-consumer playbook, focusing on sales, marketing, and LTV/CAC ratio optimization.

19:41 Why do biotechs sell to pharma in the first place?

Elliot Hershberg 00:19:32

From the investment side, it’s too early to know what to make of this. Traditionally, biotech companies primarily sell to biopharma because winning at late-stage clinical development is an uphill battle. Financing a billion-dollar science project through a registrational trial or multiple Phase 3 trials is incredibly difficult.

Even if a biotech company successfully becomes a commercial organization, Wall Street often shorts the launch. Building a commercial engine capable of hitting milestones and capturing market value is a Herculean task.

Two things could change this equation: AI and more innovation-forward regulation. Is it possible to lower the costs of early and late-stage clinical development? If we can make progress on that front, it becomes easier for new companies to reach the market with approved products.

Additionally, the distribution advantage for biopharma is hollowing out. Eli Lilly’s primary driver for adoption is LillyDirect, which feels similar to Hims. If it becomes easier to build distribution and launch successfully, the landscape changes.

This is a long-term shift, and the timing is uncertain. Historically, pharma companies took the risks to build obesity drugs and direct-to-consumer channels.

But imagine a mashup between Hims and a consumer-focused biotech franchise. If they can innovate faster and build those channels, such companies could certainly emerge.

Eric 00:22:48

Possibly.

Elliot Hershberg 00:22:49

I think there are some interesting shifts right now in terms of baseline consumer behavior, rather than just patient behavior, for being interested in these types of biological interventions.

23:03 Longevity: a science problem or an ecosystem-alignment problem?

Daniel 00:23:02

This makes me think a lot about the fact that if what we want to exist are genuine longevity therapies—like a drug I could take every day or a shot I take once a month that can extend my lifespan and delay all types of chronic illnesses—it’s easy to think of that as a scientific challenge. We need some lab to crack something.

Alternatively, it could really be a question of just the alignment of resources and efforts in that direction. We could imagine what you were just describing, where if big pharma companies like Eli Lilly figure out how to go direct to consumers and there’s consumer demand for preventative drugs like that, it reorients the entire ecosystem.

Eventually, that causes biotechs to start developing those types of therapies. To what extent do you think of longevity therapies, sleep drugs, and all these things we might want as genuine science problems versus a challenge of orienting the ecosystem toward this new opportunity?

Elliot Hershberg 00:24:07

I think there are two big angles. One is starting in a traditional, marketed, and regulated indication. For neuroprotection, we have a ton of indications where you need healthy maintenance of neurons and preservation of neural function. You could imagine tackling problems in Parkinson’s and Alzheimer’s.

Then, similar to the way the story played out for GLP-1s—starting in diabetes, moving to obesity, and then scaling into broader populations—you tackle science in indications, get to marketed products through classic randomized controlled trials for new chemical matter, and then realize that there are bigger opportunities.

I think there’s also a more bottoms-up dimension involving biohackers, Reddit forums, and all this activity around different science where it’s more about experimentation and bottoms-up research. It’s less around medical claims at the start.

You’re not going with the indication claim first; you’re really backing into just exploring different types of medicines and getting consumer-led data collection. I think there will probably be some instance of both as this plays out.

To your question, though, it does feel like capital-S science where there’s still a lot that we don’t know about partial reprogramming and these types of modalities. I think there are really low-hanging fruit solutions for longevity, like whether we should all be taking GLP-1s or PCSK9s, or doing things for cancer prevention.

But for neuroscience and core healthspan extension, that whole world of research and products does feel more nascent.

Daniel 00:26:35

One thing I would press on a little bit is that GLP-1 biology was known about for a long time, but my understanding is that the pharma companies didn’t see the opportunity to commercialize it.

I wonder if there are similar low-hanging fruits just sitting around, ignored because those consumer markets have been ignored. There are ones we know about, like statins, but maybe for these other diseases of aging like Alzheimer’s, we’ve just not been able to crack it.

27:14 Why GLP-1s came from pharma

Elliot Hershberg 00:27:14

It’s probably both. Our friend David Yang wrote a good blog post about some of the history for why GLP-1s first came from pharma. There is some level of top-down tastemaking from pharma companies, especially in a world where they have a real advantage in late-stage clinical development and distribution through commercialization.

When that’s the case, two things can change to open the aperture of what we go after. One is that just becomes less true. It becomes more possible for biotechs to do things that are not a clear acquisition target.

If you think of the 30,000-foot view, there’s a shopping list of needs and hot targets that are really viable from an acquisition perspective. Biotechs are formed to service that demand. One way to think about how biotech venture has operated for a long time is as pre-financing R&D for biopharma. It’s off-balance sheet R&D.

In that world, you shouldn’t expect things to happen that are dramatically different from what pharma wants. If it becomes more possible to make a commercial biotech because there are consumer applications or a different market setup, you can imagine companies having different incentives and behaviors.

You could see investors having different expectations. Or, even if none of those things change, the pharma companies might start playing in these different areas and undergo a mental shift due to the success and scale of these products we have been overlooking.

If the Lillys of the world are planting lots of seeds and putting out the wanted sign for more biology, then we should also see more demand that way. It’s a question of whether you are decoupling biotech from pharma to lead to more independent commercial organizations that can pick what science they really believe in, or if you are getting different top-down tastes because of this. I think both are possible.

Daniel 00:29:37

It just changes who the tastemaker is. If the biotech startups have ambitions of going public on their own and going to the market on their own, then they’re dependent on the tastemakers within VC and what the VCs are willing to back.

Elliot Hershberg 00:29:50

That’s right.

29:52 Why Elliot is Bullish on Standalone Biotechs

Eric 00:29:52

Loyal is a good example of a company that’s pushing for the disintermediation of pharma as the end customer of early-stage biotech. They’re taking on novel biology risk and chemistry risk with developing their small molecules for dogs to extend lifespan, but they’re also taking on the goal of distribution.

They are building out their own distribution channels and running the clinical trials. They’re eating their own dog food and really taking things end to end.

On the other end, we’re seeing a power law distribution of resource availability within pharma more than ever. We’re in this barbell era of life sciences and pharma where we’re going to see more acquisition and more large deals with the major heavyweight pharma players like Eli Lilly, Novo, and others.

We’re also going to see the rise of small, nimble biotechs who go from end to end on distribution and drug discovery. It’s very interesting to see both happening at the same time. If you had to put your finger on one side of the barbell, which one are you more personally interested in or bullish on?

Elliot Hershberg 00:31:04

I am interested in a world where we can have more standalone biotechs. If you change the economics of discovery and it becomes more tractable to get an approved product, you’re able to set taste and follow your own vision.

As an early-stage venture investor, I believe in small teams having answers and decentralized tastemaking. To the extent that those companies can scale and become product companies, it changes the economics of biotech venture, how it works, and what products we should expect.

I find that particularly exciting. It’s hard to say how the five-to-ten-year balance between those two will play out, but there’s a lot of progress toward making it more tractable for companies to do this.

Daniel 00:32:07

There are probably a lot of benefits beyond higher returns on biotech investing. If biotech startups don’t need an exit and can instead become platform companies, you’re not shedding all that talent. Hopefully, those teams are better set up to go create a bunch of new drugs. That would be the hope for what these biotechs can turn into.

33:03 Is Drug Development Becoming more like Software? Not if upside is capped at <$10B

Elliot Hershberg 00:32:30

At the start of Bio 1 for Amplify, it was scary putting money into the ground. You’re always told to buy when others are fearful, but in practice, it’s still scary. We strongly believed in the tailwinds for the sector and the set of impactful discovery technologies, but the fear was there.

There is a parallel story regarding my partner Sunil, who started Amplify in 2012. It was not for biotech initially; it was for complex enterprise solutions like databases, developer tools, and data infrastructure. At the time, those weren’t really categories, and people didn’t think developer tools companies could be big. The prevailing sentiment was that enterprise was dead.

In that period, YC was exploding, consumer web companies were everywhere, and that’s where the dollars were going. Enterprise was overlooked for reasons that rhyme with how biotech is set up today: there was a constrained exit market and distribution was hard. It was difficult for a new startup to compete against Oracle, IBM, and Cisco.

Sunil calls the pattern “the 32 long.” A product manager from a place like IBM or Oracle would know the roadmap and realize they could leave, start a company to build the next product feature, and basically sell it back to one of the large-scale distributors. They would hire engineers to work in the back room and have constrained exit sizes. While the web had huge exits, database land saw $200 million or $400 million bolt-on acquisitions.

That seems strikingly similar to how biotech has been. Venture sets the taste and repeat professional managers come from the effective buyers, then they leave and sell it back into those businesses. When it’s a constrained exit environment like that, you are capping upside and limiting the dynamics of the market.

What changed? The AWS Marketplace happened. There was a chance to distribute enterprise solutions much differently. Leverage shifted to great engineers, researchers, and scientists who knew how to build great products. All of a sudden, they had a different distribution channel and the ability to make large companies.

It wasn’t $400 million exits for Amplify initially; it was $50 billion companies like Datadog. No one thought those companies were going to be that big. In a world where we move away from this capped upside strategy, you will see a lot of things change about how dollars go into biotech.

Eric 00:36:44

It’s very interesting. Biotech has operated for the past 20 years in that capped upside environment. A $10 billion exit is really the peak of what pharma would pay for a molecule or a series of molecules from a biotech. While $10 billion is great, it’s an order of magnitude or two less than the best outcomes in technology companies.

There hasn’t been a breakaway story of platform biotech companies achieving escape velocity and becoming standalone companies that own end-to-end molecule development, clinical development, and commercial development.

The longstanding question has been: what is the right technology lever, market environment, or finance lever to enable that breakaway success story? Maybe we’re seeing today that because of the compressing cost of developing and distributing novel molecules, all the fundamental modules are in place for that story to finally happen. It’s the combination of all these things into a single tapestry.

Elliot Hershberg 00:37:58

There are different business models even if you don’t change any of the levers. If you don’t assume a difference in development costs or the atrophying of distribution advantages, there is still business model innovation that enables compounding and growth.

Roivant and BridgeBio are two really interesting examples outside of the $10 million cap. These are now roughly $30 billion market cap public companies that are explicitly focused on enduring. One thing they had to do was figure out how to offer a product that was interpretable by Big Pharma. They did this through firewalled subsidiary companies that are imminently acquirable, containing the actual products Pharma wants to bolt onto their distribution engines.

I’m seeing a lot of innovation on that model, where companies walk farther up the value chain. You might first sell a Phase 1 program, then a Phase 2, then maybe a Phase 3, or eventually move into co-development and build out your own scale.

One of my favorite parts of the BridgeBio story is that Neal Kumar was deeply involved in the industry as an investor at Third Rock. Interestingly, BridgeBio was not financed by Third Rock because it didn’t fit the typical business model of venture capital.

We need new capital in the market interested in underwriting these types of businesses. Neal often mentions that many asset allocators think they are better at it than BridgeBio; they want to manage the portfolio themselves. But if you imagine groups piecing together these tools in new ways to enable scale and repeated product development, the exit sizes and business scales can be much larger.

These companies reached commercial escape velocity much faster than others. Argenx took a long time to become a $50 billion company. Alnylam, once they put their strategy in place, had a “5 by 15” goal to have five later-stage products by 2015, which happened quite fast. For BridgeBio, it was about a 10-year story. By the 10-year mark, they had two or maybe three approved products, which is incredible.

41:30 The geroscience hypothesis and longevity as massive markets, medium prices

Daniel 00:41:15

Eric and I have been debating something related to how technology might change the physics of the biotech and pharma industries. There is a concept in the longevity field called the geroscience hypothesis.

The biological idea is that if we intervene in the biology of aging, we will see the benefits of delaying various chronic illnesses. The business analog is that if we make drugs tackling aging biology, the markets will be massive and lucrative, catalyzing a flywheel of investment into the field.

I’ve been wondering if that business hypothesis is correct. Do we need to stay focused strictly on aging biology, which is still somewhat immature, or is it better to brand scientifically easier interventions as longevity? This could include GLP-1s or peptides that improve sleep. Which path is more important for actually delivering longevity therapies?

Elliot Hershberg 00:42:37

Tim Oppler at Stifel has a great phrase for the GLP-1 phenomenon: “massive markets, medium prices.” Historically, we tried small markets with high prices. That is a specialty franchise model that works because it delivers high patient value. The price is justified by the fact that the drug significantly reduces the future medical burden of a rare disease.

Then there was an era where we tried massive markets with massive prices, and payers effectively said no. Early approaches for Regeneron’s PCSK9 drugs or certain Alzheimer’s treatments faced this. The math of a high price point combined with a huge market would simply break the American healthcare payer system.

The real innovation now is targeting massive markets at a medium price point. Tapping into huge health demands with a meaningful but sustainable price point is the way forward. While there will be price competition for GLP-1s over time, this model will push research toward biology that wouldn’t have been prioritized otherwise.

I don’t know how directly this is coupled to longevity, but it is certainly adjacent. If you have your metabolism, cholesterol, and sleep under control, that will have a net impact on longevity.

There are still questions regarding the tractability of using longevity itself as a primary endpoint for medicines. That feels like its own can of worms to solve. However, this focus on massive markets and medium prices is likely very good for the interventions that impact how long we live.

45:36 Treating age related diseases or aging as an indication

Eric 00:45:32

Let’s dig into that point. There’s a lingering question in the field: do you go after aging itself as a disease indication?

Do you go after the “big kahuna,” or do you steadily tackle individual subsets of chronic disease that we know are related to aging but have a different label? For example, you might deal with weight loss, chronic inflammation, or sleep management.

As an investor, do you have a preference for one approach over the other?

Elliot Hershberg 00:46:04

The primary angle so far has been businesses betting on a trajectory and rollout similar to GLP-1s—targeting diabetes, obesity, weight loss, and general health maintenance.

For example, Nulimit is a really interesting business focused on liver disease, regenerating and improving immune defense for elderly populations, and targeting T cells. There are all sorts of ways to scope it and gather data to see if there is something there.

I’m a big believer in randomized controlled trials and seeing an intervention supported by evidence. You might even learn things about what your medicine does that you wouldn’t have thought of a priori.

To your point about the massive surface area of GLP-1 medicines, they interface with the body everywhere—the CNS, the gut—and they are big levers where you can learn a lot from secondary endpoints as people take them.

I’m biased toward establishing a gold standard body of evidence for a medicine initially, then following where the science leads. Some of my friends disagree and are taking different approaches, and that’s good. It’s a free market.

47:55 How do you value curing aging?

Daniel 00:47:52

The upside of an indefinite lifespan is infinite. How do you invest in something where there’s a 99% chance of failure but a 1% chance of infinite upside?

I wonder if there’s an analog in biotech investing. If I’m developing a longevity therapy but list it for a specific indication, that might limit the drug’s value to $1 billion.

But if I’m trying to cure aging, it’s a $100 trillion drug. Is that trade-off relevant when deciding which biotech to invest in?

Elliot Hershberg 00:48:33

Upside matters. Biotech is unique because you can get somewhat granular market sizing relative to tech before the product even exists.

You can still be wrong in your commercial estimates, but you have an uncanny ability to carve up and underwrite the market size. For example, BridgeBio found it was deceptive how big ATTR-CM was.

Pfizer, Alnylam, and BridgeBio stumbled into a much bigger indication because the disease was being massively misdiagnosed. There is a coupling between infinite market opportunity and tractability within the timescale of a venture investment.

You’re constantly balancing that. You might take a $1 to $5 billion product opportunity with a tractable clinical development strategy. In those cases, you know a lot from a Phase 1B study, and the patient enrollment requirements are clear from the outset.

No one disagrees that Alzheimer’s is a huge market with significant unmet medical need. However, the prevailing investor sentiment has been that the burden of evidence required for a transacted data package—and the cost to advance it yourself—is a limiting factor.

Biotech investors are looking for the Venn diagram of what is big enough to drive returns while remaining tractable for a small biotech. This is why rare disease, oncology, and I&I have been such focus areas. The cost to proof-of-concept is manageable.

Interestingly, things are starting to change for Alzheimer’s. New mechanisms like brain shuttles and blood-brain barrier penetrant biologics are showing meaningful movement on smaller patient numbers.

The Aaliyah acquisition was around $3 billion based on a Phase 1 study, which is significant. Once something becomes tractable from a development perspective, you can go after it.

But when you weigh tractability against upside, a multi-billion-dollar research effort is hard to sign up for. Given the probability of technical success in biotech, you could outlay hundreds of millions of dollars and be left with zero. That’s the balance people try to strike.

Eric 00:52:34

One topic we touched on earlier is the interesting bifurcation of life sciences and biopharma. I’m curious to dig into the “consumer biotech” era we seem to be entering.

On one hand, we have mega-blockbuster drugs in the GLP-1 class and other potential molecules like muscle agonists and sleep drugs. On the other hand, we have N-of-1 medicines.

How do we identify the interesting bets or companies emerging from these two categories? Let’s look specifically at the N-of-1 personalized medicine side.

What is causing the current uptick in that space, and what are you most interested in seeing emerge?

53:19 Sid Sijbrandij, N-of-1 medicine and going “Founder Mode” on cancer

Elliot Hershberg 00:53:21

We have this interesting setup where it’s never cost more to get to an approved drug. We have Eroom’s Law, the inverse of Moore’s Law, which represents an exponential decline in R&D efficiency and ballooning costs for late-stage clinical development. As a rough dollar amount, it costs about a billion dollars to get a drug approved.

At the same time, it’s never cost less to get to a viable development candidate. With modern technologies, in the most extreme cases, that cost can be closer to a million dollars, whereas it used to be tens of millions. This creates an interesting disconnect: it’s incredibly expensive to produce a finished medicine, but the cost to reach a viable proof of concept is decreasing exponentially.

Over the past year, I’ve gotten to know Sid Sijbrandij, the co-founder and CEO of GitLab. I first met Jacob on his team, who is essentially the CEO of Sid’s care. We went for a walk along the waterfront in San Francisco, and he told me what they had been doing. Sid took the idea of the “million-dollar molecule” and the depreciating cost of therapeutic creation very seriously.

Sid was unfortunately diagnosed with osteosarcoma, a rare diagnosis for someone of his age and health. He quickly went through the entire standard of care for his indication and was left with his oncologist essentially saying, “Good luck,” and suggesting there might be a trial out there.

Sid proceeded to measure everything humanly possible about himself and his cancer. He used 10x Genomics technology—cutting-edge single-cell technology typically used in R&D. He used every commercially available liquid biopsy and did a bake-off between the different tools.

He used that information to take rapidly accelerating discovery technologies and build a personalized pipeline of medicines. He worked with companies and academic labs to create ten discrete medicines explicitly designed for his cancer to serve as a backstop as his disease progressed.

He found an interesting insight about FAP in his tumor from the single-cell data and received a targeted custom radiotherapy in Germany. He is now disease-free. Sid had a completely different outcome than what oncologists would have predicted for his cancer type.

You could look at this story in two ways. One is that Sid is clearly atypical—he’s a brilliant engineer who built a $7 billion public business and had the resources to do this. You could say it’s just a human interest story where the punchline is, “Good for Sid.”

The other interpretation is that while it’s incredibly hard to change the standard of care for oncology overnight, we are seeing the potential for a shift. We talked about the gold standard for getting a drug approved, but it’s also hard to find patients to test new medicines. Experimental medicines often end up in late-stage refractory patients.

While we want to provide the best standard of care, treatments like immunotherapy—including the personalized neoantigen vaccine and checkpoint inhibitors Sid used—are known to be most effective in earlier-stage patients with functioning immune systems.

We have a challenge where the tools exist, but getting them into the standard of care requires trial and regulatory innovation. The frontier of what’s technologically possible is probably decades ahead of what an average oncology patient actually receives.

Sid is thinking a lot about how to change this. He has started a fund and several businesses based on his experiences. At Amplify, we’ve invested in companies tackling this, like Decade, which is working to make personalized radiotherapies much more accessible.

I’m in the camp that Sid is not an anomaly, but a sign of where the world is going. In a similar timeframe, we saw someone make a vaccine for his dog using GPT, and Patrick Collison using AI models to interpret his genome.

These trends point toward people having more command over their health, with an emphasis on gathering information and using new tools for interpretation. Combined with faster discovery technologies, we could see a set of n-of-1 medicines leading to dramatically different outcomes for patients.

Eric 01:00:28

One analogy that comes to mind is the ability to collect patient-specific biomedical data across many modes, parse and analyze it, and act with an increasing amount of specificity and leverage. How does this rising tide of knowledge and capability start to press against rate limiters like the FDA, clinical trials, and the standard of care in the United States?

We are seeing a biohacker movement where people operating at the frontier experiment on themselves. For a small subset of those experiments, we see remarkable results.

The unanswered question is what happens next. How does this evolve from n-of-1 individual experimentation to an industrialized sea change in how people manage their health going forward?

Elliot Hershberg 01:01:32

I want to push back on one dimension because it’s worth giving some credit to the FDA. Sid filed several applications for single-patient INDs. He was essentially saying he wanted to test an experimental medicine on himself outside of an existing trial.

Considering the typical cost and time of filing paperwork with the FDA, how long do you think that took each time?

Eric 01:02:06

Eight weeks.

Elliot Hershberg 01:02:07

It took 48 hours. Every time, he was able to get a fast response from the FDA and test experimental medicine. There is a single-patient IND track that exists. As Sid put it in his memorable way, the FDA wants him to live. There is actually not a huge burden there.

There are other positive things the FDA is doing, like platform designations to approve a part of a technology that is then portable for more personalized use cases. For example, if there is a gene-editing therapy using a lipid nanoparticle, the CRISPR and editing machinery remain the same for each patient.

The only thing that changes is the guide RNA. The FDA is asking if we should have a more flexible way to regulate and approve that, since we cannot run a randomized controlled trial for every possible guide RNA. We cannot do a billion RCTs.

CAR-T therapy is another example. These are approved products where it is not just a singular discrete asset; it is a process. The FDA had to think about the quality controls for editing a patient’s own cells and giving them back.

The FDA was not the rate limiter there. That product reached the market and is having a huge impact on the lives of patients with blood cancers. When it comes to single-patient INDs, the plausible mechanism pathway is what was put forward by Macri and Prasad.

There are many tools being put in place by the FDA. I wonder if this falls closer to the GLP-1 example, where it is simply not how pharma thinks about marketing products. It requires either a company like Lilly to take moonshots on personalized approaches, or a standalone biotech to start marketing them.

That would have the downstream effect of changing investor behavior. Right now, there are decent regulatory tailwinds. The technology is getting there, and the case studies are compounding.

After I wrote about Sid, I had a flood of patients in my inbox looking to achieve similar outcomes. A surprising amount of these people end up having dramatic differences in their outcomes. These things will continue to compound.

The missing link is the first massive product success. It could be something more traditional that goes through the FDA rather than needing something dramatically different in place. A lot of the pieces are already there.

Eric 01:05:43

That is a really interesting and optimistic point. You are reframing the question: what do the actors within the industry need to do today to exercise agency?

How can they effectively make use of the technology tools and the regulatory environment that are already available? You are bridging latent patient demand into a novel company or a new way of connecting personalized medicine with patient access.

Elliot Hershberg 01:06:10

I think that is right. A lot of the core substrate and setup is there. It comes down to an integration problem and a market creation problem.

Daniel 01:06:25

Sounds like a great area to build in and to invest into.

Elliot Hershberg 01:06:28

I totally agree.

Daniel 01:06:30

Amazing. Thank you so much for joining us on the Free Radicals podcast.

Elliot Hershberg 01:06:33

Thanks, guys. It was a lot of fun.

Daniel 01:06:34

Thank you for listening to this episode of the Free Radicals podcast. If you enjoyed this episode and would like to support us, the most helpful thing you can do is share this with a friend you think might enjoy it too.

Please also leave us a 5-star review on Spotify and Apple Podcasts, and like and subscribe on YouTube. It would really mean a lot. I’m Daniel Shur, and my co-host is Eric Dai. Thanks for listening.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

2:08 History of the longevity / geroscience field
9:39 Origins of the geroscience hypothesis
14:15 How to frame gerotherapeutics as the greatest preventative drugs ever
18:11 The zone 2 in a pill drug in clinical trials
20:41 How to distinguish “real longevity drugs” from other drugs like GLP1s
25:10 Why it matters that we draw this distinction
28:56 Why it requires a quantum leap for pharma to tackle the massive longevity opportunity
32:24 What it takes to get longevity drugs approved
35:32 The importance of establishing endpoints for aging (to get the flywheel going)
40:24 Odds of getting the flywheel towards longevity escape velocity going
42:48 Most exciting companies in clinical stage today
45:49 The secret sauce of the geroscience field (and what big pharma will need to learn)
58:39 What makes this interview unique
1:05:23 The reason grifters have rushed into the longevity space (and how to tell what’s real)
1:10:47 The difference between being healthy and actually building towads longevity
1:14:30 How to contribute to the field

Transcript

2:08 History of the longevity / geroscience field

Daniel 00:02:08

James Peyer, welcome to the Free Radicals podcast.

Dr. James Peyer 00:02:11

Daniel, thank you for having me.

Daniel 00:02:12

James, you’ve been working in longevity for two decades. When you entered the field, it was a fringe idea, but today longevity is practically mainstream.

While it means different things to different people, there are currently no approved longevity drugs on the market. You’ve founded over 20 startups to change that, brought in over half a billion dollars into the field, and are positioned to soon have what may be the first drug to slow aging.

We are at an incredibly exciting moment in the field. Before getting into the details of the drugs you’re developing now, could you take us back to when you first entered the field? What was the state of things then?

Dr. James Peyer 00:03:00

I started getting into the biology of aging as a realistic set of processes that could be modified in the early 2000s. It’s crazy to think that was more than 20 years ago.

At the time, the intellectual leaders in the field included Aubrey de Grey, who had just introduced the idea of the “seven deadly things.” This was a quantum leap in thinking, viewing aging as a set of discrete processes that broke down and could be repaired.

The other leader who was very important then was Michael Fossel, a professor at Michigan State who wrote Reversing Human Aging in 1996. That book came out shortly after the creation of the first telomerase reverse transcriptase, which allowed the immortalization of human cells.

Michael Fossel and Michael West, the founder of Geron Corporation, were trying to commercialize telomerase. They believed that since we had figured out how to make cells immortal, stepping up to humans shouldn’t be that hard.

I was very excited about telomere extension as a pathway to extending lifespan. As a student, I dove in to ask those questions, but I found that almost no one in the world was ready to take them seriously.

Geron was already struggling. It had been taken public too early and remained a penny stock for the next 20 years, until they finally developed a telomerase inhibitor, which was the opposite of their original mission.

In graduate school, I proposed a plan to my boss to modify aging in the stem cells of a mouse. He looked at me and said, “This is an interesting idea and it might work, but never talk about aging biology in my office again.”

That was the general hostility toward the field 20 years ago; serious people simply didn’t do this. There were only a few labs, like David Sinclair’s, along with Brian Kennedy and Matt Kaeberlein, who were publishing work from Lenny Guarente’s lab in yeast that began to move the needle.

The field really kicked off after two big shifts. The first was the Interventions Testing Program, or ITP, led by Rich Miller at the University of Michigan. In 2009, they showed that rapamycin could extend the healthy lifespan of male and female mice. This was the first unequivocal proof that a mammal’s lifespan could be pharmacologically extended.

The quality of that data was so high that it forced people to take the field seriously. The second shift was the 2013 publication of the famous “Hallmarks of Aging” paper by Guido Kroemer and others.

Before 2013, conferences were dominated by scientists arguing over which single thing caused aging. People were in different camps, debating whether it was telomeres, mitochondria, free radicals, or sirtuins.

The “Hallmarks of Aging” paper took a “big tent” approach. It acknowledged that everyone had discovered important things that change as the body ages and that they were all relevant.

This framework brought the geroscience academic community together in a cohesive way for the first time. That foundation allowed people like me to come in and say, “Now that we’ve identified what changes with the biology of aging, let’s figure out ways to modify those processes and develop drugs.”

Daniel 00:09:36

Now that there is a consensus in the geroscience community about what aging is, it lays the groundwork for the Geroscience Hypothesis.

This is the idea that by intervening in aging biology, we can treat multiple age-related diseases. Where did that hypothesis come from?

9:39 Origins of the geroscience hypothesis

Dr. James Peyer 00:10:04

The person who deserves the most credit for this is Felipe Sierra, who was the head of the National Institute on Aging for years. He is credited with coining the term geroscience and articulating the geroscience hypothesis along with Ron Kohanski.

They established the Geroscience Working Group in the early 2010s, which is where this emerged. That conception of geroscience as a unified goal of aging biology was a powerful new idea.

The goal was to create a medicine that targets aging biology to slow down or prevent all diseases of aging at once. This was enabled by discoveries around rapamycin and the generalized slowing of aging-related diseases through mTOR inhibition in mice. Felipe gets a lot of credit for that.

Daniel 00:11:14

Looking at where we are now, how validated is the geroscience hypothesis? Do we need a drug on the market that targets aging biology and extends lifespan to truly validate it?

Dr. James Peyer 00:11:33

This is a tricky question. If we set humans aside for a moment and look at mice, flies, or worms, the answer is unequivocal. The geroscience hypothesis has been demonstrably proven correct beyond a shadow of a doubt in all those model organisms.

You can change individual genes or use drugs. There are dozens or hundreds of different approaches that prove the hypothesis in animal models. When looking at humans, a major argument is that there is more in common between a human and a mouse than between a mouse and a C. elegans worm.

If the same intervention works in a worm, a fly, and a mouse, the jump to humans should be logical. That is our solid evidence base. However, regarding evidence in humans, it is harder to draw a clear line.

I have been passionate for years about identifying and defining a gerotherapeutic—a medicine with the potential to test the geroscience hypothesis. We do not have a medicine that has demonstrated this to my satisfaction in humans yet.

But we do not necessarily have to run 30-year prospective trials watching people age and die. There is a near-term opportunity to test drugs in healthy elderly people and show disease prevention.

We have already shown that specific conditions like heart disease, strokes, or even COVID-19 infections can be prevented. We have the ability to run preventative clinical trials. That is the lens through which we will build the evidence base for the first gerotherapeutic. We just haven’t reached that point yet.

Daniel 00:14:10

Was it always your perspective that we could arrive at gerotherapeutics through preventative medicine? Was that clear to you when you started this journey 20 years ago?

14:15 How to frame gerotherapeutics as the greatest preventative drugs ever

Dr. James Peyer 00:14:27

Definitely not 20 years ago. This viewpoint was built out of scar tissue, not divine revelation. The most important realization is that anything we create as a community must be a drug.

If it is a drug, it must progress through Phase 1, 2, and 3 clinical trials at the FDA. The FDA really only does two things: it approves medicines to treat specific diseases or it approves medicines to prevent them.

There are far more treatments than preventatives. Ten years ago, I was looking for diseases we could treat using drugs that target the biology of aging. In fact, this is still what Cambrian BioPharma and most longevity biotech companies are doing.

They take something that targets aging biology and ask what diseases it can treat, because treating a disease is much easier to trial than preventing one. This framing of geroscience as prevention came from discussions with the community, the FDA, and people in pharma.

Most people coming into the geroscience space haven’t framed it this way. But if you imagine a gerotherapeutic on the market, what is that drug? It is likely something where you go to your doctor to measure biomarkers.

Eventually, every person reaches a point where those markers decline. For example, everyone’s metabolism declines as they age. Metabolism changes ahead of these various diseases. At some point, a doctor might tell you that your metabolism has declined to a point where it is hurting your body’s function.

You are then at high risk for diabetes, heart disease, or obesity. They would put you on a drug that resets your metabolism to youthful levels. You would take that pill every day, just like a statin for cholesterol, or receive an annual injection like a flu vaccine.

When you think about how a person would actually use this drug, it becomes clear that it is a preventative medicine. You are trying to prevent the things that happen with aging.

The biology underneath may be different than a vaccine, but from the perspective of a pharma company or the FDA, it is crystal clear. This creates a bridge to take a concept they don’t understand and frame it in terms they do. It makes it less scary and more predictable, making it much more likely to become a reality.

Daniel 00:18:09

This isn’t just theoretical. I believe the drug you are using as an example is ATX304. Can you tell us about that molecule?

18:11 The zone 2 in a pill drug in clinical trials

Dr. James Peyer 00:18:18

It’s always challenging to prognosticate about a molecule before it has finished clinical trials, but there is a clear reason we’re excited about ATX304. This drug is the first direct activator of AMPK to ever make it into human trials.

We haven’t announced the results of the first big clinical trials yet. Those will be shared in June of this year at the American Diabetes Association meeting. However, in animals, we’ve shown that this drug can reprogram or rewire an animal’s metabolism from a decrepit state back to a younger state.

The two biggest things that change in our metabolism as we get older are our ability to activate AMPK and our ability to activate our mitochondria. AMPK is the central sensor of ATP in our body, and its activation declines with age. Along with that, the cellular factories that make ATP, the mitochondria, also decline.

This drug targets both of those pathways and resets them to youthful levels. This produces a host of benefits, including reversing obesity, fatty liver disease, and type 2 diabetes. It can even replace insulin in type 1 diabetes by generating a massive demand for glucose.

While those effects are seen in disease states, it also works in healthy animals. If you give this drug to healthy older animals for a few months, their ability to run on a treadmill doubles. They can run twice as far with less muscle exhaustion.

That is an exciting lens through which to view this core metabolic longevity pathway. We’ve been able to target this safely for long periods to treat and prevent various diseases while restoring metabolic function to a younger state. It’s a very cool drug.

Daniel 00:20:32

It sounds amazing.

Dr. James Peyer 00:20:34

It’s still in the animal stage, so we have to wait and see how the human trials play out.

Daniel 00:20:40

Absolutely. This is a great example for us to reflect on to understand what it means to develop gerotherapeutics and what the future might look like.

Could you compare ATX304 to a GLP-1? I’m hearing a lot of similar effects, but they act on different pathways. I have heard that we probably shouldn’t think of GLP-1s as longevity drugs.

20:41 How to distinguish “real longevity drugs” from other drugs like GLP1s

Dr. James Peyer 00:21:11

There’s a really interesting conversation going on in the community right now about exactly that point: should we be thinking of GLP-1s as the first gerotherapeutics?

I did a debate for the Longevity Biotech Association with Nir Barzilai on this. He argued that we should think of them as longevity drugs, and I took the opposite position. It’s worth summarizing the two sides of the argument.

The argument for saying GLP-1s are the first longevity drugs is that they fulfill some components of the geroscience hypothesis. They aren’t just treatments for obesity; they treat type 2 diabetes, reduce heart attacks, and can be used for liver fat and kidney diseases. All of these cardiometabolic diseases seem to be changed by GLP-1s.

A sub-argument is that obesity is already a preventative medicine category at the FDA. Because obesity leads to so many different diseases, targeting it is a gateway to targeting all of these other conditions.

The argument against GLP-1s as longevity drugs, which I find more compelling, is that GLP-1 levels do not change as we age. It’s not that we lack GLP-1 as we get older or that we eat more because our appetite isn’t suppressed. In fact, people tend to eat less as they age, and their GLP signaling goes down.

The issue is that, as a culture, we overeat. Chronic overeating leads to obesity, which significantly increases the risk for all these diseases. That is a result of evolutionary programming.

I look at GLP-1s as amazing drugs for the chronic overeating that plagues modern society, but that is separate from targeting core aging biology. Reducing appetite is a great path for treating obesity.

However, ATX304 and the path Cambrian is pursuing is completely different. We are looking at the core part of metabolism that declines with aging. We want to identify what metabolically separates a healthy 20-year-old from a healthy 70-year-old and then reverse those differences. Treating something that declines with aging to set it back to a healthy state is what makes a drug a gerotherapeutic in my mind.

25:10 Why it matters that we draw this distinction

Daniel 00:25:10

I’m interested to dig into why it matters how we identify what’s a gerotherapeutic and what’s not.

The value of the drug you’re describing seems clear. A compound that works for healthy people to make them more metabolically healthy or reverse age-related decline represents a massive market. Those benefits go far beyond treating people who are already obese.

Why do we need a niche longevity community making these arguments when it seems like a logical path for a very successful drug?

Dr. James Peyer 00:25:53

It’s a fun question. Let’s take a step back and acknowledge that the obesity space opened up by GLP-1s is the biggest and best pharmaceutical market of all time. It’s estimated to be worth $100 to $150 billion a year in revenue, which is roughly the size of Google Search. This is an incredible business.

As we create these first obesity drugs, which are essentially preventative medicines, there’s a tendency to ask why we need more. The obesity pathway already exists, and drugs are getting approved. People are demanding new versions with better lean body mass maintenance, lower rates of nausea, or even greater weight loss than current GLP-1s.

Because of this, there’s a natural resistance to going beyond that established pathway. However, the geroscience community, including myself, argues that this isn’t enough. GLP-1s aren’t going to dramatically improve health for people who are already at a healthy weight.

If you take a 70-year-old who is neither overweight nor obese and put them on a GLP-1, they are more likely to become less healthy than more healthy. The grand challenge of geroscience is taking a quantum leap to do something unprecedented: helping a healthy 70-year-old acknowledge that their body has declined over the last 40 years and improving their life, even if they seem “okay” for their age.

That is what differentiates this field. It leads to increases in healthspan and lifespan that transcend drugs designed only for specific conditions like obesity. That requires another quantum leap. Keeping our eye on that prize is a big deal for me, and it distinguishes the geroscience community from the broader pharma world chasing the obesity market.

28:56 Why it requires a quantum leap for pharma to tackle the massive longevity opportunity

Daniel 00:28:57

As someone who hasn’t worked in pharma and doesn’t understand all the nuances of drug approval, this feels like such a massive opportunity. I hear you saying that it requires a quantum leap because it’s so different from our current approach.

What is the disconnect? Where do pharma companies disagree or lose faith?

Dr. James Peyer 00:29:24

Endpoints. It’s all about endpoints and insurance company reimbursement. This is where the technical and nuanced discussions meet reality.

Ten to fifteen years ago, there was a major push to reframe obesity. Instead of viewing it as a lack of discipline, it was categorized as a medical condition. While that movement started with a small group of advocates, it was essential for the world we live in today.

Reframing obesity as a disease—whether or not we should do that with aging is a separate discussion—allowed the FDA to establish an approvable endpoint. If a drug safely reduced weight by 5% or more in one year, it could be approved for “weight maintenance,” as they call it.

That pathway had been sitting unused for decades. Once obesity was focused on as a disease, pharma was convinced they could create drugs for it. GLP-1s jumped through that hoop.

The next challenge was insurance companies. There is still a fight over how long and for whom insurance will pay for these drugs, but the demand is clearly there. The 5% weight loss threshold gave the industry a safe target.

When the field jumped from diabetes to obesity, it felt relatively secure. If you try to jump to the prevention of aging, the question becomes: prevention of what? There is no baked-in endpoint for the FDA to point to. Without a clear clinical trial target, it feels like the Wild West.

Daniel 00:32:23

It seems like there are two paths for people in your field to push this forward. One is working within the current paradigm by finding an existing indication and expanding from there.

The other approach is trying to change endpoints through government policy, which would then set a target for the entire industry to chase.

32:24 What it takes to get longevity drugs approved

Dr. James Peyer 00:32:50

I agree that there’s been a general dynamic between changing the system versus using the system. My take is that real progress only happens if both are done at the same time within the same organization. This is informed by a look at history.

Let’s look at two categories of preventative medicines approved in the last 75 years: cholesterol-lowering drugs and HIV drugs. It wasn’t that the FDA requested new HIV drugs and asked for data on lowering viral loads while the pandemic was raging.

Instead, researchers had to create the first drugs, like AZT. AZT was the first HIV drug, a reverse transcriptase inhibitor. It was approved based on showing improvements in survival for people with HIV.

That set off a cascade: how can we improve the next one faster? How can we improve the dose or treatment regimen to make it work better? This created a positive feedback cycle that led to lowering viral loads as the way of creating better anti-HIV cocktails.

But you needed a drug first. No one created the endpoint and then asked for a drug. The same applies to cholesterol. We had the statins and the push that cholesterol was important, but cholesterol wasn’t an approvable endpoint on its own until statins came along.

Researchers tested long-term survival and the risk of heart attacks and strokes associated with high cholesterol, and then lowered them using the statin drug. This allowed the FDA to see that lowering heart attacks and strokes worked.

Once the drug was approved, the next generation of statins could be even better. Now, you can get a drug approved just by lowering cholesterol. These lenses are very important.

35:32 The importance of establishing endpoints for aging (to get the flywheel going)

Daniel 00:35:32

Why do we need an initial drug approved in order to establish an endpoint? Is it because no one is going to pay to run a large trial to establish endpoints without a marketable product at the end?

Dr. James Peyer 00:35:43

For an endpoint to function for a regulatory agency, it needs to fulfill two jobs. First, it needs to fairly accurately predict the risk of some downstream event. In the case of cholesterol, it needed to predict the risk of heart attacks or strokes. In the case of HIV, it needs to predict the risk of AIDS or death.

Second, the endpoint has to move in response to an intervention. You can’t really prove that second criteria until you have an intervention showing that you’ve moved it.

Without an intervention, it’s hard to show that moving the marker actually changes the risk profile. In the longevity space, there are many possible biomarkers that fit some, but not all, of these criteria.

A famous one is GDF-15. GDF-15 levels are highly correlated with mortality risk. If your levels move in a certain direction, you are either in trouble or relatively safe.

However, the things we know extend healthy lifespan—exercise, a good diet, weight loss, and good sleep—don’t seem to move GDF-15 that much. It’s a general indicator, but it’s a poor surrogate endpoint for the FDA to use for a clinical trial of a preventative medicine.

It’s important to think about those nitty-gritty details when considering how to get a drug approved.

Daniel 00:37:54

This is helpful to understand. You can’t know that something is a surrogate marker unless you move it and establish causality. That’s the value of the first gerotherapeutic; it allows us to establish that causality.

Dr. James Peyer 00:38:13

Exactly. Have you had Andrew Brack from ARPA-H on the podcast yet?

Daniel 00:38:19

Not yet, but we would love to.

Dr. James Peyer 00:38:20

Okay.

Daniel 00:38:21

We’ll want to bring him on at some point to talk about ARPA-H and PROSPER.

Dr. James Peyer 00:38:25

The smartest implementation I’ve seen so far is what Andrew and the PROSPER teams have been doing. They have a three-tiered approach to create the dataset needed to meet the criteria for a new biomarker.

They are measuring several markers in healthy older people and doing an initial sensitivity analysis by intervening with exercise. We know exercise should help, so they are seeing what moves in response.

Then they will test those markers in three clinical studies using an SGLT2 inhibitor, an mTOR inhibitor, and a GLP-1 to see if the same things move. Those are all imperfect but possible gerotherapeutic drugs with shared mechanisms.

From there, they will use a composite endpoint based on what moves successfully in those pilot studies with novel gerotherapeutics designed to target key aspects of aging biology.

Cambrian BioPharma is a participant in that third part. We’ve created safe mTOR complex I inhibitors—essentially rapamycin without the downsides.

This will be funded as part of the PROSPER program to test this biomarker endpoint for improving the intrinsic capacity of healthy elderly people. This is the closest project that exists to establishing biomarkers for aging that could be the foundation for future preventative medicine approvals.

Daniel 00:40:23

There’s a clear vision of both establishing endpoints for aging to spur drug development and getting the first gerotherapeutics out there to start that flywheel. How do you feel about our odds currently? Do we have enough shots on goal?

40:24 Odds of getting the flywheel towards longevity escape velocity going

Dr. James Peyer 00:40:49

It’s hard to talk about this in an unbiased way because Cambrian is now six years old. We’ve had the incredible opportunity to survey the field of what was bubbling up from academic groups all over the world and have harvested what we think are several extremely good shots on goal.

These include the AMPK and mitochondrial activator ATX304, as well as an mTORC1 selective activator, which is a safer version of rapamycin. As you look across the rest of the global pipeline in geroscience, there is a lot out there.

The key is making sure these programs are well-funded, not just through their initial stepping stone indications—whether it’s obesity, diabetes, or inflammation—but also supported to move into preventative medicine trials.

We have a dozen really good shots on goal across the world right now. Unlike the generation from five years ago, the biology, chemistry, and quality of the drugs in these clinical trials are miles better.

This gives me a lot of confidence that current clinical-stage companies in the geroscience space are going to unlock the flywheel effect.

42:48 Most exciting companies in clinical stage today

Daniel 00:42:48

Which companies are you most excited about besides your own?

Dr. James Peyer 00:42:52

A few deserve special mention. One is BioAge, which is taking an NLRP3 inhibitor forward in cardiometabolic disease to target chronic inflammation.

This is their second shot on goal. They had an Apelin program that had a setback, although I think they’ll be back in the clinic with Apelin soon. NLRP3 is a great inflammaging marker.

I would also point to Life Biosciences, co-founded by David Sinclair and led by Gerry McLaughlin. They have taken a big shot on goal with epigenetic reprogramming, using three Yamanaka factors to work on optic nerve regeneration. They recently got an IND after showing positive results in primates and are going into humans this year.

Another one I’m personally excited about is the work of Céline Halligouët and Loyal. They have shown successive metabolic reprogramming and healthspan extension in dogs.

They are likely to get the first preventative medicine-like approval for metabolic healthspan extension in canines sometime in the next year. I’m not a spokesman for Loyal, so I can’t represent their timeline, but it looks very promising.

The last one I will mention is Insilico, led by Alex Zhavoronkov. He has been a huge advocate of aging biology for a long time. Insilico has many different targets, and I know Alex thinks about advancing things beyond just the first indication.

Daniel 00:45:13

We’ve had Céline Halligouët and Life Biosciences COO Michael Ringel on the podcast previously for anyone who wants to go back and listen. Those are all amazing companies.

If we look earlier in the pipeline at basic aging biology research and the identification of new targets, how many are we identifying? Even if current clinical trials are successful, I hope we have many more things coming down the pipeline.

45:49 The secret sauce of the geroscience field (and what big pharma will need to learn)

Dr. James Peyer 00:45:48

I have a pessimistic view of the new targets approach to geroscience. While it’s great academic work and big discoveries will come from it, as a drug developer, I think it takes us in the wrong direction.

The targets we want to tweak with gerotherapeutic interventions are not new. mTOR was not discovered by aging biologists. Neither was AMPK, the Yamanaka factors, or NLRP3.

The brilliance of the geroscience community is not finding new targets that no one appreciated until now. The genius has been finding ways to say a pathway used to work well, and as we aged, it started working poorly.

Let’s figure out not how to hit it with a hammer, but how to use a dimmer switch to increase or decrease the brightness and bring it back to a functional state. This dimmer switch approach to biology is unique in drug development.

It allows us to target pathways like mitochondrial function, AMPK, and mTOR that are vital to our biology. When they go from young to old, they might change by 50% rather than turning completely off or on.

Adjusting that 50% is the genius of what the scientists in our community have discovered. My prediction is that the most important drugs will target these well-known pathways.

Pharma never figured out how to drug them correctly because they weren’t using the lens of natural aging biology. They were using the pharma lens, which asks only if we can turn something entirely off or on.

Daniel 00:48:22

That’s really interesting and not something I’ve come across before. Can you give me some examples of what it means to turn the dimmer up or down versus turning a target off or on?

Dr. James Peyer 00:48:31

Let’s use mitochondrial function as a simple example. A person’s ability to respire—their resting metabolic rate—declines every decade starting in their early 20s. Mitochondrial function declines along with it. As you get older, you use fewer calories, and your fat becomes less metabolically active.

This decline is not absolute. If a 25-year-old has a 2,000-calorie-per-day resting metabolic rate, it doesn’t drop to zero by age 75. It might decline to 1,200 or 1,000. To correct that, you only want to increase that resting metabolic rate by about 20% or 30% to return it to a youthful level. It’s a dimmer switch, not a hammer.

This applies in the other direction as well. mTOR is the best example. Tissues in young people have low levels of mTOR complex 1 activation. As they age, you see a lot more activation, but the difference is only about 50%.

If you gave giant doses of rapamycin and knocked out all mTORC1 all the time, cells couldn’t grow because you need mTORC1 to grow and divide. You just need a dimmer switch to turn it down. These core pathways at the center of the biology of aging don’t go from off to on; they move from 3 to 7 or 7 to 3.

Daniel 00:51:14

This represents the challenge pharma has in managing chronic illnesses, which are largely diseases of aging. It is fundamentally different from treating infectious diseases, where you are trying to kill an organism, or cancer, where you are trying to kill a tumor. It isn’t a simple matter of replacing a missing protein. Chronic diseases of aging require this dimmer switch approach.

Dr. James Peyer 00:51:44

Exactly. You’ve stumbled across what makes the geroscience approach so special. Modern pharma was built on the success of penicillin and vaccines. In the 1970s, we even thought cancer was caused by viruses.

The mentality of turning things off—the bacteria, the virus, the cancer cell—penetrates every part of pharma. We saw this with cholesterol; the idea was to turn off cholesterol to stop heart attacks. It turned out to be much more complex than that.

The geroscience viewpoint is not about off versus on. It’s about the slow, steady degradation of a pathway that causes a complex decline in a cellular system. We have to change the dial to run it back.

That level of complex systems thinking represents the most fundamental change in how we think about treating chronic diseases in 100 years. As a drug developer, I find this the most compelling and differentiating aspect of our field.

Daniel 00:53:34

This is fascinating. You’ve pointed out the incentives that prevent pharma companies from developing these drugs, but there is also a deeper point about how pharma companies are structured, trained, and equipped. They are hammers, and aging is not a nail. That is a massive issue to solve, and it requires a major change.

Dr. James Peyer 00:53:56

That’s a great way of putting it. They have powerful hammers for turning things on and off, but almost all the pathways that have come out of the geroscience community are not suitable for that hammer.

People often ask why Cambrian doesn’t have many antibody programs. Antibodies are a popular approach right now because they are easy to scale and attract a lot of money, but they are generally designed to turn pathways off.

I haven’t found many targets in aging biology that I would want to block entirely with an antibody. That is why you don’t see many antibodies in the geroscience community.

Daniel 00:55:00

I want to press on one area of this. I imagine that, ultimately, there is something binary happening somewhere in the cell that you could target. If a process is ramping down or up too much, there is probably a specific target responsible.

Perhaps we just don’t know what those targets are yet. In the meantime, we need to tackle what we currently understand to get those first drugs out there.

Dr. James Peyer 00:55:31

The closest things that exist to those binary changes are almost certainly genetic and epigenetic changes that happen within somatic cells. If my ATGCs are mutated in a cell, that is a binary, permanent change. It is a piece of damage within the cell that could cause some downstream problem.

There is widespread disagreement even within the geroscience community about whether those DNA changes even contribute to normal human organismal aging outside of the context of cancer. Even if they did, how would you fix it?

I remember discussions 25 years ago about perfect nanomachines that knew your DNA sequence, could get into every cell, and spell-check all four billion DNA letters to replace any that went off. That is just an unrealistic viewpoint right now.

Instead, in 20 years of chasing these pathways, we have found downstream events like mTOR overactivation, mitochondrial repression, and rising chronic inflammation. We know the key nodes that integrate these signals. Wherever those binary damage signals happen, they are integrated by a sensor in the cell.

That sensor changes 10%, 30%, or 50% up or down to change the behavior of a cell. This is where geroscience has produced miraculous data, like lifespan and healthspan extension and the reversal of various diseases of aging. That is where the data really lives.

I am open to thinking about long-term approaches. LifeBio, with their epigenetic reprogramming approach using Yamanaka factors, is the closest thing we have to these early modifications. However, it is still a blunt instrument.

Yamanaka factors change an awful lot of things at once. They are not just removing specific methylation marks; they are resetting the cell to a much earlier state. Biology is orders of magnitude more complicated than any other technological system humans interact with right now.

Daniel 00:58:37

This is super interesting. We went down a long tangent there, but that was a really interesting point that I had never fully appreciated before.

58:39 What makes this interview unique

Dr. James Peyer 00:58:43

These podcasts are fun for me because they are an opportunity to preach to my own choir. My normal job involves preaching to pharma and getting them excited about geroscience and aging biology.

With you, I get to have a different conversation about why our community is special. Oftentimes, people who are buried in this day-to-day do not even understand it.

We all intuitively feel there is something special about the fact that we can manipulate the biology of aging. Articulating what actually makes it different from other existing fields is a valuable thing for our community.

Daniel 00:59:49

You have articulated a particular capability of the geroscience community: a way of looking at biology and designing interventions that can touch the dimmer switch in a process.

When I look at the work happening in the longevity community, you feel like you belong to a different generation. You have been developing small molecules, which is standard pharma in a certain way.

Meanwhile, talking to partial epigenetic reprogramming people feels much more next-gen, and cryostasis or replacement research is even further out there. How do you think about the landscape and where the focus is in the field today?

Dr. James Peyer 01:00:44

I am fully supportive of research and clinical efforts across the spectrum. When we started Cambrian, we took a mechanism-agnostic approach.

In fact, one of our programs is a recombinant protein. It is a replacement for a gene that gets turned off when we move from fetal development to adulthood. It allows fetal hearts to regenerate and can be added back into adult pig hearts to allow them to regenerate.

We also had a women’s health company, Aviva, which we sold to Granada last year. That was also a recombinant protein molecule that gets turned off as part of the menstrual cycle and could, in theory, prevent menopause from happening.

However, when I study the historical precedence of preventative medicines for chronic diseases of aging, it is hard to find one that is not a small molecule. I ask what the experience is for a person taking a drug to treat a declining pathway of aging.

Historically, it is one of two things. Either they have a pill they take every day to improve the functioning of a pathway, like statins, antihypertensives, or HIV medications. Or they get a very cheap, one-and-done injection given widely to a huge population to condition the immune system, which is the vaccination parallel.

I do not think the vaccination parallel is helpful for most geroscience interventions. Small molecule daily dosing can be very cheap to manufacture and deliver to a wide number of people without interrupting their lives.

The first of these medicines are going to be cheap, extremely safe small molecules. Past that, the “weird risky stuff” can be part of the flywheel. The more work we do now to figure out those next-generation technologies, the faster they will arrive.

If I were watching this space, I would look for cheap-to-manufacture small molecules that can get approval through existing pathways and then be expanded into preventative uses. That is my ABCD for the geroscience space.

Daniel 01:04:14

You previously described the hope for this space: producing preventative medicines that everyone takes. It should hopefully be the next GLP-1, but ten times better. Everyone would take it, not just people who are overweight.

Dr. James Peyer 01:04:37

That’s exactly it. GLP-1s are a great precedent because the first gerotherapeutic should do for everyone’s preventative medicine what the GLP-1s have done for people with overweight or obesity.

If you are not on that medicine, people will ask why. It is going to make you healthier, make you live longer, and make your life better. We can’t say that for GLP-1s for people without overweight or obesity today, but it will be a characteristic of the first gerotherapeutic.

1:05:23 The reason grifters have rushed into the longevity space (and how to tell what’s real)

Daniel 01:05:23

I mentioned at the beginning that there is a lot of mainstream interest in longevity. We have huge health influencers in the space.

I meet people all the time who say they are interested in longevity, but when I start talking about pharma, they don’t understand. That’s not what longevity means to them. What is your take on why the culture got so excited about longevity, and what effect does that have on the field?

Dr. James Peyer 01:05:50

All great snake oil and great lies are built on a seed of truth. Once something sounds plausible, it creates a place for a grifter to come in and ask how they can make money on it.

In my view, longevity has exploded as a wellness category with a snake oil problem. This involves peptides, clinics offering cash-pay services, and a giant group of social media influencers. They typically sell things not approved by the FDA.

The reason there is so much excitement is that we are able to modify aging biology in animals. There are drugs in clinical trials that look really promising for humans. That’s the core of it.

However, consumers are not patient. Grifters step in and claim to understand this complex biology better than you do. They sell products today that didn’t need to go through clinical trials.

I really think the reason the space is exploding is because of the success of mouse geneticists from ten years ago. Once there are bona fide successes in the space—both within the wellness community and the broader pharmaceutical community—the grifters will fall away.

We are seeing this now with weight loss fads. Grifter diet pills have fallen away because the GLP-1s actually work. Now the grifters are just trying to make fake GLP-1s to undercut the prices of companies like Novo and Lilly.

An example of how we’re trying to put our foot on the scale is a cool side project that spun off from Cambrian BioPharma called Rapalogix. We took one of our novel mTOR complex I inhibitors—a safe version of rapamycin—and brought in Rahul Mehta.

Rahul is an expert in pharmaceutical-grade cosmetics who previously worked at Allergan and AbbVie to create the brand SkinMedica. He was able to take this mTOR complex I inhibitor and put it into a cosmetic product called ReQ.

We ran clinical trials showing it restored skin collagen and elasticity. It has been winning cosmetic competitions around the world since it launched six months ago because it actually has a novel ingredient that affects aging biology.

That is my answer to addressing the wellness space: go create something new that works and put it there. Not many things fit into that category. Most things have to go through long Phase 1, Phase 2, and Phase 3 trials, which is most of what we do.

However, it is interesting to tear down the house of cards built on the poor data packages the rest of the wellness community stands on.

Daniel 01:10:17

I started laughing as you explained this because I had a joke ready: if I eat that cream, will I live longer?

Dr. James Peyer 01:10:26

The answer is no. The amount of mTOR inhibition is much too low.

Daniel 01:10:33

If I eat a lot of the cream, maybe? But then there would be other issues.

Dr. James Peyer 01:10:38

This is where my legal counsel would throw a fit. But sure, ten grams of it a day and you’re fine.

1:10:47 The difference between being healthy and actually building towads longevity

Daniel 01:10:48

On the non-grifter side of the world, there are decently well-informed health recommendations, like doing Zone 2 cardio and keeping your cholesterol low. That is often called longevity medicine.

Dr. James Peyer 01:11:09

That is exactly the distinction I would create.

Daniel 01:11:13

That is the distinction between the grifters and the non-grifters. However, I worry that the space can get problematic when they tie those habits to aging clocks.

They claim you are slowing your aging because of better habits. That can make people think they are genuinely slowing the process, whereas those habits are likely just causing you to age at a normal pace.

We don’t have anything available today that is going to extend a human past the expected lifetime. I’m curious how you think about that.

Dr. James Peyer 01:11:48

I completely agree with you. Movement toward better diet, exercise, zone 2 training, and better sleep are things we know work, but people aren’t doing enough of them today. The trend toward doing more of that lives within the wellness longevity brand, which is great.

In fact, it is laying the foundation for people to think about their health. If someone is already doing all of this to stay healthy, they will naturally want something that resets their biology once it is shown to be safe and effective at targeting a key aging pathway. I love the enthusiasm people have for this.

Regarding aging clocks specifically, you have to distinguish between their practical and theoretical states. Practically, they are toys right now. None of the aging clocks have been validated for the biomarker endpoints we talked about earlier.

They don’t predict outcomes, and moving them hasn’t been shown to mean anything in any clinical trial. They just provide a number. Even as a toy, it can help drive behavior. People like to see a number moving when they go to the gym or eat better.

While they aren’t always predictable, as they improve, they can provide a gamification of good habits. There are also groups trying to build datasets to validate these clocks over long-term observations.

This is an interesting idea because there is powerful evidence that these clocks correlate with multimorbidity risk and aging. However, putting that data together in a way that tells us something concrete is a challenge no one has surmounted yet, despite how compelling the initial data is.

1:14:30 How to contribute to the field

Daniel 01:14:30

Are there any other topics we haven’t touched on or anything you wanted to add?

Dr. James Peyer 01:14:36

We’ve covered a lot in the last hour. I don’t have any other major talking points, except for one point of clarification. Throughout this conversation, we have been using “geroscience,” “gerotherapeutics,” and “longevity” somewhat interchangeably.

Regarding this last part of our conversation, it is important for members of this field to distinguish between the geroscience hypothesis and the “longevity wellness” category. Geroscience involves the drug development efforts aimed at slowing aging pathways for FDA approval.

The longevity wellness category has essentially taken over for the discredited “anti-aging” label from 10 or 15 years ago in cosmetics. Instead of anti-aging creams, we now have longevity creams, longevity ice baths, red light therapy, and longevity peptides.

We are starting to distinguish ourselves and pull away from those wellness terms. If you are interested in the biology of aging and what academics are researching at the cutting edge, the keyword you should follow is geroscience.

As longevity becomes a massive category, there will be a separation as players in this field build a tent under that name. For your audience, who cares deeply about the science, geroscience should become part of your vocabulary if it isn’t already.

Daniel 01:16:53

And Free Radicals is now rebranding to the Geroscience Podcast.

Dr. James Peyer 01:16:57

Exactly.

Daniel 01:16:59

Are there any last things you’d like to leave the audience with? We have many students who want to work in this space. Do you have any advice or places you recommend they explore?

Dr. James Peyer 01:17:13

Getting started in the aging biology space right now is unique because it’s not a consolidated industry like pharma. It is a collection of companies within the biotech space, most of which are small.

For college students interested in this space, go work in a lab studying aging. Compared to 20 years ago, geroscience is now the fastest-growing specialization in academia worldwide. Everyone wants to study aging biology.

Professors who previously focused on other areas are moving into this space because they are excited about it. Go work in a lab and get exposure to the real science if you haven’t already.

Regarding the job market, drug development is incredibly difficult. Working at a longevity company as your first gig can be a great opportunity, but the sector is still small.

Most people who play important roles in these companies bring skills from other aspects of drug development. They may have worked on clinical trials for cancer or obesity at a pharma company or another biotech.

Getting into the drug development world and understanding how to do it right is key. You can then bring all of that knowledge back to the geroscience world. Whether it is your first, second, or third job, that is the lens I would take.

The more people we have in this diaspora of aging biology enthusiasts within the broader drug development and academic worlds, the stronger the field becomes.

Daniel 01:19:28

That makes me think of one last question. Outside of the aging biology and geroscience space, who are our biggest allies? Which adjacent spaces should people look toward for collaborators?

Dr. James Peyer 01:19:41

I am extremely encouraged that big pharma companies are trying to figure this space out. Eli Lilly, for example, has announced a therapeutic area focused on healthspan.

Big pharma companies are beginning to engage, even if they do not fully understand it yet. As we discussed, aging is a hammer that does not quite have a nail to hit, but there is a real opportunity to work with these organizations to make an impact.

There is also the public sector realm. Advocating for NIH funding and engaging with Congress and the FDA are valuable actions that drive not just geroscience, but every field rooted in fundamental science.

As geroscience becomes a larger part of the scientific tapestry in medicine, we must determine how to leverage these scientists and their discoveries. When I consider our allies, I think of big pharma and the government entities that fund research.

The third group is insurance companies, both in the US and globally. Many of these overlap with governments that bear the cost of aging and its associated diseases. They are highly incentivized to keep people healthier and to find ways to slow or prevent those conditions.

These are the allies I have spent the last 10 years trying to cultivate. Many have joined our side and are excited about this mission. I see momentum growing in these three sectors, which are all vital for making new medicines a reality.

Daniel 01:22:10

Amazing. James Peyer, thank you for joining us on the Free Radicals podcast.

Dr. James Peyer 01:22:15

Daniel, it was fun to be here. Thanks for the wide-ranging and very technical, but hopefully helpful conversation.

Daniel 01:22:22

I think it was super helpful. Thank you, James.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

2:35 Why did Nathan dedicate his life to defeating aging & death? (And the absurdity of asking that question)
8:59 Why has Daniel been on the fence, despite being longevity-pilled
14:09 The agency and bravery of founders to tackle the hardest problems (like aging)
18:46 Many ways to contribute to longevity besides entrepreneurship
19:41 Is longevity underinvested into?
24:48 What does it mean to treat aging itself as the issue (vs other diseases)
32:23 Do we need more scientific arguments, or philosophical arguments?
39:50 Nathan’s experience longevity-pilling influential people
47:03 Billionaire paradox
53:18 The challenge in going from viewing aging as inevitable to malleable
57:29 What needs to change about our culture to save us from aging & death
1:16:55 Should we be intolerant of pro-death & pro-aging views?
1:22:59 Join us at Vitalist Bay!

Transcript

2:35 Why did Nathan dedicate his life to defeating aging & death? (And the absurdity of asking that question)

Daniel 00:02:35

Nathan Cheng, welcome to the Free Radicals podcast.

Nathan Cheng 00:02:38

How’s it going?

Daniel 00:02:39

Doing well. We’re excited to have you on. Nathan, you were a huge part in getting us to start this podcast. Eric Dai and I met at a longevity biotech fellowship retreat almost exactly a year ago. You played a huge role in helping me learn about this space and encouraged me to start this, so I have been very excited to bring you on for a fun conversation.

In this conversation, I want to help people understand what it really means to work on aging. How can people work on this problem? How does it differ from other things people could be doing? Why is it so important?

Before we dive into that, I would love to hear your personal story. How did you end up dedicating your life essentially to solving aging and death?

Nathan Cheng 00:03:40

It depends on how far back you want to go. If we take a step back and recognize how crazy of a question that is: how did you get interested in trying to solve the thing that is going to kill you and everybody that you love? It kills hundreds of thousands of people per day.

It seems like that should be the default thing people work on, but it isn’t. It wasn’t for me for most of my life, either. I grew up outside of Toronto and eventually started a PhD in physics at the University of Toronto. It was pure physics, not biophysics. I didn’t even take biology in high school.

I got maybe two years into my PhD and realized I was having a sudden existential crisis. I realized I’m actually going to die. This is not a theoretical thing. I think most people learn about this when they’re young and then spend the rest of their lives trying to put it at the back of their minds.

They try to deal with the fact that life is ephemeral and dedicate their lives to something meaningful. For me, that hit me like a ton of bricks. This is actually going to happen. It is not just me who is going to die; it is my parents and everybody I love. It is a bummer. What can I do about this?

At the time, I didn’t know anything about longevity, aging, or biology. I couldn’t deal with this incredible existential reckoning. I ended up dropping out of my PhD and doing a lot of random things for a couple of years, including backpacking around the world for a year.

Right before the pandemic, I decided I wanted to do something about this problem. I decided to learn biology and try to somehow contribute to the space. Around July 2020, I started a Substack newsletter called Longevity Market Cap.

For me, it was an exercise to learn more about the space and teach others about my journey in figuring out what’s going on in longevity. It started to gain a bit of a following. Eventually, by chance, Balaji Srinivasan discovered my Substack and we connected.

He connected me with Eric Torenberg, who was running a startup back then called On Deck. On Deck was building out founder communities in different verticals, and they were very interested in longevity. We teamed up and I built out the first Longevity Biotech Fellowship at On Deck.

The idea was to get people who were interested in building in longevity connected. Eventually, we spun that out as a nonprofit after a year and a half. That became the Longevity Biotech Fellowship, which is where you met.

At the same time, I started a fund called Healthspan Capital. We’ve been operational for about four years, investing exclusively in longevity biotech startups. We’ve made about 32 investments to date.

I also started the Vitalism Foundation with Adam Gries, who you’ve had on the podcast. That’s focused on the social and political influence network to essentially increase the amount of attention and resources towards trying to solve the problem of aging. That has been my trajectory.

8:59 Why has Daniel been on the fence, despite being longevity-pilled

Daniel 00:08:59

Something that’s been on my mind is a good framing for this conversation. As I mentioned, the work you were doing with LBF played a really important role in getting me to start this podcast. The podcast is doing something for the community, but could I be doing more?

Certainly, I could be doing a lot more for longevity. I would consider myself pretty longevity-pilled. I recognize that I’m going to die unless we solve this problem. Yet, I can’t yet bring myself to dedicate every waking hour to this mission.

I only dedicate a certain percentage of my time per week. Why is it that I can’t get myself there? One argument that goes around in my head is: do I need to do this? Maybe we’re going to solve it anyway because Big Pharma is going to go there.

That’s where the money is. The biggest market is preventative medicine for everybody, so they’re going to solve it anyway. Or, on the flip side, whether they’re going to solve it or not, how can one person make a difference? How do you respond to those counterarguments?

Nathan Cheng 00:10:28

Daniel, I’m glad you’re bringing up this conversation. I guess this is an intervention. Eric and I are going to try and get you back onto the straight and narrow path of fighting aging and death.

This is a great question because it is something we deal with at the Longevity Biotech Fellowship. For context, with LBF, we run cohort programs where we specifically try to find people who are mission-aligned and talented. These are people who have self-professed that they want to get involved and solve aging.

Despite that fact, not everybody ends up devoting their life to this mission or working full-time on it. Even if you ask them how long they want to live and they say indefinitely or as long as possible, they don’t always make the leap. There are many reasons why that happens.

You touched on some common themes. For people out there right now, there’s a sense that with AI or big pharma, someone else is going to figure this out. They feel optimistic that we’re already on track. Other people are extremely pessimistic. They feel it’s impossible, that it’s just a part of life.

Peter Thiel has written about this. He says if you’re an ultra-optimist or an ultra-pessimist, it’s basically the same thing. You’re going to do nothing. You have no agency.

For us, it’s no longer just about finding people for LBF who are mission-aligned or who want something. It’s very easy to say you want something. At the end of the day, the only thing that matters is what you are willing to do. What effort are you willing to sacrifice to actually make it happen?

It’s hard to say who those people are. Sometimes it’s people who are extremely intrinsically motivated. They have had a deep aversion to death and an existential crisis for as far back as they can remember, and it never left them.

Otherwise, you have to have a certain amount of risk-taking and courage because this is a very underdeveloped field. It’s not like AI where people are being paid tens of millions of dollars. Longevity biotech is still very new.

To really get into this, you need agency. This has to be extremely important to you at a visceral level, but you also have to have courage because success is not guaranteed. You might not be able to find a job since there aren’t that many companies.

In most cases, you have to have the agency to make your own opportunities in the space. Let’s drill down further, Daniel. Tell me, why aren’t you working on this full-time?

14:09 The agency and bravery of founders to tackle the hardest problems (like aging)

Daniel 00:14:18

One element likely comes down to risk-taking and courage, as you said. There is a piece of belief required that my role in it would matter.

Dedicating my life and career to longevity could be individually fulfilling and also save me from death, which would be great. But the scary scenario to imagine is making all these sacrifices—taking a huge pay cut and grinding—and then still aging and dying anyway.

That is a very sad future to imagine. Of course, that is what any entrepreneur has to imagine: the scenario where they fail. It’s not unique to this problem, but it is uniquely absurd.

If you don’t do anything about it, you will definitely age and die. That’s some of the thinking I go down.

Nathan Cheng 00:15:31

That’s pretty much spot on. Everybody knows the success stories like Elon Musk, entrepreneurs who took the risk and made it. But for every Elon, there are tons of people who tried and failed. You never hear about them, but they sacrificed everything and potentially ruined their lives.

However, aging is going to ruin your life too. At some point, it is the ultimate extinguishing of all your optionality and future possibility. You have to have an all-or-nothing mindset.

Daniel 00:16:09

Eric, let’s get your opinion. You’ve gone through the rollercoaster of becoming a biotech entrepreneur. I imagine you’ve dealt with some of these same concerns. How do you think about this issue?

Nathan Cheng 00:16:24

Hmm.

Eric 00:16:28

If you need encouragement to become an entrepreneur or to take massive risks, you should not be doing it. I take the opposite point of view from where I think you’re going, Daniel. I don’t think people should need more encouragement.

If you break society down into different subsets, there is a slice of people who are normal—let’s call them “normies.” Those normies constitute 60% to 80% of all people. Then you have 5% of people, probably less, who are extremely risk-taking.

While social changes might slightly increase the ratio of risk-takers, it is a stable equilibrium. There may be a small percentage more entrepreneurs at any given point, but it’s not a huge amount.

If you narrow that down further to people who are risk-taking and also deeply care about longevity—those willing to commit everything to this work—that’s a very small slice of the pie. If you need to be convinced to enter that slice, it’s not for you.

It is extremely painful and highly unlikely to work exactly the way you think it will. If you’re worried about what you’re giving up all the time, it’s just going to make it even more painful.

People who are crazy enough to do it are so obsessed with the problem and so excited by it that they literally can’t imagine doing anything else.

An individual can get to that point through a trigger event or a philosophical awakening, but it doesn’t usually come from external conviction. It comes from something ineffable that just happens to you.

Nathan Cheng 00:18:45

Not everybody has to be an entrepreneur to work full-time in the space. People can join companies. There are definitely more companies now than there were ten years ago, so it is possible to find a job, even if it is difficult.

Beyond that, people don’t necessarily have to work full-time on this. It depends on your relative skill sets. The Effective Altruism movement did a good job of funneling people toward certain causes without requiring them to quit their jobs and build wells in Africa to have an impact.

People can “earn to give” in their own way, or they can invest in the space. There are other ways to contribute besides getting a full-time job or starting a company in longevity.

18:46 Many ways to contribute to longevity besides entrepreneurship

19:41 Is longevity underinvested into?

Daniel 00:19:42

We live in a society with a division of labor. In theory, some people should work on this problem while others work on other problems. If everyone started researching aging biology, we would die much quicker of starvation.

There is an optimal allocation, which raises the question: how do you infer that we have a misallocation currently? There are people working on longevity, and billions of dollars are being invested into biotech and pharma broadly.

Why do you believe there is an issue? In what way is longevity underinvested?

Nathan Cheng 00:20:30

I could play devil’s advocate and say that markets are mostly rational or efficient, meaning we’re putting the right amount of resources in today given the prospects and current understanding. One could make that argument.

Another way to look at it is by examining the actual numbers compared to other important causes. If we restrict ourselves to the biomedical space, do you know how much money is spent globally on cancer research and development every year?

Daniel 00:21:20

We’ve probably heard that stat twenty times on this podcast. I just remember it’s a very big number.

Nathan Cheng 00:21:26

It depends on how you estimate it, but the figures are in the hundreds of billions of dollars—low hundreds of billions per year.

By comparison, do you know how much we spend on longevity R&D, aging biotech, or trying to therapeutically treat aging?

Eric 00:21:52

I don’t remember exactly, but I’m going to guess $5 billion.

Nathan Cheng 00:21:57

It depends on how you count it and which companies are included, but funding for aging research is in the low to mid single-digit billions per year. There is roughly a 100x order of magnitude difference compared to other fields.

This is despite the fact that the number one risk factor for cancer is actually age. Your risk of cancer goes up exponentially as you get older. Aging is by far the most dominant factor—more than your diet or whether you smoke.

If you look at the number of deaths, age-related diseases make up between 70% to 80% of deaths in most countries. Cancer is just a subset of those diseases. It is insane that there is this thing killing so many people, yet we allocate such a small amount of resources toward it.

Even without knowing the numbers, any person on the street would understand there is a mismatch. Most people know about cancer research and charities; they are familiar with that as something society cares about. But the average person does not think about aging in the same way.

Most people see aging as natural and not a malleable process. There is also an aversion to thinking about the arc of life and death. People are wed to the typical narrative: you are born, you grow up, you have a career, you get older, and you die.

That is the accepted plan, even if the plan sucks. It reminds me of the Joker’s line in The Dark Knight: as long as everything goes according to plan, nobody panics, even if the plan is horrible. People have just accepted aging as the normal state of affairs.

24:48 What does it mean to treat aging itself as the issue (vs other diseases)

Daniel 00:24:48

Who cares what people think about aging? Maybe it is okay if people deny it or do not see it as valuable, so long as the biotech machinery is running.

You mention a huge mismatch in investment, but the counterargument would be that people do not die of aging; they die of heart disease, cancer, and Alzheimer’s. We are spending a lot of money to treat those things.

In my mind, that raises a scientific question: is aging actually a process in and of itself that can be targeted? I am convinced that it is, but what do you see as the most compelling argument for that being the case?

Nathan Cheng 00:25:50

You are alluding to the idea that if people are researching age-related diseases, they should be able to find a solution that way. The question is why we think modifying the aging process would be more effective than just targeting specific diseases.

If you followed the breadcrumbs for these diseases rigorously, you would realize that only older people get Alzheimer’s. If you tried to prevent these diseases, you would come to the conclusion that we should target the changes happening to cells, tissues, and organs throughout life.

That is not the consensus view within pharma, though people are starting to wake up to it. Generally, people think about diseases as separate things. If you go to a hospital, the wards are split up by organ systems, like a cardiology unit.

The geroscience view recognizes a universality to these diseases. If you plot the risk of all these diseases on a graph, they have the same shape. The risk goes up exponentially as you get older.

From studying cell biology and the extracellular machinery, we know there are certain changes universally conserved across model organisms. We believe these are correlated to the molecular mechanisms that drive aging at the molecular level.

To prevent these diseases and the suffering they cause, it makes sense to target that underlying process. The best evidence that we can intervene is mostly in model organisms.

We do not have as much evidence in humans because it takes so long to run a clinical trial to determine if you have slowed aging. But in mice, there are a number of compounds which—

Daniel 00:30:11

Oh, I lost my water bottle.

Nathan Cheng 00:30:15

For reference, we give these water bottles out at the Longevity Biotech Fellowship cohorts. These are called the ITP molecules. ITP stands for Interventions Testing Program.

This is the gold standard research benchmark performed at the NIA, the National Institute on Aging, to determine whether a molecule extends lifespan in mice. They conduct triplicate studies using diverse mouse strains to account for genetic backgrounds.

They give some of the mice a molecule—rapamycin is a favorite example—and give a placebo to others. They then track them over their lifespan to see if the treated group lives longer.

For a certain number of molecules, there is a fairly robust, repeatable effect. Rapamycin extends the maximum lifespan in female mice by roughly 14%, and the median lifespan by about 20%. Sometimes you can add acarbose to rapamycin to get a synergistic effect.

We know we can extend lifespan pharmacologically in model organisms, and people are trying to extend this to other organisms we care about, such as dogs. There is the Dog Aging Project and a startup called Loyal.

Of course, researchers are also trying to develop therapies, small molecule drugs, gene therapies, and cell therapies to target aging in humans. That is a bit more complicated to demonstrate than a lifespan study in mice.

32:23 Do we need more scientific arguments, or philosophical arguments?

Daniel 00:32:25

This raises several questions for me regarding the biggest leverage point for bringing about longevity escape velocity on a cultural level. I wonder about two different trajectories.

The first approach is proving the geroscience hypothesis, which is the claim that there is a fundamental process of aging that, if intervened in, will prevent a whole host of diseases. There is already a lot of evidence behind that.

Do we just need to ensure more biologists understand and recognize that claim so they are motivated to work on it? In other words, do we need to convince them of the science, or do we need to do what Vitalism is doing, which is waking people up to the threat of aging and death on a philosophical level?

Nathan Cheng 00:33:35

It should be a combination of both. You need real results because some people are not convinced by philosophical, economic, or ethical arguments.

At the same time, we have a massive mismatch in the human and capital resources put toward this problem. There is a lot of leverage in trying to move the needle on the funding picture, which includes finding the right people who are aligned with the mission and coordinating them to push things forward.

On the aging biology front, a lot could be done. A low-hanging fruit is the fact that people don’t even learn about aging biology as part of a standard biology degree. They certainly don’t learn about it in high school.

It is crazy because aging is the number one killer by far. Increasing exposure for scientists and early-career people would be huge. To change the system, you have to increase the amount of attention on this space.

The education system and grant funding follow societal priorities. Scientists gravitate toward fields that have more funding relative to others. These efforts are essentially the same thing, but one is downstream of the other.

Daniel 00:36:07

If I look at the leverage points you have been focusing on, the Longevity Biotech Fellowship focuses on the talent pipeline—activating people who already believe in the mission to go work on it.

Vitalism works on broader cultural awareness. It is an attempt to create an ideology and a movement that people can rally behind.

Nathan Cheng 00:36:37

Fundamentally, when you look at the Longevity Biotech Fellowship or the work I am doing at the Vitalism Foundation, they have a similar thesis—just two different sides of the coin. That thesis is essentially that people are important.

Everyone is obsessed with AI, and AI will certainly be an interesting tool and source of leverage for solving aging. But at the end of the day, what really matters right now is people. Someone has to go out there and do the work: the science, the company building, and so forth. That is more of the LBF side.

On the other side, it is also important to find people who are mission-aligned and have specific skills, expertise, or influence. For instance, a lawmaker might be aligned with the mission. I think many people already exist in the world who are naturally aligned with this mission. They logically come to this conclusion on their own. You do not have to do much convincing; it makes sense to them the first time they hear it.

If that is the case, perhaps we already have enough people. This is not a new ideology where you have to start with one person and slowly convince the entire world. There are tons of people who naturally believe this. Our job is to find those people wherever they may be and coordinate them more effectively.

This is especially true if those people are in positions of influence, such as business leaders or cultural influencers. There are people in important cultural institutions like YouTube. MrBeast, for example, has said he hopes someone solves aging. He has expressed a belief that someone will solve it and that it would be cool to live longer.

Nathan Cheng 00:39:22

The Vitalism side is really about finding people who are already mission-aligned and bringing them together through coordination. Instead of thousands of disconnected people pushing on their own, we can bring them together into a network where they can coordinate and push in the same direction. That is much more powerful.

39:50 Nathan’s experience longevity-pilling influential people

Eric 00:39:50

Who are some of the most influential people you have had the opportunity to connect with over the topic of longevity, and how did you find their alignment? How aligned were these individuals with your view of longevity?

Nathan Cheng 00:40:06

That is a good question. I am not sure what I am allowed to say. I will mention the ones who are fairly public.

Bryan Johnson was a speaker at Vitalist Bay last year. He also attended a two-day LBF workshop that we ran in conjunction with the Foresight Institute a couple of years ago. He is very much aligned with what we are doing. In the health influencer space, he is probably the most well-known, especially regarding longevity. Being able to connect with people on that end is useful.

Vitalist Bay has historically been great at gathering those people, essentially throwing up the “bat signal” that solving aging is the most important thing. People naturally gravitate toward each other at these events.

If you ask how I normally find these people, the internet is still the best network in the world. I always tell our LBF fellows to write more on the internet, even though I no longer take my own advice. Do as I say, not as I do.

My entire career in this space, starting from zero, came from writing. Putting yourself out there and sharing your ideas on X is one of the best ways to connect with people. Most of the people I end up working with or co-founding things with are internet connections, either directly or through an introduction from someone I met online.

If you are starting out today, another good way to meet people is through the Longevity Biotech Fellowship or events hosted by the Foresight Institute.

One person we connected with through Vitalist Bay was Jaan Tallinn. He did a fireside chat during AI/Bio Week at Vitalist Bay last year. That was interesting because the rationalist community he funds is its own unique area.

There is a lot of overlap between people concerned about existential risk from AI and people who are Vitalists concerned about aging. At the end of the day, what motivates both groups is the desire to live and survive.

Daniel 00:44:27

Nathan, I’d love to pry a little bit. You mentioned there are names of people you have interacted with through Vitalism where you are not sure if you can share them.

Perhaps you can give us a flavor by anonymizing them. You could describe their roles to add a bit of mystique to the work you are doing behind the scenes. You are in these smoke-filled rooms with billionaires, longevity-pilling them.

Nathan Cheng 00:44:59

I don’t want to make it sound like that. “Smoke-filled rooms” makes it sound a certain way.

Daniel 00:45:07

What is the opposite of a smoke-filled room? A place where people plan good things in the daylight?

Nathan Cheng 00:45:15

That’s a good question. The people we connect with don’t have to be billionaires. I don’t know why people automatically assume billionaires are the go-to people.

While many have funded interesting projects—like Yuri Milner and Jeff Bezos backing Altos Labs, Brian Armstrong co-founding NewLimit, and Sam Altman with RetroBio—you can find high-impact people in all sorts of different ways.

People with expertise in policy, politicians, or former politicians can be very useful. Of course, scientists are also essential. Beyond doing the actual work, their credibility is vital for public communication.

We need scientists from well-known institutions to be more vocal about the fact that aging is a solvable problem. A combination of diverse experts is what really makes a movement or a network like this powerful.

47:03 Billionaire paradox

Daniel 00:47:04

I’d also love to emphasize the point you made about billionaires. There is a myth that billionaires are secretly investing in longevity and trying to live forever.

In one of your presentations, you shared a great stat: out of the 6,000 billionaires in the world with a combined net worth of $10 trillion, only 30 have invested anything into longevity. Even then, the amounts they have invested aren’t actually that large.

Billionaires aren’t funding this as much as people think. Furthermore, why would it be a bad thing if billionaires were trying to cure the thing that plagues every single human on Earth? Nobody gets mad at Mark Zuckerberg for funding cancer research.

Nathan Cheng 00:47:57

This is a whole can of worms. You’re likely referencing a presentation given by Adam Gries. He calls it the “billionaire paradox.”

Billionaires should have the most motivation to fund this because they have the resources and the most to gain. Even if they just put their money in an interest-bearing account, they would have far more money if they had more years of life.

Historically, the data suggests that billionaires don’t fund longevity much more than the average person does. People have a perception that it’s all billionaires because they are the only ones who can afford to fund a billion-dollar project. They are overrepresented in the pool of potential funding sources for mega-startups, but very few of them actually participate.

There is a strange stigma. For example, Altos Labs has never publicly stated who funded the startup, despite the fact that it was one of the biggest funding events in biotech history, launching with $3 billion.

That silence is evidence that some people believe it’s not socially acceptable for billionaires to be publicly associated with longevity biotech. It sucks that we are in a situation where people feel ashamed to fund research into the thing that causes the most suffering.

This sentiment likely stems from a general dislike of billionaires and an unfounded fear that these treatments will only be for the rich. Historically, most technologies and medicines go to wealthier nations first.

We saw this with the COVID vaccines. They were rolled out to wealthier nations first, but were very soon democratized. Capitalism works by taking a product with a large market and getting it to as many people as possible, which drives down the price.

Those fears are visceral but largely unfounded. Regardless, having more advocates who are not billionaires would help a lot. It’s powerful to have average people and normal scientists saying this is a problem they are dedicating their lives to.

53:18 The challenge in going from viewing aging as inevitable to malleable

Eric 00:53:18

I think we’re striking on the root etiology for why longevity isn’t more prestigious or normalized as a movement. The core issue is that it’s seen as too fringe and radical. People accept death not only as inevitable, but as a necessary part of life.

There is a philosophical belief that extending life indefinitely would be bad. Most people immediately worry about the negative first and second-order effects of radical life extension.

Other guests have addressed this. Rayni Romani spoke about the positive economic impact of extending lifespan, and Adam Gries has spoken to this as well. But even then, it’s a hard question to answer in the moment.

How do we rewrite society when aging is eliminated? What happens when aging is gone, and why is it a good thing?

Nathan Cheng 00:54:45

You have to provide a world model or articulate a positive worldview for when this happens. Most people don’t associate longevity with a positive future.

While there are obvious benefits to not dying of decrepitude and age-related diseases, not everyone understands that the same way. We need better arguments against certain pushbacks.

For example, Elon Musk’s main objection is that if people live longer, they get set in their ways. He argues society will ossify because we won’t change our collective mind about things. That’s an interesting point, but those objections are very theoretical.

They are conjectures about what might happen in the future. We have to balance that against the gravity of the current situation. People are dying and suffering today. We have 100% confidence in that.

Most people don’t want to get physically older or more prone to debilitating diseases. What is the alternative?

57:29 What needs to change about our culture to save us from aging & death

Daniel 00:57:30

This conversation has been compelling because I’ve wondered how much the cultural work matters. I’ve questioned whether doing the YouTube channel and podcasting really moves the needle when scientists are already doing the work.

This conversation shows why it does matter. People get stuck early in their intellectual development regarding longevity. They quickly fall into arguments that it’s an “evil billionaire thing” or that it’s simply inevitable.

There are so many roadblocks. There was an MIT Technology Review article earlier this year that was an exposé on Vitalism. The reporter reached out to two bioethicists for comment, and their responses were the best they could do.

It showed me that not enough good bioethics work has been done in this space. One bioethicist argued that defeating death would be bad because of our funeral procedures. They claimed death is important to us because we find so much meaning in the rituals surrounding it.

My reaction is that just because we have processes around something doesn’t make it a good thing. The Romans used to go to the bathroom in the same room together as a social gathering.

That doesn’t mean we shouldn’t have invented indoor plumbing. It’s absurd. The lack of people working in this space means we don’t have enough good ideas, which stops other people from getting involved. There is just so much nonsense.

Nathan Cheng 00:59:54

We can always bring back shared latrines as your next project after we solve aging. Regarding that article, journalists often feel they have to be balanced.

Even if the opposite view is extreme, they find someone to represent it. I don’t know where they found those bioethicists, but the argument that we’d be robbed of the cultural experience of a funeral is insane.

Who would prefer that cultural experience over being alive? Your grandfather would probably rather be alive than have a funeral held for him. I think most people would agree.

That might be a minority view, though. The bigger objections are that death is just part of life or the concerns about social ossification. Those are more representative of the actual hurdles we face.

Daniel 01:01:34

I would push back on that. I agree that the more representative take is that people are blindsided by aging. They see it as natural and inevitable. However, the role of philosophers is to help us understand it better. When bioethicists say these things, they are defaulting on that responsibility.

People struggle with this topic because there isn’t enough good intellectual work being done to help them. Mortality is hard to grapple with. We need experts and compelling individuals to help us cope with it, but not by burying our heads in the sand. They should help us face it head-on and decide what we want to do with that information.

Nathan Cheng 01:02:31

Fair point. If you look at successful recent movements like Effective Altruism (EA), they often start with a philosophy book. Someone has to write the core ethics to explain why the cause is good. It starts with moral innovation and then spreads.

There needs to be a foundation before a movement can be accepted culturally, especially if you think of it as a moral or social movement.

Eric 01:03:16

Nathan, do you feel the tides are changing for society’s views on aging and longevity? Beyond where society is today, what is the velocity and acceleration of this change?

Nathan Cheng 01:03:39

I wish I had better data. I’ve only been working in this space since 2020, so I only have five years of data points. However, I’ve seen an uptick in interest in longevity. It might not be “hardcore” longevity yet, but there are now longevity influencers.

Whoop, one of the most popular wearables, now includes a biological age score. People are thinking about this on a practical, day-to-day level. Regarding the idea of solving aging and death, Bryan Johnson is probably making the most noise and bringing this to the mainstream.

His influence on the space is likely a net positive. He has popularized a principled, “Don’t Die” view. I don’t know anyone else who has reached as many people with that specific philosophical view on solving aging. There has definitely been progress.

Scientifically, you would want to use opinion polls. Many people quote a Pew Research poll from about ten years ago asking if people want to live longer. It would be worth repeating that to see if views have changed.

Those polls depend heavily on how you ask the question. Many people haven’t thought about it at all. If you ask if they want to live longer without qualifiers, they imagine their grandparents’ health trajectory extrapolated. Naturally, they say they want to live 80 or 90 years.

I don’t have hard data, but it would be great to see more research. Emil Kenziora, the co-founder of Tomorrow Bio, conducted a poll that found roughly 30% of people want to live significantly longer than the current average. I’d have to look closer at how they phrased that.

Daniel 01:07:29

You mentioned that Bryan Johnson is a net positive for the field. This brings up something that makes me nervous about working in longevity. I worry about doing more harm than good for the field.

Personally, I like Bryan Johnson and I’m grateful for his work. He was a big part of getting me excited about longevity again. I had the seed planted as a teenager by Aubrey de Grey’s book, but Bryan brought it back to the present.

However, I’m not sure if he is a net positive. I’ve seen polls showing he is very alienating to middle America. I worry he might turn a large portion of the country against Silicon Valley and longevity research. Could this lead to more restrictions on biotech?

I have a similar fear regarding how we brand longevity. If we grow the audience for this podcast and brand it in a certain way, could it end up being a net negative? It might be a silly thought, but I’m curious about your reaction to that fear.

Nathan Cheng 01:09:06

Most people wonder if they will have a big or small impact, but they rarely consider whether they might have a negative impact. It really comes back to courage. You need the self-awareness and confidence to believe you’re doing the right thing, even though you can never be certain.

Nobody can predict the future, and nobody has the intellectual capacity to know for sure if their work will be net positive or net negative. You just have to get in the arena and see where the chips fall.

Regarding Bryan Johnson, I’ve heard people take the complete opposite view of him. I wholeheartedly endorse his “don’t die” philosophy. Giving people the choice to not be forced to die is incredibly powerful. Thinking about my own parents, I certainly hope they don’t have to die.

I don’t personally care much about the supplements or the biohacking, but his mainstream audience is often the reverse. They discover him through YouTube looking for diet, nutrition, or exercise tips. If you look at the Blueprint subreddit, that is what most people are interested in.

There is a dichotomy between the goal of solving aging and the practical, day-to-day things people are actually doing. I don’t think those daily rituals will have a major impact on the development of longevity science, nor will they make a huge difference for individuals beyond the basics like getting enough sleep.

There is tension within the scientific community because many scientists are against using things like supplements or biological age clocks for personal health decisions. They feel the technology isn’t ready yet.

I think Bryan has a meta-view of biohacking. He likely knows it isn’t the direct solution to aging; we will require better technologies for that. But he treats it as a cultural ritual that reinforces the idea of taking care of yourself so you don’t die.

Eric 01:13:31

Peptides for biohacking are exploding in popularity today. It is a highly contentious topic spanning legality, safety, and efficacy, and it’s worth a deeper debate in the future.

The origin of peptides and molecular biohacking actually started in a niche, uncool part of the internet: bodybuilding forums. Men were discussing creatine, testosterone, and growth hormone stacks to maximize their gains.

It was a group of very online, highly dedicated men obsessed with maxing out their physical statistics. We are now seeing that same mindset infiltrate a broader ecosystem of educated, socially accepted people in New York and San Francisco.

They are obsessed with “peptide maxing”—using things like tirzepatide or retatrutide to look and feel their best. It took about 20 years for this to move from bodybuilding forums to the mainstream.

Those cycles are compressing. The time between a niche trend and the mainstream is getting much shorter. Once the longevity interventions we discuss in these weird pockets of the internet show even a hint of working, it won’t be long before they go mainstream. I’m very bullish for the field.

Nathan Cheng 01:15:41

Once the “looksmaxers” get a little older, they will basically be forced into longevity maxing.

Daniel 01:15:48

That brings up a frustrating counterargument I often hear: “Why do you need drugs for your lifespan? Why can’t you just be healthy, exercise, and eat well?”

People don’t seem to realize that even if you are perfectly healthy, you are still going to die. There is such a blind spot regarding aging.

Nathan Cheng 01:16:23

It blows my mind when people think that optimizing their diet will solve aging. They think if they just get rid of all the microplastics, they’ll be fine.

While microplastics might not be good for you, removing them isn’t going to be the limiting factor for living beyond 120 years.

Daniel 01:16:54

One other thing I want to dig into is your theory of change. In one of your X posts, you made a comment about needing to be intolerant of people who are pro-death or pro-aging, and you brought up Nassim Taleb’s idea of the “dictatorship of the small minority.”

I’d love to hear what that is and how it influences your thinking.

1:16:55 Should we be intolerant of pro-death & pro-aging views?

Nathan Cheng 01:17:26

That was an off-the-cuff response to Brian Armstrong posting that he thinks aging is a disease. I wholeheartedly agree and loved that he posted it, despite the negative comments he received.

The idea of being intolerant of pro-death statements relates to Nassim Taleb’s “tyranny of the minority.” Taleb recalls an anecdote where he realized most soft drinks in the US are kosher by default.

Even though most people are not observant of kosher rules, a small minority adheres to them so strictly that it causes others to follow suit, provided it doesn’t cost them much. You can sway an entire group with just a small minority.

We see other examples of this in cultural movements, such as land acknowledgments. At a conference or a government building, someone stands up to acknowledge that the land was stolen. It doesn’t cost the audience anything to listen, and if one person is insistent, they can sway everyone else to observe that practice.

In the context of longevity, if you take the statement that “aging is a disease” to its logical conclusion, you realize it is a disease that kills the majority of people and causes immense suffering. These are chronic diseases that debilitate people over decades.

If you truly believe this, you should be compelled to stand up when someone says aging is a good thing. If you view it as something that causes mass suffering, you would be intolerant of someone calling it a good thing.

Most people in the longevity space will say they support research, but they won’t go to the extreme of calling out people who say death is a good thing. We haven’t progressed to that level yet.

Bryan Johnson and others in our community call it “deathism,” but it’s a weird situation. If any other cause were killing this many people and you believed it was a moral issue rather than just a technology being developed, you would stand up and say it is bad.

My hypothesis is that this is due to the Overton Window. Most people think of solving aging as a “nice to have” or a gift, rather than viewing death as something that takes life away—something we should stop just like we stop malaria or other diseases. We haven’t quite moved aging into the same category as other horrible diseases yet.

1:22:59 Join us at Vitalist Bay!

Daniel 01:22:59

Nathan, we’re at the end of the conversation. I’d love to leave our audience with actionable things they can look up or conferences they could attend. Where can people get involved?

Nathan Cheng 01:23:16

The next thing people should check out if they’re interested in longevity and solving aging is Vitalist Bay. We’re hosting a four-day conference on behalf of the Vitalism Foundation in Berkeley from May 14th to 17th.

We have about 120 speakers lined up, including Joe Betz-Lacroix from Retro Bio, Eric Verdin from the Buck Institute, Morgan Levine from Altos Labs, and many others. I’m especially excited about our two specific tracks.

The first track focuses on the replacement side of longevity biotech. This involves replacing tissues and organs as a way to potentially solve aging, an area I’m very bullish on. We have a great lineup of speakers for that.

The second track is biostasis, which covers the cryopreservation of organs. This has high potential impact for transplant medicine. Our speakers include Greg Fahey, John Bischoff from the University of Minnesota, and Hunter Davis from Until Labs. You can find more information at vitalistbay.com.

Another resource is the Longevity Biotech Fellowship. We run this program twice a year for people who are hell-bent on solving aging and want to build new things in this space.

We run a curated cohort program that connects these individuals, gets them up to speed on frontier research and technology, and connects them with mentors and collaborators. Our goal is to get people building new projects, organizations, and startups.

Over 500 people have gone through LBF, resulting in many startups, projects, and even this podcast. You can apply at longbiofellowship.org. Our next cohort retreat kicks off on August 24th.

Daniel 01:26:21

Amazing. Eric and I will be at Vitalist Bay recording episodes of the Free Radicals Podcast, so everyone should come. It will be awesome.

I also want to emphasize the pitch for the Longevity Biotech Fellowship. It was an amazing experience for me when I attended a year ago. It was my first time being surrounded by people who believe that aging and death are bad and who actually think about solving it.

Being in that environment was incredible. I had never met so many like-minded people before, and it was extremely energizing. I’ve been buzzing with excitement about working in this field ever since.

Thank you, Nathan, for everything you do and for coming on the podcast.

Nathan Cheng 01:27:08

Thanks for having me.

Kexin completed his PhD in Computer Science at Stanford, where he was advised by Jure Leskovec on artificial intelligence for healthcare and biology. As a PhD student Kexin garnered over 10,000 citations, published in Nature and earned six best paper awards at leading machine learning conferences, all this before the age of 30.

Kexin has earned a reputation as one of the brightest minds in AI for life sciences. You’ll see why in today’s conversation.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

2:47 Phylo’s vision to transform how biologist work
6:35 Phylo’s research areas
12:48 Why it’s harder for AI to do biology rather than software, and how to get around that
18:05 Building a virtual cell is a data problem
22:31 What it means for a biotech to go fully AI-native
27:26 The role of a human scientist & the importance of taste
32:22 Experiments in teaching AI scientific taste
36:57 How agents are transforming biology
42:53 The dream of the one-man biotech
47:21 How AI agents can find low hanging fruit through indication expansion
51:51 Can AI agents inflect biotech progress to give us radical life extension?
1:00:47 POPPER model for hypothesis validation
1:04:17 AI research institutes

Transcript

2:47 Phylo’s vision to transform how biologist work

Daniel 00:02:47

Kexin Huang, welcome to the Free Radicals Podcast.

Kexin Huang 00:02:51

Thank you for having me.

Daniel 00:02:51

Your company, Phylo, recently raised $13.5 million from Andreessen Horowitz and Menlo Ventures to build a research lab studying agentic biology. Your first product is Bio OmniLab, the world’s first IBE, or integrated biology environment.

Can we start by hearing about the vision for Phylo and what you are building?

Kexin Huang 00:03:16

At Phylo, we’re motivated by a problem we see in how biologists currently work. If you look at any figure in a paper or any result scientists have generated and trace back the process, it actually involves multiple scientists and a variety of fragmented tools, software, and databases. It is manual, complex, and time-consuming.

We want to fundamentally change how scientists work by letting agents handle the execution. These agents automatically connect with fragmented tools, software, and databases so biologists can focus on the science. If you provide a question, the agent handles the execution and directly gives you a result.

We are building an integrated biology environment to achieve this vision. This IBE contains multiple layers and components, including infrastructure for computing, an agent layer, and a biology software database. We put everything together into an AI-native, biologist-friendly interface. This enables scientists to move much faster than before.

Daniel 00:04:49

We definitely want to dive more into the details of what an IBE enables and how it will affect the future of biology research. But first, let’s talk more about the vision.

I noticed that you describe Phylo as a research lab. To me, that seems analogous to AI companies like OpenAI, DeepMind, and Anthropic. What does that say about your vision? Do you see yourself building foundational models in the same way?

Kexin Huang 00:05:21

We are a research lab that does both research and product. We want to build a product that is immediately useful to scientists while we continue pushing the frontier of agentic biology through research.

This creates an organic positive loop: research translates into product, and then product users provide interesting new problems and ideas for future research.

Our overall bet is that agents will fundamentally change how biology is done. There are tons of research and product problems to solve, and we want to build an environment that can support both.

Daniel 00:06:09

To make sure I’m framing this correctly: is the research focused on the AI agent in Bio OmniLab that orchestrates different pieces, or is it on specific tools like AlphaFold? Is the idea that Phylo is developing those orchestrator agents?

6:35 Phylo’s research areas

Kexin Huang 00:06:35

Our research spans three categories. The first is the evaluation of agents—determining how good a biological agent really is. Once we have the right evaluation, we can improve the agent harness or the underlying LLM. There is a lot of intricacy in building a good evaluation, and we spend a lot of time on that.

The second category is unlocking novel capabilities for agents in biology. We’re thinking about using agents to build things like AlphaFold or virtual cell models so that any biologist can become an AI researcher. We’re also exploring agents for curating massive datasets, which is currently a huge problem. We want to push forward completely novel applications.

The last category is the agent harness. We want to improve both the agent and the LLM to achieve expert-level performance across biomedical tasks.

We are currently taking two approaches to this. One is training the underlying base model using reinforcement learning. We released a preview version called BiOMNI-R0 last December and are currently training BiOMNI-R1. The second approach is focusing on a scalable agent harness to improve performance.

Daniel 00:08:54

How do you tune agents to make them better at biology research? What is the general framework?

Kexin Huang 00:09:09

There are several ways to improve. You can operate on the underlying LLM using reinforcement learning. Given the right evaluation or verifiable reward, you can train the model through backpropagation.

The constraint in biology is that, unlike coding, you don’t have endless verifiable rewards. Each reward must be expert-curated or come from expensive experiments. Because the amount of reward data is small, we need a much more token-efficient reinforcement learning method.

The second approach is to fix the LLM and improve at the agent harness level. This involves adding different tools or skills. One method falls into the category of system prompt learning.

We look at the traces where an agent fails, have another agent identify the issues, and then have another agent distill those lessons. We can then feed those lessons back into the main agent as skills. This allows the system to improve systematically.

Eric 00:10:49

If I’m understanding correctly, you have three primary research areas at Phylo: evals, agentic capabilities, and agentic harnesses. This creates an ecosystem where, with a harness, you can utilize any number of different skills or capabilities from other service providers.

You’re developing your own in-house agentic capabilities while ensuring you can build the appropriate evaluation criteria to determine when agents are working well in the context of biology research.

Kexin Huang 00:11:25

Exactly.

Eric 00:11:26

How did you arrive at these three areas? Was this something you had been thinking about for a long time, or has the field already converged on these as the right areas of focus?

Kexin Huang 00:11:37

It is primarily problem-driven. To create a product that is truly useful for scientists, we first need to know if the agent performs well across a wide range of biomedical tasks, such as single-cell omics or binder design.

Since biomedical tasks are so scattered, we need an evaluation framework to prove the agent is effective. Second, we realized agents aren’t proficient at every task, so we want to improve them. We are innovating on the agent harness and training LLMs to improve quality.

Finally, we see that agents have unlocked new capabilities. It’s a gold mine where people have only explored a small portion. We want to discover these hidden capabilities and surface them to scientists so we can move much faster.

12:48 Why it’s harder for AI to do biology rather than software, and how to get around that

Eric 00:12:48

This raises a question about developing rigorous and trustworthy evaluation criteria in life sciences. Agents have shown the most rapid growth in software and coding, largely because that field offers immense training data and fast feedback loops inherent to the nature of code.

In that domain, you can quickly iterate and improve. Biology is the exact opposite. We operate in a messy, real-world domain with many unknown variables regarding how we collect data, what it means, and what we should be collecting in the first place. How do you build evaluations given that messiness?

Kexin Huang 00:13:50

Unlike coding, where you can have an agent generate testing scripts for automatic verification, biology lacks a verifiable reward system. Many people are pushing the idea of using experiments as rewards for reinforcement learning by having a “lab-in-the-loop.”

In that model, the agent proposes something, it is performed in the wet lab, and the results are used as a reward signal to train models. However, reinforcement training currently requires multiple steps, and those rewards must be stable.

In biology, experiments are very noisy due to batch effects and other variables. This makes it a very difficult task to build a good reward for reinforcement learning if we focus on the biological experiment itself.

Currently, we focus more on computational tasks like data analysis, single-cell omics, human genetics, and survival analysis. For these, we work with experts, such as lead authors of major papers or industry veterans, to curate tasks for us.

We ask them to perform the tasks they encounter every day and establish the ground truth. A significant challenge with data analysis is that if you give the same task to ten different people, you often get ten different answers because it’s very open-ended.

These analysis tasks involve multiple steps, each with many degrees of freedom. When judging if an answer is correct, experts often look at the intermediate traces rather than just the final result. They check if the right statistical tests were used and if the steps were logical.

For our evaluation, we focus more on the intermediate process. We use a framework where scientists curate rubrics for the process, and we judge if the agent is meeting the criteria at each step. This provides a more realistic and faithful evaluation.

Eric 00:16:41

One of the most promising areas for agentic biology so far was the breakthrough with AlphaFold. That success benefited from the Protein Data Bank (PDB), a richly annotated dataset of static protein structures collected over many decades with hundreds of millions of dollars in research grants.

AlphaFold demonstrated that you can interpolate those static maps to predict novel protein structures for new sequences. However, biology becomes a thousand times messier the second you step out of those well-constrained problems.

Is that a correct interpretation? Where do you imagine things go from here as we continue to move into more complex and poorly defined problems?

18:05 Building a virtual cell is a data problem

Kexin Huang 00:18:05

There are two different data modalities in biology. The first involves biological data like proteins, sequences, and molecular structures. The second is natural language and biological reasoning data.

These are very different modalities. The biological data space is hardly mapped out; there is so much missing data in that distribution space. Protein structure prediction is one area that is relatively well-mapped, which is why models can interpolate well and achieve high performance.

However, in many other areas, such as building virtual cell models with perturb-seq datasets, we have only covered the tip of the iceberg. It is very hard for a model to generalize or interpolate when it only sees that small fraction of the space.

In contrast, the natural language data space is fully mapped out because it uses internet-scale data. There is an enormous amount of biological text and literature available, which gives models a much better sense of the data distribution. This is why agents are so effective and can make such accurate inferences.

By viewing these fields through the lens of data distribution, we can identify which problems are currently solvable. Because the natural language distribution is mapped out, we can easily extend agents to new capabilities.

Daniel 00:20:15

In terms of the data distribution, there are many levels to consider. Within the virtual cell, there is a lot of data missing just to understand cellular behavior. When we look toward the idea of a virtual human, we are missing the mapping of preclinical data to clinical data.

How do you see that issue being solved, and what is Phylo’s role in that?

Kexin Huang 00:20:41

Building a virtual cell or a virtual human involves significant modeling, but the primary hurdle is data. We simply don’t have enough of it to map out the space. It is almost impossible to build a high-quality model if the data doesn’t exist.

One speculation is that we wouldn’t even need complex new model architectures if we had sufficient data. While some argue AlphaFold’s success was due to its amazing architecture, I believe the balance between architecture and data depends on the modality. For a virtual cell or human, it is primarily a data problem.

The challenge is curating data in a way that significantly improves the model. A few years ago, I interned at Genentech where we worked on a “lab-in-the-loop” approach. We let the model select which data was needed to fill the data space more smoothly. That selected data was then used to further train the model, which could then select more data in a continuous loop.

At Phylo, we are focused on a different angle. Our goal is not to build a virtual cell or a virtual human. We want to help biologists accelerate their work so they can make new discoveries much faster.

22:31 What it means for a biotech to go fully AI-native

Daniel 00:22:31

You mentioned on Phylo’s blog the idea of what happens when organizations go fully AI-agent native. If everyone is doing their work in Biomni, from preclinical research through to clinical trials, they have all their data in one spot. Biomni seems to be the tool that enables that future.

Kexin Huang 00:23:04

Exactly. As more scientists in an organization use Biomni Lab for tasks ranging from target discovery and molecule design to clinical trials, we capture all the discovery traces. We record the entire path from the initial question to the final result.

This becomes an incredibly valuable system of record for a team. For the first time, it captures multimodal data alongside reasoning traces. We view this as a new data modality—a mix of biological data, human reasoning, and agent activity.

There are many ways to utilize these records. For instance, you could file an IND much faster because all the necessary information is already centralized. It could also assist in making high-level portfolio decisions.

With a complete system of record, you can trace failures back to experiments conducted ten years ago. You might even find insights like GLP-1 earlier. Findings like GLP-1 were hidden in the datasets of various organizations for twenty years. An agent could automatically identify these hidden findings and surface them much sooner. Once we capture this as a system of record, the applications are unimaginable.

Eric 00:25:02

I’ll speak to two things. The first is an article published in December 2023 by Jaya Gupta and Ashutosh Garg from Foundation Capital titled “AI’s Trillion Dollar Opportunity: Context Graphs.” That article went viral because of its thesis on agentically native organizations.

The major opportunity lies in shifting how we think about storing decision traces and the context graph behind business decisions. By building a new agentic native system of record, agents can store and retrieve data from all that context to perform context-specific inference.

That is exactly what you are describing with Biomni and Phylo: building a unified system of record and a new context graph that stores all decision traces for the life sciences.

Kexin Huang 00:26:12

Reading the paper really resonated with me. An interesting byproduct is that it actually addresses the reproducibility crisis.

If you look at the majority of papers in journals like Nature, Cell, or Science, and try to recover a specific figure, the entire process from the initial question to the final result is often unrecorded.

Now, since the agent handles everything from the question to the result, all execution and discovery steps are recorded and saved in a Jupyter notebook. You can always go back and reproduce it. It’s an interesting way to solve that old problem.

Daniel 00:27:05

This reminds me of how big companies, not just pharma, struggle to capture the implicit knowledge stored in people’s heads. I ran a small business in the past, and you end up being very dependent on certain employees who have implicit knowledge about how things need to be done.

Kexin Huang 00:27:25

Exactly.

27:26 The role of a human scientist & the importance of taste

Daniel 00:27:26

If AI can increasingly capture this information in large companies, it raises the question of whether it can also automate those processes.

Once implicit knowledge moves from someone’s head into a system, is capturing that data and those records a prerequisite for automating scientific work? If so, what core elements of scientific research will remain unautomated, where humans still make the biggest contribution?

Kexin Huang 00:28:03

Biology has an interesting property: different people prefer different ways of doing the same task. For single-cell annotation, some people prefer Seurat while others prefer Scanpy. They might use different thresholds or different sets of marker genes. It’s a very personalized process.

Users often don’t want the “best practice”—they want their specific way of doing things. This is especially important for organizations that have established protocols for particular tasks. We want to design the agent so we can codify this task-specific knowledge, making it a “second brain” for AI employees and scientists.

Increasingly, this knowledge will be distilled into agents and made available. However, some parts will still be very hard to codify. These generally fall under the category of “taste.” For example, in portfolio management, some people just have a sense that a specific target is a good idea. That subtle intuition is very hard to codify right now. Coming up with the right idea and having good taste is something agents currently lack.

Daniel 00:29:44

The argument that taste cannot be replaced is common across many fields. I see it in my own work with business strategy. AI’s intuition on business strategy isn’t great because it doesn’t have taste. I feel like I know better.

Why do you think an AI scientist lacks that intuition or taste? Do you think Biomni can eventually develop that over time?

Kexin Huang 00:30:21

That’s an interesting question. I’ve thought a lot about how people develop taste. No one is born with the scientific taste to judge which papers are good.

Why can certain Nobel Prize winners consistently identify great ideas? I think it is a data distribution problem. The more you read about a space, the more you map out what is available. You see what works and what doesn’t, and you get feedback from the community. You keep building this internal data distribution of taste. Eventually, you can generate a new idea and map it against that distribution.

The prerequisite for taste is having a sense of what is good and what is not, and then being able to sample from that. In theory, an AI scientist could do this. It just needs to see enough examples and receive enough feedback to develop its own taste. Right now, AI scientists are static. They are snapshots of “average” taste rather than evolving systems that explore and learn what truly works.

Daniel 00:32:12

So, if an AI agent had an objective function and received feedback on what is working, it might develop a better sense of taste and a better way of achieving the goals you give it over time.

32:22 Experiments in teaching AI scientific taste

Kexin Huang 00:32:33

That is one hypothesis I have.

Eric 00:32:40

This raises an interesting question about how to codify, quantify, and store data regarding human cognition and taste. It’s a hard problem to map.

You can ask a human why they made a particular decision or judgment, but it’s like the Heisenberg uncertainty principle—measuring the system changes its nature. What different ways have you explored for measuring human cognition and taste?

Kexin Huang 00:33:27

We explored one idea in a side research project. We looked at whether we could use a citation predictor as a proxy for how impactful an idea is. Since citations are a proxy for impact, they correlate somewhat with taste.

We collected many papers, looked at their citation counts after three years, and distilled their core hypotheses. We then trained a large language model to take an idea and predict how many citations it would receive in three years.

We eventually realized that taste is highly time-dependent. If you proposed CRISPR in 2000, it was an amazing, visionary idea. If you propose it today, it doesn’t show good taste because it’s already established. Because of that time dependency and the difficulty of avoiding data leakage, we moved away from that direction. Quantifying the quality of an idea is a fascinating but incredibly difficult challenge. It probably requires a more orthogonal approach.

Eric 00:35:32

That is very interesting. It correlates to the question we had earlier: how do you use agents to determine the whitespace of available data to traverse, and then design experiments to efficiently collect data in those spaces?

In theory, if you have a reward function or objective function tuned around taste—where you prioritize areas considered high-taste versus low-taste—it would make those sorts of agentic knowledge base traversals more efficient. How do you approach that problem right now?

Kexin Huang 00:36:07

Right now, we perform this traversal by defining a goal. Currently, that goal is maximizing a property of interest. If we have a proxy for taste, the agent can also try to maximize that as well.

Once we define a goal, the agent is able to perform exceptionally well. For example, it can intelligently select which gene to target in perturbation outcome prediction models. Because the LLM reads the entire internet, it has a good estimation of the optimal next step within the constraints of the goal.

36:57 How agents are transforming biology

Eric 00:36:57

Let’s take a step back into the broader AI and bio ecosystem. You’ve been part of this ecosystem in some of the top institutions, working with impressive practitioners like Aviv Regev, Jure Leskovec, and many others.

Can you tell us how the ecosystem has evolved since you first joined the AI for bio mission and where you think it sits today?

Kexin Huang 00:37:23

I started my AI and biology efforts seven or eight years ago. Back then, BERT—the early large language model—had just been released. There were no agents yet, so we focused a lot on modeling biological and chemical datasets, molecular structures, and clinical trials using AI models.

This also included perturbation models, perturb-seq, and human genetics data. For the first five years of my research, I focused on building AI models that could accurately represent this data.

Three years ago, I realized that while these models provide great predictions and interesting results, they are not immediately useful to scientists. If we look at the day-to-day life of my biologist friends, they don’t use these models. They are focused on analysis, statistical tests, and reading papers. These are standardized but manual and time-consuming tasks.

When ChatGPT was released, our entire computer science department was in panic mode, doubting our research directions. We realized that agents could potentially fundamentally change how biomedical research is done and provide immediate value to biologists.

Traditionally, single-cell annotation takes a week; now it can take a few minutes. That is a dramatic productivity boost. This immediate usefulness is what drives me, so I shifted my entire research direction to agents three years ago. As large language models improve, they are increasingly able to deliver expert-level performance to scientists.

Daniel 00:39:54

My understanding is that one of the big bottlenecks in biomedical research is simply that biology takes a lot of time. If you want to run an experiment on cells, you have to culture them.

The feedback loop could be two weeks or more, and in the case of clinical trials, it can be a decade. I imagine many scientists are doing this analysis while waiting for their cells to culture. How much do you think automating those activities can actually speed up scientific progress?

Kexin Huang 00:40:36

There are two levels to this. First, there are tons of underutilized data out there. Using agents to analyze this existing data can already generate significant new insights. GLP-1 is a perfect example; the data existed 20 years ago, yet it took two decades to bring it to market. Finding hidden discoveries in existing datasets is already a huge win.

On the other hand, wet lab experiments remain a major bottleneck in the day-to-day life of a scientist. Agents are already revolutionizing the dry lab part of research, but I can also imagine them integrating with wet lab components in a unified environment.

Last week, we announced a partnership with Adaptive, a robotics lab that performs protein experiments on the cloud. We integrated their API, and now our agent can perform protein expression and affinity experiments. The agent waits two weeks for the results and then automatically returns the data.

The scientist doesn’t need to do anything; they just use natural language. The full loop from dry lab to wet lab is encompassed in a single environment. We can automate this wet lab component in the future as well.

The current bottleneck is that wet lab experiments are diverse. High-throughput experiments are easier to automate, but low-throughput ones are very flexible. We don’t have the hardware or robotics to automate that part yet. Once we can, we can wrap it as an API and integrate it into our environment to automate the entire process.

Daniel 00:42:35

There are companies working on automating all types of wet lab experiments. As those companies go to market, you can pull those in as API calls into your AI agent.

Kexin Huang 00:42:51

Exactly.

42:53 The dream of the one-man biotech

Daniel 00:42:53

If you had to guess, when will I be able to sit at my laptop and start discovering drugs end-to-end on my own?

Kexin Huang 00:43:04

The bottleneck is definitely the wet lab right now. The majority of wet lab work is still not automatable. While the idea of a one-man biotech is super interesting, we shouldn’t underestimate the complexity of the real drug discovery process.

Coming up with a drug is relatively easy now, but making it clinical-trial-ready involves many steps that are not yet automated. In our vision for building a one-man biotech, there are several stages.

The first step is to automate all individual non-wet lab tasks to an expert level. We can already do a lot with data analysis, but many questions remain regarding things like modeling the ADME properties of a drug. Agents are not yet able to do that very well.

Our idea is to automate the non-wet lab parts and wait for wet lab technology to mature. Once it does, we can integrate it via APIs or wrap CROs as a service. We can then piece them together to create a one-man biotech. It will take a while—I estimate we are about ten years away from achieving that vision.

Daniel 00:44:50

If you get to the point of a one-man biotech, are you essentially at the no-man biotech level? I wonder if that “one man” is someone with exceptionally good taste, like a pharma executive, or just a random person at a laptop. In the latter case, what do you even need that person for?

Kexin Huang 00:45:15

That is a very interesting perspective. If you can achieve a one-man biotech, then you could potentially have a meta-agent that launches 100 biotechs automatically.

In biology, there are endless questions. When I was doing work on rare diseases, I found that 92% of diseases still have zero treatment. If we can achieve this one-man biotech model where a single person with good taste selects the right targets, augmented by agents, we could dramatically mass-produce therapeutics. That would be a very exciting future.

Eric 00:46:12

I’m inspired by what you’re describing. It reminds me of something my prior colleagues at Andreessen Horowitz wrote about years ago: the idea of the “beach biotech.”

The concept is that you could be sitting on a beach with your laptop and a Starlink connection, typing into a chat interface to run a team of agents that robotically implement all aspects of drug discovery.

When that was written in 2022, we thought it might take ten years. We are likely still ten years away from the full end-to-end scope, but the first steps are moving faster. Identifying the right target and designing chemical entities as hits or leads already feels like it’s only a year or two away from being almost entirely automatable.

47:21 How AI agents can find low hanging fruit through indication expansion

Kexin Huang 00:47:21

The traditional timeline of drug discovery is going to be changed dramatically by agents. Beyond de novo drug discovery, there is also indication expansion.

This is already an approachable problem for agents because there is so much data available. If you connect agents with real-world evidence, you can expand indications much faster through drug repurposing.

Daniel 00:48:02

Help me understand that more. Where is that data available? For example, GLP-1s like Ozempic are seeing a lot of indication expansion. Where are we getting that data? If people are taking it in the real world outside of a clinical trial, is that being captured anywhere?

Kexin Huang 00:48:24

Real-world evidence companies collect data from insurance and electronic health records (EHR). There are many signals in this data.

If someone takes Ozempic and happens to have other conditions, you can perform statistical analysis and compute p-values to identify if the drug correlates with a treatment effect. That is how Sildenafil was originally repurposed—by mining through real-world evidence.

Daniel 00:49:03

Can you actually expand an indication for a drug based on statistical analysis alone?

Kexin Huang 00:49:11

There are tons of signals that help you select the right indications. Indication expansion is a data harmonization and synthesis problem. There might be hundreds of pieces of evidence pointing toward a new indication.

The challenge is aggregating this information in an intelligent, scalable fashion. Previously, this was a manual process. Now, you can let an agent loose on EHR or insurance datasets to compute p-values for all drug-disease pairs.

You can then add that evidence to your portfolio for review. You still need to back it up with more evidence, but you end up with a much better picture to make informed decisions.

Eric 00:50:13

This relates to your work on TxGNN, where you map the existing space of FDA-approved therapeutics against the tens of thousands of diseases that lack effective therapies.

The idea is to take existing drugs, determine the network of pathogenic molecules they modulate, and see if those drugs are well-suited to be repurposed for other diseases.

Kexin Huang 00:50:49

Exactly. The idea for TxGNN is to model biology as a knowledge graph, where the nodes represent genes, diseases, variants, and pathways.

If we know an FDA-approved drug targets a specific gene, and that gene is in a pathway related to a certain disease, we can map out those paths. If there are many paths between a drug and a disease, we have higher confidence that the drug can be repurposed.

We are modeling the biological side of the evidence, but you can also aggregate EHR data and other sources to help you make informed decisions on whether a drug should be repurposed for a specific disease.

51:51 Can AI agents inflect biotech progress to give us radical life extension?

Daniel 00:51:51

This conversation makes me wonder how AI agents in science might expedite discoveries. This includes repurposing existing drugs, which can bring significant benefits to patients, as well as speeding up the discovery of new drugs and targets.

On the Free Radicals podcast, we are very excited about cutting-edge biotech, especially in the realm of longevity. We want to find drugs that can completely rejuvenate us, pause aging, and give us indefinite lifespans.

Do you envision these AI agents providing an incremental benefit, or do you think this is a breakthrough that can provide a 10x or 100x magnification of biotech progress?

Kexin Huang 00:53:06

There are two ways agents can be helpful. The first is dramatically improving the speed of our current drug discovery process. The second is generating innovative ideas and completely different ways of approaching these problems.

We had a collaboration with Kejun, a postdoc at UW, who is working on an aging agent. We built an agent that mines through Gene Expression Omnibus (GEO) data. We used mouse data to compute biological age and studied how various interventions change transcriptomic nodes.

By measuring which interventions reduce biological age, we can identify treatments that reverse aging. This is unconventional because no human scientist could process that much data. It was only unlocked by the scalability of an agent.

I imagine we will see more innovative uses of agents that lead to new discoveries much faster. In summary, the two paths are automating existing processes and discovering completely novel insights.

Eric 00:55:02

That reminds me of our last episode with Jonathan Guttenberg and Omar Abudayyeh from Harvard. We discussed what it takes to reach a true acceleration point in life sciences.

The main bottleneck the industry must grapple with is the length of the feedback cycle. It can take a decade or more to move from a therapeutic hypothesis through preclinical work and clinical trials to FDA approval.

No matter how efficient you are before that point, you still have to wait years for clinical trials to run. AI might allow us to accelerate past that bottleneck through the identification of digital biomarkers and other interim readouts of therapeutic efficacy.

In longevity, platforms like Phyllo or Bio-Omni could be extremely valuable in creating a rigorous context graph for therapeutic hypotheses and interim data. This would allow us to determine if a therapy is working as intended before completing a 50-year randomized control trial.

We need to move away from using natural language to communicate biology and toward a machine-learning-native language. This would allow us to parse complex datasets and find the hidden signals embedded within the latent space of biology.

Kexin Huang 00:57:05

There is so much data out there that no one is currently analyzing, and it contains many hidden insights. Regarding clinical trials, agents can make the process faster and significantly lower costs.

Biomarker discovery is a major topic, but many companies are also working on the administrative side of clinical trials. Much time is spent on patient recruitment, trial matching, and establishing eligibility criteria. Agents can help automate these individual steps.

If we collectively automate everything from the preclinical to the clinical stages, we can squeeze the development window by several years.

Daniel 00:58:14

If AI can parse through massive datasets to identify trends that humans cannot, how does that impact the value of human research? We are moving from focusing on theory and mechanism to statistical analysis.

Does this mean humans should focus on providing AI with the data it needs? Perhaps the highest leverage point for humans is mapping out the landscape of data and using our taste to determine which data is most important.

Kexin Huang 00:59:09

In biology research, there are still many open questions. An agent is a very powerful machine, but it still requires a human to start the process and provide the input prompt. Because the biological space is so large, it pushes humans to identify which questions have yet to be mined or surfaced.

Instead of focusing on execution, we can focus on which ideas are worth pursuing. Asking the right questions is incredibly difficult, and that is where the human’s job lies.

I think a lot about the fundamental shifts required to make biological discovery much faster. There is the AI component, but there is also measurement technology like single-cell RNA-seq and perturb-seq. These help us get a much higher resolution on biological cells.

One idea is to ask an agent to improve that measurement technology. There are tons of ideas like this, but finding them requires a lot of time and taste.

1:00:47 POPPER model for hypothesis validation

Eric 01:00:47

You recently published research on Popper, which matches PhD-level scientists by reducing experiment time tenfold. Help us understand the source of Popper and its purpose.

Kexin Huang 01:01:05

Our idea centers on hypothesis validation. Large Language Models can generate tons of hypotheses, and I believe that generation is not the bottleneck. The real bottleneck is validation. Validating a hypothesis is extremely time-consuming. We wanted to use agents to rigorously automate that process.

We were inspired by Karl Popper’s school of the philosophy of science. His approach is to try to falsify a main hypothesis by constantly designing experiments to disprove it. If you falsify it, you iterate on the hypothesis. If you cannot falsify it, it is tentatively accepted, though it is never considered the absolute ground truth.

Our key idea is to use agents to design these falsification experiments based on a main hypothesis, such as one gene regulating another. The agent uses various datasets to design these tests.

For example, using GTEx data, the agent can compute p-values to see if a gene is expressed in a specific disease context. We can continuously generate different falsification experiments from different data sources and aggregate those p-values in a statistically rigorous, sequential fashion.

This results in a validated hypothesis. It is a powerful tool because you can screen any hypothesis you have, and the system provides a rigorous score based on the available data. Any organization or lab with massive datasets could use this. You provide a hypothesis, wait a few minutes, and receive a score along with all the supporting evidence and falsification results.

Eric 01:03:38

The idea is that this would eventually serve not only as a copilot for human scientists, but it could become a way for fully autonomous agents to devise thoughtful, high-taste experiments.

Kexin Huang 01:03:57

Exactly. It ties back to the “system of record” idea. Once you have a system of record with massive data, you can test new hypotheses against it. The system will rigorously generate a p-value to help you validate your main hypothesis.

1:04:17 AI research institutes

Daniel 01:04:17

Zooming out to how science is done broadly—peer-reviewed journals, scientists debating on Twitter, and conferences—there is a massive infrastructure for the scientific method and how it is communicated. Which aspects of that ecosystem do you think will transform as we move toward agentic science?

Kexin Huang 01:04:50

Many startups are currently trying to automate the process from idea to published paper. It is already feasible, which means we will see a massive increase in the number of papers being generated. This poses a strong demand for rigorous peer review.

I can imagine a new genre of peer-review agents designed to evaluate these agent-generated papers. In the future, there will be an ecosystem of agentic science where papers are generated and reviewed by agents, and humans decide which ones are worthwhile.

At the end of the day, it is still a human society. Humans need these scientific results to develop drugs or build new research. I see a future with a human-agent co-existent peer-review infrastructure.

Currently, agents are not listed as authors; humans use them to publish faster. Humans also use agents to review papers. In the future, I imagine this becoming more autonomous, even moving toward agent-to-agent communication.

Daniel 01:06:25

The economist Tyler Cowen says that if we want to understand the impact of AI agents, we should imagine them setting up institutions analogous to human ones. Humans form universities to collaborate. What organizational systems will these AI agents form?

Kexin Huang 01:06:44

Many members of our team are fascinated by the idea of an agentic research institute. This would involve an army of agents collectively working toward a high-level goal. I believe that is feasible.

This fits the “one-man biotech” idea where a human defines the overall goal and lets an army of agents work together for a year. The agents generate results while the human is on the beach.

Eventually, an agent might gradually come up with its own high-level goals and launch its own research institute to achieve them. The exact shape of these institutes is uncertain, but they will likely be able to achieve much higher complexity than individual agents.

Eric 01:08:02

Another topic we would love to discuss is the Therapeutics Data Commons, or TDC, which is an open-source data repository for therapeutics. What is the TDC, how did you get involved, and what is its role in advancing science?

Kexin Huang 01:08:24

TDC started almost six years ago while I was working with Marinka Zitnik at Harvard. At that time, I was building AI models for biological data. We noticed that we were manually building and curating machine learning datasets, including training and testing sets, to make them both realistic and machine learning ready.

We saw a clear demand for this. In computer vision, there is ImageNet, but in machine learning for drug discovery, no such standard existed. We wanted to create an ImageNet for drug discovery—a standard dataset infrastructure to build models and monitor progress.

This motivated us to build the Therapeutic Data Commons (TDC), which covers the entire drug discovery pipeline, from target discovery to molecular design and clinical trials. It encompasses more than 50 machine learning ready datasets and tasks. Scientists can use these datasets to build their own models. It is essentially an infrastructure for machine learning in drug discovery.

Daniel 01:09:42

We’re almost at the end of our conversation. Kexin, an exciting way to wrap this up would be for you to describe how you see the next 10 to 20 years of biological scientific progress.

Where do you see the world going, how do you think Phylo will contribute, and what gets you most excited about that future?

Kexin Huang 01:10:11

We are moving toward an era of discovery abundance. Previously, discoveries were scarce and infrequent. In the future, they will be abundant.

If we look at an x-axis for time and a y-axis for the number of biomedical tasks—whether that involves omics, single-cell analysis, or wet experiments—human progress is bottlenecked by bandwidth. However, for an AI agent, the number of tasks will grow exponentially.

Since a discovery is a compilation of many different tasks, we can hypothesize that as agents handle tasks exponentially, we will see significantly more discoveries. That is the future we are excited about at Phylo.

If we can provide the infrastructure that enables this, we will be part of every discovery. Whether it is a Nobel Prize-winning breakthrough or day-to-day work, we want to be part of it. An abundance of discoveries is a truly exciting future.

Daniel 01:11:33

Just as Claude is a co-author on a high percentage of GitHub commits, hopefully in the future, a Nobel Prize will be co-authored by Phylo or a biomedical AI.

Kexin Huang 01:11:42

Exactly.

Daniel 01:11:45

Awesome. Thank you so much for joining us on the Free Radicals Podcast, Kexin.

Kexin Huang 01:11:49

It was a very engaging discussion. Thank you for having me.

And best of all? They’re totally longevity-pilled. People want to feel younger and healthier, and big pharma is waking up to it.

In this week’s episode, we discuss the biology of aging, the future of longevity therapeutics, and how Harvard can bring legitimacy to the supplements industry.

As PhD students at Harvard in Feng Zhang and Aviv Regev’s labs, Jonathan and Omar pioneered research on programmable targeted genetic engineering with CRISPR. Now Jonathan and Omar lead a joint lab at Harvard, where they develop molecular and artificial intelligence based approaches to program biology, unravel cellular aging, rejuvenate hair, develop next-gen therapeutics, and decode biological complexity with virtual cells.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

0:00 Intro
2:47 What it means to program biology
9:57 The shift towards longevity & preventative medicine within Harvard & Big Pharma
18:55 How simulations of biology can accelerate drug development
26:49 The data that will make biology readable by machine intelligence
33:00 Defining aging & thinking about what the customer actually wants
42:26 New biomarkers and the future of aging interventions
51:15 The rise of proteomics as a health & aging biomarker
56:01 Building the virtual cell
1:00:40 Disrupting and democratizing the supplements industry
1:07:51 Genetic engineering, small molecules, peptides... what are the right tools for aging intervention
1:13:45 How longevity pilled is everyone here?

Transcript

2:47 What it means to program biology

Daniel 00:02:47

Jonathan and Omar, welcome to the Free Radicals podcast. You guys run a joint lab at Harvard Medical School where your tagline is programmable biology.

What does it mean to program biology like a computer? How does that guide your lab’s work?

Dr. Omar Abudayyeh 00:03:01

Programmable biology is an interesting buzzword that has joined the zeitgeist. I come from an engineering background and grew up as part of the personal computer and internet generation.

I was coding even in middle school. To me, it was an incredible experience to design digital systems. You could write code and it would do something almost instantaneously. If it wasn’t quite what you expected, you would debug and iterate until you eventually came up with something deterministic that could do useful things.

When I encountered biology in college and graduate school, it did not fit that phenotype. Biology felt like a black box. You can’t see what you’re working with, and it’s hard to even think about how you would design something.

A lot of the time, it feels like you’re grasping in the dark, trying to uncover new biology. It is difficult enough to design a system to do exactly what you want by specification, let alone simulate those systems ahead of time.

Airplanes operate with incredibly high accuracy; they aren’t falling out of the sky. A big part of that is the ability to simulate designs ahead of time to ensure they have a reasonable chance of working before they ever leave the ground.

This applies to all sorts of technology. When Nvidia was getting started, they wanted six-month design-to-product cycles. What saved them was the acquisition of a company with simulation software for chips. They could simulate their designs before manufacturing, ensuring the chips worked on the first try.

When I think about programming biology, I think about having enough data to make models and tools that allow us to simulate and design something useful that actually functions as intended.

During graduate school, we came up through the CRISPR world and developed some of the first CRISPR technologies. That was my first taste of achieving this vision. You could go into a DNA sequence editor, design a bunch of CRISPR guides, and order them.

Within a week, one or two out of three guides would work well, and you would have a genome-edited cell. That was the first time we could go from the digital world to the physical world at record speed.

When we think about programming biology, we think about taking that paradigm and applying it everywhere: to cells, tissues, and even humans.

Could we have a world model for humans where we simulate my health 20 years into the future? I could see which lab results are on a bad trajectory and what diseases I am at risk for. Then, we could simulate hundreds of interventions in real time to change those trajectories. That is my definition of programmable biology. Jonathan, do you want to give yours?

Dr. Jonathan Greenberg 00:06:10

Obviously, we are pretty aligned. Simply put, it’s an aspiration. We want the capability to manipulate biology.

The ultimate reason for this is that we are biology. Perhaps there is a bit of anthropological bias there, but we want the capability to manipulate our biology in reproducible, predictable, and complex ways to cure diseases, enhance people, and accomplish amazing things.

If you are in the medical or biotech sphere for more than two minutes, you understand that biology is incredibly complex. Most drugs fail. It takes a long time to develop them. We don’t even have drugs or putative targets for most interesting conditions.

How does neurodegeneration work in Alzheimer’s? We don’t really know. There are great examples we have stumbled into, like GLP-1 receptor agonists. When we actually understand how to intervene in biology in an impactful way, we can change the lives of billions of people.

The question is: how do we find a more reproducible, time-efficient, and capital-efficient way of doing that? Biology is one of the most complex natural systems in existence because of evolution, selection, and drift.

How do we understand how to intervene and create medicines predictably? In biotech, people are always biting their nails over whether a clinical trial will succeed or fail. It’s always a binary event.

When the Artemis rocket launched recently, no one was truly worried about it blowing up. There was a level of predictability there. How do we get biology toward that?

How do we predict what my disease risk will be in five years and determine how to intervene to change that? How do we create those interventions so they don’t take ten years to develop?

This is especially critical now because everything else is speeding up so much. If we develop a drug, it realistically won’t reach patients for seven to ten years. I don’t want to bury my head in the sand while every other field moves quickly.

We have to be able to program things at every level, from the molecular nucleic acid and protein level to the cellular and human level. We want to use that to translate quickly into an impact for patients and consumers, reducing the cycle time on this flywheel.

9:57 The shift towards longevity & preventative medicine within Harvard & Big Pharma

Daniel 00:09:57

You’re both clearly focused not just on publishing, but on impact. You’ve discussed curing diseases and human enhancement, and you’re also doing work on aging interventions.

These are exciting topics to us, but intervening in aging is not exactly a mainstream way of thinking. Similarly, human enhancement is a topic that many avoid.

In my experience, even friends at Harvard Medical School don’t want to talk about aging biology; they see aging interventions as having a “fringe” vibe. Since your work is so cutting-edge and futuristic, how does it feel working on it at a place like Harvard Medical School, which I imagine is a bit more conservative?

Dr. Jonathan Greenberg 00:11:02

You’ve outlined what is essentially a communications problem. How do we communicate this in a way that makes sense?

Regarding aging, we can debate whether GLP-1 receptor agonists are aging drugs, but for those taking them and losing significant weight, it is extending their lifespan. Almost everyone is on Ozempic, Tirzepatide, or even gray-market Retatrutide right now. When people discuss Chinese peptides, they are usually talking about Retatrutide.

It’s all about how you frame it. Everyone wants to be healthier and live longer. For those of us in our 30s, we notice we don’t have the same energy we had in our 20s. We want to improve that.

People go to the gym, which is a form of enhancement. They take protein powder or creatine. People want to improve their daily lives even in the absence of disease. When they do have a disease, they want relief.

These are different ways of looking at the same fundamental problem: understanding what happens to the human body as it ages from a cellular and medical perspective. We need to frame our work in a way that answers those questions and addresses the innate desires people have.

Much of our work involves figuring out how to communicate these ideas and identifying the societal impacts if we are successful. We are clearly moving toward something that will materially improve humanity.

Dr. Omar Abudayyeh 00:13:19

Consider Jeff Bezos. He built a company that has created more value than almost any other in history. To do that, he endured a decade of tremendous societal pressure and articles predicting Amazon’s failure.

He survived by focusing his entire team on what the customer wants and how to create value for them. Science often suffers from being disconnected from what patients and the general public actually want.

Every single person is aging. Whether you are five or 60 years old, you want to be able to intervene in your health. Look at companies like Hims; despite the controversy, they are an incredible company. They recently landed a historic deal with Novo Nordisk.

Hims has become a face of pharma because they communicate effectively and build a compelling story about how people can take better care of themselves. Traditional pharma is starting to catch up with direct-to-consumer channels like LillyDirect, but it’s still early.

Despite the conservatism of academia, we recognize that a shift is coming. We need to move from waiting for someone to have a disease before treating it to understanding preventative medicine. We can potentially predict what is going wrong with a body ten years in advance and intervene today.

That is our vision. While other great groups at Harvard, such as Vadim Gladyshev or David Sinclair, are doing excellent aging work, my North Star is understanding what eight billion people on this planet need and working toward that.

Dr. Jonathan Greenberg 00:15:52

Aging can sometimes rub people the wrong way. For a long time, the field has struggled because it is a very tough problem. We do not yet have a consensus on what aging is or how to characterize it.

If I may make a humble analogy, aging is very much like artificial intelligence before its recent rebrand. AI went through many generations as a theoretical concept, centered on the question of how to make computers think like people. There were many epochs of development where people questioned if the field would ever have a material impact.

We even went through an “AI winter” about 15 years ago, where it was seen as something of a fringe interest. Then we hit a moment where everything aligned. We had the compute and the architectures that were able to scale. Now, the impact of modern machine learning is incontrovertible.

Aging has suffered from similar issues. We haven’t had the tools, the understanding, or the data. Data is a huge factor. AI could not have trained these modern models without the internet, Wikipedia, YouTube, and Reddit.

We need more data and better tools to approach these problems. We need new perspectives. We are likely at the point where we can unlock this field. It can transition from something without a solid foothold to something everyone thinks about.

One indicator of this shift is that Eli Lilly, the biggest drug company in the world, is formally investing in aging. They are acquiring aging assets and putting their money where their mouth is.

In the next five years, this will become much more mainstream. We will see more effective ways to understand aging and intervene in longevity. It will become a topic people talk about with confidence, with less snake oil and more real impact. That is where we want to push the field.

18:55 How simulations of biology can accelerate drug development

Eric 00:18:55

You talked about the epochs of artificial intelligence and how machine learning moved from the world of science fiction toward the realization of programmable, commoditized intelligence. We are seeing the beginnings of that takeoff today.

When it comes to aging, there is a fundamentally different challenge with programming biology. Feedback cycles are rate-limited by real-world checkpoints rather than strictly technological innovations. It takes time to turn a hypothesis into an experiment and get data back.

How do we move past the idea that programmable biology is just a meme that will take several decades to realize? What can we work on today to make the vision of programmable biology a reality?

Dr. Omar Abudayyeh 00:20:00

One way to speed up the cycle is to simulate biology. If you break drug development into two phases—preclinical and clinical—the process can collectively take 10 to 15 years.

The preclinical phase comprises the science of target identification and the pre-IND work required to demonstrate that an asset is safe. The clinical stage involves phase one, two, and three trials. We should consider if we can break each of those phases into simulatable constructs.

For science and target identification, we could create virtual models of cells to run in silico screens of every known small molecule or conduct genome-wide CRISPR knockout activation screens.

If you are looking for a better drug for Alzheimer’s and you have the right data, you could run those screens in silico. Instead of looking at billions of things experimentally, you could test the ten best candidates in a single month. You feed those results back into the model and quickly identify a drug candidate.

If you add automation where AI actually runs the experiments, you could compress years of work into months. Furthermore, if you had virtual tissues, mice, and monkeys, you could do much of the pre-IND work computationally to show something is safe. This would further filter the funnel and compress timelines.

We could do a similar thing with clinical trials. One reason trials move quickly in China is that they have a larger population, which makes recruitment easier. If we had virtual humans to run virtual trials, we would only need a handful of humans to validate the in silico findings. This would provide much higher statistical power because of the simulations behind it.

I have heard officials from the FDA and other countries envision a future where trials could be conducted entirely in silico. If you show a drug is safe computationally, you could bring it to market and then collect real-time data in a phase four setting to evaluate it.

The current bottleneck for this vision is the lack of data in biology and medicine. We need better data to build these models. That is a primary area of focus today.

If we can build these models and showcase these workflows, it could completely change how we think about drug development. It would look more like a tech company, where you might only need one or two years of work to build a product and bring it to market. That would be incredible for the world.

Dr. Jonathan Greenberg 00:23:33

Data is an accelerator for development. Understanding more about patients and humans as they take interventions allows us to understand how to improve healthspan and lifespan.

Dr. Jonathan Greenberg 00:24:00

If I were to run a naive longevity trial, I would develop a drug, run the trial, and wait to see if the subjects’ lifespans increased. That is an incredibly hard trial to run, and the FDA doesn’t recognize aging as an indication.

However, if we frame it around how we affect people in specific ways—such as increasing muscle mass, improving DEXA scans, or increasing VO2 max—we can find better indicators. We now have amazing tools like epigenetic age and plasma proteomics to predict health and age. These markers allow us to see changes quickly.

If a drug is super-rejuvenative—like a “Tuck Everlasting” drug where you take it and immediately look ten years younger—it’s easier to run these trials. But even without that, better data collection fed into the models Omar outlined allows us to get rapid feedback on these approaches.

We could take existing interventions like metformin, SARMs, ACE inhibitors, or SGLT2 inhibitors and observe their multimodal effects on the rate of aging. We can see if they quickly change these features and how that translates into clinical manifestations and how people feel.

In an ideal scenario, we have the ability to simulate and use data to find a signal very rapidly. This provides an optimistic path to test new interventions for longevity in broader populations for safety and efficacy. This will be a huge unlock for discovering and iterating on ways to extend healthspan and enhance the human experience.

26:49 The data that will make biology readable by machine intelligence

Eric 00:26:49

There are different languages we can use to describe biology. Traditionally, medicine and molecular biology relied on brute-force macroscopic views derived from microscope slides, physician visits, or molecular levels from a blood draw.

In reality, biology is a complex, interconnected system. The ability to decipher huge troves of multimodal data allows us to uncover non-obvious signals decades before the onset of disease. This requires rethinking biology from a human language to a machine language.

This is the intersection of AI and life sciences. We have to reimagine how we collect data so it becomes native to the way we unravel the complexity of biology, deciphering digital biomarkers through the machine language of life sciences.

Dr. Omar Abudayyeh 00:28:00

It’s not just about collecting the right data, but discovering new types of data. The UK Biobank has done a tremendous job by collecting metabolomics, proteomics, epigenetic data, and wearable data.

Wearable data is a really underappreciated segment. One of the challenges is that large datasets are often not collected in tandem with other types of data.

The UK Biobank is unique because you can see EHRs alongside lab results, proteomics, metabolomics, genomics, and wearable data. Seeing those multimodal connections is where AI models really thrive.

Eric 00:28:47

Your lab has gone through several epochs. It started as a genetic engineering lab, moved into protein engineering and RNA editing, and now focuses on the intersection of aging, biomarkers, AI, and futuristic longevity.

What is the core vision of your lab? How did you transition from genetic engineering to where you are now?

Dr. Jonathan Greenberg 00:29:27

The vision remains consistent: we want the ability to program, simulate, and intervene in biology. Early in our scientific careers, we both worked on bioengineering and tools.

I focused on proteomics and mass spectrometry, while Omar worked on peptide engineering and sensing. Our work is about manipulating systems to create interventions and collect data.

Dr. Jonathan Greenberg 00:30:07

One amazing thing that emerged during graduate school is that we were working on ways to generate large amounts of data. Genome editing tools like CRISPR enable screening and genetic methods to produce vast datasets.

As we performed more screens, we figured out how to build virtual cell models. This is why our lab produces these models; there is a clear connection between the data and the modeling. Human data is a great complement to this cellular data. We can gather information from sources like the UK Biobank and obtain tissue-level and histology-level information.

Now that we have the ability to generate that data, we need to build models that have utility and identify what new data we need to collect. One great value of building a model is that it identifies the lacunae in your data corpus. This points us toward building new tools to measure or collect that data, such as finding ways to measure plasma proteomics with much higher throughput.

It is all connected to the concept of what we can fundamentally engineer and build to achieve the long-term vision of programming biology. This effort is supported by how the field has evolved. Machine learning and the ability to perform complex simulation analysis didn’t exist ten years ago.

We are standing on the shoulders of giants, and we are taking advantage of these advancements to build systems and iterate. While we are currently focused on building models, we are also searching for new tools and data sources to improve them. We are agnostic about the specific tools, whether they involve proteins or measurements.

You have to keep an open mind. You can’t just stick with protein engineering because you have done it for a decade. You need to follow the North Star of what we are trying to accomplish and figure out what to build to get there. That is one of the guiding messages of our lab and our work.

33:00 Defining aging & thinking about what the customer actually wants

Daniel 00:33:00

We would love to get into the specifics of your aging research and the tools you are building in that space. Before we dive into that, aging can be a murky topic. In the mainstream, aging is like water for a fish; it is all around us and almost fades into the background.

Some people don’t even think of it as a specific process. They just see different events happening, like cancer or heart disease. However, for those working in longevity and aging biology, there is a distinct phenomenon to understand. How do you conceptualize aging, and how would you define it?

Dr. Omar Abudayyeh 00:33:41

That is a good question. There are different ways to think about it, but you need to understand both the systems level and the organ-specific level of aging. You have to identify where things go wrong and what ultimately matters.

To use a Jeff Bezos analogy: what does the customer actually want? People care about what they feel. That can include a loss of energy, focus, or memory, as well as sleep problems, muscle health, weight, or mobility issues. It also includes actual diseases.

Many people don’t realize that hundreds of millions of individuals suffer from liver or kidney fibrosis in older age. There is also dementia, skin problems, and muscle loss. In the lab, we consider each of these phenotypes and what would be most useful for helping people live healthier, more productive lives.

At the scientific level, we look at both systems and specific organs. At the systems level, chronic inflammation affects everything. It is a driver of fibrosis, atherosclerosis, heart failure, and dementia. You can target those systems-level mechanisms to have a broader effect.

We have chronic inflammation models and screens in the lab to understand how macrophages and monocytes age. This helps us better predict and develop interventions. We also look at tissue-specific areas, such as skin, muscle, and neurocognitive screens for memory and focus.

We even look at how to enable better sleep. It would be amazing if you could sleep only four or five hours but feel like you slept eight. We have a spreadsheet that breaks down everything people care about to determine where there is the most interest for a potential product.

Daniel 00:36:29

Regarding specific intervention points, you mentioned intervening in chronic inflammation at the systems level or focusing on tissue-specific levels like sleep. Sleep is so high leverage and it degrades with age.

I have often wondered what would happen if we could fix sleep in the elderly. Would that provide massive downstream benefits or solve several issues related to aging? Do you have hypotheses about where the highest leverage points for intervention will be?

Dr. Omar Abudayyeh 00:37:04

If you are sleep-deprived or have low energy, you will likely be more stressed. Stress increases inflammation, which can affect your lifespan. That is one angle, but it also has a significant effect on your general ability to function.

If you could give people the ability to sleep four hours and feel like they slept eight, you are giving them back four hours of their life. You spend a third of your life sleeping, so that is a form of life extension in itself. There are many different axes for the benefits you can achieve here.

Dr. Jonathan Greenberg 00:37:44

Sleep and cognitive performance are fascinating areas that have been rightfully difficult to study. We have no great ways to measure sleep. You can use trackers or specialized beds to measure duration and quality, but we don’t have a great way to measure how rested you actually are.

You could perform a battery of cognitive tests, but it is unclear if those results reflect acute or chronic states. Self-reporting for sleep is also incredibly unreliable. If you were to invent something that allowed someone to sleep four hours less, how would you ensure they function as well or better, both acutely and chronically?

You don’t want someone to be high-performing in the short term only to suffer neurodegeneration in twenty years. We must also ensure we aren’t simply inducing insomnia. We need better ways to measure these outcomes.

Dr. Omar Abudayyeh 00:38:43

Right.

Dr. Jonathan Greenberg 00:38:44

Better measurement methods are emerging. We can look at biological signals from wearables, epigenetics, or plasma proteomics. We are seeing significant progress toward better proteomic signatures of sleep deprivation.

The ability to measure these things allows us to understand how to intervene, whether in animal models for discovery, in human trials, or through prospective studies. We want to collect and understand more of this data, which then feeds into the ability to build in silico models of sleep.

Dr. Omar Abudayyeh 00:39:34

Right.

Dr. Jonathan Greenberg 00:39:35

Cognition is a similar challenge. It is very hard to measure acute performance, and it is not one-dimensional. Many different aspects of cognitive function work in both the short and long term. As these tools mature and we develop new ones, we can start to think about how to intervene and improve quality of life.

Sleep is tremendously correlated with cognitive function as we age. One of the biggest factors in aging is whether you socialize enough, and if your cognitive function declines, you will socialize less. This creates a problematic feedback loop. We want to intervene to break those unwanted cycles.

As we improve our measurement capabilities and run more trials—such as randomized controlled trials on whether a cognitive enhancement actually has an effect—we will see huge impacts. The field of nootropics has historically suffered from poor science because we lacked the means to measure and intervene effectively.

As we develop these tools, we will take areas that have been “fluffy” and crystallize them into rigorous interventions. Twenty years ago, weight loss was limited to Weight Watchers or amphetamines. Now, we have reliable and diverse targets for obesity. These other fields will follow that same path.

Lilly’s acquisition of Sentessa is a prime example. Orexin agonists will likely have a material impact on sleep and may become sleep-enhancing drugs. We need to know the acute and chronic effects, and we need better ways to measure them. These fields are moving toward real science, and we need to push them in that direction.

Interventions in sleep and cognition will fundamentally change how people age and how even younger people function. That is the goal.

42:26 New biomarkers and the future of aging interventions

Eric 00:42:26

I’ll touch on a broad principle inspired by that. I’m reminded of a paper from James Zou’s lab at Stanford, published earlier this year, regarding sleep foundation models. They collected data from around 65,000 patients and looked at polysomnography (PSG) data from their sleep.

They mapped that PSG data against long-term health outcomes and found it was highly correlated—in the 0.78 range—with various diseases of aging, including Alzheimer’s, chronic kidney disease, heart failure, and strokes.

This speaks to the fact that signals of future health or mortality are embedded early in life within the latent space of multimodal data. Whether it is sleep, heart rate, or proteomic data, the information is there.

The grand mystery of biology today is not which questions to ask, but how to feasibly collect that data and apply it to modulate a person’s health trajectory. We already have converging evidence for therapeutic interventions like GLP-1s and Orexin agonists. Now, the question is what else we can develop to modulate biology in a meaningful way today.

Dr. Omar Abudayyeh 00:43:57

There is so much data you can collect about a person, and it is super counterintuitive how predictive that data can be. I’ll give you an example. We had a paper published last year where we built biological age and mortality models on top of face images.

We scraped tens of thousands of celebrity, Wikipedia, and politician images where we knew when the person died and the age when the photo was taken. Those models were super accurate. You can get 90% accuracy for biological age and literally predict the years they have left in their life. It is all from signals you get from their face.

This data is predictive of cancer risk and all sorts of different diseases. Voice is another interesting one. Voice can predict your diabetes status because there is something about the muscles in your vocal cords that connects to metabolic health.

I think we should be collecting this every day, building models, and leveraging it. The FDA is thinking about bringing in hundreds of new types of biomarkers and readouts for trials, and machine learning models should be considered part of those markers. A model that measures sleep data to predict disease risk could be a new type of biomarker for reading out a trial.

This connects to aging because we have all these biological clocks for biological age, organ age, and rate of aging. These should become biomarkers for trials that we can actually rely on. When that happens, we will be a step closer to building interventions for general aging phenotypes.

Eric 00:45:48

That answers our earlier question about how to compress the duty cycles of life science intervention and biological engineering. It’s by getting to these intermediate readouts like aging.

Dr. Omar Abudayyeh 00:45:59

Absolutely.

Daniel 00:46:01

On that point of having readouts for aging, you mentioned many different biomarkers, such as people’s faces or their voice, that can indicate health status.

Currently, many people default to methylation clocks as a way to measure aging. However, methylation clocks have a lot of drawbacks and are not really well understood. Where do you see methylation clocks sitting within the toolkit of biomarkers for aging?

Dr. Omar Abudayyeh 00:46:32

Jonathan loves methylation clocks, so I’ll let him take that.

Dr. Jonathan Greenberg 00:46:36

I love methylation clocks so much. They have been the historical clock and founded the concept that we can measure aging, the rate of aging, and the aging of different subsystems.

As you said, we don’t really have a good sense of interpretability. Despite immense data from people like Steve Horvath, who has done really great work, we don’t have a great mechanistic concept even across species.

There are really cool things you can do with methylation clocks. Work from True Diagnostic is amazing because they showed that using this methylation data, you can actually impute many typical blood tests. You can impute vitamin levels, inflammation levels, CRP, testosterone, and exposure to different toxins or plasticizers.

You can actually buy that test commercially. Instead of having to use one of those phlebotomy companies like Function Health or Superpower, you can do a simple blood draw with a Tasso and get all those imputed values. While they aren’t exactly the same, there are some very cool aspects to it.

All that being said, I think methylation is going to be overshadowed by proteomic tests. The Biomarkers of Aging Consortium runs these comparisons every year, and it looks like the proteomic tests are actually more predictive than the methylation-based tests.

They are also much more intuitive. I understand what a protein biomarker is doing. If your IL-6 goes up, you know what that is; it’s not some random SNP. We love proteomic tests. They are also much cheaper than some epigenetic tests, and you can get very cool data out of them.

There is work showing you can impute tons of diseases, inflammation, and your VO2 max. If you run an analysis, you can get your grip strength from your proteomics test pretty easily. It is well correlated, and grip strength is a major aging marker along with VO2 max.

If you look at the UK Biobank, they have 50,000 proteomic samples and are expanding that further. It is an incredible resource to be able to look into health. We view this new technology as something that can complement existing technologies in a much more understandable way.

Finally, we think about inflammation and interventions. There was the IL-11 story from Stu Cook’s lab about two years ago. IL-11 is an inflammatory interleukin that has been studied as a mechanism promoting fibrosis.

In a landmark paper, they showed that if you block IL-11 with targeting antibodies, you can extend the lifespan and health span of mice. Could we discover that with plasma proteomics?

The short answer is that IL-11 isn’t in a lot of the plasma proteomic panels because it’s hard to measure. But there are other biomarkers to discover using these approaches. We want clocks that can tell us what interventions or lifestyle changes we could implement. I think epigenetic clocks, and especially plasma proteomic clocks, are moving in that direction.

Daniel 00:51:12

Regarding the value of proteomic clocks versus methylation clocks: in my mind, proteomics are downstream of epigenetics. The epigenetics determine which proteins are being expressed.

Is there just too much noise in the process from methylation to proteomics to make the direct epigenetic signature useful for this purpose?

51:15 The rise of proteomics as a health & aging biomarker

Dr. Omar Abudayyeh 00:51:41

Proteins are a more direct reflection of the biology occurring inside a cell. They map the active pathways to the engaged cell circuits. With plasma proteomics, you are looking at thousands of proteins secreted from every organ in the body, which tells you exactly what they are actively doing in real time.

You can deconvolve that data back to the underlying biology. To me, that offers a much higher-resolution picture of what is happening with your body. Because you can see those proteins and the associated genes, you can connect them back to specific pathways and interventions.

Methylation doesn’t change as quickly. You might miss acute developments within the body that proteomics would immediately reflect. There is significantly more power and resolution in this approach. The UK Biobank is even releasing 500,000 proteomic signatures, covering almost the entire biobank.

Dr. Jonathan Greenberg 00:53:16

The link between CpG methylation, gene expression, and protein expression is incredibly complicated. We don’t yet fully understand how specific CpG sites affect expression directly.

When you calculate your epigenetic age through a clock like TruDiagnostic, you aren’t getting a universal epigenetic age. If you do a Tally Health test or a buccal swab, you are getting the epigenetics of your fibroblasts. If you get a blood draw, you are getting the epigenetics of your PBMCs.

You can’t easily get the epigenetic age of the neurons in your brain because most people wouldn’t want to undergo that biopsy. You can’t abstract this away because every tissue and cell type in your body has different patterns of methylation. They have different gene expressions, just as you can see post-mortem across different tissues.

While Steve Horvath or Morgan Levine have shown how different tissues age, you can’t sample those in living people. At the end of the day, you won’t be able to access most tissues in your body to get their epigenetic age. There are correlations, but they aren’t direct.

Blood is not just one big homogeneous soup, but it is—

Dr. Omar Abudayyeh 00:55:04

—a little bit more.

Dr. Jonathan Greenberg 00:55:04

You can sense where these different features are coming from. Hamilton Oh’s work from Tony Wyss-Coray’s lab is a classic example of how we can identify where proteins are being expressed and map that back to specific tissues. In that way, we are getting a much more accessible measurement than epigenetics.

As Omar said, it is also much more biologically interpretable. The 2024 biomarkers of aging competition showed that proteomic clocks were beating GrimAge v2. There is more signal there, and it is cheaper and easier to measure than an Illumina BeadChip.

56:01 Building the virtual cell

Eric 00:56:01

Let’s wrap this back into the idea of the virtual cell. The virtual cell is just one layer of the hierarchy of biology. You can think about a virtual human, a virtual society, or even a virtual protein interaction map.

How do we build the intersection between these data collection concepts and interweave them with AI analytics to build virtual models of biology?

Dr. Omar Abudayyeh 00:56:30

With the virtual cell, we’ve been exploring two different routes. One involves gathering various types of data and aligning them into a similar embedding latent space to perform useful tasks.

We have focused on single-cell RNA sequencing because it is the largest dataset available, with hundreds of millions of data points and researchers now aiming for billions of cells. These models are in their early stages, similar to the early days of protein folding.

While currently inaccurate, there is significant work underway to make them performant and to combine them with metabolomics, epigenetic information, proteomics, and genomics. The other route is teaching language models how to understand this data and then using reinforcement learning to make predictions.

Typical virtual cell models are difficult to interact with; they simply spit out a vector of gene expression. In contrast, language models have biomedical literature baked into them, allowing for a conversational interface where you can see the reasoning behind a prediction.

This is very useful for understanding why a prediction was made and the level of confidence in it. We convert genomic variant data, methylation, proteomics, and single-cell RNA sequencing into text to teach the model.

This approach is quite performant. The model can explain that if a certain gene is turned on, it affects specific pathways and gene expressions, resulting in a new cell state. This allows us to leverage human knowledge in making predictions in a way that was previously difficult.

Whether using embedding space models or reasoning models, the goal is to build a world model that represents this data in a jointly embedded space and can move through that space across any axis.

In robotics, Yann LeCun has been leading this with models like JEPA. These models embed various data types, like video or motor data, to think through time and determine the next state. We can think about a human in the same way.

By putting proteomics and genomics into a joint embedding space, we can simulate long-term health outcomes. We can simulate thousands of scenarios, such as the effect of a new drug on a cell or a liver, to build a comprehensive world model of simulations.

It is still early, but seeing the current performance levels makes me confident we will get there.

1:00:40 Disrupting and democratizing the supplements industry

Daniel 01:00:40

How do you envision the tools you’re building will eventually impact the consumer health market?

How do you see your work leading to medicines and other interventions reaching the market at a faster pace and with more impact than we’ve had in the past?

Dr. Jonathan Greenberg 01:01:04

The ability to nominate interventions quickly, whether through drug repurposing or identifying natural metabolites that hit specific pathways, is a major accelerator.

While we still need to run clinical trials and collect data before marketing a product, these in silico methods allow us to understand safety profiles and move more rapidly into phase one or phase two trials.

There is a clear demand for these interventions. Looking at GLP-1s, people are buying them from compounding pharmacies because the demand for weight loss, muscle mass, and improved sleep is so high.

If we can find effective interventions using these methods, there is an incredible opportunity to translate them quickly. On the diagnostic side, people want better ways to measure their health and aging.

Companies like Function, Superpower, Inside Tracker, Hims, Whoop, and Oura are already providing blood testing and data that help people understand their bodies better than a primary care physician might.

If we can create better tests with solid scientific backing and communicate those results clearly, there is a massive revealed desire for that. These approaches will help translate those desires into reality.

Dr. Omar Abudayyeh 01:03:43

Virtual cells, virtual tissues, and virtual people are incredibly useful for drug development. One less appreciated area, especially among scientists, is the supplement space. Jonathan and I both love that space. We take many different things to optimize our health, whether it’s fiber or supplements for better focus and energy.

The issue with the supplement space is that it is often rich with scams, fraud, or products that aren’t real. Virtual models and world models can help bring substance to that industry. There are tens of thousands of natural products, Chinese herbal products, and generally recognized as safe (GRAS) molecules.

On their own, these molecules can be useful, but in combination, they can have synergistic effects. It’s traditionally very hard to apply rigorous science to that space, but world models can accelerate and democratize that process.

This is something we’ve been looking at: where could you screen millions or billions of combinations of these compounds to find new phenotypes to test? We’ve been doing this in the nootropic space.

For example, we’ve been trying to bring a product like Rune to market for better energy and focus. It’s a combination of four actives that work together to elicit a stronger energy, focus, and mood phenotype. This is an interesting angle where we can add more science to the consumer health space in a way that hasn’t been done before.

Daniel 01:05:26

Would you like to tell our audience a little bit more about Rune? Where can they go to learn more or purchase some?

Dr. Omar Abudayyeh 01:05:34

Sure. Jonathan, do you want to take this?

Dr. Jonathan Greenberg 01:05:37

You’re better at ad reads than I am.

Dr. Omar Abudayyeh 01:05:41

We are fans of products like Zyn. The whole nicotine space has been interesting to watch. Palantir just announced they’re putting Zyn vending machines in their offices to boost worker productivity, which is amazing to see.

However, nicotine does have downsides. It can cause vasoconstriction and have various cardiovascular effects. We were interested in whether we could recreate a Zyn-like product using safe compounds.

Rune is a Zyn-like pouch, but it doesn’t have nicotine. Not only do you feel a buzz of energy within the first ten minutes, but you get three to four hours of lasting energy and focus. It really allows you to lock in. You can check it out at takerune.com.

Daniel 01:06:32

I’ve been wanting something for a long time that I could take in the evening to give me more productivity and focus without impacting my sleep. Caffeine doesn’t work for that. Would Rune work for that purpose?

Dr. Jonathan Greenberg 01:06:43

If you take a small amount, perhaps. But we tuned the formula specifically to get you ramped up. This is something you should take early in the day to lock in and get a lot of productive work done.

We’re always exploring and iterating on the formula, but I would advise taking it earlier rather than later, especially if you’re sensitive to caffeine.

Daniel 01:07:14

We have about 15 minutes left. How are you guys feeling on energy? Are you tired, or should we keep chatting a bit more?

Dr. Jonathan Greenberg 01:07:22

Well, we took Rune! I’m not kidding, this has been very fun. The way the field is moving and the way you guys have directed this conversation is super exciting. I could go for another hour.

Eric 01:07:43

I can’t, so we have to stop the interview at some point.

Dr. Jonathan Greenberg 01:07:46

We’ll mail you some pouches.

Daniel 01:07:49

I would love to try it.

1:07:51 Genetic engineering, small molecules, peptides... what are the right tools for aging intervention

Eric 01:07:51

I think we should go back to genetic programming, because that’s where a lot of your lab’s work started. I’m curious to hear how you think the concepts we’ve been talking about today intersect on a ten-year timeline.

Can we deconvolute biology to a point where we understand at the genomic level the specific mechanisms by which we develop disease and age? Can we identify those mechanisms and find short-term interventions that partially reverse the course of disease?

More interestingly, can we actually go in and genetically program disease out of our trajectories altogether? Is that something that will happen in the next ten years? How do you view the intersection of genomic engineering, longevity, and AI?

Dr. Jonathan Greenberg 01:08:40

If I can push back on the assumption, why do you think genetic programming is the best intervention for this?

Eric 01:08:48

I don’t necessarily think it’s the best.

Dr. Jonathan Greenberg 01:08:49

We are very intervention-agnostic.

Dr. Omar Abudayyeh 01:08:52

Right.

Dr. Jonathan Greenberg 01:08:52

Genome editing and genetic programming make perfect sense in specific areas. If someone has a congenital defect or an inherited metabolic defect, these tools are a perfect match for treating and changing those conditions.

For longevity, we need to find the right tool for the right job. We need a toolkit of interventions, whether that involves nucleic acid genome editing, gene therapy, small molecules, antibodies, peptides, or RNA interference.

We should view the space of potential fixes holistically. If an ideal tool doesn’t exist, we look for the closest alternative and iterate from there. We aren’t biased toward one specific modality.

There is significant promise in using genome editing to engineer cells that target senescent cells, such as the uPAR targeting work from Scott Lowe’s lab. However, you have to think across all possibilities.

Personally, I am very focused on small molecules, though I also value antibodies and peptides. You really need to consider the target and the biology involved rather than trying to over-engineer a solution.

Engineers often have a bias toward over-engineering, but simplicity is key. A solution should be only as engineered as it needs to be.

One project I am very excited about involves Alzheimer’s disease. We are investigating whether we can change all relevant cell types in the brain to ApoE2 and install other protective mutations. While that is a longer path, it is a very exciting and practical direction. You have to pick the right tool and iterate as quickly as possible.

Dr. Omar Abudayyeh 01:11:52

I have two other issues with using gene editing for longevity. First, many mutations are double-edged swords and are not always well understood.

A classic example is a mutation common in people of African descent that increases resistance to malaria but also increases susceptibility to HIV. Correcting a known genetic mutation that causes a rare disease makes sense, but adding new mutations into heterogeneous genetic backgrounds is different.

You won’t fully understand the consequences of augmenting what is considered “normal.” For example, some South Asians have PCSK9 mutations that lower cholesterol, but that exists within their specific genetic background. We don’t know exactly what that mutation would do in a totally different genetic background, yet people are already trying to apply it broadly.

My second issue is the “iPhone 1” problem. Imagine if you were given the first iPhone in 2008 but were stuck with it for the rest of your life. While everyone else has an iPhone 17, you are stuck with ten apps on a device that barely works.

That is the equivalent of using mutations for longevity. You are locked into a specific modification, but what happens if we discover something better? You would have to get another genetic mutation to upgrade.

Look at GLP-1 drugs. I used semaglutide for a few years, but then I upgraded to Zepbound, which is a dual agonist. It’s amazing, and now there is a triple agonist coming out that I can’t wait to try. I don’t want to undergo genetic engineering every time I want an upgrade.

1:13:45 How longevity pilled is everyone here?

Daniel 01:13:45

We are all clearly longevity-pilled here, but I’m curious just how far that goes. How old would each of you like to live to?

Dr. Jonathan Greenberg 01:13:57

It depends on what my life would be like. I don’t want to get into “Methuselah” territory with this question.

Personally, I don’t have strong priors on how long I will live, but it would be great for humanity to live long, fruitful lives into their hundreds.

My North Star is reducing the chronic diseases of aging that cause so much suffering. Diseases like dementia and neurodegenerative conditions are massive issues. Eliminating those or reducing their severity in the population would be a huge achievement.

Dr. Omar Abudayyeh 01:15:03

I’ll quote someone I look up to, John D. Rockefeller. He was once asked how much money was enough, and his answer was, “Just a little bit more.”

Daniel 01:15:15

Jonathan, you gave a pretty good answer, but Omar gave the correct one. You’ll receive your prize after the podcast.

On the Free Radicals podcast, we are big fans of total biological control and indefinite life extension, provided we remain healthy. Life is good and we enjoy living. We’d like to keep it going until it’s no longer good.

Dr. Jonathan Greenberg 01:15:43

I won’t debate that.

Daniel 01:15:46

Is there anything else you’d like to leave our audience with? Would you like them to check out your lab website, Rune, or anything else?

Dr. Jonathan Greenberg 01:15:59

Check out both of those. We are excited about the field and we are starting to release a lot of new work, so keep your eyes peeled.

You can follow us on X or follow Omar’s TikTok. This has been really fun.

Daniel 01:16:19

Thank you both for joining us on the Free Radicals podcast. This was a really fun conversation, and we’re very excited about the great work you’re doing.

Dr. Omar Abudayyeh 01:16:28

Thanks for having us.

In this episode, we discuss how Dylan gets bipartisan support for healthy life extension initiatives, what it’s like selling politicians on longevity, and A4LI’s role in accelerating longevity towards its ChatGPT moment.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

1:52 Origin story of the first longevity lobby
10:30 State of government support for longevity
14:49 Trump administration’s view on longevity
17:30 The unique policy needs of longevity biotech
23:42 A4LI’s success with right to try
30:30 Why lobbying is so effective for accelerating longevity
34:32 Case for optimism in longevity & the need for more investment
40:04 What arguments resonate with congresspeople
49:23 Bryan Johnson’s impact on the perception of longevity
52:15 Who are the most important people to influence

Transcript

1:52 Origin story of the first longevity lobby

Daniel 00:01:52

Dylan Livingston, welcome to the Free Radicals Podcast.

Dylan Livingston 00:01:56

Thank you so much for having me, guys. I’m excited to talk.

Daniel 00:01:58

Dylan, you’re 28 years old and you run a lobbying group for longevity. Can you tell us how you got here?

Dylan Livingston 00:02:05

Yeah. I think that’s part of the surprise when I reach out to people about getting involved with A4LI. They ask, “Why are you a 28-year-old?” I’m 28 now. When I started this, I was 23. So, at 28, I’m a little older now, making it less shocking. When I first started at 23, it was even more shocking to a lot of people why a 23-year-old was so passionate about the ideas of longevity.

However, I think seeing a younger guy advocate for things that most people don’t start thinking about until they are 40, 50, 60, or 70 shows that this issue is not age-dependent or part of life-dependent. It’s something that everybody should be invested in, thinking about, and pushing for a better future. It has definitely helped.

I’ll give my background quickly. I was first introduced to the idea of longevity, or “defeating aging” as it was called back then, when I was about 12 years old. It was just about the time I finished the Harry Potter series. My father went to a regenerative medicine conference. He runs a broker-dealer on Wall Street, investing in emerging technologies. He came back and showed me a video of a guy he met. I had just finished Harry Potter, and he showed me a video of Aubrey de Grey speaking about potions to defeat aging. I thought, “Who is this guy? What’s going on here?” I felt like I was dreaming.

That’s when I first discovered the field. I did an honors biology paper on it, but I tucked it away as I grew up. Then I got involved in politics during college. I went to school in southeast Pennsylvania and started college in 2016, when Pennsylvania flipped from a traditionally blue to a red state. The Democrats became very involved in organizing in Pennsylvania and recruited people my age to join the campaign because we finished college in 2020, which was the 2020 election.

That’s how I got my start in politics. I started working in politics throughout college, then worked for the Biden campaign in Pennsylvania after I graduated college. During the campaign and my senior year of school, everything was shut down because of COVID. I went from pitching in my senior year of baseball and doing my physics thesis to sitting in a basement with my grandfather, who was 92 at the time. I was walking on eggshells, putting cleaner on the groceries, and freaking out that we were going to kill Grandpa.

Life at that point seemed like a coming-of-age moment for me. It was a time when I was seeing my grandfather lose hope in life and the prospects of living longer, seeing his friends get COVID and die. The mental toll that took on him was significant. I got freaked out about my own aging and the prospects of aging for people in my family and the people I love.

I remembered back to when I was 12 and started becoming a fanboy of the longevity field while I was doing my phone calls and door knocking for Biden. I realized at the very end of my Biden time that this longevity mission is what I wanted to dedicate my life to. I think this is the most noble cause: helping people live as healthy as possible for as long as possible. I also noticed a constant theme from people like Aubrey de Grey and David Sinclair: the field is mature. We have companies, longevity companies, researchers, labs, and investors, but we don’t have any government action. So, at the ripe age of 23, I took it upon myself to make it all happen.

Daniel 00:06:07

Thank you, Dylan, for the backstory. Could you tell us a bit about the beginning of A4LI? You mentioned you started it when you were 23. You didn’t have lobbying experience before that. What did that first year look like?

Dylan Livingston 00:06:24

I had the idea mid-2021. I was trying to figure out what I was going to do after the campaigns, and I knew I was passionate about longevity. I was still living at home because of COVID, and I went to my dad. I asked, “What do you think about this idea for a PAC or a lobbying organization for longevity?” He said, “You should do that. That sounds like a good idea.”

I got the idea mid-2021. As I mentioned, my dad had some connections. He wasn’t super involved in the longevity space, but from years past, he knew people and where to start. He gave me a list of people. I went on LinkedIn and messaged about 1,000 people, saying, “Hey, let’s talk. I want to do this.” I found that if you bother people enough, they eventually return your phone calls.

As a side note, I do think the experience with Biden and the DNC was very useful for getting A4LI started. If I can go into rural Pennsylvania and knock on doors for undecided voters to get them to vote for Biden in the most contentious political election in our lifetimes, I think I can message people on LinkedIn and take a rejection if it happens.

That’s how I got it started. I eventually got connected to the right people. I started plugging into the Foresight Institute community and met people at Lifespan.io, connecting with players in the field. Then I asked people for money. We eventually raised enough money to officially launch in 2022 with activities. We did some polling in the beginning and started our initiatives.

The official start of A4LI was 2022, but that first half-year before we launched was dedicated to outreach and understanding what biotech CEOs and researchers actually wanted in a political advocacy organization. I had a general idea—obviously funding, expedited approval pathways, etc.—but the details of that were unclear to me. The details of why that is necessary were also unclear. That’s something the researchers and biotech people know the answer to. So I spent a lot of time that first “minus six to zero” period before we officially launched, understanding what people actually wanted to see out of an organization like ours, and then finding funding to hire a small team to get it going.

Daniel 00:08:57

When you were doing your initial exploration of the space, messaging everyone on LinkedIn, did you already have the hypothesis that you were going to create a lobbying group and wanted to figure out the angle? Or were you more broadly just exploring, figuring out what the leverage point was?

Dylan Livingston 00:09:12

I knew this was necessary, whether people agreed with me or not. I’ve had people say, no, actually not really. Most people understand that this is necessary. It’s not a hard sell.

You guys are doing all these things, and no one is talking to the people who regulate what you’re doing about what you’re doing. So I’ll go talk to them on your behalf. When you approach them as a supporter and an advocate, that’s really, to this day, obvious.

At the time, I was just a complete fanboy. I thought this stuff was so cool. The implications of a longer, healthier life for me personally, and for everybody, are just so fascinating. The universe is so big; I want to explore as much as I can. That’s my impetus for longevity. Also, I don’t want to get old. That probably doesn’t seem too fun.

Coming to them as an advocate and a supporter made it easy for them to be nice. They all wanted it. I also mentioned before that Aubrey de Grey and David Sinclair and all these people had specific talks. I didn’t even come up with this idea myself; I kind of did, but I heard other people vouching for it and asking for it. So there wasn’t too much pushback.

10:30 State of government support for longevity

Eric 00:10:30

Dylan, I’m curious to dig a little deeper into the current state of government funding for longevity. When you started A4LI five years ago, at 23, what was the state like then, and how have things changed since?

Dylan Livingston 00:10:47

Things have changed in good and bad ways since 2022. The topic of longevity, or aging, is far more integrated into mainstream discussions. There is generally support in the current administration. Part of my job is talking with them, and there is at least passive support for it. The goal of A4LI is to get their support in legislative situations.

We are starting to see that happen mainly in state governments. There is also interest now in federal initiatives, where there wasn’t even a year ago. When I talk to people on the Hill, this is usually the first time they’ve heard of this, or they’ve seen Bryan Johnson and ask, “You mean that guy Bryan Johnson on Twitter?” There are a ton of misconceptions. There are 535 congressional and Senate offices. You need to talk to a lot of people. It’s not just, “Here’s a packet and here’s a 20-minute pitch; do this.” They get that 100 times a day. It’s a relationship game here.

In 2022, we had no relationships. I’ll give a funny story. The first person we got on the Congressional Caucus was Congressman Paul Tonko. We got him first because of my grandmother, who was involved with his campaign in the early and mid-80s when he was running for local office. I went to a fundraiser of his and said, “Congressman Tonko, great to meet you. Do you know my grandmother Sue?” He replied, “Oh my God, yes, I know Sue!” That was where I started, because I didn’t know where else to begin. Building relationships, forging and building these connections, is part of the game. A constant flow of information about updates in the field is also needed. That’s what we’ve been providing, and it didn’t exist in 2022.

Now, one negative change is funding cuts, but that is NIH-wide and part of a broader reorganization this administration wanted to implement. Beyond that, we have two program managers in RPH now who didn’t exist four years ago. They are focused on the longevity mission, on crucial parts of it, and are from the longevity field. Andrew and John are people you’d see at a conference. It’s nice to see that the mindset is transferring to these positions in government.

Jim O’Neill, Dr. Oz, who spoke at our conference last year – these are all people who understand the issue. Unfortunately, politics is very messy; it is more of an art than a science. The appetite is there far more than it was in 2022. I would like to think that A4LI was a driving force of that.

The last thing I’ll say is the Congressional Caucus we put together has been growing. We didn’t have a base of support in Congress until 2023. Now we have a Congressional Caucus, which started with four members and is now at ten. There are metrics we can look to, like that, to say there’s a growth in interest in this field because we have people willing to sign their name and hear reports on the longevity field from the A4LI group.

14:49 Trump administration’s view on longevity

Eric 00:14:49

Thinking a little more about what specifically has changed under the current administration, especially considering key players like Jim O’Neill and Alyssa Jackson: What specifically have you seen that’s been promising in terms of this administration’s support for longevity, and what more could they be doing?

Dylan Livingston 00:15:13

Jim and Alicia are the two big bright spots, I would say. The general Maha movement aligns with what I call bucket one of longevity. I break it down into diagnostics and lifestyle in bucket one, and therapeutics and devices in bucket two. They are fully invested in bucket one: diagnostics and a healthier food-as-medicine mantra. This has made it easier for us to have conversations about bucket two because they are already bought into bucket one.

People like Dr. Oz and Kali Means are generally aware of this and interested. Jim and Alicia are probably the two big highlights here. Jim, who was Deputy Health Secretary under RFK, is now moving over to the National Science Foundation, which is equally exciting in my opinion.

A lot of foundational research is needed. I will never forget the first briefing we did as an organization. Matt Kaberlein spoke and showed a picture of the map from 400 BC that Greek shippers used. He said this map represents what we have in aging right now. We have enough to get around, trade, and know a little bit, but the true map is so much greater than what we currently possess. We need foundational research to better map the rest of the aging world.

At NSF, that could definitely be a priority. That’s very exciting. Alicia, as the head of ARPA-H and someone with a longevity mindset, can continue to fund projects like Prosper and Front. Those two are truly exciting. The field has expressed this; there have been articles about it. It is a very exciting development. Jim and Alicia are in significant positions. Jim’s term at the NSF is six years, so he will be there for a while, which is beneficial. He will hopefully be a mainstay in the government.

17:30 The unique policy needs of longevity biotech

Daniel 00:17:30

There are many existing relationships between pharmaceutical and biotech companies and the government. Can you say more about the unique needs of the longevity field that created the need for A4LI? What makes it so important that people like Alicia Jackson are in charge of ARPA-H?

What does it mean for these people to be “longevity-pilled”? Why does that matter? What is different about the longevity field, from a policy standpoint?

Dylan Livingston 00:18:00

Nice. We are throwing out the pills. I love it. This is so Gen Z of us. On a lighter note, I use “longevity pill” when talking to staffers on the Hill, and it works. People understand the concept of the pills; it is a good term.

Longevity in some ways breaks the current healthcare model. Preventative, proactive health breaks the reactive sick care model that we currently have. Specifically, aging therapeutics break the paradigm of receiving treatment only once you are sick. The vision for the longevity field is for someone to take a therapeutic at 40 that restores them to 30, for example, and continuously does that over time. Ultimately, they would never have to deal with any age-related diseases because they maintain the immune and cellular function of a younger person. That is the goal.

First, funding priorities must shift. If we can find a drug that targets aging effectively, this is clearly the more promising model for achieving meaningful healthspan gains. A number I always highlight is that if we cured all forms of cancer right now, healthy lifespan would only increase by about 2.5 years. This is shocking because the venture capital world for cancer biotech is around $20 billion a year, and the National Cancer Institute receives $9 billion a year in funding.

This is not a unique need; every field of research needs money. However, the uniqueness of longevity makes funding even more necessary. The amount of dollars that go towards longevity and aging research at the National Institute on Aging (NIA) is so fractional it is almost a joke. It is less than 1% of the NIH’s total budget, going towards something that is ultimately responsible for 90% of the things that kill people. It is laughable. In my ideal world, every area of disease and biomedical research would receive infinite money. But if there is a finite amount of dollars, or if new appropriations are to be made, they should probably be allocated to the area that will yield the most healthy lifespan gains.

Second, there is the entire FDA issue of how we actually measure whether an aging therapeutic is working. Conferences are organized around this topic. The Biomarkers of Aging Consortium holds the Biomarkers of Aging Conference at Harvard Medical School each year. This is something that needs to be solved if we are going to have proof that an aging drug actually works.

The natural way to determine if an aging drug works is by looking at a longitudinal study and seeing if a person lives healthier into their 90s and 100s, without disease until they are 100. However, you would probably have to start that person on the drug at 50 years old. Such studies are expensive and time-consuming, and our “Reels minds” do not like things that take too long; everything is short-term these days. So, the Biomarkers of Aging Consortium is trying to find surrogate endpoints that could be used in an aging clinical trial to give quicker feedback on whether a therapeutic works.

This is one part of it. There needs to be more foundational research to discover a set of biomarkers that the FDA and the industry can agree on. In the meantime, we should be incentivizing drugs that target multiple diseases, which is a proxy for the aging field. The implication of a senolytic is that it will affect multiple disease pathways. If we created a multimorbidity accelerated approval pathway, similar to a white paper we wrote for AAPLM, we could give these aging companies a leg up and get them to market sooner. This is not exclusive to aging companies. I do not know if I would consider a GLP-1 to be acting on the biological mechanisms of aging, but it very clearly has a multimorbidity effect.

Therefore, the second area of focus. First, the more important thing for this field is standardizing and discovering these surrogate endpoints, getting FDA and industry buy-in, and having them validated. Second, in the meantime or in parallel, we need to set up these accelerated approval pathways to get better, more effective multimorbidity drugs on the market as soon as possible. This is a longevity initiative. Even if it is not an aging initiative, the ultimate mission of the organization is to advocate for legislation that increases healthy lifespan. We still want to push initiatives that will ultimately increase that number.

23:42 A4LI’s success with right to try

Daniel 00:23:42

You described three key areas of policy that you are trying to advance. One is increasing funding for aging research, given the huge gap relative to other funding. The second is the validation of surrogate endpoints for aging that the FDA could approve for use in clinical trials. The last piece was expedited approval pathways. Are these the big three things you are focused on, or is there anything else in terms of policy you would mention?

Dylan Livingston 00:24:13

Those are the big three at the federal level. You could also consider other longevity initiatives like Medicare coverage for certain generics, age testing kits, or insulin devices. We are a team of three and a half people, so there is limited bandwidth for what we can push for.

The other significant initiative we have undertaken, and where we’ve had the most success at the government level, is advocating for Right to Try. 2023 was a significant year, marking the organization’s first true year of operation after 2022 was spent establishing ourselves. In Q1 of 2023, we launched our Longevity Science Caucus. By Q3 of 2023, we successfully passed a bill in the state of Montana that expands eligibility under the Right to Try Act.

Right to Try is a law passed by the first Trump administration. It states that if you are a terminally ill patient, defined as someone with six months or less to live, you have the right to access a therapeutic that is not yet fully approved. This is a sensible law. If I had terminal cancer, I would want to try every available option. However, in my opinion, it is not conducive to actually improving health outcomes. At the point of terminal illness, Right to Try has not proven effective. You haven’t seen any press about someone miraculously cured under Right to Try because no one has been. While there are isolated cases, it is not a common occurrence. My take is that once a biological system is in terminal decline, it is extremely difficult to halt that decline.

Therefore, people should also have the right to try not to become terminally ill, which means starting treatments earlier and being proactive with their health. Specifically, individuals being proactive with their health should consider drugs and therapies rooted in the hallmarks of aging. These address the foundational drivers of various diseases. If you’re taking something at 40, you wouldn’t take a cancer drug; you would take a senolytic or similar preventative therapy.

In 2023, we expanded Right to Try eligibility to include all patients, not just those who are terminally ill. Another impetus for this expansion was the prevalence of economic zones worldwide where people seek experimental therapies. These are often in places where I personally would not feel comfortable seeking experimental treatment. Let me share a story my mom doesn’t particularly like, but it illustrates the point. When I was in high school and started drinking beers, my mom would always advise: “If you’re going to drink, do it with your friends in the basement and be very careful. Don’t go into the woods, fall, cut yourself, and die from bleeding out as a 16-year-old.” The same logic applies here. If you are undertaking something risky, why would you do it in a risky environment? If I’m getting gene therapy, I want to do it in a place where I speak the language, and I’m confident about the medical standards, not in some questionable Tijuana clinic.

There’s that angle. My opinion is that at those Tijuana clinics, you don’t receive the best therapeutics because reputable drug developers, biotechs, and pharmaceutical companies will not participate in such environments. So, Montana’s approach elevates the standard of care for patients receiving experimental treatments. It also encourages better therapeutics to be provided. The longevity companies I work with will not operate in Tijuana, but they might consider Montana. That’s the other angle.

The ultimate idea behind Right to Try, and why we believe it’s a longevity initiative, is that this version specifically allows people to be proactive with their health. It also enables the gathering of data in a more decentralized way, which can inform investor and government decisions on better aging therapeutics. Even if it’s not a standard Phase 2 clinical trial, if a company generates a data set from 30 people who received their drug under the Right to Try regime, and there’s positive data, I would be shocked if an investor or governing body didn’t take note.

It’s a way to generate more data. It also provides an avenue for biotech companies that might have raised a $10 million Series A for a Phase 1, showing promising data, but are unable to raise the $60 million needed for a Phase 2. This gives them another option to generate more data without significantly impacting their finances, as they can charge people for these therapeutics. This is completely optional. It targets individuals who would likely seek experimental treatments at places like Tijuana clinics anyway, but now they can come to Montana and pay for it there. It’s a way for these companies to continue their runway, generate data at no cost to them, and give smaller biotechs a chance to bring something truly promising to the market. I see many major benefits of this Right to Try regime for the longevity industry and the biotech industry in general.

However, that’s not necessarily longevity-specific. The big three initiatives remain what I mentioned earlier. This is more of a passion project on the side.

30:30 Why lobbying is so effective for accelerating longevity

Daniel 00:30:30

We’ve discussed some of A4LI’s specific achievements, the relationships you’ve built with the government, and your future initiatives. I’d like to take a higher-level view and discuss the strategy and theory of change. When aiming for change, there’s government policy, popular support, and private investments in the field. Is there a flywheel effect at play? How do you view the relationship between all these factors, and what is the most effective way to apply leverage?

Dylan Livingston 00:31:03

A lot of people in the longevity field talk about the ChatGPT moment. It’s hard to get buy-in for a social movement around something that doesn’t really exist. Unfortunately for the aging field, we don’t have a drug or even a device at the validated level of ChatGPT. We don’t have that ChatGPT moment yet. It’s hard to make a social movement around something that doesn’t exist yet.

A4LI’s approach recognizes that a social movement requires general public buy-in. That’s a lot more people needed to buy into an idea. Ultimately, the point of a social movement is to change policy and government structures. You can straight-shot it by creating a 501(c)(4), skipping the social movement aspect, and lobbying the government. This involves selling a group of 538 people on something rather than 330 million.

There are many advocacy organizations doing great work on the social movement side. When I looked at the field, I saw other groups doing that. I didn’t want to be duplicative. However, no one was doing this. I was, and still am, the only person lobbying for the longevity field in DC.

To me, it’s the highest leverage place to put money and time. The value of getting one member of Congress versus one person in the general public on board with this is so much higher. Every congressperson is around 70 years old and wants to live longer, just like everyone else secretly does, even if they don’t admit it. They are people too. There are fewer of them to get on board, and they hold tremendous power over what we are ultimately trying to influence: policy.

That’s my take on it. I don’t think we will get that social movement until we have a ChatGPT moment. Fortunately, the work we do as a lobbying organization and the initiatives we are pushing can bring us closer to that ChatGPT moment. Once we get that ChatGPT moment, we will see what happened with AI. Nobody was talking about AI; it didn’t even exist in 99% of people’s minds. Then an amazing product came out, everyone started talking about it, the government responded, took action, and is now investing heavily in AI. This leads to better AI, creating a self-serving cycle.

It will be the same when we have that ChatGPT moment. We just need to achieve it. A4LI is doing the work to straight-shot that as much as possible. It will be someone from the private industry who does this. Hopefully, it’s someone soon, someone we know. Hopefully, one of the companies gets it done. There are enough companies out there that I feel good about the odds. There are probably 250 longevity companies with a clinical trial. Based on FDA averages, a couple of them will likely pass.

Hopefully, it happens soon. Once that happens, the social movement starts, and we are trying to directly accelerate that timeline to the ChatGPT moment as much as possible.

34:32 Case for optimism in longevity & the need for more investment

Daniel 00:35:01

That’s a pretty optimistic take. How many biotechs have there been working on cancer? We still have many types of cancer.

Dylan Livingston 00:35:10

If you look at the numbers, the effectiveness of cancer therapeutics has definitely gone up. Cancer in the ‘70s, ‘80s, and ‘90s was essentially a death sentence. Every life is a death sentence. We all have an expiration date at some point. Cancer, despite being beaten, is still probably the thing that will come back. But at least we can delay and keep people alive and healthy far better than we could even 10 or 20 years ago. We are making progress in that space.

Alzheimer’s might have been a better example, as we don’t have anything there. I think the best example is GLP-1s. We’re seeing that GLP-1s are probably the closest moment to a ChatGPT moment. It’s not quite aging, but it’s the first time people are saying, “I’m going to take this shot, and I’m going to be healthier as a result of medicine, not just lifestyle.”

Maybe the ChatGPT moment has started. It just hasn’t been in the aging thinking space yet. People are still thinking of that as the fat drug, not the healthspan-extending drug. It’s both. We need one of those drugs to be the aging drug for the cultural zeitgeist to change, and to have an aging movement rather than a GLP-1 movement. Maybe I’m an optimist.

Daniel 00:36:45

To be clear, we’re all optimists, but I was reacting to just how optimistic it was.

Dylan Livingston 00:36:50

You’re right. You could say the same thing about stem cells. I would hope that one of the 250 longevity companies with a clinical trial will get it right in the next 10 years. That’s my hope. With that success rate, I feel like it will happen. We’ll see. Fingers crossed. Knock on wood.

Daniel 00:37:17

A thesis we have on the Free Radicals podcast is that we need more people working on this. It’s easy to hear optimism and think, as I have thought this before, “Okay, they’re going to handle it.” No. There are still perhaps a thousand people really working in this space? Maybe 2,000. It’s a very small space.

Dylan Livingston 00:37:40

Very small.

Daniel 00:37:41

That’s actually tackling aging as a disease.

Dylan Livingston 00:37:45

It is. You’re definitely right. When I was figuring out what I wanted to do with my life, I could have tried to go into investment or whatever. If my dad was a broker, I probably could have figured something out there and just waited. I could have assumed that the big companies then would eventually figure it out, like the Jeff Bezos companies of the world.

I agree with you. There needs to be work done. There needs to be more people in the field. I hate it when people say AI will figure everything out. Maybe, I don’t know, but you’re just going to run the risk. You’re not even going to try. You’re just going to give responsibility to robots that we’re creating to figure it all out. Okay.

In my opinion, this is the biggest issue. Nothing kills more people than aging. It’s not even close. War, communicable diseases, murder, accidents, suicides, drug use, whatever. It’s not even close. If we want to alleviate the most suffering in the world, this is clearly the path to tackle.

I don’t know how people aren’t inspired by that idea. This is what gets me up out of bed every day: thinking about this grand problem and making life better for as many people as I can. I’m not saying this to be corny; this is actually how I think. I don’t understand how that’s not enough to inspire people to go into this field.

There are a lot of great organizations doing the work, Free Radicals included, that are getting the message out there and making it a popular topic to discuss. So I commend you, and I encourage people who are listening to get involved.

Eric 00:39:36

I want to dig into that more. There’s this question: how do people not get inspired by that?

Eric 00:39:41

I’m curious to hear from your perspective. You’ve spent time in the DC circuit and met with many legislators, investors, and company builders. From people who control and guide policy interventions, what is their response? How do people in Congress and various regulators respond to the longevity pitch?

40:04 What arguments resonate with congresspeople

Dylan Livingston 00:40:04

There is always general support. The big thing for us as an organization is hiring more people. As I said, it is a relationship game and constant communication. Without getting too political, there are different groups, lobbying groups, and political organizations that raise enough money to assign someone to each district, to each congressional person. They communicate with constituents of that representative, acting on behalf of an interest group. They are paid to oversee that Congressperson to ensure they support a specific issue.

When a Congressperson hears something repeatedly and is in constant contact about it, that is when they act in that field. We are a team of 3.5. The relationships I have built are mine, but we need 5, 10, dozens, or hundreds of other people doing this. I spend a lot of my time fundraising with the goal of growing this team, and hopefully, we will achieve that in the next couple of months.

The issue is not a lack of interest; everyone supports it. The issue is follow-up and engagement. This is why we started a caucus: to create a central node for disseminating information. When a constituent reaches out versus the co-chair of a caucus, it is completely different. We want to hire more lobbyists for A4LI. We need people who can build a network of constituents to keep pressure on their representatives and establish systems to not only spread awareness but also reinforce the message.

Legislators on both sides of the aisle are very supportive of this. We conducted polling to understand how to communicate this issue to each side of the aisle. You can tie it into Republican talking points and Democratic talking points. DOGE was ineffective because it did not tackle Medicare spending, which drives a majority of our deficit. If we had a longevity drug, the biggest part of Medicare spending—services used by Medicare participants—would decrease. The way to get them to stop using Medicare services is to give them a longevity drug so they do not need it at 65. For someone fiscally conservative, that is the talking point.

If you are talking to someone who is hawkish on the military, we have a recruitment and retention problem in the military these days. You want to keep your most experienced warfighters as healthy as possible for as long as possible: the generals, the commanders, etc. This is clearly a useful tool for the military in that regard. On the Democratic side, health equity is a frequent topic, often from the standpoint of Medicare for All. Metformin and rapamycin cost cents on the dollar to produce. A pill is a cent or something nominal like that. A great way to achieve health equity is by potentially covering rapamycin and metformin on Medicare and making them accessible to as many people as possible, thereby raising all boats.

The point is that longevity fortunately ties into all these different interest groups. We have been successful, as evidenced by our bipartisan caucus, in getting both sides of the aisle on board with this idea for different reasons, but they all come to the same conclusion: this is a good thing. The second layer is constant follow-up. Quite frankly, there is another layer of political advocacy that needs to be achieved, which I am pursuing. As you said, this field only has about 1,000 people, so it is difficult to properly establish a PAC because you need all 1,000 people to contribute money.

The second layer is contributing to candidates who advocate for this as an industry and helping them get reelected. We want to make our voice heard in elections. We want pro-longevity candidates. I do not want a politician who doesn’t support this. Do you support people dying or getting sick? I do not want a politician who does not support this. As a field and an issue space, we should have a voice and a way to support those who advocate for the right things. All of these things combined represent a big effort. As the head of it all, I am focused on long-term planning, putting systems in place for sustainable growth, and trying to get appropriations requests and right-to-try bills passed.

Daniel 00:45:43

What I am hearing from you is that a lot of this is people are supportive, but you need the resources to keep them focused on it and to work on it. It needs to be the squeaky wheel. Is that the correct framing? I imagine that many people, especially non-politicians, object to treating aging as a disease. Do you not encounter that with politicians?

Dylan Livingston 00:46:10

The most relevant experience I’ve had to answer this is the testimony and bill process in Montana and New Hampshire. In both cases, the bills were pretty split across party lines. They were both carried by Republicans in both state governments. Even though they were carried by Republicans, a lot of Democrats supported them in both states.

Politics usually follows party lines. However, there were crossover votes and people told us they supported these ideas and longevity in general, but ultimately voted with their party. We’ve done polling that shows about 70% of the population supports longevity. 30% either has no opinion or is against it.

There will always be naysayers on every issue. You see those ridiculous polls where, for example, 5% of people will respond yes to supporting murder. There are always people who give shocking responses.

This is a popular issue. We have polls and legislation to show it, and I’ve had conversations confirming it. This is more popular than people in the field give it credit for. The issue comes when it’s left to the imagination of the media and the general public. Stories can be twisted about Jeff Bezos, Elon Musk, Peter Thiel, and Sam Altman all putting money into living forever. That’s a very easy story to twist.

When it’s laid out, as we did in our polls, explaining how aging biology targets a different pathway and the shared common cause of diseases, and asking if people would be on board with preventing all these diseases, those questions get 70-75% support. The other aspect of this, going back to 501(c)(3) advocacy organizations doing social movement work, is the need for a more coordinated effort on messaging.

It needs to be very explanatory because this is a brand new field, a paradigm shift, unlike established issues like gun rights. There needs to be a better approach and effort to coordinate the field’s messaging. If we can get the right messaging in the social movement area of the field, the polls will consistently show 70%. A few Gallup polls show the field is divided 40-60 for-against in public opinion. However, those questions were written by people not from the field and in a non-explanatory way. That’s my take on it.

49:23 Bryan Johnson’s impact on the perception of longevity

Daniel 00:49:23

A big risk for the messaging would be Bryan Johnson’s vision of longevity. For the non-Silicon Valley population, that’s likely very alienating. There’s a need to moderate that messaging as we near a ChatGPT moment for longevity.

Dylan Livingston 00:49:49

I debate Bryan Johnson all the time: is he a net positive or negative for the field? I think he is a net positive, but there are definitely negatives. I see it more than anybody. As I mentioned earlier, when I speak with members of Congress or their staff, they’ve either never heard of this field – which is the best scenario – or they say, “Oh, you mean Bryan Johnson, who’s doing mushrooms on Twitter?”

I don’t care about that part, but it definitely puts a certain image in their head that is hard to break out of. You’re right. There is a certain audience, specifically Silicon Valley and cities like New York, for whom it’s totally normal. There’s a lot of country out there other than the cities, and they all have senators and members of the House. They all have influence over how this is done.

That’s why I debate this question. It’s a good and interesting one for me because, speaking to someone from Alabama, if they’ve come across Bryan Johnson, that’s likely a ‘no’ from them, or a lack of interest. It’s a tricky question.

The good news is the field is shifting more towards ‘healthspan’ talk. Personally, I want to see lifespan and healthspan gains for myself and everyone else, and I believe that should ultimately be the goal. It’s good that we phrase it so that healthspan and compressing morbidity are more palatable to everybody. We already know of things that increase healthspan, like exercise and diet.

I appreciate the field making this shift. My preferred term is ‘healthy lifespan’ because it touches on both concepts. The field is naturally moving towards this healthspan messaging. That’s probably the key to get those red states that might be freaked out about living to 200 on board.

52:15 Who are the most important people to influence

Eric 00:52:15

One more question on the importance and strategic directions of lobbying. Who are the most important people you’d like to influence over the next six months?

Dylan Livingston 00:52:26

Obviously, Donald Trump and JD Vance. At this point, if you get Trump to do something, it will probably happen in some way, shape, or form.

Daniel 00:52:36

On that point, why is Trump not all in on figuring out the longevity thing? Is it not in his orbit, or is it nonsense to him?

Dylan Livingston 00:52:47

I’m yet to meet President Trump, so I’ll let you know when I meet him. I’ll give you an answer to that. There are people in the administration who know about it. Jim O’Neill was the second in command and he was the CEO at SENS Research Foundation. I would be shocked if he hadn’t heard about it in some way, shape, or form.

It’s hard when you’re the president. What we’re trying to do is pursue a direct path to get meetings with him and pitch this idea: “Maybe you should do a national initiative on longevity.” The president is hard to get a hold of. This is also why we do a lot of our work on the congressional side. First, that’s where the budget is always set. Second, there’s a lot more staying power in Congress. The reelection rate for incumbents is about 98%. When you’re in Congress, you’re there for as long as you want to be. That doesn’t change. The president changes seemingly every four years since 2016.

Trump would be great because, in the short term, if you could make the pitch to him, the whole thing is settled. Obviously, he’d be ideal. However, members of the Senate Health Committee, where we have connections, are also important. Getting them to truly do something big is the next step. Getting them interested has been the goal until about 2026, primarily through our summits, fly-ins, meetings, and caucus. Now, 2026, 2027, and 2028 are the years we want these health committees, the Longevity Science Caucus, and the administration to actually sign and pass legislation.

Daniel 00:54:38

We’ve reached the end of the interview. Is there anything else you’d like to leave our audience with?

Dylan Livingston 00:54:44

This is a small field. If you are engaged and interested in seeing the longevity field accelerate, check out our work and follow us on social media. We’re doing our Georgetown H-SPAN Summit at the end of June. We’ll have a lot of great elected officials coming to speak, and many prominent longevity figures. So check out a4li.org and get involved.

The last thing I’ll say is this: when a small group of people are truly engaged in something, things get done. It’s true. The example I’ll give you is the rare disease advocacy groups. Rare disease by definition is rare. There are not many people who have them. But the advocates for that field are incredibly vocal, organized, and engaged. It’s not just a couple people; it’s thousands who are afflicted with this and deal with it.

While we are a small field, we need it to be as active as possible. That’s how we’ll gain momentum now and ultimately reach that ChatGPT moment I was talking about. So get involved at a4li.org. Follow us on Twitter and LinkedIn. Let’s make this all happen.

Daniel 00:56:09

Dylan Livingston, thank you for joining us.

Dylan Livingston 00:56:11

Thanks everyone. Thanks.

Michael is now the Chief Operating Officer at Life Biosciences, a company founded by Dr. David Sinclair, to treat aging with partial epigenetic reprogramming. Earlier this year, Life Bio earned FDA clearance to proceed into human clinical trials with their lead candidate, marking the first time that humans will be administered an epigenetic reprogramming drug, which may be the holy grail of longevity.

Michael also sits on the US board of the Hevolution Foundation, an organization that is investing up to one billion dollars per year into longevity research, and is also on the board of the American Federation for Aging Research.

In this episode, we discuss Michael’s insightful paper on why aging is an optimization by evolution, why that means it’s malleable, how Life Bio is going into the clinic with the first epigenetic reprogramming therapies to reverse diseases of aging, and how to push the longevity field forward.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

0:00 Intro

2:30 Why aging is an optimization by evolution

15:10 Why scientists miss the role of evolution in aging

19:31 Evidence that aging is a regulated process

24:40 Hormesis and adaptive stress responses

28:55 The promise of partial epigenetic reprogramming in humans

35:08 Life Bio’s first therapy proceeding to clinical trials

47:47 Why targeting the eye, particularly glaucoma and NION, is a strategic choice

52:03 The pharma industry’s perspective on anti-aging research

56:39 The importance of public engagement for funding and policy change

1:05:42 Reasons for optimism in longevity

1:08:26 Advice on how to have an impactful and interdisciplinary career

1:11:54 The ethical case for longevity

1:17:34 Concrete steps for audience members to support longevity research

Transcript

2:30 Why aging is an optimization by evolution

Daniel 00:02:30

Michael Ringel, welcome to the podcast.

Michael Ringel 00:02:32

Thanks for having me. I’m a big fan of what you guys are doing, so thank you for having me on.

Daniel 00:02:36

Great to have you here. We’d love to start by talking about the theory of why we age.

You recently published a paper titled “Why We Age,” in which you argue that many theories of aging ignore evolution by natural selection. Instead, you suggest that aging is the result of an optimization by evolution.

Can you walk our audience through the core of your argument about why we age?

Michael Ringel 00:02:58

It’s a great first question. The best thing anyone watching or listening can do is read the article. You can easily find it by searching “Ringel, Why We Age.” It’s open access and published in a journal called Biological Reviews. Everything in that journal is written for generalists, so anyone can understand it with just a little bit of background.

The paper is incredibly comprehensive. It looks at all the different theories and the evidence used to assess them, and it comes to a very clear conclusion. I will try to describe it here, but the depth I can provide verbally is not the same as what you can get from the paper. Since this is of interest to so many people, I really encourage you to read it.

To bring this logic to life, let’s look at one of the most illuminating puzzles in biology. Many unicellular organisms, such as certain bacteria and archaea, do not age. However, there are also unicellular organisms within those same groups that do age.

Conversely, when we look at animals, most of us age, but there are some—including sponges, certain jellyfish, Hydra, and potentially flatworms—that do not. How do we make sense of this? Why do some unicellular organisms age while others don’t, and why do most animals age while a few remain biologically immortal?

Many popular and academic theories suggest that aging is an inexorable process of wear and tear that we simply cannot avoid. These theories argue that entropy, DNA damage, or oxidative damage eventually overwhelms the system.

Other variations include the idea of antagonistic pleiotropy, where genes that are beneficial early in life cause unavoidable problems later on. A related idea is mutation accumulation, which suggests that because there are fewer individuals left in older age cohorts, there is less selection pressure, allowing aging to creep in as a negative effect that evolution is unable to eliminate.

What all these theories share is the view that aging is a dysregulated accident. The assumption is that evolution simply cannot overcome barriers like entropy. However, these explanations fail to explain why non-aging species exist.

If aging is an unavoidable accident of physics or biology, why do these species not succumb to it? They still have DNA, metabolism, and mutations, and they still face entropy. These theories also fail to explain the specific patterns of which species age and which ones do not.

There is significant evidence that aging is actually a regulated process. For example, it is relatively easy for different species to evolve vastly different lifespans. Pacific rockfish shared a common ancestor just eight million years ago, yet some of those species evolved to live 10 years while others live 200.

An even starker example is found in eusocial insects like ants. Worker ants typically live about a month. This isn’t just due to external wear and tear; even if you isolate a worker ant and keep it safe, it still ages at that rate.

In contrast, a queen ant of the same species can live for 30 years. These individuals are extremely closely related—often sisters or a mother and daughter—and share almost the same DNA. The only difference is which genes have been turned on or off through epigenetic regulation. This is a clear example of the biological ability to regulate aging and lifespan.

If aging is a regulated process, why does it exist at all? This is where the “disposable soma” theory becomes essential. It is predicated on rock-solid theoretical foundations, starting with the premise that energy and resources are finite.

As you know from your physics background, Daniel, work requires energy. An organism’s work can be divided into two main categories: maintenance and repair of the soma (the body), and growth and reproduction.

You can prove mathematically that it is never evolutionarily optimal to invest 100% in the maintenance of the soma. While I won’t go into the equations, I can provide an intuitive understanding. If you invest maximally in somatic maintenance, you are wasting energy that could have been diverted to growth and reproduction.

Evolution does not maximize the individual; it maximizes fitness, which is the long-run rate of increase for a line of descent. It is a process of selection, not forward-looking design. The organisms we see today are the result of combinations of investment that allowed for the highest proliferation rate.

As long as there is some extrinsic mortality, there will eventually be an age where marginal investment in the soma is less valuable than investment in reproduction. This means evolution actually selects for aging as an adaptation.

The clue lies in the fact that the most primitive organisms—primitive animals and single-celled organisms—are the ones that do not age. In these organisms, the soma and the germline are commingled. If the soma were allowed to age, the germline would age as well, and the lineage would go extinct.

The disposable soma theory predicts that whenever the germline can be separated from the soma, aging will evolve. This is exactly what we observe. Most unicellular organisms haven’t evolved to separate the two. However, some bacteria and archaea have evolved a trick: they divide asymmetrically.

This allows one cell to become a “mother cell” where aging accumulates and is eventually selected for, while the other cell remains rejuvenated. Similarly, multicellular animals that do not age, like the Hydra, are those where the soma and germline remain commingled. They can revert their entire bodies back to the germline state.

Humans cannot do this; we cannot turn ourselves back into an egg or a sperm. Because our soma is distinct and disposable, we age. This powerful theory explains why certain organisms age, why others don’t, and why specific species fall into each category.

Because the amount of investment in somatic maintenance is a regulated process, it implies that aging is manipulable. This is great news for those interested in adjusting healthspan—the period of life lived free of disease and disability.

We already have the biological mechanisms within us that completely rejuvenate the germline, as that part of our biology must remain ageless for the species to survive. If we can tap into those existing mechanisms and apply them to the soma, we have a very powerful way to modulate both lifespan and the incidence of age-related diseases.

15:10 Why scientists miss the role of evolution in aging

Daniel 00:15:24

I suggest everyone read the paper. You only touch on a portion of the rich evidence, and I believe your paper provides one of the clearest and most structured explanations of the theory of aging I have seen. It was extremely helpful in orienting myself in this space.

I want to dive into a few specific things you mentioned, but first, I have a meta-question. You noted in your paper that we should obviously use an evolutionary lens, citing the quote: “Nothing in biology makes sense except in the light of evolution.”

At a meta-science level, why have there been so many non-evolutionary ways of looking at aging? Why are so many people potentially making a grave error in how they think about this?

Michael Ringel 00:16:15

That is essentially a psychology question, which is slightly outside my expertise. However, I would say this approach is not uncommon and isn’t limited to aging.

I’ll walk through an illuminating example from the paper regarding allometry, which is totally unrelated to aging. Allometry refers to the scaling relationships of organisms; as an organism gets bigger, certain things scale differently.

A classic example is that an organism’s weight increases to the power of three. It is a volume function, assuming constant density. However, the ability to support that weight is a two-dimensional function based on the cross-sectional area of the legs.

Grounded in physics, you might expect an allometry ratio of two to three. But that isn’t actually what we see. We observe a ratio closer to three to four. For a hundred years, people struggled to understand why observations differed from predictions.

Then, a great paper in Science by White and colleagues in 2022 proposed an optimization algorithm. Similar to what I described regarding aging, energy must be allocated between different needs, including scaling and growth.

When you run that optimization algorithm, the result is very close to three over four. They derived 0.76. More importantly, they derived several other unexpected features of allometry that matched observation.

This is a powerful example of why thinking about biology as an optimization algorithm is so important. Unfortunately, that optimization is not about you living longer; it is about maximizing fitness. Those two goals are not tantamount to each other.

Daniel 00:18:37

You also mentioned that because aging was optimized for, it is highly regulated, which implies we can find ways to intervene. We will discuss partial epigenetic reprogramming as a specific example later.

On a theoretical level, I’m having trouble fully understanding that. You mentioned the rockfish—different but closely related species with very different lifespans.

I can see how evolution optimizes different niches for various lifespans. But how do we transition from that broad evolutionary scale to a single species in a single lifespan? How does that optimization imply that we can tap into something in our own biology?

19:31 Evidence that aging is a regulated process

Michael Ringel 00:19:31

This understanding also informs why many current interventions are effective. Let’s look at caloric restriction and other known interventions that work in model organisms and likely have some value in humans.

Why would starving an organism extend its life? Mechanistic theories struggle to explain this. For instance, mutation accumulation has nothing to do with the rate of mutations at older ages.

Some mechanistic theories suggest that we are slowing down metabolism or hyperfunction. However, at a single-cell level, we see that caloric restriction does not slow overall metabolism; energy per unit of mass remains the same or even increases.

Exercise also challenges these mechanistic explanations because it speeds up function. Under those theories, exercise should make you die faster, yet it fails as an explanation for why exercise is beneficial.

The disposable soma theory offers an elegant explanation that matches our observations. Imagine the allocation between maintaining the soma and investing in growth and reproduction. In a famine, which is essentially an organism under caloric restriction, it is not an ideal time to reproduce.

Offspring are unlikely to survive or thrive during a famine. The biological strategy is to reduce investment in growth and reproduction even further than the total loss of energy and redivert that energy to maintenance and repair. This stretches and maintains the soma in hopes of a better season.

We see this clearly in model organisms. A mouse that typically lives two years can stretch its lifespan into a third year if the second year is a famine, waiting for a better year ahead.

Various insults, such as hypoxia, lack of energy, or a lack of sufficient amino acids, hit a switch that tells the organism to wait for a better time. This fits the theory and matches reality perfectly.

However, the disposable soma theory also predicts that this diversion matters more in short-lived species. For a mouse, one extra year is a massive increase in breeding opportunities. For yeast or worms waiting for rain, an extra month is a significant deal.

In long-lived organisms that already average out natural variability, caloric restriction has a diminishing impact. If you graph the percentage of lifespan increase against the underlying lifespan of the organism, you see a log-linear decline.

By the time you look at rhesus monkeys, which are relatively long-lived and closer to humans, the predicted effect is very small. In a noisy experimental environment, you may or may not see a signal.

This is exactly what happened in experiments conducted by the NIH and the University of Wisconsin. One showed a result while the other didn’t. It was a 6% increase compared to a 50% increase in mice, which is exactly in line with the predicted trend line.

Extending that trend line to humans suggests a lifetime of caloric restriction is probably only worth a couple of extra years of maximum lifespan. It is not going to radically change our maximum age.

While I am not suggesting everyone should practice caloric restriction, managing diet, exercise, and sleep remains highly valuable. These habits greatly reduce the chance of early mortality and morbidity. You will be healthier along the way, even if it’s unlikely to make you live past 120.

Eric 00:24:39

There is an interesting convergence of evidence across various conversations, papers, and companies. Exposure to acute or manageable levels of stress, known as hormetic stressors, is consistently linked with better mortality and health outcomes.

Exercise, cold exposure, heat exposure, and caloric restriction are all forms of hormetic stress. Recently, research on various mammalian species showed that exposure to low-grade radiation actually resulted in a reduced incidence of cancer.

Our bodies have adaptive mechanisms to increase stress responses, such as improved DNA repair, when faced with increased baseline hazard rates. This recurring theme is beautifully explained by the disposable soma theory.

24:40 Hormesis and adaptive stress responses

Michael Ringel 00:25:46

There is a molecular biology analog to everything I described in evolutionary biology: a master switch that senses upstream indicators of whether the conditions for growth are in place. It monitors factors like sufficient amino acids, oxygen levels, and energy conditions.

This switch turns various downstream processes on or off, favoring either growth and reproduction or maintenance and repair. It activates protein, lipid, and nucleotide synthesis and mitochondrial function while turning off repair and recycling pathways like autophagy and the proteasome-ubiquitin system.

This switch is mTOR. It sits in the middle, regulated by many upstream factors. The most important are along the energy-sensing dimension, as that is typically the constraint.

Sensors like REDD1 and HIF capture upstream indicators, while GATOR and Sestrin sense amino acids. These converge on mTOR, which then directs downstream mediators like TFEB for autophagy or various sensors that turn on protein and nucleotide synthesis.

If you look at this to predict the best interventions, the most effective one found to date by the National Institutes of Health’s Interventions Testing Program is rapamycin. It works on FKBP12, which directly binds mTOR.

The next most effective interventions are in the upstream nutrient-sensing pathway: 17-alpha estradiol and acarbose. This matches the story well at both the evolutionary and molecular biology levels.

Daniel 00:28:28

As you mentioned, we think there is a relatively low upper bound on the impact nutrient-sensing pathways can have on lifespan because we expect a bigger impact for shorter-lived organisms.

However, you’re working on something at Life Bio that we believe doesn’t have that upper bound: epigenetic reprogramming. You are about to start the first human clinical trials for this.

Could you tell us more about that and the specifics of Life Bio?

28:55 The promise of partial epigenetic reprogramming in humans

Michael Ringel 00:28:59

There is powerful biology already within us that completely rejuvenates. Our germline is rejuvenated every generation. Babies are born young, even though every baby starts as a single old egg cell and an old sperm.

We can measure them and confirm they are old epigenetically, transcriptionally, and functionally. You see a decline in fertility and other factors as these cells age.

Babies themselves are not old. Even after fertilization, the zygote remains old for the first few days. However, between days seven and nine of embryogenesis, a miracle occurs and the embryo is made young again.

If you measure the age of an embryo after day nine using an epigenetic clock, it reads zero. This was shown by Steve Horvath back in 2013. It remains close to zero on transcriptional or functional clocks as well.

During that period, epigenetic marks—methyl and acetyl groups on the DNA and histones—are wiped back to their original state. The cell is told it is young and begins to behave accordingly.

Consequently, babies do not inherit their parents’ age-related diseases. They are not born with Alzheimer’s or cardiovascular disease. They may inherit a genetic disorder, but they don’t inherit these diseases of aging.

We have already tapped into this biology artificially through cloning. Every clone starts as a single old cell from an old individual. By taking a fibroblast from an old mouse and treating it with four transcription factors—Oct4, Sox2, Klf4, and c-Myc (OSKM)—we can recapitulate the rejuvenation process seen in babies.

Those cells are then born young. They live normal lives and can reproduce normally. The hardest test of aging biology is not whether you can shift aging slightly, but whether you can completely reset it. This is the only thing we know of that can do that.

You can’t do this in vivo because you don’t want your cells to become stem cells; they would cease to function, and there is a high risk of cancer. However, a powerful insight emerged at the end of 2020 from the Sinclair lab at Harvard Medical School.

They showed that by treating with just three of those four proteins—Oct4, Sox2, and Klf4 (OSK)—and leaving out c-Myc, you don’t take cells all the way back to a primordial state. They do not become stem cells.

Researchers measured this by looking at Nanog expression, a marker of stemness, and found no increase in tumor formation. In fact, there was a decrease, along with profound therapeutic consequences.

In that first study, they restored vision in mice that were blind for three different reasons: a model of glaucoma, normal aging, and a physical injury where the optic nerve was crushed.

In all three cases, they were able to restore vision completely, making the eyes function like those of a young mouse. This demonstrates the profound value of tapping into this pathway. It has been shown to be safe in preclinical models, giving us the confidence to proceed with human clinical trials.

Daniel 00:34:17

Great.

Eric 00:34:18

How does your view of partial epigenetic reprogramming compare to others in the field, like the work happening at New Limit led by Jacob Kimmel?

Jacob has previously argued that partial reprogramming is potentially not a totally encapsulated age-reversal mechanism, but rather one of many levers that can meaningfully reduce the symptoms we affiliate with aging. He believes a whole suite of different interventions will be necessary to reverse aging. Does that square with your understanding?

Michael Ringel 00:35:06

We are very much in favor of there being many different approaches to ameliorating age-related diseases. It’s a small field right now, and we need much more work using the principles of aging biology.

We welcome the existence of other companies in this space, including those working on partial epigenetic reprogramming. It is a validation of the quality of the science. There are certainly hundreds of age-related diseases for us to tackle, so there is a very broad white space here. In biopharma, the enemy is not another company; the enemy is disease.

There is a lot of evidence that partial epigenetic reprogramming using OSK has therapeutic value in preclinical models. In addition to the first study, there have been a number of others, both at Life Bio and in academic labs, showing the power of this platform to rejuvenate cells and thereby ameliorate age-related disease.

At Life Bio, we replicated the work out of Harvard Medical School in mice and other rodents. Then we replicated it in non-human primates, which is a really important step that gives us a tremendous amount of confidence that this is likely to translate to humans.

Many things work in mice that unfortunately don’t work in humans, but non-human primates are much closer to us and much more predictive of what actually works in humans, particularly in the eye. That gives us a lot of confidence that these treatments should be safe and effective.

We need to prove that through clinical trials, ensuring safety first and then exploring efficacy. However, many other labs have shown the safety and efficacy of OSK treatment in preclinical models across a range of diseases just within the last year or two. The field is blossoming quickly.

A recent study on engram cells showed cognitive rejuvenation. When treating the brain with OSK, it not only improved the ability to learn but actually restored lost memories in mice, which was profound.

Another study by Zhang and colleagues showed that partial epigenetic reprogramming in the liver restored hepatocyte function and reversed features of fatty liver disease. Specifically, it addressed metabolic-associated steatohepatitis, which is one of the most common high-burden diseases in humans.

A recent study showed that using OSK to rejuvenate chondrocytes—the cells in your joints responsible for cartilage formation—led to the restoration of both the cartilage and the bone. Another study on nucleus pulposus cells in the vertebrae showed that it reversed features of aging and restored back pain symptoms to baseline levels.

Back pain is one of the highest causes of morbidity in humans. People often don’t think about it until they experience it, but addressing it provides huge value.

Additionally, a preprint from Pinheiro and colleagues showed that the restoration of endothelial cells, which line your arteries, reversed vascular disease and hypertension. Vascular disease remains the largest cause of mortality in humans, so addressing that is of massive value to society.

There is a large body of evidence that epigenetic changes are implicated across the full breadth of age-related diseases. If we can reverse them, we can ameliorate those diseases. This growing body of evidence showing that OSK works across a wide range of disease states gives us the confidence to bring this platform forward to address multiple conditions. Our goal is to start with blindness but certainly not stop there.

35:08 Life Bio’s first therapy proceeding to clinical trials

Daniel 00:41:04

Since you mentioned not stopping with blindness, where do you see Life Bio’s platform developing in the future? You provided evidence for many tissues where OSK can be rejuvenative. Can you tell us about some of the things you’re already working on that extend beyond the eye?

Michael Ringel 00:41:21

Let me start with the eye because that is our primary focus in the near term. These are incredibly high-burden diseases. Blindness obviously carries a high direct burden, but there is also a significant follow-on burden. When age-related blindness occurs late in life, there are many sequelae that follow.

It increases the risk of falls, which leads to high-burden injuries like hip fractures. It is also associated with an increased risk of institutionalization in nursing homes, which places a high cost on the healthcare system. Furthermore, it increases the risk of social isolation, which drives depression and neurodegenerative diseases like Alzheimer’s. There are many reasons to be concerned about the impact of blindness on society.

We are targeting two specific conditions in our first trial. The first is glaucoma, the number one cause of blindness in people over the age of 60. Most people have heard of it. It is generally caused by fluid buildup that increases pressure in the eye. However, some patients develop glaucoma even with normal pressure because their retinal ganglion cells are particularly susceptible.

The second disease we are targeting is NAION, which stands for non-arteritic anterior ischemic optic neuropathy. It is the most common acute cause of blindness in people over the age of 50. It is essentially a stroke that damages the optic nerve at the back of the eye.

Unfortunately, there is no approved treatment for NAION. A person might wake up one day blind in one eye, rush to the doctor, and receive a prognosis that no therapy exists. Additionally, there is a 30% chance it will happen in the other eye within five years. People live in fear of these consequences.

Bringing relief to these patients is hugely valuable. Glaucoma affects roughly 60 million people worldwide, and more than 3 million are legally blind because of it. While some drugs can treat glaucoma early, there is no treatment for many patients who have progressed past that point. Even for those who start medication or undergo laser surgery, the disease often continues to progress. A medicine that can ameliorate this age-related blindness is incredibly valuable.

That is our starting point, but we have also released data showing that we have restored function in a preclinical model of MASH, a fatty liver disease. Using OSK in a mouse model, we were able to rejuvenate hepatocytes and restore function across many clinically meaningful measures.

Crucially, this is independent of body weight; it is not a weight loss drug. It rejuvenates liver cells to restore function, meaning it could potentially be synergistic or additive with weight loss drugs. There is still much to do. We are just starting clinical trials in the eye and have not yet reached the clinical phase for the liver or other organs.

We have other undisclosed targets as well. We are finally at the point where we can test this in humans and prove whether this technology will work for people.

Daniel 00:46:03

It is amazing. To ensure our audience fully appreciates what is happening: this is the first human testing of epigenetic reprogramming.

While it aims to cure certain types of blindness, which are major contributors to disease burden, it also serves as the first proof point for what could be the holy grail of longevity.

Michael Ringel 00:46:26

That’s right. It is a proof point. We don’t know the outcome of the trials yet, but there is a range of possibilities. Currently, nothing improves function in the eye; you can only slow the rate of decline.

If we can do better than the current standard of care, we have created immense value for patients. We are hopeful for results similar to what we saw in non-human primates, where we observed clinically meaningful improvements in function. Even a result less dramatic than that would still be very valuable as a first therapy.

Eric 00:47:12

When you look back at the strategic considerations around starting with glaucoma and NAION—moving from the eye into the liver—target and indication selection are the deepest levels of strategy for a biotech company. This is an area where you have spent decades crafting your skills.

Walk us through the thought process of how you arrived at the eye first, specifically with glaucoma and NAION. More broadly, were there other tissues or indications you considered, and how did you decide to discard those for the first phase?

47:47 Why targeting the eye, particularly glaucoma and NION, is a strategic choice

Michael Ringel 00:47:47

We believe a broad swath of age-related diseases is in scope. This covers an enormous spectrum, representing more than 90% of human mortality and a similar percentage of morbidity. Most diseases are potentially within our scope.

When considering which diseases to target, one lens is identifying what brings the most relief to patients. We focus on indications with the highest societal burden, whether through healthcare costs or patient morbidity and mortality. Even with that lens, the list of serious diseases remains long.

We also consider where we can prove the technology and where the science is most likely to work early on. We have a strong conviction that OSK works broadly, but delivery remains a challenge. Getting these three proteins inside a cell is not trivial.

While there are general solutions for delivering proteins extracellularly or small molecules intracellularly, we lack a general solution for delivering proteins inside a cell. There are various delivery methods available. You can deliver it as a DNA segment that expresses mRNA and proteins, as RNA, or as the proteins themselves.

These can be delivered via viral vectors like AAV, lipid nanoparticles, or emerging technology like exosomes. We are flexible about which technology delivers the OSK, as long as it gets inside the cell. We estimate that 10 to 15% of age-related diseases are currently accessible with existing delivery technology.

That figure will grow as researchers develop better AAVs and other methods. We want to avoid safety and efficacy risks associated with the delivery technology itself. The eye is an excellent starting point because AAV2s are already on the market.

Delivery into retinal ganglion cells is a validated process, which gives us confidence. This is also where the original work was done. While we have a vast field of potential targets, only a few “lampposts” have lit up to show what works in mice.

We don’t want to limit ourselves to those areas, but they offer a solid starting point for translating mouse data into human results.

Daniel 00:51:37

I’d like to shift gears and discuss your experience as a managing director at Boston Consulting Group. Since you advised top pharma companies, you know what they care about regarding R&D.

Longevity and aging biology are still somewhat outside the mainstream. How are top pharma companies thinking about these fields today, and how has that changed over the last 25 years?

52:03 The pharma industry’s perspective on anti-aging research

Michael Ringel 00:52:17

I have many conversations with CEOs and R&D heads at top pharma companies. There is awareness and interest, but the proof is ultimately in the pudding.

It is up to biotech companies to bring forward treatments that have passed safety and efficacy hurdles in clinical trials. Once we reach “proof of concept” and show that a treatment works in humans, pharma’s interest in addressing age-related diseases will be immense.

It is a straightforward process. There is often concern about a lack of interest, but these companies are motivated to bring medicines to market that help people. As soon as they see data confirming a treatment’s efficacy, they will be interested.

Biotechs simply need to reach that proof of concept. From there, pharma companies will want to acquire those assets and bring them to as many patients as possible.

Daniel 00:53:47

That makes sense. One of the reasons for starting this podcast is the sense that the early-stage biotech ecosystem needs more people.

We need many shots on goal to generate the proof points that trigger pharma investment and create a virtuous cycle. What is your perspective on the current ecosystem? Are we taking enough shots on goal?

Michael Ringel 00:54:16

This may be the most important question of the podcast, and it is why I am so grateful your show exists. If we want to work backward from the goal of getting medicines on the market, we have to look at the core source of activity in this space.

For medicines to reach the market, we need pharma companies to be interested. That interest will happen as soon as we bring assets to the proof-of-concept stage. For biotechs to bring assets to proof of concept, they need funding from investors, but they also need the foundational ideas to move forward.

Almost all biotech ideas originate in academia, including our own. Academics rely on funding to explore aging biology and make discoveries. Without that funding, the discoveries I described would not be possible.

A significant amount of public funding comes from the NIH. In the U.S., where that funding is directed is heavily driven by policymakers, Congress, and the President. Those policymakers are ultimately answerable to public opinion. Therefore, everything comes back to building a grassroots movement and making it known that people are interested in this space.

I will give you two examples of how powerful public interest is in changing this chain of activity. One is the campaign run by Mary Lasker in the 1940s, 50s, and 60s to raise awareness for cancer. At the start of her campaign, cancer was a taboo subject. People wouldn’t talk about it because it was viewed as a death sentence.

That campaign changed public opinion over a 20-year period. It shifted from a majority of people avoiding the subject to a majority engaging with it. This led to action. In 1971, the National Cancer Act was passed, launching the “War on Cancer.” This led to an initial $1.5 billion in NIH funding.

56:39 The importance of public engagement for funding and policy change

Daniel 00:57:20

More.

Michael Ringel 00:57:20

That funding led to a more than tenfold increase in the annual rate of publications in the space. By 1980, roughly a decade later, there were 30 cancer drugs on the market. They were not great drugs, but they were a starting point that proved we could make progress.

A more recent example is the 2014 Ice Bucket Challenge. Something like 20 million people participated, raising more than $200 million directly. It also galvanized government action. NIH funding for ALS increased fourfold following the challenge.

This led to an increase in publications, and there are now at least three drugs on the market for ALS. Again, they are not perfect, but they are better than no treatment and represent a vital starting point. Both of these success stories began with public interest.

The work you are doing to build a community is incredibly important. None of this happens without community organization. We need to frame aging biology as a tool that helps us live longer and healthier while addressing age-related diseases.

Daniel 00:58:48

That makes a lot of sense. Another piece of the puzzle is getting the right talent into the space. That is often downstream of a grassroots movement, which increases interest and secures funding.

A few years ago, I thought about starting a podcast about longevity biotech. I spoke with a friend at Harvard Medical School who didn’t want to touch aging biology with a ten-foot pole. I think about that a lot.

Even at conservative institutions like McKinsey, BCG, or Harvard Medical School, parts of aging biology are still seen as “kooky.” While there is great scientific work being done, there are also weird elements on YouTube and elsewhere that affect the field’s reputation. What will it take to ensure people see this field as legitimate and prestigious?

Michael Ringel 00:59:50

We are very focused on data-driven paths. There is a reason we go through a methodical clinical trial process with the FDA. It is essential that claims are supported by evidence and that this evidence is gathered transparently. Having a regulator validate those claims is the right direction for the field.

There are many claims out there regarding things that may or may not work. We won’t know until we test them. It is vital that products go through a testing process so that information is available for people to make informed decisions.

Whether a decision is made by an individual or in consultation with a physician, people should have the data. I support the empowerment of individuals, but they need the information to decide. We should require companies to validate their claims in a transparent way that is evaluated by a regulator.

This is the path the pharmaceutical industry follows. This process protects us from safety issues, products that lack efficacy, and contamination. There is a long history of those exact issues occurring before these regulatory approaches existed.

Daniel 01:01:30

One takeaway is the absolute focus on evidence. If I talk to a pharma executive, I don’t need to talk about living forever or meeting aliens. That isn’t a compelling investment case. However, I can tell them about studies that have identified specific mechanisms of action and targets.

Focusing on the evidence is clearly the way to build interest. I’m curious about another angle. When you found consultants at BCG who were interested in joining your interest group, what got them excited?

The evidence is part of it, but was there something else? Was it a realization of how limited life is or how damaging aging can be? I’m curious what has resonated most with people at BCG or within pharma companies.

Michael Ringel 01:02:27

Ultimately, as my mother would say, what’s more important than your health? A lot of people care about their health from a personal standpoint. This is not about living forever—unfortunately for your aspirations—but it is about living longer, living healthier, and addressing age-related disease. That is still tremendously valuable.

It starts with personal interest, but also the realization that there is very cool science and a real chance to help other people. We can build something here that is of tremendous value to society.

Eric 01:03:08

Another aspect we’ve discussed with prior guests on the Free Radicals podcast is how we create more prestige and legibility in the biomedicine and biopharmaceutical fields. This industry has, for various reasons, received a lot of bad public press over the past few decades.

There is a challenging, fundamental PR problem for healthcare, drug discovery, and pharmaceuticals that we haven’t cracked yet as an industry or as a society. It is not only unprestigious in some ways to enter drug discovery, but it is actually considered morally wrong. How we solve that issue is a big question we have to answer.

Michael Ringel 01:03:59

I agree with that concern. If you look at public trust surveys regarding the pharmaceutical industry over time, it has declined precipitously. This is a shame because that reputation is 180 degrees from the truth. This is one of the few industries that really helps people.

There are legitimate concerns around pricing, accessibility, and ensuring equity, but these medicines are miracles. They are incredibly valuable, as is the industry that helps create them. In my experience, the people involved are really motivated by trying to improve human health; it is something they truly care about.

It is frustrating that there is such negativity around the industry. I don’t have the answer on how to change it, but I do think addressing questions around equity and access is helpful in that regard.

Daniel 01:05:09

We were talking about investment in the longevity space. You sit on the U.S. board of the Hevolution Foundation, and you did work at BCG that helped lead to its creation. This organization has up to a billion dollars annually to deploy into longevity science.

Can you tell us about the work you did at BCG that convinced them this was worth an investment of a billion dollars per year? Secondly, what is the Hevolution Foundation currently up to?

Michael Ringel 01:05:40

This is one of the few projects I can talk about from my time as a consultant because it is a matter of public record. The CEO has discussed it, and it was a public tender because it was government work. The idea for Hevolution already existed; we were brought in to help think about what it should look like.

We were involved in that strategic work and some of the early hiring, which was fantastic. I cannot take credit for what the organization has done, but it has been great to see the willingness of a private foundation to augment government spending in this area.

I sit on the board of the U.S. foundation. While the board isn’t involved in making management decisions, we have been involved in discussions around strategy and approach. The focus is on things that unlock the field and support its growth.

This fits the spectrum I described earlier: if we want medicines, we work backward through biotech formation, academic support, and all the way to public interest. The foundation considers activities on each of those dimensions to build that full chain.

Grantmaking to investigators is part of it, but they are also involved in company formation. Another goal is to address problems the field faces as a whole that a single company cannot solve, perhaps through a “Manhattan Project” approach.

For example, the lack of validated biomarkers that can make quick predictions of long-term effectiveness is a potential unlock where the foundation has focused.

1:05:42 Reasons for optimism in longevity

Eric 01:07:56

Let’s take a step back and look broadly at the longevity industry today. You have spent time as a consultant, a company builder, and a leader in this space. On a scale from 1 to 10, how would you rate the progress of longevity sciences?

How far along are we in validating the core hypotheses the field has raised? What does the outlook look like, and which readouts are you personally most excited about?

Michael Ringel 01:08:24

There is a lot of reason for optimism, but also a reason to push harder. We published a piece in Nature Aging looking at the acceleration of the field relative to the background rate of biomedical research. It is typically outpacing the general field by between 2% and 7% a year.

I would rather be growing faster than the background rate, but we are starting from an incredibly small base. Only about one percent of the total NIH budget goes toward the fundamentals of the biology of aging.

Given that age-related diseases account for over 90% of mortality in developed nations, one percent of funding seems disproportionately low. I’m glad we are on the cusp of translation, which will help push the field forward, but I would like to see much more action and support.

This is where the value of podcasts and listeners comes in. Each of you has a role to help build this if you are interested. All the most successful efforts have been grassroots, and the power really lies in action at the local level to drive overall progress.

1:08:26 Advice on how to have an impactful and interdisciplinary career

Daniel 01:10:10

I have a selfish question about your career, which I think is extremely cool. You have a PhD in biology and a law degree, you spent 25 years at BCG, you’ve published scientific papers, and you have become a thought leader in longevity.

How have you managed to maintain that interest and expertise across domains over your career? It’s very exciting that you haven’t stayed in a single box.

Do you have advice for how I might do something like that in my career, or how listeners who aren’t scientists might make contributions?

Michael Ringel 01:10:56

It is really sweet and a nice ego boost to hear all that. Obviously, there were many years in that process, so for younger folks listening, there is plenty of time to try different things.

I think it comes down to working on things that fit your passion but also where you can see a path to success, and then pivoting as your interests change. Maybe one year you are interested in astronomy and the next year bird watching. Be willing to change your path based on your interests.

Eric 01:11:50

Michael, the Hastings Center Report paper you co-authored with your wife Carolyn and your colleague Arthur Caplan is, in some ways, a direct rebuttal to Ezekiel Emanuel’s famous argument for accepting death at 75.

What is the strongest anti-longevity argument you’ve personally encountered? Where does it break down, and how have you formulated a counter-argument to it?

1:11:54 The ethical case for longevity

Michael Ringel 01:12:11

Scientists don’t always think about the ethical underpinnings, but it is important for society to consider interventions that could extend healthspan and lifespan. I would direct people to read the paper for a more in-depth discussion. It is open access and available in the Hastings Center Report, which is a gold-standard bioethics journal.

You can Google “Ringel” and the title “Why We Can Thrive Past 75.” In this case, the author is my wife, Carolyn, not me. The paper walks through eight different arguments raised against longevity and effectively debunks them.

I recruited Carolyn to the cause because she is a bioethicist who teaches at Harvard Medical School. I felt that a scientist or someone in industry writing about this wouldn’t carry the same weight as a bioethicist. Careful review by someone like her or Art Caplan, a leading bioethicist who recently retired from NYU Langone, carries a lot of credibility.

The starting point is that health and life are generally recognized as good things. It is difficult to argue that they are good, but not past the age of 75 or some other arbitrary number. When you look at the arguments people raise, probably the most common one is around equity. People ask if this will only be for the wealthy.

While there are inequities in healthcare and new treatments start out expensive, the right answer is not to stop the research. Instead, we should work on improving equity and access. We’ve seen this with Hepatitis C drugs. They were originally very expensive, but prices have come down and access has improved. The goal is to reach a point where everyone has access to the cure, rather than no one getting it.

Interventions related to maintaining healthspan will likely be cheaper and easier to democratize than late-stage care. We spend a lot on the last year of cancer care, for example. As the old adage goes, an ounce of prevention is worth a pound of cure.

Small changes earlier on can avoid significant downstream costs and be relatively inexpensive. If you’re interested in the ethics of this, I encourage you to read that paper. If you have a question, one of those eight arguments likely addresses it.

Daniel 01:17:09

So working on longevity is the right thing to do and extremely valuable. You also made a strong case for why we need a grassroots movement.

What are some concrete things our audience can do to get involved or support this work, besides subscribing to the Free Radicals podcast?

1:17:34 Concrete steps for audience members to support longevity research

Michael Ringel 01:17:35

There are organizations working on this, such as the American Federation for Aging Research (AFAR). I’m on their board, and people can donate or get involved through them.

Many people are also interested in what they can do for themselves right now. An observational study by Nguyen and colleagues at the Veterans Affairs in the U.S. identified eight healthy habits associated with 24 extra years of life. These habits are hugely meaningful for preventing early mortality.

First is what I call the holy trinity: diet, exercise, and sleep. For diet, eat a mostly vegetable-based, Mediterranean-style diet. Regarding exercise, even a little goes a long way. While more is typically better, even five minutes a day makes a huge difference in mortality studies. The benefits appear to accrue asymptotically, so a small amount gets you most of the benefits.

Then there are the three “thou shalt nots”: no smoking, no drinking, and no drugs. The safe dose for all of those is zero.

Lastly, there are two points around mental health: managing social connections and managing stress. Those eight things combined make a massive difference in early morbidity and mortality. There is a lot you can do today through lifestyle choices without any medicine involved.

Daniel 01:20:07

Hopefully, we will soon be able to add more to that list of eight through new discoveries and the new drugs and therapies being developed by people like you.

Thank you for joining us on the podcast, Michael.

Michael Ringel 01:20:18

Thanks for having me.

In this interview, you’ll hear how insightful Jose is about deeply technical topics in biology, and you’ll see why Retro Bio was eager to bring him on as their head of theory (the only role of its kind in the entire biotech industry).

Our conversation is wide ranging, spanning a deep dive on Retro’s work to replace and engineer microglia to rejuvenate the brain and how our cells have the ability to turn back the aging clock but choose not to. We also covered the technological stagnation and why biological engineering is the new frontier of progress, as well as philosophical topics like transhumanism and how a future of total biological control might impact our values and way of life.

Retro Biosciences was seeded with $180M by OpenAI CEO Sam Altman to develop therapies to prevent and reverse age-related disease, and is widely recognized as one of the leading AI for longevity companies. Previously, we hosted Rico Meinl, the head of Applied AI at Retro, so make sure to give that episode a listen as well.

Watch on YouTube. Listen on Spotify or Apple Podcasts.

Chapter Markers

0:00 Intro

2:49 What is aging & why cells have a tough choice to make

9:02 When cells choose to reverse aging themselves

12:43 Cellular vs Organismal Aging & the magic wand experiment

18:31 What is reprogramming

22:40 How reprogramming plays a role in DNA damage repair

25:42 Do we already know how to cure aging? FOXO3!

28:37 How to cut through the complexity of interconnected biology

32:32 Why transcription factors are so great for intervening

36:49 Does a rejuvenation program exist already in the genome

38:51 Michael Levin: from thinking in terms of genes to morphogenesis

48:14 Tech stagnation and why physics is cooked

55:03 Why doesn’t the world look more futuristic

57:26 Transhumanism & asking ourselves what we want out of life

1:05:34 Do we need war for technological progress

1:09:31 Government role in science funding

1:15:06 How Jose became the Head of Theory at Retro

1:24:53 How AI might put software engineers out of a job, and push them towards biotech

1:27:28 What it takes to get a flywheel in biotech

1:29:04 Rejuvenation vs Prevention

1:34:23 Aging is the coolest hardest problem to work on

1:36:09 What does it take to cure aging

1:42:25 Delivery mechanisms for genetic therapies

1:48:53 Retro’s work to replace microglia and engineer them outside the body

1:59:10 Consciousness

2:00:39 Jose’s Origin Story

Transcript

0:00 Intro

[Montage]

Daniel 00:00:54

Welcome to the Free Radicals podcast, where we interview the scientists and builders working to dramatically extend human lifespan and bring about a sci-fi future where humanity has full control over biology.

Today’s guest is Jose Luis Ricon Fernandez de la Puente, author of the popular blog Nintil and Head of Theory at Retro Biosciences. Jose is a prolific blogger covering a wide breadth of topics, including economics, philosophy, progress studies, science funding, and longevity.

His writing has been published in A16Z Future, Works in Progress, and by the Adam Smith Institute. His work previously earned him a fellowship with Emergent Ventures, Tyler Cowen’s competitive program for intellectually ambitious projects.

In this interview, you’ll hear how insightful Jose is about deeply technical topics in biology. You’ll also see why Retro is eager to bring him on as their Head of Theory, the only role of its kind in the entire biotech industry.

Our conversation is wide-ranging, spanning a deep dive on Retro’s work to replace and engineer microglia to rejuvenate the brain and how our cells have the ability to turn back the aging clock but choose not to.

We also covered technological stagnation and why biological engineering is the new frontier of progress, as well as philosophical topics like transhumanism and how a future of total biological control might impact our values and way of life.

As context, Retro Biosciences was seeded with $180 million by OpenAI CEO Sam Altman to develop therapies to prevent and reverse age-related diseases. They are widely recognized as one of the leading AI-for-longevity companies.

Last week, we hosted Rico Meinl, the Head of Applied AI at Retro, so make sure to give that episode a listen as well. I am your host, Daniel Shur, and my co-host, though he couldn’t make it for this episode, is Eric Dai.

I hope you enjoy this episode of the Free Radicals podcast. Let’s start with an easy one: What’s aging?

2:49 What is aging & why cells have a tough choice to make

Jose 00:02:51

One definition of aging I like to use is a decline from a state of fitness. If you look at the process that gets you from age 30 to 80, you can see that it begins even before you are born. It is a slow accumulation of damage.

In parallel to that, there is also the process of development. You are building your body and becoming more than just a single cell.

From a biological perspective, you could say aging is the accumulation of cells falling apart due to entropy. In that sense, aging is the same thing that happens to a building or a car. The car ages as it falls apart.

The interesting part is that, unlike those systems, our cells are constantly trying to fight entropy. They are trying to repair themselves all the time. We get damaged and we fix ourselves.

From that process, we eventually get diseases and the phenotypes of aging, such as the loss of muscle mass or hearing, which end up resulting in death.

It doesn’t necessarily have to be this way. Many species don’t seem to age much. Axolotls and naked mole rats still die, but they don’t seem to decline like we do.

Some cells in culture, like induced pluripotent stem cells, are able to stay in culture while keeping a similar state for a very long time—much longer than normal cells. If you can do more repair than you receive in damage, you can stay unaged for longer.

Daniel 00:04:44

One of the things you’ve argued on your blog is that DNA damage is central to aging. Can you talk about what that process is on a cellular level?

Jose 00:04:54

All cells in your body have roughly the same DNA. It is constantly floating inside the cell across different chromosomes. Sometimes the double helix gets a nick on one side, or you get a double-strand break.

This happens from UV radiation hitting your skin or from toxins like cigarette smoke. The cells have to repair that damage.

When that damage to the DNA causes the way the DNA is folded to change slightly, it accumulates over time into aging. In a way, DNA damage is what makes the clock tick.

DNA damage is unavoidable. You can repair it faster or slower, and better or worse. Our cells can actually do more repair than their baseline level.

Many of the things people point to for extending lifespan, like calorie restriction or rapamycin, ultimately upregulate DNA damage repair. Across every species, better damage repair boosts lifespan.

Even in species that live longer than humans, like bowhead whales, they repair damage much better. There is a strong correlation: the longer you live, the better damage repair you do.

Intuitively, it makes sense. If you get this damage and the DNA has to be repaired, it acts a bit like a scar. There’s another theory on top of that put forth by David Sinclair and Lenny Guarente.

They argued that the very same proteins involved in DNA damage repair, like sirtuins or the polycomb repressive complex, are also involved in keeping the chromatin in its proper state.

The cell has to choose between repairing damage or preventing aging. The more damage there is, the faster the clock ticks.

If there were no damage, those proteins could stay in that healthy state. But because there is damage, they have to go do their job. This was mostly found in yeast, but in humans, there seems to be evidence of a similar process. Those chromatin support proteins move to different regions when there is damage.

Daniel 00:07:24

You mentioned a really interesting trade-off. When DNA damage occurs, the cell needs to decide how to allocate its energy between repairing the DNA itself—avoiding a mutation—versus correcting the chromatin structure.

Is it right that if there’s a lot of DNA damage, the cell might focus on correcting the DNA damage itself, causing the chromatin structure to degrade?

Jose 00:07:53

You can think of it as short-term gain for long-term pain. It is unclear why we evolved this way. There are theories in yeast regarding nutrient sensing, but it remains ambiguous because you can have both.

Another argument compares this to sleep. Why do we sleep? From an evolutionary perspective, it is risky; you have to shut down the entire system for hours. Perhaps you need to shut everything down to perform certain maintenance.

Maybe the reason we cannot stay young forever is that entering a rejuvenation or chromatin-fixing mode prevents simultaneous damage repair. This might make an organism highly vulnerable.

This could explain why species that live longer are typically those that are not heavily predated; they can afford to spend more time on repair. It is puzzling because our cells have the capacity to roll back their clocks, but they often choose not to.

Daniel 00:09:01

Tell me more about that. When do our cells choose to turn back the clock?

9:02 When cells choose to reverse aging themselves

Jose 00:09:05

The classic example is during embryogenesis. We have the egg and the sperm. Eggs are created at the beginning of a woman’s life and stay relatively stable. Sperm, however, accumulates more mutations and ages over time.

When they combine, that accumulated age is wiped out. This process, which some papers call “ground zero,” is where aging truly begins. Shinya Yamanaka discovered that certain factors involved in embryogenesis can also revert any cell in your body. You can turn on that program and essentially wipe out the age of those cells.

Other cells can do this to some extent as well. If you remove a chunk of the liver, it regenerates. The regenerated area actually behaves and appears younger.

We see this in other organisms like planaria. These are strange worms that, if chopped into pieces, can grow each chunk into a whole new worm. This regeneration seems to bring rejuvenation with it.

There appears to be a connection between regeneration, rejuvenation, and cancer resistance. Beyond that, in vitro experiments show that the niche where you place a cell can rejuvenate it. Fibroblasts, which are skin cells, seem to de-age when placed in a more elastic or youthful niche.

A similar mechanism occurs in the intestine. The intestinal crypts contain stem cells that constantly produce new cells to regenerate the lining, which is damaged whenever we eat.

When damaged, some of these cells can transdifferentiate or dedifferentiate. We see this more clearly in axolotls. If you remove their limb, the cells at the site dedifferentiate into a stem-like state. They then regrow the entire arm, and that new arm appears younger.

Daniel 00:11:34

This connection between regeneration and rejuvenation is very interesting. For something to be able to regenerate, you essentially need cells to change their identity.

Jose 00:11:47

Right.

Daniel 00:11:47

Cells need to sense what they need to turn into, and then they transform. Rejuvenation feels similar because as we get older, many of our cells lose their identity. We need to restore it. Is that the right way to think about it?

Jose 00:12:00

Yes. If you think of DNA damage as making the clock tick, then aging itself is the state of the chromatin. It is a matter of how that chromatin is folded.

When you look at patterns of gene expression in old cells, they are doing less of what they should. Fibroblasts in your skin produce less collagen as you age, and cells in your liver produce less albumin. They are failing to perform their specific jobs.

If you express transcription factors for that specific cell type, they appear rejuvenated. In that sense, the phenotypes of aging are essentially a loss of identity for that cell type.

12:43 Cellular vs Organismal Aging & the magic wand experiment

Daniel 00:12:44

When we think about aging, there seem to be multiple levels. There is cellular aging, which we just discussed, but then there is also the aging of the entire organism. I think of things like atherosclerosis, which is a higher-order natural occurrence. How do you think about these different levels?

Jose 00:13:04

Classical aging theory—focusing on pathways like IGF1, AMPK, rapamycin, and mTOR—was largely studied in unicellular life like yeast. Those pathways translate all the way to humans, but we also face emergent properties that arise when many cells work together.

This leads to issues like cancer, senescence, and atherosclerosis. C. elegans worms, which are commonly used in research, do not have stem cells, senescent cells, or cancer. As a result, they do not suffer from heart disease. Their aging is almost purely cellular.

In humans, we have these emergent forms of aging. In a recent blog post, I proposed a thought experiment called the “epigenetic magic wand.” Suppose you wave a wand and every cell in your body becomes 19 years old. Would you live forever, or would you only live for another few decades?

The answer is unclear. Plaques might still accumulate in your arteries. You could suffer a hydraulic failure where blood flow stops, resulting in a stroke and death, even if your brain and other organs are cellularly young.

We are still trying to understand how much cellular rejuvenation contributes to organismal rejuvenation. As a heuristic, if a certain type of damage already accumulates in children, it probably won’t be reverted simply by rejuvenating cells. You would have to remove those deposits directly.

Daniel 00:14:42

And that is potentially the case with atherosclerosis.

Jose 00:14:45

I’m not an expert in atherosclerosis in teenagers, but I know that people with familial hypercholesterolemia get plaques very early on. I think of these things as a balance: damage is constantly being produced and removed. If the rate of removal or fixing is higher than the rate of generation, you are in good shape.

For some diseases, the rates of degeneration are very high. In children with familial hypercholesterolemia, they will develop issues earlier. For most people, the rate is lower, so they may not develop plaques.

However, for some things, damage is permanent. If you lose your teeth as you age, they don’t grow back. You have to do something specific about it. The hair cells in your cochlea that allow you to hear eventually die. Even if you were to rejuvenate everything else, if those cells are gone, there is nothing left to rejuvenate. You have to put them back or create them.

The same applies to neurons. When they die, you aren’t making new ones all the time, so they won’t come back. Even the lens in your eye doesn’t turn over much. It is a protein structure that is made when you are born and stays there. If it gets damaged, that’s it.

There are many long-lived proteins, like elastin in the skin, that may have to be addressed separately from cells. We don’t know yet if rejuvenating an entire tissue will allow the cells to fix the structures in between. It might happen to some extent, but maybe not 100%.

Daniel 00:16:25

What do you think the priority should be? Should we be focused on curing cellular aging through things like reprogramming, or should we focus on fixing macroscopic damage and replacing missing cells?

Jose 00:16:42

We should focus on everything; there is room to do it all. At Retro, we are pushing a strategy of cell replacement while also trying to fix things in situ.

The field of longevity recently found its “this is it” moment with reprogramming. It is the one thing that seemingly and robustly reverses age. Many things can potentially slow down aging, but very few things actually reverse it in a way that remains after you stop the treatment.

In the past, people talked about antioxidants and telomeres, but this time is different. Those things never did what reprogramming does. Given that we know how to do it, we should focus on it now. Pragmatically, you can test these things in vitro and see if they are working.

Other things might be harder. For atherosclerosis, there are already companies working on removing plaques and damage. At Retro, we work on what we do because we saw an opportunity to impact disease by targeting aging biology in a way that no one else was.

We didn’t want to just make another marginal benefit. If we can replace all of your blood cells or the microglia in your brain, that is a unique contribution. Someone should do it.

18:31 What is reprogramming

Daniel 00:18:32

We’ve brought up reprogramming a bunch of times. For people who don’t know, can you explain what reprogramming is and why it is so amazing?

Jose 00:18:40

Reprogramming typically refers to Yamanaka or OSKM-based reprogramming. You take a normal cell, like a skin cell, and turn it into an induced pluripotent stem cell. This is similar to an embryonic stem cell; it is a cell that can become almost any other cell type.

The interesting part is that it also has its age wiped out. You can take a skin cell from an old person and a young person, reprogram them, and the results are very similar. The damage is gone.

The mutations remain, but cells have a remarkable ability to work around DNA mutations to some extent. This de-aging of the cells is a genuine bounce back. It stays that way. It’s real.

Daniel 00:19:40

You can even grow a whole new organism from that cell, and they will be young and then age like a normal organism.

Jose 00:19:46

That goes back to the early experiments by John Gurdon. In reprogramming, you add molecules or transcription factors to cells to get stem cells. Before that, experiments showed you could take a nucleus from the skin of an old mouse or a frog and put it into an egg.

This is essentially cloning. Dolly the sheep, for example, came from one such nucleus. Dolly wasn’t born old, so we knew that something happened to wipe out the age of the nucleus. Yamanaka found a way to make that process scalable and possible for other cell types without having to use eggs.

Daniel 00:20:33

The amazing thing with reprogramming is that we can turn a specialized cell into a stem cell, which can then turn into anything. It turns the age back to zero. This is really what spurred the newfound interest and funding in the longevity space, because it is so impressive and real.

Jose 00:20:54

We have known how to make induced pluripotent stem cells (IPSCs) and understand these processes for a long time. The idea of taking young cells made from IPSCs and putting them back into people is called regenerative medicine. This concept existed long before companies like Altos, Calico, and other aging-focused startups.

The field became much more exciting recently due to partial reprogramming. If you were to fully turn on these factors in a mouse, for example, the mouse would develop tumors and die. You don’t want to create stem cells of age zero; you want to maintain the cell’s identity while making it younger.

Alejandro Ocampo, one of our advisors and a pioneer in the field, found that if you express these factors briefly and then stop, you can achieve a rejuvenation effect without the cell losing its original identity. This suggested we could potentially apply this to all cells in vivo.

Companies like New Limit and ourselves began searching for safer ways to reprogram various tissues and cells once it was shown to be doable. Prior to reprogramming, there was no way to reverse aging. We have discussed slowing aging for decades through methods like autophagy, rapamycin, or genetic knockouts in worms, but reversing it was the “wow” moment that created all this excitement.

22:40 How reprogramming plays a role in DNA damage repair

Daniel 00:22:41

I want to understand more about what is happening with reprogramming and how it relates to DNA damage repair. With reprogramming, you are restructuring the chromatin to express the right genes for a specific cell identity. But then there is this other piece involving damage repair.

Jose 00:23:04

Mechanistically, it is still unclear exactly how reprogramming works. You can think of these factors as keys that enter the DNA to turn on a specific program. Instead of having to turn on 100 genes individually, you can turn a single key.

For example, turning on NF-kappaB activates the inflammation program. Turning on Oct4, Sox2, and Klf4 activates the pluripotency program, which shuts down the somatic program and establishes pluripotency.

During this process, there is a drastic increase in DNA damage repair capacity and an initial burst in mitochondrial function. IPSCs perform at least an order of magnitude more DNA damage repair across every pathway than somatic cells.

Is this DNA damage repair what causes rejuvenation? It is possible that if you induce enough repair, the cell can catch up and return to a youthful state, but this remains unproven. There is likely also something involved with reopening and loosening chromatin that was previously closed. We don’t yet know the exact sequence of events.

Daniel 00:24:33

Is it correct to assume IPSCs have higher damage repair because they are embryonic cells about to divide many times, which is stressful for the genome? Is it increasing its repair ability in preparation for all that mitosis?

Jose 00:24:51

Not quite. It is difficult to say because IPSCs inherently want to divide. If you take normal cells in culture that are also dividing every 20 to 24 hours, they still age. IPSCs do not age in vitro in the same way.

While their chromosomes can eventually become unstable, you can keep passaging them much longer than normal cells. I once tried to find the longest someone has kept an IPSC in culture, and it was several years.

The key lesson is that cells already possess the machinery for this repair. We didn’t have to edit any genes; the capacity is already there. The challenge is simply learning how to switch it on.

25:42 Do we already know how to cure aging? FOXO3!

Daniel 00:25:43

Sometimes when I look through longevity literature, it seems so complicated that we understand nothing. But then I find something that makes me wonder if it’s actually simpler than we realize.

You mentioned that we already know some longevity genes, like Foxo3, which extends the lifespan of various organisms. Is the primary challenge simply figuring out the right way to activate Foxo3 in humans?

Jose 00:26:11

Foxo3, also known as DAF-16 in worms, is like a key that turns on programs for DNA damage repair, autophagy, and antioxidant responses. Instead of targeting many individual genes, you can activate Foxo3 to achieve these benefits.

In humans, genetic studies show that individuals with specific Foxo3 SNPs tend to live longer because they have more active Foxo3. There was a recent paper from China by Guanghui Liu, who was a student of Juan Carlos Izpisua Belmonte—a fellow Spaniard in the longevity field who is now at Altos and was also Alejandro Ocampo’s PI.

They overexpressed Foxo3 in mesenchymal stem cells and injected them into monkeys. The monkeys appeared phenotypically younger, though not necessarily intrinsically de-aged.

If you were to maximally activate Foxo3, it might reverse aging, but there are likely more variables at play. Researchers have found that overexpressing Foxo3 in mice can lead to muscle atrophy because excessive autophagy causes the muscles to shrink. You could have maximal Foxo3 activity and still die.

However, if you had to choose one gene to increase throughout the body to slow or reverse aging, Foxo3 is the candidate that most experts would agree on.

Daniel 00:27:57

There is a process in our cells effectively trying to prevent the damage of aging through DNA repair. These processes maintain the correct genome and chromatin structure. We need to find ways to turn that up and manage any side effects that occur.

Jose 00:28:18

To be clear, there isn’t just DNA damage in the nucleus; there is also damage to the mitochondrial DNA, which causes them to function less effectively. We also see the accumulation of protein aggregates, and autophagy is another process that works less well with age.

DNA damage is a major component, but everything is connected. One of my pet peeves in biology is the idea that if you fix one thing, you fix others, and if you break one thing, you break others. It becomes difficult to determine what is causing what. To me, the right question is where the best point to intervene is, not simply what the cause is.

28:37 How to cut through the complexity of interconnected biology

Daniel 00:28:49

A lot of people, when they see how everything is connected, throw their hands up and say it is impossible to isolate anything. But you make the point that a hierarchy still exists and we can give primacy to certain things. What do you prioritize in aging?

Jose 00:29:08

DNA is a primary one. Consider a heuristic: the longer something remains unchanged, the more important it is. My analogy here is gravity. Gravity extends everywhere, so everything in the universe is attracting everything else.

While we can still use Newtonian mechanics to simplify things, consider the solar system. You can think of the Earth orbiting the sun, but that is just a mental model. You could also imagine the Earth as fixed with the sun orbiting it. The math is the same, but it makes more sense to have a hierarchy where the largest structures are most important. Galaxies function in a similar way.

This applies to turnover rates. If you have a mutation in your DNA, it is essentially permanent unless the cell itself is removed. In contrast, proteins turn over constantly. They are produced and cleared easily.

For example, liver fibrosis or a scar is constantly being created and destroyed. It is not a static mass of collagen; it is being added and removed continuously. This gives you a clue that if you alter the balance of cells involved in that process, the fibrosis would go away because the cells are capable of removing it.

The same is true for the chromatin state. For a cell to function healthily, it must have specific regions of chromatin open and others closed. This is determined by epigenetic marks, such as methylation groups attached to the DNA and histones.

Nucleosomes are the structures around which DNA is wrapped. Depending on the state of the histone tails, the DNA is more or less folded. These markers are constantly being added and removed stochastically.

Imagine a stretch of DNA for a fibroblast containing the collagen gene, COL1A1. The cell makes collagen because that area of DNA is open, allowing RNA polymerase to transcribe it. It remains open because of specific histone marks and a lack of methylation.

Transcription factors bind to keep it open, but they only stay for a few seconds at a time. They are not covalently bound; they move back and forth via hydrogen bonds. Histone marks are added and removed by histone deacetylases, and DNA methylation is managed by DNMT enzymes.

The epigenome is turning over, but it does so very slowly. If you alter the epigenome, everything else downstream conforms to it, rather than the other way around.

Conversely, if you put one broken protein into a cell and wait, it will eventually be degraded and replaced. It doesn’t matter as much in the long run. DNA and chromatin are the primary drivers of everything downstream, with the exception of proteins that do not turn over, like those in the crystalline of the eye.

32:32 Why transcription factors are so great for intervening

Daniel 00:32:33

Everything is connected. We need the right chromatin structure for the cell to have the correct identity. That structure leads to the translation of the right proteins, but you still need the right transcription factors to exist in the cell to bind to those chromatin sites.

I think about the path dependency of aging. Eventually, a cell could get so unhealthy that even if we restore the right chromatin structure, you might not be able to recover the correct cell identity.

Jose 00:33:09

If you fix the chromatin structure or these transcription factors, you have to consider where they come from. They are produced from open DNA being read. If you were to fix the chromatin itself, you would likely address everything else as well.

Cells exist in an equilibrium state. If you compare a young cell to an old cell, they differ in many ways. The metabolites, protein abundance, cell shape, stiffness, and chromatin are all different.

This is all part of an equilibrium that slowly shifts toward an aged state over time. The question then becomes how to shift it back to youth. In principle, there are many ways to accomplish this. In fact, there are many ways to reprogram a cell without using Yamanaka factors.

If aging is characterized by a loss of identity for a specific cell type, you can try to reinforce that identity. For example, hepatocytes in the liver are defined by a transcription factor called HNF4alpha.

A study in mice showed that expressing HNF4alpha in the liver makes the organ younger. Fibrosis disappears and it becomes more functional. You can achieve these effects without Yamanaka factors simply by helping a liver cell be more of a liver cell.

In that case, transcription factors have great causal power to rejuvenate the cell. In contrast, a random protein like albumin does nothing if you introduce it to a cell; it just makes more albumin.

Transcription factors are the programming code for cells. While there are feedback loops, TFs are the tools that are truly able to shift a cell’s state to other places.

Daniel 00:35:36

Is that because TFs act at the epigenetic level?

Jose 00:35:40

It is because they act in a coordinated way to shift entire programs across the cell. When a cell undergoes a process like inflammation and needs to produce cytokines like IL6, IL1, and TNF alpha, it would be inefficient to build separate regulatory elements for every single component.

Instead, you have programs where you turn one thing on. Transcription factors bind to specific motifs in the DNA—almost like a CRISPR guide—to turn on those genes. It is a very clean process.

This means you can identify which genes are controlled by a specific TF by finding those motifs in the DNA. You are switching on an entire genomic program.

If you wanted to activate 100 genes with CRISPR, you would naively need 100 CRISPR guides, which is too much work. With one TF, you can turn everything on at once.

36:49 Does a rejuvenation program exist already in the genome

Daniel 00:36:51

Do you think a rejuvenation program exists naturally—a set of transcription factors that will turn on the program we care about—or will we need a synthetic transcription factor to drive a new kind of program?

Jose 00:37:10

It depends. We seem to have a regeneration program that carries some de-aging effects. Normal cells don’t perform Yamanaka reprogramming on their own; it typically only happens in the uterus in a very controlled environment for the egg and sperm. We are forcing it to happen elsewhere.

In principle, we could force a program from the liver to happen somewhere else because the DNA is the same everywhere. The transcription factors we have could work, but some are short-lived or ineffective when introduced to a cell.

This is where modifying TFs becomes interesting. Instead of using four Yamanaka factors, you could modify them to be more potent so that you only need two or three.

You can either combine existing TFs or use AI to make the ones we have more potent. These proteins have specific domains; we know what makes them enter or exit the nucleus.

We can rationally design them, and with AI, we can go even further. You can use existing factors for a proof of concept and then tweak and improve them when creating a final medicine for humans.

38:51 Michael Levin: from thinking in terms of genes to morphogenesis

Daniel 00:38:53

You brought up planaria earlier, and I’ve noticed several references to Michael Levin on your blog. Can you talk about how Michael Levin influenced your thinking about biology?

Jose 00:39:02

I think I discovered Michael Levin before everyone else. Perhaps people got into his work because of me.

Daniel 00:39:08

Now even normies are into Michael Levin.

Jose 00:39:10

I discovered Michael Levin on YouTube a while back and was very impressed. I suggested his work to Allison Duettmann at the Foresight Institute, and it spread from there.

Michael Levin thinks in terms of higher-level structures rather than just genes. Initially, he focused on bioelectricity, which led to the ideas of morphogenesis and morphostasis. He views cells as existing in a state where they react to and adjust their environment in a sort of dance. When that dance is broken, you get aging.

Levin points out several experiments that reframe how we think about biological systems. One way of thinking is to look for a broken gene to fix or determine which genes to overexpress. However, Levin highlights many phenomena that involve no genetic changes at all.

For instance, if you take a cancer cell and put it into an egg, you can grow a healthy mouse. The mutations are still there, but the cancer disappears. While the mouse may eventually develop cancer because the mutation “primes the pump,” the outcome is not deterministic. Destiny is not written in the genes.

He also points to the strange nature of limb regeneration in lizards and axolotls. When they lose a limb, it grows back. How does the organism know to grow exactly five fingers or reach a specific length? Where is that information encoded?

While the genome created the embryo, using it to explain regeneration is like trying to use quantum mechanics to do mechanical engineering. It is technically true, but not useful. We need to look at a higher level of structure, specifically the signaling between cells.

This takes us back to development—how one cell becomes an entire being. Cells signal to each other to determine their position, whether they are in the head or the legs. From this perspective, aging is a disorder where cells lose coordination.

Levin recently published a paper describing aging as a “lost astronaut.” Initially, every cell is working together on the project of building and developing the being. Once development is finished, that shared directionality is lost.

Challenging the primacy of genes allows us to think beyond genetics when fixing biological systems. This is important because gene therapy, while powerful, cannot yet be delivered everywhere. Working at the cellular level offers other paths.

Take fibrosis, for example. If you ask a group of scientists how to cure it, they will offer different ideas based on their approach. One might suggest a GWAS to see who is prone to fibrosis, or use single-cell RNA sequencing to find a gene to inhibit.

Another way, inspired by systems biology experts like Uri Alon, is to look at cell turnover. This shifts the focus from genes to cells. You could potentially cure the condition by depleting myofibroblasts, which produce collagen, or by boosting the activity of macrophages.

These approaches don’t touch the genes; they target the cells. We see similar systemic perturbations in experiments where young blood is exchanged for old blood, or in bone marrow transplants. These methods change the system without having to edit genes one by one. It opens up new frontiers of thinking.

Daniel 00:43:47

The way I initially learned biology was through mechanisms and complicated flowcharts of protein activations. While those are extremely useful, they don’t provide an abstraction layer that helps you theorize more broadly about biology.

Jose 00:44:08

The genetic view is powerful and easy to explain. In many diseases, if everyone with a specific mutation gets the disease, you can trace that chain step by step. That is very clean and satisfying to the human brain, but aging isn’t like that.

There is no single gene you can turn on to live forever. Biology is a “hot mess” compared to computer science, math, or physics, where entities and laws are crisply defined. In biology, things are rarely that clean.

You can do two things: you can try to find small areas where you can impose a model, or you can give up and decide we need AI to solve everything. Alternatively, you can decide that we don’t need to understand every small detail.

We can build another level of abstraction on top and observe what is happening there. We don’t always need full mechanistic evidence for everything.

It is similar to how we don’t need to know how every individual molecule of water is bouncing around to see that a pot of water is boiling. We can see the bubbles and understand the state of the system without knowing the exact path of every molecule.

Daniel 00:45:51

Levin emphasizes that one of the unique aspects of biology is the nature of repair. When you’re repairing a car, the vehicle is inert; you’re simply trying to fix it.

But when you’re trying to repair an organism, you’re dealing with individual agents at every level. Each cell and each tissue is trying to accomplish its own goals and will respond to your actions.

Levin frames it as thinking of the body as an alien we’re trying to communicate with. How do we give it the right goal?

Jose 00:46:24

In the car analogy, a vehicle takes damage passively until it simply breaks. Cells experience random damage as well, but they don’t act randomly.

If a human being is punched, they react in a predictable way to protect themselves regardless of where the damage came from. Cells do the same, using identical mechanisms. Even though damage is entropic, cells react predictably.

They can also enter states that resemble cellular trauma. Much like people with trauma who are locked into self-harming behaviors, aged cells suppress the mechanisms that would fix damage.

Chronic activation of inflammatory pathways, for instance, downregulates repair mechanisms. These cells become trapped in a state where they are damaged and can’t fix it because they are actively stopping themselves from doing so. They are trapped in a local minimum instead of jumping to a better equilibrium.

Rejuvenation involves improving coordination within and across cells to give them a high-level goal: to play better together. Aged cells are almost trying too hard to maintain a faulty identity. They become locked in and lose their flexibility, adaptability, and resilience.

48:14 Tech stagnation and why physics is cooked

Daniel 00:48:17

Stepping away from molecular biology, you’ve shown a lot of interest in the roots of progress in science. Can you talk about your views on progress? Are we currently in a state of technological stagnation?

Jose 00:48:35

I was thinking about this even before the progress studies movement. My interest came from two conflicting narratives: Ray Kurzweil’s singularity, where everything is accelerating, and Peter Thiel’s stagnation.

I wanted to find the individual truth between those sweeping statements. I’ve spent time looking at various areas of technology to see if Moore’s Law is slowing down or if the cost of electricity is decreasing.

Biology is making progress, but in areas like fundamental physics, not much has changed in a long time. We might be nearing the end of discoveries in physics. While some things remain to be resolved, they likely won’t lead to technologies like faster-than-light travel or time travel.

Daniel 00:49:41

Wait, really? Even given infinite time, you don’t think there are deep truths about physics left to discover that will enable incomprehensibly advanced technology?

Jose 00:49:56

I don’t think so. Sean Carroll has a paper on this topic attempting to prove this formally. Physics is more likely to be completed than other fields.

The unification of quantum mechanics and general relativity has not been done, so there is something left to do there. There are phenomena involving very small things moving very fast or massive objects that require that explanation.

However, for the phenomena we observe in daily life that could be used for useful purposes, what we see now is likely all there is. There are still things to unfold in fields like material science—we have better battery chemistry now, for example.

There was no fundamental physics discovery involved in that, but we still achieved improvements. At some point, we discovered electromagnetism and the nuclear force, giving us nuclear energy and magnets. I don’t believe there is a new force waiting to be found that would allow us to do something radical.

Daniel 00:50:58

That’s very disappointing.

Jose 00:51:00

It is sad.

Daniel 00:51:00

I don’t believe you.

Jose 00:51:01

It’s like the Metallica song, “Sad But True.”

Daniel 00:51:06

It’s funny because my initial interest in longevity stemmed from a childhood fascination with physics and science fiction. I wanted to visit other planets, meet aliens, and discover the secrets of the universe.

Since we don’t have faster-than-light travel and space exploration takes a long time, I realized we would all have to live much longer. Many people in the longevity field share this motivation.

They want to see everything in the universe and understand deeper truths. It’s ironic to think that the fabric of reality might not actually have that much more to discover.

Jose 00:51:48

There are certainly strange things about the ground truth of reality yet to be figured out, such as how consciousness works. Perhaps meditation will eventually provide that answer.

But the physics portion seems very stable. I’m open to changing my mind, but I occasionally see claims about things like reactionless engines, such as the EM Drive.

These claims arrive like comets, then they pass away. The likelihood of them being true is small because our current laws of physics fit our observations incredibly well.

Where could new phenomena be hiding? Before we understood magnets, there were observations of strange magnetic stones and people wondered why they moved. That led to new technologies and the discovery of a fundamental force. Today, is there any mystery of that scale left?

Daniel 00:52:57

What about dark matter? It’s a huge portion of the universe.

Jose 00:53:01

I’m open to exploring what dark matter actually is. Some claim it doesn’t exist as a substance but is actually a force. Either way, can we use it for anything? There might even be some dark matter in the room with us right now.

Daniel 00:53:19

Is the dark matter in the room with you right now?

Jose 00:53:22

Dark matter interacts very weakly with normal matter. What are you going to do with it? I guess you could use it for large-scale interstellar galaxy engineering, but short of that, its utility seems limited.

Daniel 00:53:39

We often think that we’ve gotten most of the good stuff out of physics. There is obviously a lot left to do in engineering, but what about other fields? When it comes to stagnation, how do you see the rest of science?

Jose 00:54:00

Biology is currently the endless frontier. There is so much happening there that we don’t yet understand because it is so intricate. It isn’t simple, and you cannot easily break it into broad categories.

At the same time, it isn’t just “stamp collecting.” If you take stars, for example, you can essentially catalog them and say, “Yes, there they are.” That isn’t particularly exciting; it’s just one more star.

In biology, the interactions between all the small components are incredibly intriguing. There is still plenty of ground to cover because we know how much we don’t yet know. We see this with longevity research. Some animals can regrow limbs.

Daniel 00:54:47

Limbs.

Jose 00:54:47

We cannot grow limbs yet, but there is no physical reason in principle why we couldn’t eventually figure it out—how to grow an arm, or even a third arm. That remains to be done.

Aside from that, we have AI, which is also advancing rapidly these days.

55:03 Why doesn’t the world look more futuristic

Daniel 00:55:05

Another claim Peter Thiel makes is that we’ve basically only seen progress in the world of bits versus the physical world. In the physical world, everything looks the same.

We didn’t get flying cars or major breakthroughs in manufacturing; we just didn’t get much. Perhaps now we are finally seeing cool things happen because of AI.

Jose 00:55:26

To some extent, yes, but I’m not entirely sure about that. There is an idea that the future should look like the future. If you look at cities like Chongqing, Shenzhen, or Singapore, they actually look like the future. That is because they chose to pursue it. The US could look like that if people really wanted it to.

If you look at this room, for example, the camera looks like it could be from fifty years ago, but it has much better optics now. We often arrive at convergent shapes that are simply optimal for what they do.

I touched on this in my metascience article two years ago. If you take the laws of physics and the constraints of being human, some designs follow semi-deterministically. Take a restaurant, for example. Human beings like food and gathering together. We are a certain height, so we need tables and surfaces that are easy to clean.

Restaurants haven’t fundamentally changed in a long time, and as long as human beings remain as they are, they likely never will. You can add bells and whistles like conveyor belt sushi, but the basic idea remains the same.

Stagnation doesn’t mean no one is trying; it might mean we have reached an equilibrium or an optimum for many things. Take flying cars. People have been trying to build them for a long time. It is only recently that the power-to-weight ratio of motors has reached the point where flying cars can be affordable or practical.

The same applies to self-driving cars. I arrived here in a Waymo, which we didn’t have ten years ago. It is still a car, but it is new. Why does a car look like a car? Wheels are a great invention, and you need space for people. People like comfortable seats. When you add those constraints together, you end up with the designs we have.

57:26 Transhumanism & asking ourselves what we want out of life

Daniel 00:57:29

We have a lot of constraints because of our biology. What do you think of transhumanist ideas? Do you think fundamental aspects of human biology will change in our lifetime?

Jose 00:57:42

Aging is the part I’ve thought about the most. If we could live radically longer lives, how would that change people? Right now, everyone follows a specific life arc: you grow up, find a career, settle into an identity, and eventually die.

If you could live much longer, you could have multiple life arcs. You could be a Buddhist monk for ten years, then a movie director, then go back to software engineering.

Some people say that death gives life meaning, but I don’t believe that. If we lived longer, people would just have more fun and take things more lightly. Instead of hyper-optimizing everything because you only have so many years, you could just chill and live life.

You could finish one life arc and then start another, changing over time. For women, for example, it would remove the pressure of the biological clock.

Life would be less driven by external constraints imposed by biology. It would feel more in the moment because you wouldn’t have the existential fear that time is constantly ticking away.

Daniel 00:59:09

I’ve always disagreed with the argument that death gives life meaning as well. Life itself gives life meaning. We find things we enjoy and we create meaning.

However, it does raise an interesting question: the constraints of our biology give rise to many of our values. We like food because we need food, and then we develop fancy cuisine on top of that. Everything stems from the identity of our organism.

As we adjust those biological boundaries, what do we turn into? What do we end up caring about in the end?

Jose 00:59:54

I’ve thought about this occasionally, though I haven’t written about it at length. We have preferences—things we like and things we don’t like. Some of those are informed by our biology.

Take modern art, for example. I used to think it was just nonsensical lines. Then I tried to understand it and eventually saw what the artist was trying to achieve. Suppose you generalize this.

Suppose you had a magic wand that could change any preference you have. You could decide to dislike peppers, choose to like all foods, or stop liking food entirely. Some people wish they liked food less so they could be thinner; they wish they had that preference.

Imagine someone wishing they were bisexual so they could be attracted to everyone, or someone wishing they were more introverted or extroverted to suit their work. You could change these things instantly.

What does that do? We often take ourselves for granted and make choices based on who we are. But if who we are is up for grabs, what happens? On what grounds do you choose?

The classic conservative argument for many things is that human nature is incompatible with certain structures. People choose what to do, or even set policy, based on those constraints. If you don’t have those constraints, the question of what you should want becomes quite intriguing.

I don’t have a good answer for that yet. There are things I would not want to change. I very rarely lie. I have strong values around being honest and having high integrity. I would not want to change that. Even if I could press a button to lie 50% more, I would not press it.

You have core things that define how you want to live your life. Everything else—even having four arms—wouldn’t change the fact that you’re the same person. Those core things I would not change.

Maybe the answer is that you keep your core values and then stay playful with your biases. You could decide to become the kind of person who is into wine and develop an extreme taste for fine vintages.

Or you could choose to become a monk, throw away all earthly interests, and meditate all day. You press a button and you don’t like food anymore because you want to do a “monk arc” for ten years.

I don’t think people will all convert to being the same. They will have arcs, much like in songs or dances. There are sequences and moves, with micro-moves within them, but they don’t all result in the same thing.

Daniel 01:02:48

When I think about the end goal of biomedical technology, I see it as giving us increasing freedom to make these choices. With an indefinite lifespan, time is no longer as limited in terms of all the different things you want to experience.

You’re no longer limited by disease or your health. There might even be another layer involving psychological interventions. People make choices, but they often have trouble implementing them.

What if there was a drug, an electrical stimulation, or some other method that allowed you to choose? If you wanted to be bisexual, you could receive electrical stimulation and now you are. We can find ways to give ourselves more leverage over our choices.

Jose 01:03:34

My ultimate take is similar to what I said about cells earlier. You exist in an environment, and ideally, you are well-adjusted. You fit in where you are, and as the world pushes on you with incentives, you can push back.

For example, if you’re in venture capital, you will probably start a podcast. If you have a podcast, you’ll probably do VC. Those things go together. You could fight it and wonder why you want to do it, or you could just go with it.

Hopefully, the result is that people become better adjusted to their environment. If someone wishes they were more of a certain trait to fit in, they will be able to achieve it.

Ozempic is the first button to press: “I wish I was thinner.” Granted. Next might be, “I wish I was smarter” or “I wish I was more reliable.” People will be able to try these changes and see how they react and how they fit in their environment. Eventually, they will just be happier. Ultimately, that is what everyone wants.

Daniel 01:04:44

I agree with your point that we wouldn’t all converge to the same thing. People simply want different things. Giving people more ability to choose the things that make them unique would be amazing.

Jose 01:04:54

I like using dance metaphors. If you’re doing a partner dance and you make a move, maybe you repeat it because it’s fun, but eventually it gets boring. Something else arises and you do something new.

Even with the same person, you could go for hours and new things come up. If everything were the same, someone would decide they don’t like being the same and they would become different. You end up with a constant flux between sameness and difference.

Some people might end up being the same. Cults form when people want to be like the Borg and become identical. That would be interesting to see.

1:05:34 Do we need war for technological progress

Daniel 01:05:37

Regarding progress, I saw you wrote some counterarguments to the claim that government pushes technological progress forward, specifically the idea that we need war for technological advancement. What do you think of that?

Jose 01:05:55

That argument is made in various contexts. For example, some say that because so much money went into the Second World War, we got many innovations out of it. Therefore, it must be true.

But we should consider the state of those innovations prior to the war. Many were already in development. Additionally, we have to ask: what would those scientists have been doing in the absence of war?

There are ways to look at this using econometrics. A researcher named Alexander Field looked at productivity growth and found that war does not necessarily lead to an increase in productivity, despite what you might imagine.

I prefer looking at both aggregate statistics and the history of individual technologies like radar or lasers. Statistics come from models of the economy, not the real economy, so you have to look at both simultaneously.

There are cases where the argument might be true. Nuclear energy, for example, happened when it did because of the government. However, you could argue that because it was driven by the government, we actually have less nuclear power now than we would have otherwise.

Daniel 01:07:13

And we had a ton of nuclear weapons.

Jose 01:07:15

Exactly. Nuclear power and nuclear weapons became forever tainted. They are viewed as scary, explosive, unsafe, and military-driven because they were pushed into production so quickly.

In a world without war or nuclear weapons, the development might have gone slower, resulting in safer reactors. People might not have panicked about nuclear energy, and we could have more of it now than we do.

This is a hypothetical, but given that much of the opposition to nuclear power stems from the mental image of a powerful bomb, it is plausible that people would be more accepting without that association.

The same applies to disasters like Chernobyl, which involved a bad reactor design. These catastrophes happened because the technology was rushed. I don’t know for certain if things would have gone more safely without the war effort, but you can imagine it.

Another example is the iPhone. Economist Mariana Mazzucato wrote a book making the claim that the government essentially invented the iPhone, or at least funded many of the technologies leading up to it.

When you examine those claims, they aren’t quite true. Because of its size, the government is involved in many things. For instance, the U.S. government was involved in Apple’s early days because they gave a guaranteed loan to a bank that then gave money to Apple. But Apple already had money; would they have failed to exist without that specific loan? Probably not.

Siri did come from a government grant to the Stanford Research Institute (SRI), but is Siri really the core part of what makes the iPhone successful?

Daniel 01:09:08

It’s the worst part of the iPhone.

Jose 01:09:10

You can trace back many things, like LCD screens, in a similar way. Perhaps a charitable reading of Mazzucato’s book is that it highlights the many ways the government was involved, but she overstates the case by claiming that role was central. It is very hard to determine the counterfactual of what would have happened in its absence.

1:09:31 Government role in science funding

Daniel 01:09:35

This reminds me of when I was young and read Frédéric Bastiat, the French economist. He has a quote about the costs that are seen and the costs that are unseen. This comes up constantly with government intervention.

You don’t know the counterfactual. When the government is this large, they are involved in everything. Take the NIH, for example. Virtually every drug developed is a result of government funding, but that’s because we live in a paradigm where the government is the primary driver of fundamental research in biology.

What do you think of the current state of biology research?

Jose 01:10:16

Because we lack easy natural experiments, it is difficult to measure productivity growth in many areas. One impactful example is stem cell research. Japan leads in this area partly because the US banned embryonic stem cell research at one point. Even with private funding, that blunt prohibition significantly hindered research.

Basic research is often exploratory and random. You never know what will come out of it, and much of it doesn’t seem to lead anywhere immediately. It can be very niche and specialized, which leads to questions about its immediate value.

There is also the ongoing discussion about the reproducibility crisis in science. If a study doesn’t replicate, it isn’t necessarily because the data was false. It could be that the replication attempt was poorly executed or that the original experimenters omitted a crucial detail in their documentation.

If we slashed the NIH budget by half, we would eventually see the impact in patents and trends. It might be similar to Twitter under Elon Musk—the company was cut in half, but it still functions. We might find we don’t need the same level of overhead.

For problems like cancer or Alzheimer’s, we could fund research more directly. Instead of just throwing money at various projects, we could use Focused Research Organizations (FROs) with specific goals. The government has done this successfully in the past with moonshots.

SpaceX is a perfect modern example. They set out to make reusable rockets and solved the technical problems to get there. In contrast, the “War on Cancer” hasn’t felt like a moonshot. It hasn’t been an iterative process of building toward a specific goal; it’s often just a series of isolated research papers.

We should also experiment with how we structure research institutions. Currently, the average age at which an academic receives their first grant is around 40. Historically, scientists like Louis Pasteur became professors in their early 20s.

We should try giving fresh PhD graduates their own labs. Youth might bring more creativity and a lack of fossilized ideas. While they might lack the wisdom of older scientists, we won’t know the impact unless we give them the opportunity. Success in such an experiment wouldn’t fix everything, but it would provide valuable data for specific fields.

1:15:06 How Jose became the Head of Theory at Retro

Daniel 01:15:11

How do you think about your role as Head of Theory at Retro? It is an incredibly cool title. You mentioned earlier that there was a backstory to how you ended up with it.

Jose 01:15:28

Before joining Retro, I was considering my next move after my previous project, Rejuvenome, ended. I was on my second O1 visa and needed to find a new role to stay in the country.

I have a diverse background in data science, electric cars, AI, and biology. I even worked at Twitter for a period after it was acquired. I eventually decided that I should be a founder and build a biotech company based on what I had learned.

I was specifically interested in Alzheimer’s disease. I wanted to understand why antibody drugs kept failing. After developing an idea for an Alzheimer’s company, I sought advice from Joe Betts-LaCroix, the CEO of Retro. We had met previously at a conference.

When I visited the Retro space, I was blown away. The labs were built inside shipping containers with custom HVAC systems designed by Joe himself. It felt like the biotech version of seeing the Tesla tent assembly line or the SpaceX facilities.

It was unconventional and inspiring. I asked if I could join the team, and that was that. I just wanted advice initially, but I ended up finding a home there.

Daniel 01:17:42

The experience of showing up and seeing where the rubber meets the road with building stuff is remarkable. It’s funny how much work goes into actually doing biology.

Jose 01:17:59

When we started Rejuvenome, Adam Adelstone and I were thinking about how to start a lab. We considered outsourcing to a CRO, partnering with an academic lab, or building our own, which seemed complicated.

The idea that you could take a warehouse and turn shipping containers into labs never crossed my mind. That’s what agency is: thinking of the unthinkable. I realized I needed more of this in my life.

Daniel 01:18:43

I’ve met people who have performed genetic engineering on themselves. They use electroporation to put plasmids into their muscles, specifically for follistatin.

There is a company doing that in Prospera now. That is true agency. People in Silicon Valley are electrocuting their muscles to insert plasmids.

Jose 01:19:10

That was the original story. When I considered what I could contribute to Retro, I thought about my ability to synthesize literature and my background as a software engineer. The CEO suggested I become the Head of Theory.

We have since semi-abolished job titles, using them only externally to explain what we do. When I joined, I helped modernize our computational biology infrastructure and established better tools and libraries.

Recently, I’ve been building internal tools like Retro OS. It’s a tool that handles data visualization, label making, sample management, and more.

Primarily, I help people think through their options for experiments. We discuss whether a program is worth starting, the market for osteoarthritis, or new opportunities from other companies. I provide the information they need to do their jobs, either by building visualization tools or doing the research myself.

Daniel 01:20:58

Once superintelligence arrives, will you be the most replaceable person or the least replaceable person?

Jose 01:21:06

I’m not sure. So far, I haven’t found ChatGPT or similar tools very helpful for my work. I mostly use them as a “second Google.”

My work involves thinking about things that aren’t currently on our radar. Once you know which question to ask, an LLM can give a decent answer.

However, how do you know which questions to ask? If you already knew you had a problem, it would be easy to solve. I spend my time thinking about the things we, as a company, are overlooking.

Daniel 01:21:49

We talked about agency before. It seems the most powerful aspect of thinking is having the agency to choose the right questions and decide what to focus on.

Jose 01:21:59

I recently wrote a blog post about how to be more agentic. Sometimes it involves a bit of serendipity.

Regarding the Retro OS tool I mentioned, there was a week where I felt like things were going fine and I wasn’t sure what to do next. While walking around the block, I realized I could just build this tool.

It started as a small solution to a minor problem and eventually grew into something much larger. You have to be in the right environment and pay attention to what is happening. Sometimes, you have to stop trying in order to get what you want.

Daniel 01:22:42

You mentioned Retro OS, which makes me wonder if we will one day have a human OS. Will we have a software layer for our biology?

I wonder if we will reach a level of engineering control where we can alter our biology as easily as writing code—curing cancer or even giving ourselves wings. What do you think that would look like?

Jose 01:23:16

It’s unclear. A “human OS” implies personalization, but deep down, human biology is quite universal. One version of that question is how much we could eventually change ourselves.

Aging might be the easiest target because there is a natural path from young to old that we can try to reverse. Something like growing a third arm is much harder because it doesn’t happen by default.

If we were to attempt that in the future, you would likely need to attach a bioreactor with the right growth factors to trick the environment into growing a new limb.

Biology is somewhat plug-and-play, so nerves might eventually connect and make the limb usable. Michael Levin published a paper on trying to regrow limbs in animals using bioreactors and growth factors. We are in the very early stages of that research.

1:24:53 How AI might put software engineers out of a job, and push them towards biotech

Daniel 01:24:54

We’ve covered a lot of ground. Is there anything else you’d like to talk about?

Jose 01:24:58

Before working in biotech, I was working in AI. I left just as everyone else was joining because it felt too crowded. There are so many smart people in AI, and it feels like the money, attention, and talent are being absorbed by that field. These days, it feels a bit lonely in biotech. I want everyone to come join us.

Since so much AI development is focused on automating software engineering, and there are so many talented people in that sector, I predict that software salaries will eventually decline as the market crunches. Those people will then have to find work elsewhere. I hope that “elsewhere” is biotech, because we truly need more people working on these problems.

Daniel 01:25:47

That is one of the primary goals of this podcast: to reach a wider audience and showcase the exciting research happening in longevity biotech. We hope to inspire software engineers to enter this space.

It would be amazing if economic factors pushed people into biotech. It would be even better if it wasn’t due to declining software salaries, but because bioengineering salaries were increasing.

Jose 01:26:19

Biotech has different dynamics than software. Software often has power law dynamics where you can hack something together and have an unbounded upside. Even successful companies like Eli Lilly, despite the GLP-1 boom, are only a fraction of the size of Nvidia.

If the software market becomes smaller, biotech becomes relatively more attractive. Salaries are determined by productivity, but it is difficult to identify a “10x” or “100x” biologist because the feedback loops are so long.

In software, you can identify a talented engineer within a few months, allowing them to command a higher salary. In biology, even the smartest person on earth has to wait years or even a decade for a drug to be approved to prove their value. While biotech doesn’t always offer software-level money, it offers deep meaning.

1:27:28 What it takes to get a flywheel in biotech

Daniel 01:27:33

What needs to happen to create the same flywheels in biology that we see in tech? One obvious factor is the need for faster feedback loops.

Jose 01:27:44

We need to work on the right problems. For a long time, biotech and pharma focused on specific drugs for small patient populations that couldn’t be scaled. GLP-1s represent a revolution because they work for almost everything; obesity affects so many other conditions, from arthritis to muscle loss.

Since you can sell those drugs to everyone, you have much higher margins. To me, the most obvious target is aging. Most health complaints are age-related, so if we had an aging drug, the market would be everyone on earth. You would no longer be limited to a specific cancer population or a small genetic subset.

At Retro, our strategy is to build something that works for a specific condition, but can then be applied everywhere else without changing the fundamental approach. The only way to make a biotech company that looks like a tech company is to target the biggest possible market. Aging is that market.

1:29:04 Rejuvenation vs Prevention

Daniel 01:29:09

You brought up an interesting point: there is currently no clinical pathway within the FDA to get an aging drug approved.

Jose 01:29:18

I’m not so sure about that. If you actually had a drug that worked, I think there would be a way to get it through.

Daniel 01:29:22

If we had something that treated aging, we would certainly find a way to get it approved, but it is incredibly hard to demonstrate. That is why companies like Retro focus on treating a specific disease and then extrapolating that toward rejuvenation.

It makes me wonder if there is a fundamental choice between rejuvenation and prevention. Is it possible that prevention would actually be easier, and we are simply barking up the wrong tree?

Jose 01:30:01

Prevention is potentially easier to achieve, but it is much harder to prove. If you give a treatment to someone with osteoarthritis and the condition disappears in a month, you can get approval very quickly. If you tell someone they are aging 10% slower, a trial might require 5,000 people and many years to show results.

Reversal is very appealing because you can see a step-function change. If someone has Alzheimer’s and then they don’t, that is much easier to measure than a change in the slope of decline. From a company-building perspective, reversal is more attractive.

At Retro, we have a treatment involving the replacement of microglia in the brain. In theory, applying this to someone with Alzheimer’s might stop the disease in its tracks. Personally, I prefer taking big swings. They may be radical, but if they work, the results are apparent quickly.

By aiming for large effect sizes, you avoid the need for massive clinical trials. If a treatment doesn’t work in a few people, you don’t need a large trial to tell you it failed; you just move on to the next big idea. While these radical approaches add scientific risk, they allow for smaller trials with much stronger signals.

Daniel 01:32:02

The concept of rejuvenation is very appealing. If we can figure it out, it would be better than anything else because we could save everyone.

However, I keep thinking about the damage of aging. Many issues are emergent from cellular aging. While atherosclerosis might be a mechanical exception, fibrosis is an example of an outward manifestation of the loss of function in individual cells.

Even if we rejuvenated individual cells, we might not solve these emergent phenomena. Conversely, if we could stop cellular aging in a 25-year-old, would we prevent almost everything from happening?

Jose 01:32:52

We don’t know how to do that yet. At Retro Biosciences, we focus on interventions that genuinely target aging, have a viable market, and are technically feasible.

We don’t know how to cure cancer yet, but we might be able to cure Alzheimer’s. If you had a potential cure for Alzheimer’s, would you not pursue it?

Our treatment involves creating young cells from iPSCs and placing them in the brain to replace old cells. Reversal is a powerful approach. We are applying this to the brain and to blood stem cells, replacing a patient’s entire blood supply with young cells.

The scientific risk is low because the research and patents already exist; we know how to produce these cells and administer them to patients. The real challenge is the grind of setting up a pipeline to manufacture and quality-control them cheaply.

We believe in the vertical integration of manufacturing. Cell and gene therapies are powerful, but people are often deterred by their million-dollar price tags.

These therapies can be much cheaper, but you must innovate on the manufacturing process yourself. You may need to build your own bioreactors or clean rooms if you want to make these treatments accessible to everyone. Much of the current high cost is driven by profit margins rather than the intrinsic cost of production.

1:34:23 Aging is the coolest hardest problem to work on

Daniel 01:34:29

I need to take a quick break.

Jose 01:34:30

I missed that poster: 160.

Daniel 01:34:34

That refers to Omri’s blog post about the raise to 160. It’s a cool room.

Jose 01:34:40

It is such an interesting problem to work on. While there is a lot of excitement about making this a reality, it is also a tremendous amount of work. It feels like we are building the track while running on it.

We don’t have all the answers yet. We have a potential path for Alzheimer’s, but we’re still figuring out how to grow the company, fund research, and solve the broader problem of aging.

I wish we had scaling laws like in AI, where more compute consistently leads to better results.

Daniel 01:35:16

We had Martin Jensen on the podcast recently, and he wrote an article with a point that really stuck with me: biology is harder than rocket science.

In many ways, that’s true. You’re trying to work with human biology, but you can’t even touch a human subject until you’ve validated your work in countless other ways.

Jose 01:35:40

Even in vitro cell culture is somewhat insane. You can take cells and keep them alive in a broth, inside a plate and an incubator, completely removed from their natural environment.

Daniel 01:35:52

The environment where we conduct these experiments is highly unnatural. We were discussing recently how mice in cages are exposed to harsh fluorescent lighting, and the mice in the top cages are constantly shaken by the movement of the mice below them.

Jose 01:36:06

We try to control those variables; for instance, our cages have individual lighting. However, they are still mice, not humans in a cage.

1:36:09 What does it take to cure aging

Daniel 01:36:16

What do you think will be the major unlocks that make biology more like engineering? You mentioned scaling laws, and the AI models we apply to biology certainly have them.

Jose 01:36:33

There are papers on scaling laws for protein design models, though they aren’t as robust as those for text models. Many people are currently trying to create virtual cells by measuring outputs after various perturbations.

The problem is that these models are often built on RNA-seq data, which is only one modality, and they typically rely on cancer cells. We have to ask how much we are actually learning from that. Furthermore, if the solution to Alzheimer’s involves replacing cells, you wouldn’t necessarily see that path from a virtual cell model. You might see it from a virtual human, but not a virtual cell.

Cells also behave differently based on age. You don’t see age reflected strongly in RNA-seq, but you see it in the chromatin state. You also see it under perturbation. If you give alcohol to both a young and an old person, their livers react differently.

Is there anything like a scaling law in biology? If you plot lifespan extension against time, you don’t see a smooth scaling law. It looks more like a period of little change followed by a massive improvement. I suspect we will eventually figure it out, likely through cellular reprogramming.

We have had most of the necessary tools for a while. It is primarily a matter of executing the process in a fast iteration loop. We need to try something, learn why it didn’t work, and try again.