Could they be left holding the bag if chip demand slows?

We also discuss how Anthropic’s founders are negotiating “supervoting shares” as they prepare for the IPO, so they can’t be overruled by activist investors demanding profits above all else.

Plus the Bending Spoons SaaSpocalypse strategy: buy SaaS companies on the cheap, jack up prices, slash headcount, and get out as much value before they go to zero.

Vibes & Benchmarks is a weekly AI news podcast hosted by Ali Rohde (Outset Capital) and Josh Albrecht (Imbue + Outset Capital). Each week, they pick apart the latest news, from startups to SpaceX to OpenAI. They argue through what is actually new, what is overfit to Twitter, and what might still matter 6 months from now.

You can also listen on Spotify, YouTube Podcasts, and Apple Podcasts.

Subscribe now

So the near-term threat to knowledge workers might not be losing their jobs. It might be getting handed 5 jobs at once.

Plus why Google/hyperscalers spending its free cash flow on AI is a GOOD thing.

Why NVIDIA, Google, Palantir — basically all the companies besides OpenAI and Anthropic — joined the new open-source AI coalition. Why this is incredibly self-serving AND the right things to do.

New Vibes & Benchmarks with Josh Albrecht.

Vibes & Benchmarks is a weekly AI news podcast hosted by Ali Rohde (Outset Capital) and Josh Albrecht (Imbue + Outset Capital). Each week, they pick apart the latest news, from startups to SpaceX to OpenAI. They argue through what is actually new, what is overfit to Twitter, and what might still matter 6 months from now.

You can also listen on Spotify, YouTube Podcasts, and Apple Podcasts.

If the latter, then technically this is preventable. But humans are going to continue to be fallible. Weak systems are all around us, just waiting to be discovered by AI.

So what are we to do? Hunker down? Batten the hatches? Just expect more hacks, rogue agents, and leaked data the new norm? All of the above?

Latest episode with me and Josh out now:

Vibes & Benchmarks is a weekly AI news podcast hosted by Ali Rohde (Outset Capital) and Josh Albrecht (Imbue + Outset Capital). Each week, they pick apart the latest news, from startups to SpaceX to OpenAI. They argue through what is actually new, what is overfit to Twitter, and what might still matter 6 months from now.

You can also listen on Spotify, YouTube Podcasts, and Apple Podcasts.

Subscribe now

Josh had just sat down w/ one of his interns and gone through PRs line by line. One used a custom scheduler.

His advice: delete it. Use cron or anacron. Pick the boring, stable tool.

The higher-level lesson was that good engineering often means writing less software.

Claude can handle syntax. It can even help you learn Rust faster. Someone still has to know which tradeoffs matter, which abstractions are a trap, and when a missing feature is actually what’s keeping the system simple.

AI makes code cheap. Taste becomes the bottleneck.

New Vibes & Benchmarks w/ Josh Albrecht.

Vibes & Benchmarks is a weekly AI news podcast hosted by Ali Rohde (Outset Capital) and Josh Albrecht (Imbue + Outset Capital). Each week, they pick apart the latest news, from startups to SpaceX to OpenAI. They argue through what is actually new, what is overfit to Twitter, and what might still matter 6 months from now.

You can also listen on Spotify, YouTube Podcasts, and Apple Podcasts.

Subscribe now

and I discuss this potential bleak future.

Teams ship more. Faster. So do their competitors.

But you also can’t not use AI, because then you fall behind.

So companies are forced to spend ever more on tokens, work harder and harder, and still not capture the benefits.

The game is simple and horrible: run harder, pay the toll, the only ones that benefit are those providing the models.

Agree? Disagree? Watch for full ep.

Vibes & Benchmarks is a weekly AI news podcast hosted by Ali Rohde (Outset Capital) and Josh Albrecht (Imbue + Outset Capital). Each week, they pick apart the latest news, from startups to SpaceX to OpenAI. They argue through what is actually new, what is overfit to Twitter, and what might still matter 6 months from now.

You can also listen on Spotify, YouTube Podcasts, and Apple Podcasts.

Subscribe now

No company has ever gone public in the US at a $1T+ valuation.

This year, we’re likely to get three.

SpaceX. Anthropic. OpenAI.

I asked Josh: does this represent the pinnacle of capitalism, or its Achilles heel?

You can guess which he chose.

His take: retail gets invited in only after others have captured most of the upside.

We dug into SpaceX, the company with the most egregious valuation as a revenue multiple. It did roughly $18.7B of 2025 revenue and is aiming to go public at $1.75T, a ~94x multiple.

“Doesn’t really make sense, but it’s AI, yay.”

We did identify a potential bull case. Forget data centers in space or selling enterprise AI. What if SpaceX becomes a chip company?

We also discussed runaway recursive self-improvement. Josh’s view: RSI only gets scary if the improvement curve is both meaningful and accelerating.

But that runs into all kinds of physical bottlenecks: chips, compute, memory, electricity, the actual hardware supply chain. Even recursive self-improvement has to run on scarce GPUs, which will prevent us from ever getting there.

And finally: this week’s “loop engineering” craze, ie the idea that engineers will design systems that create loops, and then the loops will do the prompting, replacing the human at the center of it all.

Josh was skeptical.

1) We’ve known about while loops and for loops for a long time, this is not new.

2) “Loops don’t magically solve anything. They just make you waste a ton of tokens if you don’t know what you’re doing anyway. Most people probably shouldn’t be using loops.”

Vibes and Benchmarks is a weekly podcast where Outset Capital GPs Ali Rohde and Josh Albrecht discuss what’s top of mind in AI. You can also listen on Spotify and Apple Podcasts.

Subscribe now

Josh and I debated that question and more in our latest weekly podcast, including what AI unemployment actually looks like.

His (depressing as hell) argument:

Right now, you become more valuable if you can leverage AI. Every company is looking for AI-native talent. These jobs are paying more, not less.

Then the company uses that talent to build the AI systems, incl extracting the domain knowledge that actually makes these systems work.

The AI-native worker is the hot hire and the extraction layer.

Over time, the company is able to replace more of these roles, and hire fewer and fewer people.

Companies get smaller, jobs get scarcer, and the value of AI accrues to a smaller and smaller number of people.

Part of why I like this argument is that it explains why we’re not seeing AI job cuts show up cleanly in unemployment stats.

The first stage might be smaller companies that never hire the extra people, and larger companies spending more (on tokens and talent) to make internal knowledge usable by AI.

So maybe the question is not whether AI takes my job tomorrow.

It is whether I am getting more valuable, or helping train the system that makes me less necessary.

Push back if you disagree. I would love for this to be wrong.

Are you working less hard?

Or are you just expected to do more?

Watch the full episode here.

Vibes and Benchmarks is a weekly podcast where Outset Capital GPs Ali Rohde and Josh Albrecht discuss what’s top of mind in AI. You can also listen on Spotify and Apple Podcasts.

Series A startup Daytona had a real business: customers, revenue, enterprise interest.

Ivan Burazin walked away anyway.

He was going all in on agents.

I interviewed my friend Ivan, CEO of Daytona, on this week’s podcast. Most podcasts are an hour or more. This one is 10 Minutes or Less.

Daytona builds sandboxes, basically composable computers for AI agents. When an agent writes code, installs packages, browses, tests, retries, or keeps state, it needs a machine to do that work in. Safe, fast, able to spin up on demand.

But Daytona did not start there.

The company arc:

In 2023, Ivan started the company as an on-prem product for human engineers. It had traction. It also had enterprise sales, a slog even in 2026.
On a call with one of the biggest banks, they ask how big the company is. Ivan says 15. The bank asks, “Five-zero?” Ivan says, “No, one-five.” Conversation over.
At the same time, AI-agent companies were pulling Daytona somewhere else.
Ivan showed the new product around and felt real product-market fit for the first time. Every demo got the same reaction: “I need this right now.”
They started handing out alpha API keys. The pull felt different from anything Ivan had experienced before.
Ivan decided to shut it down completely and sent those customers to competitors anyway.
Then the numbers went vertical: $1M ARR 50 days after launch, $3M ARR 45 days later, $24M Series A led by FirstMark’s Matt Turck. The company keeps growing like mad.

That’s the clean version.

The messier version is Ivan on planes between the US and Croatia, sleeping around five and a half hours, taking maybe two days total off since, and doing weekly IVs because you can’t drink the equivalent of eight liters of water.

He went viral for tweeting that if you take a break these holidays, you’re not gonna make it. His point was narrower than hustle theater: in this supercycle, every moment you’re not working, you’re ceding it to someone else.

Someone replied that they’d never heard of his company.

Ivan’s answer: exactly. That’s why I can’t take a break.

Asked if it lets up anytime soon, he said: “I don’t know the answer to that.”

I found the candor refreshing. This is the level of speed many founders are choosing while the agent infrastructure market changes every day.

Subscribe now

Ivan is the Co-Founder and CEO of Daytona.
• X: https://x.com/ivanburazin
• LinkedIn: https://www.linkedin.com/in/ivanburazin/
• Daytona: https://www.daytona.io

Josh’s latest hot take: if you’re a programmer and you’re not getting a huge lift from AI coding tools, it’s probably you, not the AI.

Cult-favorite programmer George Hotz said he’s joining the Gary Marcus / Yann LeCun camp: AI agents can’t program.

Josh’s response was basically: skill issue.

“You were already doing it wrong. Doing it wrong even faster doesn’t help you.”

Other spicy takes:

• macOS: “Do you think macOS will get better or worse in the next 2 years? For sure worse.” Big orgs ship the slop first.

• The Pope on AI: “Seems like he knows what he’s doing more than Gary Marcus or Yann LeCun.”

• Maybe we should hope AI gets worse for society in the short term, because we seem to only react once things get really bad: “We don’t react, we don’t react, and then we way overreact. When it crashes 50% of the cars on the road, we’ll have a reckoning, and we’ll do computers differently.”

• And on the SpaceX S-1: “It’s not a $2 trillion business. It’s a business that loses tens to hundreds of billions of dollars.” The whole story rests on a launch market that’s ~$3B a year. The same TAM math that kills a seed pitch doesn’t stop being true at $2T.

Watch the episode on YouTube.

Watch the episode on YouTube

Vibes & Benchmarks is the podcast where Outset Capital GPs Ali Rohde and Josh Albrecht discuss what happened in AI this week.

Check out the other Vibes & Benchmarks episodes.

Subscribe now

Often, these founders end up losing days or weeks getting intros to funds that turn out to not even invest in their stage or sector, and miss out on the VCs that could actually be the best fit.

It doesn't have to be this way. That's why we've built VC Sheet, a new platform to help early-stage founders identify the VCs that are the right fit for them.

VC Sheet provides both curated lists and a search-able database that:

• Filters funds by stage, sector, geography, and check size

• Dives deep into investor profiles with bios, YouTube videos, personal websites, and notable investments

• Shares the best way to get in touch with each fund, including email addresses and whether investors have open DMs

Some of our most popular curated lists include:

• Funds that lead pre-seeds

• Funds that lead seeds

• The most prolific pre-seed investors

• Early-stage funds focused on NYC

• Early-stage funds focused on AI

Already, VC Sheet was featured in TechCrunch, and as the #2 Product of the Day on Product Hunt!

If you are an early-stage founder starting to plan your raise, start with VC Sheet.

Check out Outset Capital at VC Sheet

As both AI practitioners and investors, we believe this revolution will enable a new wave of game-changing companies. Already, we’ve seen entrepreneurs build on top of these technologies, using GPT-3 to generate marketing content and DALL-E to generate stock images.

However, the obvious applications of the technology barely even scratch the surface of what is possible. Instead, the biggest companies of the future will be formed by looking at what is only just becoming possible, and applying these technologies in ways that showcase their strengths while working around their (likely temporary) limitations.

Below, we've highlighted four example areas we’re interested in funding. If you’re working on something related, get in touch!

Generating code

One of the areas that has generated the most attention is using AI to generate code, especially through GitHub Copilot.

We do believe that AI has the potential to radically improve developer productivity, potentially by more than 10 or 100x within the next few decades.

However, the most obvious application today – using generative AI to generate large swaths of entire functions – is not necessarily the best pathway forward. Just try spotting the bug in the Copilot-generated code below. And unfortunately, as Jeremy Howard points out, there are plenty of other issues as well -- Copilot-generated code fails to use the standard library, is trained to imitate average (generally bad) programmers, could create legal and copyright issues, and more.

That's why, though Copilot is impressive, we're more excited by specific, nuanced, and narrow applications of generative AI to software engineering.

For example, generative AI should be able to automatically detect (and correct) coding style violations – a kind of Grammarly but for code. They should also be able to generate additional test cases, help suggest names when refactoring, or apply a large and complicated refactoring operation more broadly across a large codebase. They can also likely find places that seem likely to contain security flaws (e.g. buffer overflows), or help break apart really large functions into smaller, easier-to-use and understand functions. Just last week, Series B startup Replit released a product that moves more in this direction.

We spent only a few minutes coming up with those initial examples. We’re much more excited to hear ideas from entrepreneurs building in this area, so don’t take these examples as an exhaustive list.

Generating video content

Someday, neural networks may be able to generate entire, realistic videos – perhaps entire movies or TV shows.

Right now, that’s very far away. But today's networks are already capable of generating a very particular type of video content: simple animated "videos," such as content like anime. Some anime is little more than the frames of a comic book stuck directly into video form. Given tools like DALL-E, we're getting close to a world where those original comic frames could all be generated from text, and even the text for each frame could be generated by GPT-3! Such a composite system could, in theory, allow the creation of an entire "animated" video from as little as a single sentence, and in such a way that each individual piece of the result could be easily edited.

Generative AI can also already be used to extend animated video. Existing techniques can turn simple animated movies into something much more richly animated, such as by making the leaves on trees blow in the wind, by having characters talk or move, or even by transferring expressions acted out by humans into the animated style, as in the "MegaPortraits" work below.

We see an opportunity for entrepreneurs to build new companies on top of this initial video technology, and then to continuously incorporate new animation techniques as more research develops. It’s unclear exactly whether this new company should be a new movie studio, a set of tools for existing artists, a community of consumers creating a new type of content, or something else entirely. If you have opinions, we’d love to hear!

Generating game content

While the gaming industry has grown rapidly over the past decade, the process of generating "art assets" – the images, textures, and 3D models actually used in-game – can be extremely slow, with game designers painstakingly setting each pixel and tiny movement for every animation.

This won’t always be the case. The latest generative models enable tools that could dramatically reduce the cost for creating these assets, as well as open up the creation process to millions of people who could not otherwise build assets themselves. These tools can not only be used for generating realistic animations and textures, but also for generating entire worlds, dialog trees, and other aspects of gameplay.

In fact, generative models could open up entirely new gameplay possibilities. Projects like AI Dungeon are just the very beginning. While characters in games have always been fairly static, startups like Inworld AI are enabling users to create non-playable characters (NPCs) with the appearance of memories, personalities and human-like behaviors.

By building games around the capabilities of our latest generative models, entrepreneurs could create whole new forms of games or experiences that resonate with completely new audiences. It's always hard to tell what will resonate with people ahead of time, but we're excited to see more experimentation in this area.

Generating realistic human interactions

Humans are hard-wired to interact with other humans. Whether it's debate or negotiation or just casual conversation, most of us spend huge chunks of our time talking to other people.

As language models get better, it might be possible for an AI agent to, at least in a limited setting, provide an actually enjoyable, interesting conversational partner. We could imagine games built around interacting with other human-like agents – imagine a murder mystery where you have to actually interview the witnesses, or an improv game played with a mix of virtual and human players, or an AI that can hold its own against your most opinionated, argumentative friend. These types of interactions stretch the definition of what we would consider a "game" today, and could potentially open up entirely new markets.

And these types of interactions could be used for more than just entertainment. Imagine getting to learn a language in the context of a more game-like setting where you get to have conversations with an AI agent that sounds like a Parisian local. Or getting to train your sales teams by having AI agents that raise common customer objections and where you can choose how difficult for them to be.

For some of these applications, you might want to take advantage of technology like the "MetaHuman Creator," where the user can interact with a photorealistic human face. For other applications, however, you may want just voice, or just text, in the same way that we switch between video chat, voice only, or text when interacting with other people today.

Finally, rather than just generating new interactions, generative AI can also be used to augment existing interactions. Already, startups like Sanas use generative AI to change the accent of a customer service agent's voice in real time, while Intel has generated some controversy by building technology to detect when students look bored, confused, or distracted.

Where do we go from here?

The technologies we've seen today, while impressive, are only the beginning of what is possible.

We believe the next few years present a unique opportunity for founders to use the rapid advances in generative models to disrupt some of the largest companies in existence. The founders that succeed will be those that figure out how to both take advantage of the capabilities available today and position their companies to grow with the technologies as they continue to develop.

If you’re building something new, regardless of whether it is taking advantage of the latest advances in generative AI or something else entirely, get in touch!

Subscribe now