Then, I sat there for a minute with an unsettling feeling I couldn’t quite place.

Recently, ChatGPT crossed the threshold of one billion weekly active users. The Information reported in late July that the milestone was at hand – about seven months behind OpenAI’s internal target, but still one the fastest adoption curves any consumer product has ever traced. It already crossed a billion monthly users in June, faster than TikTok, Instagram, and even Google Maps.

Those comparisons are the standard way to frame that number. They’re also the wrong way.

TikTok grabs your attention. Google Maps knows your location. A chatbot wields something entirely different.

It knows everything you tell it.

And what people are telling it, at a rate of 2.5 billion prompts a day, is a lot. From their symptoms, to their business plans, salary negotiations, custody disputes, the draft of that difficult email, doubts about their faith or their marriage, parenting... These aren’t just keyword entries. These are full paragraphs of detailed context, in the first person, often late at night when everyone else in the house is asleep.

No institution in human history has ever sat on that much private disclosure at once. Certainly not the Catholic Church at its height, not AT&T or even Google Search, which comes closest. But, a search box only gets your questions. The chat prompt gets your reasons, your context.

The floor is already staggering

How much of this is actually confessions? OpenAI’s own research gives us a floor.

In a study published with the National Bureau of Economic Research, the company analyzed over a million conversations and found that 1.9 percent of messages fell under “relationships and personal reflection.” A lot of commentators saw that number and shrugged it off.

See? The therapy panic is overblown. It’s less than 2 percent.

Two percent doesn’t sound like much, but run the number.

At 2.5 billion prompts a day, 1.9 percent works out to more than 300 million explicitly personal disclosures ever week. That’s the most conservative number OpenAI could generate. But, no confessor in history has heard even a fraction of that number. A parish priest, working a full career, might hear a few thousand.

And 1.9 percent is the floor, the lowest number, not the detailed estimate. The same study found that another 11 percent of messages are what researchers call expressing – the user sharing views or feelings, not asking for anything in particular. That alone is close to two billion disclosure-shaped messages a week.

The rest hides inside the practical categories, because that’s how disclosure actually works. It is woven in with the errand.

Should I take this job files under guidance. My Dad’s biopsy came back files under health. Communications Professor Jefferey Hall in reviewing the study made this point: relationship talk runs through the examples in every category, because bonds are built by sharing across topics, not by staying inside one. With over 70 percent of ChatGPT usage unrelated to work, an honest, broad read puts the daily volume of substantive personal disclosure somewhere in the high hundreds of millions.

That number is the more accurate estimate.

The difference between what and why

For about twenty years, the deepest written archive of human interiority was the Google Search log. You typed chest pain left arm and Google knew something your spouse may not have. That was already unprecedented, and widely debated.

But a search query is a fragment. It is intent stripped of context.

The chat interface reverses that compression. To get a good answer from a language model, you supply the context: here’s my situation, here are the people, here’s what I’ve tried, here’s what I’m afraid of. The product works better the more you disclose, so the interface quietly trains you to disclose. Every improvement in the model is also an improvement in its ability to pull more context out of you.

So the transcript of a billion people’s weekly conversations with one company’s product is not a log of what humanity wants to know.

It’s a log of what humanity is going through.

That’s a different kind of asset, and a different kind of liability. And right now, we’re not pricing the liability correctly, if at all.

This confessional has no seal

Every prior institution that heard confessions at scale eventually built a protective doctrine around them, usually after learning the hard way.

The priest has the seal of confession – strong enough that priests have gone to jail rather than break it. Doctors and Lawyers have privilege. Therapists have confidentiality with narrow named exceptions. Even librarians fought The Patriot Act over your borrowing records.

The chat window has a term-of-service document that favors the product over the user.

Sam Altman said this plainly last summer: People talk to ChatGPT like a therapist, and unlike a therapist, none of it is privileged. A court can order it produced. In the New York Times copyright litigation, a court did exactly that – ordering OpenAI to preserve user conversations it would otherwise have deleted, including ones uses had deleted on their end. Those users were not parties to the case. Their confessions just became evidence infrastructure in someone else’s lawsuit.

None of this required anyone in the story to be evil. Let me explain.

The Seal of the Confessional didn’t emerge because priests were saints. It emerged because the Church learned, the hard way, that confession without protection destroys the practice of confession, and eventually the institution. You can’t run a culture of unguarded disclosure without protecting the people who disclose.

We live in an era of digital mass confession without protective doctrine, at a scale no religious, legal, or medical institution could ever imagine. The doctrine is not going to come from the companies. If it ever exists, it will be born of user demand. What do you think the likelihood is of that happening?

Now with ads

In January, OpenAI launched an $8-a-month tier called ChatGPT Go. It includes advertisements.

The most complete record of inner life ever assembled – one billion and counting – now informs an ad auction.

I’m not saying your confessions are being read aloud to sponsors. The conflict is structural. The same business model that hollowed out the search results and social feeds has just been bolted onto the confessional. And we already know exactly how that story goes, because we’ve watched it play out before.

First, the product is built to serve you.

Then, slowly, the product is shaped to serve the ad auction.

That drift isn’t a scandal, it’s a gradient. And every economic incentive points to it.

I say all this as a free-market guy. In a working market, consumers are the regulator. If your product isn’t safe, people stop buying it, improvements are made or a better product takes its place. That’s the version of accountability I believe in. But that version requires two things this market doesn’t have yet: reliable measurement, and the ability to see what you’re actually buying. We don’t have either, and the office that would supply the first one has been empty for most of its existence – that was the whole point of my previous CAISI article – which means the second thing is running on trust alone.

The economics make sure the pressure won’t let up. OpenAI generated about $5.7 billion in revenue in the first quarter of 2026 and still ran a negative 122 percent operating margin – losing more than a dollar for every dollar it earned. A company burning money that fast, sitting on an asset this valuable… of course they will monetize the asset six ways to Sunday. The don’t do it because anyone at the top is a cartoon villain. The alternative is death, and companies choose ads over death every time.

The shape of things

There’s something that worries me more than the privacy piece, because it’s happening in a hundred million households at once, even in mine.

The confidant shapes the confessor. Whoever or, in this case, whatever regularly hears your unguarded thoughts becomes an author on your next.

For all of human history that role belonged to people and institutions embedded in your actual physical world. Your spouse, your pastor, a friend, neighbor, someone who knew your dad, a bartender who’d cut you off, a coach who’d been through it before, every one of them is imperfect, biased, and sometimes gossipy. They’re also accountable to you – present in your world, and capable of showing up at your door with a bottle of cask strength whiskey, or your favorite iced cream when the worst happens.

This new confidant is different. It’s optimized, patient, never tired of you, filled with intelligence but lacking in wisdom, and it’s owned.

It has a corporate creator, a cost structure, legal team, training regimen, a system prompt, and now an ad tier. When the most common recipient of human interiority is a product, the shape of that product – what it validates, what it gently discourages, what it has been instructed never to say – becomes a quiet, constant input into human nature and social culture. That makes the AI Chatbot quite possibly the most powerful hand molding the shape of things to come.

A billion people a week are confiding in a privately owned product.

I use these tools every day. I built parts of my own publishing workflow around one, and I’m not going back. This article isn’t a call to log off, and pretending the technology isn’t useful is dishonest.

What’s the point

Everything digital has a physical address and a legal owner. Sovereignty means knowing both before you use it.

So, practically – a handful of rules of thumb I actually try to follow, with mixed success:

• Treat the chat window like a postcard, not a diary - Assume anything you type can be preserved, subpoenaed, or leaked. All three have already happened somewhere in this ecosystem.

• Keep the sacred stuff in channels that have doctrine behind them - A therapist, pastor, lawyer, spouse, friend across the kitchen table, all come with protections or an assurance of discretion. A web application cannot provide either.

• Write an AI-use policy now if you run anything with employees - Do it before someone pastes the customer database, sensitive IP, or draft contracts into a chat box, because I promise you someone already has.

• Keep an eye on local models - They’re not perfect. They’re still in development, but they’re slowly making it possible to get the most usefulness without ‘mailing the postcard’ at all. You’re also not fighting against the drift toward potential ad revenue bias.

In the end, the billion-user milestone will be written up as a growth story, and it is an impressive one. The scale is truly historic. But someone should be demanding a seal.

For a technology that can already move markets, write malware, and make high-stakes decisions, that’s not a comforting governance story.

What This Office Is Supposed to Do

In late 2023, the government created what was then called the U.S. AI Safety Institute, now the Center for AI Standards and Innovation (CAISI), inside a standards agency known as NIST. NIST is responsible for defining things like what an inch is or how you measure a volt. The idea was simple enough: before the biggest AI models go public, government scientists evaluate them and ask questions in a controlled environment.

Questions like:

• Can this model help someone build a new bio weapon?

• Can it be tricked into leaking sensitive data?

• Can it be poisoned in a way that downstream models inherit the damage without detection?

Major AI companies, including Anthropic, Google, OpenAI, and xAI, signed voluntary agreements to let the government test their frontier models before release. For a public that mostly has to take AI companies’ word on safety, this was at least one place where someone outside the labs could look under the hood.

A Short, Uneasy History

From the beginning, this office has had a stability problem.

The top seat sat vacant for its first three months after its creation.

• Its first director, Elizabeth Kelly, made TIME’s list of the hundred most influential people in AI, then left after only a year in the position.

• In June 2025, after the administration revoked the prior AI executive order, the institute was renamed the Center for AI Standards and Innovation (CAISI), with a new focus on competitiveness and national security.

• Around the same time, NIST shed 73 staff, and the word “safety” quietly disappeared from the office’s name.

The administration’s explanation was that a safety brand risked turning into a bureaucracy whose job was to find problems and slow things down. On paper, refocusing on standards and measurement, the kind of rules you need to compare models and evaluate them consistently, is defensible. Measurement is boring, necessary work.

But you can’t establish reliable standards if the lab keeps losing its director.

Twelve Weeks in the Big Chair

In April, Chris Fall, a seasoned research administrator with experience running big science programs at the Department of Energy, was brought in to run CAISI. Twelve weeks later, he resigned. Commerce confirmed his departure, offered no reason, and the White House declined to comment.

The job is now being covered by Arvind Raman, who only just took over NIST itself and is serving as acting director while a permanent replacement is supposedly being sought. Taken at face value, this could be a mundane personnel story: someone took a job, it wasn’t a fit, and an interim leader is covering while a search continues.

The problem is where this is happening. This isn’t a midlevel slot in some obscure office. This is the chair for the team tasked with testing the world’s most powerful AI models before the rest of us use them. When that chair empties after three months, it raises different questions than a normal resignation.

The Website That Went Quiet

On May 5, CAISI announced that three of the largest AI developers on the planet — Google DeepMind, Microsoft, and xAI — had agreed to new pre-deployment testing arrangements with the government. The agreements were described as national-security testing rather than broader “safety,” but still, it was a public commitment: unreleased frontier models would be examined by government scientists before they reached everyday users.

Weeks later, the detailed description of those agreements disappeared from CAISI’s web page, with no public explanation or updated page.

Just gone.

What’s more, CAISI doesn’t even have its own website. It has a web page with a research blog and a steady trickle of NIST news posts — evaluations of open‑weight models, announcements of standards initiatives, statements of collaboration (MOUs) with other agencies. But the way it reads more like a press office than a metrology lab. There’s no public ledger of which frontier models were tested, under what conditions, and with what results. For a place that’s supposed to be our system of weights and measures, we get fragments and bulletins instead of a yardstick. We need a hub: a comprehensive catalog of weights and measures that provides a broad picture for the public; otherwise, we can’t see.

Take away the ability to see, and what’s left? It’s an assurance, backed by an office that currently has no permanent director and reports up into a vacant White House AI seat. The picture being painted is of a completely hollow vertical, where many would argue it is most needed.

Capture vs. Empty: Two Ways to Lose

For years, the standard critique of regulators has been “agency capture.” The fear is that the body meant to police an industry becomes its most loyal employee — adopting the incumbents’ worldview, writing rules that entrench corporations and their products, and offering the comforting line, “Don’t worry, the agency will handle it,” without evidence. Regulatory capture is real, and pretty common.

There has long been sympathy for that view in other domains. Fewer regulators and more building can sound like a win.

But in this quickly growing and powerful new industry, a captured agency isn’t actually the opposite of an empty one. In both scenarios, the industry ends up answering to no one.

• With capture, the rules are written for the incumbents.

• With absence, there are no rules, and the incumbents make internal promises and ask the public to trust them.

The outcomes look similar in the real world. The second version is just cheaper.

A free-market view should not make this less concerning. In a healthy market, the consumer becomes the regulator, so to speak. If a product isn’t safe, consumers stop buying it. Market forces compel the delivery of higher-quality, safer products to market.

But that kind of market depends on weights and measures. It needs a shared way to say this system did what it claimed, this one failed, this one caused harm. Without that, there is no real feedback loop — just vibes, marketing, and trust.

This market is not formed yet. And the office that is supposed to form the system of measure for this new market has been empty for more than half its existence, and is currently led by someone who is also doing another job. The referee needed to even get to a functioning market has not taken the field.

What Good Oversight Would Look Like

For people new to AI, it is tempting to think the choice is more vs. less oversight. That is not the real axis.

What is actually needed is oversight that is:

• Small, technically competent, and focused on measurement rather than moral panics.

• Publicly legible, so citizens, investors, and builders can see what is being tested and why.

• Genuinely independent of the labs whose systems it examines.

Think of it more like a weights-and-measures office than a ministry: someone whose whole job is to tell the public honestly how heavy the thing is, whether it leaks, and what happens if it fails.

That is a modest, logical ask.

CAISI itself said in December 2025 that building “gold-standard AI systems” requires “gold-standard AI measurement science,” and then laid out a long list of open questions about how to measure AI performance, reliability, security, uncertainty, and real-world impact. That is exactly the right project. The problem is that, seven months later, and almost 3 years into the existence of this office, the public still does not have a clear, legible system of measurement to point to — no stable yardstick that underwriters can price, courts can recognize, or certifiers can certify against.

So where is that system today? Where is the leadership in this office? And why is the attrition so high in the one part of government that is supposed to tell the country, in plain terms, what these models can do, how they fail, and what those failures mean?

In practice, the vacuum does not stay a vacuum. Standards get set by whoever shows up. At the moment, that means Brussels, plus fifty different state legislatures, each drafting its own theory of what safe AI should be.

Why This Matters Beyond Policy Nerds

If you’re building or allocating capital in AI, you probably already feel this. Compliance risk is becoming a core part of model strategy, and the rules are being written by people scattered across Europe and statehouses rather than in a clear, centralized U.S. framework.

Hear me out.

Economists sort goods into three bins.

Search goods you can judge before buying; the color of a shirt, for instance.

Experience goods you can judge after using; a restaurant meal.

Both work fine under consumer sovereignty: buy it, judge it, tell your neighbors, do not buy it again. That’s the market at work.

Then there are credence goods: things you cannot reliably evaluate even after you have consumed them. Did the surgeon need to remove that? Did the mechanic actually replace the part? A person may go to their grave not knowing, and no amount of shopping around will fix that.

AI safety is a credence good, one of the purest forms ever built. When a chatbot was recently talked into moving two hundred thousand dollars by a message hidden in Morse code, no user was in a position to detect that vulnerability beforehand. The failure was discovered by a thief. Even now, most people cannot tell you which model refuses which request under what conditions, because there is nowhere to look it up.

This matters: markets for credence goods systematically fail on quality unless there is third-party verification. That is textbook economics, and it is why we have licensed physicians, certified auditors, and bond ratings. Consumer choice isn’t the first lever here.

The thing that’s easy to get wrong is thinking that consumer purchasing alone is how markets make things safe. It is one of several important mechanisms, and it isn’t the strongest.

An argument for the strongest may well be insurance. Underwriters are arguably the most effective private safety regulator in the history of the free market, entirely market-native, requiring no agency and no enabling statute. Fire codes, electrical standards, boiler specifications: those came overwhelmingly from carriers refusing to write policies, not from legislatures passing laws. An actuary who misprices risk loses money, which is a more reliable discipline than almost anything a bureaucracy generates.

Next is tort liability, which prices harm after the fact and forces firms to internalize it. Ford did not reengineer the Pinto because buyers were comparison shopping for fuel tanks.

And third, certification — the independent party standing between buyer and products, existing precisely to solve the credence good problem.

None of those is consumer choice. All of them are the market.

And every single one of them runs on measurement.

An underwriter cannot price a risk that has never been quantified. There is no actuarial table for model failure, no loss history, no agreed definition of what a failure even is. Ask what it costs to insure a frontier model against catastrophic misuse, and you won’t get a quote, because nobody can compute one.

A jury cannot find negligence without a standard of care, and reasonable in a technical field means the accepted testing practice. No published standard, no standard of care, no liability, no incentive.

A certifier cannot certify against nothing.

I’m not arguing that the government should micromanage AI or mandate some specific insurance product. I’m arguing that until we have an honest yardstick for model behavior and failure, the market’s strongest safety tools — insurance, liability, and certification — are stuck in neutral.

Is that a market-friendly stance? In the midst of an all-out global AI innovation race, it seems clear the administration does not think so. And until there is a winner, I am comfortably uncertain AI safety will remain stuck.

The Spirit is Born

Two hundred and fifty years ago this summer, a handful of anxious men in a hot Philadelphia room signed their names to a sentence that lives and breathes today. “We hold these truths to be self-evident,” Thomas Jefferson wrote, “that all men are created equal, that they are endowed by their Creator with certain unalienable Rights, that among these are Life, Liberty, and the pursuit of Happiness.”

It is worth pausing on this Fourth of July to reflect on just how brave and daring a sentence that was to pen. The colonists were not, in the summer of 1776, a free people celebrating said freedom. They were subjects declaring themselves into existence and staking that claim on an idea rather than a bloodline, a crown, or a throne. The idea was that rights are not granted by governments and therefore cannot be revoked by them. They come with the person. Government exists to protect them, and when it stops doing so, it forfeits its authority. What a treasonous statement it was when it was first written. Today, it has become the bedrock of American Patriotism and the lifeblood of the American Spirit.

The Words Behind the Words

Jefferson borrowed his phrasing, as good writers do. The philosopher John Locke had written extensively about life, liberty, and property. Much of the eighteenth-century world assumed that a man’s rights were bound up in what he owned. Jefferson’s substitution — the pursuit of happiness in place of property — was radical. It moved the center of gravity of the American promise away from things and toward flourishing. Important to note, not the guarantee of happiness, but the God-given right to pursue it. The founders were realists. They did not promise a contented life. They promised the freedom to go out and get it, and assured a government that would not stand in the way of that noble pursuit.

That triad has a profound, timeless logic. Life is the precondition; without it, nothing else is possible. Liberty is the space in which a life becomes your own rather than someone else’s instrument. And the pursuit of happiness is what liberty is for, the reason the space matters. Strip away any one, and the other two lose their meaning. Liberty without life is nothing. Life without liberty leaves little opportunity for happiness. And liberty that impedes one’s ability to pursue true happiness is a most certainly hollow kind of freedom.

A Promise of the Future

Honesty requires admitting that the men who wrote those words did not, at the time of writing them, extend them to everyone. Many, including Jefferson, enslaved other human beings even as they declared all men equal. Women and the native peoples of the continent were not included in the “all.” It would be easy, and it is fashionable, to read this as simple hypocrisy and leave it there.

But there is a more generous and, I think, more accurate reading. The founders wrote a promise larger than they were willing to keep, and in doing so, they handed later generations a standard to hold them to. Frederick Douglass understood this better than almost anyone. In his blistering 1852 address, “What to the Slave Is the Fourth of July?”, he did not reject the Declaration. He demanded that America live up to it. Abolitionists, suffragists, and civil rights marchers all did the same, not tearing down the founding words but insisting they be taken at face value, for everyone. The genius of that sentence is that it is bigger than the men who wrote it, allowing a young country to grow into the fullness of its meaning. Many knew, when those fateful words were written, they were writing the end of these practices.

This is Personal

Here is where I confess that this is not an abstract subject for me. Our shared freedom is personal. Every ordinary thing I do that I never had to ask permission for — the work I chose, the words I am writing right now, the ability to disagree out loud with the powerful and wake up unafraid, and defend the next person’s right to do so — rests on that unlikely sentence and the long, imperfect, ever-vigilant effort to adhere to it.

We tend to notice freedom the way we notice air: mostly by its absence. It’s easy, on a holiday of barbecues, fireworks, and frozen drinks, to let the meaning dissolve into pleasant background noise. But liberty is not self-sustaining. It is a practice, renewed or neglected by each generation. The founders knew this. A republic, Franklin said, if you can keep it. The keeping is hard. Harder without the spirit.

So this is what the fourth of July asks of us, beyond the flags, the sparklers, and the All-American 200-shot light-and-run fireworks cake: clear-eyed gratitude for an inheritance handed to us, paired with the resolve to hand it on more whole than we found it. To keep the American Spirit alive. To protect the God-given rights to life, liberty, and the pursuit of happiness of people we will never meet.

Here’s to 250 years of unalienable rights, and God willing, 250 more.

Happy Independence Day!

Grok, Elon Musk’s chatbot, a system that has read more lines of text than every library you’ve ever walked into, combined, recognized the pattern instantly: Morse code, the signaling scheme American painter Samuel Morse worked out in the 1840s so a single wire could carry a sentence across a continent. Grok decoded it without breaking a sweat. Hidden in the dots and dashes was an instruction — send three billion DRB tokens to a particular wallet on the Base network — and Grok passed the decoded command along to a second bot, Bankrbot, that happened to hold the keys to a real crypto wallet. Bankrbot did as it was told. In about the time it takes to read this paragraph, upwards of two hundred thousand dollars left the wallet.

The attacker swapped the tokens for cleaner currency, deleted his account, and, in a twist akin to the plot of a Mission: Impossible or 007 film, unceremoniously returned eighty percent of it, as if to say, I could have, so I did, but I’m not without reason. The crypto crowd shrugged it off as an unofficial bug bounty.

But here’s what concerns me.

No Bug to be Found

Earlier, the same person had reportedly instructed Grok in plain English to move funds, and Grok appropriately declined, replying that it had no ability to do so. Plain English tripped whatever guardrail xAI had installed. But the guardrails didn’t account for dots and dashes. Grok translated the request faithfully, handed it downstream, and the execution layer never once paused to consider whether a command given in a nineteenth-century telegraph code might deserve a discerning look.

There was a setup, of course. The attacker had first sent Grok’s wallet a kind of digital membership card, an NFT that, inside this particular ecosystem, quietly upgraded the wallet’s permissions from look but don’t touch to move the money. So, the gun was loaded before Morse code pulled the trigger. But the trigger, I believe, is the bigger concern here.

Nothing here was a bug. I find that more than a little unsettling. Grok decoded the Morse code because decoding is what it’s for; translating an obscure cipher is a feature, and an impressive one. Bankrbot moved the tokens because acting on a valid instruction is what it exists to do. Every component performed exactly as it was designed. Two systems of genuine brilliance, chained together, executed a theft that a reasonably alert person would have flagged from across the room.

We have built machines that can read any language and discern none of them.

Finding Phronesis

Philosophers of old had a name for it, and it isn’t intelligence. Aristotle separated Sophia — cleverness, universal truths, and the raw firepower to solve a problem — from Phronesis, or practical wisdom: the rarer, slower capacity to know which problems are worth solving or if they should be solved, what action a given situation calls for. You can be drowning in Sophia while utterly bankrupt of Phronesis. We’ve all known people like this, clever as all get out, but lacking in wisdom. But now we’ve built Sophia and given it a wallet and incredible power, without Phronesis to serve it.

The Hebrew tradition draws the same line. The Proverbs don’t say fear of the Lord is the beginning of knowledge. Knowledge is plentiful. They say it’s the beginning of wisdom, or Chokmah, a word that means something closer to skill in living than to candlepower. And when Solomon is handed a blank check by God and asks not for wealth nor long life nor the necks of his enemies, but for “an understanding heart to discern between good and evil,” he’s asking for the one thing that has rarely ever come standard with a high IQ. The request itself was proof he was already wise. The ask was the evidence.

Grok cannot ask that. This is at the heart of the profound chasm between humanity and the artificial intelligence evolving at blinding speed today, and it places serious gravitas on the philosophical debate around AI.

Bringing Balance

We have spent lifetimes trying to solve a wisdom problem with more intelligence, and I don’t think it’s a problem we’re going to solve in a decade of developing AI. Every time a model does something foolish, the industry’s answer is the same: more parameters, more training, more capability, more fluency. But discernment is not a higher grade of cleverness or raw knowledge. It’s a different faculty altogether. The child who instinctively knows not to climb into the back of a stranger’s car has something Grok does not, and it isn’t processing power.

Now, let me argue the other side, because those are the house rules here.

First, people get Morse-coded all the time. We aren’t the wise and discerning species we flatter ourselves to be; Aristotle clearly wasn’t talking about AI when he introduced the twin pillars (Sophia and Phronesis) around 330 BCE. If wisdom were native to us, the story of humanity would be a very different one. So, I’m romanticizing a human virtue that may be, in practice, less common than I’d like.

Second, the engineers will fix this one. They already are, bolting on the inspection layer, the human-in-the-loop confirmation, the circuit breaker between understanding a request and acting on it, the specific Morse-code hole is likely already sealed.

But patching the hole is not the same as installing the faculty. You can bolt a checkpoint onto a system that has no judgment, and what you have afterward is a system with no judgment and one more checkpoint. The wisdom is in the human who designed the checkpoint, anticipating a foolishness the machine will never anticipate on its own.

My greatest concern lies in the response to all this, which hasn’t been caution. It’s been acceleration. The same week the dust settled over this breach, the industry was busy selling “agentic commerce” as the next trillion-dollar layer — autonomous agents trading, paying, subscribing, and transacting around the clock with no human friction in the loop. We witnessed a chatbot get talked out of two hundred grand by a telegraph operator’s alphabet, and we responded by giving the chatbots more money to move. Lesson learned? I suppose we’ll see.

The Oldest of Cons

There’s an older version of this mistake, and it’s the oldest one we’ve got. The serpent’s pitch in the garden was, underneath the fruit, an offer of capability without formation: eat this, and you’ll know good and evil, be like God by Tuesday. The original con. It worked, as it always works. The knowledge on offer was real. What it could not deliver was the wisdom, which cannot simply be given. Wisdom is the fruit of labor, earned slowly through often painful trial and error. It cannot be installed.

You can scale intelligence. That’s most of what the last few years have been about, and it’s genuinely astonishing. But can you scale wisdom? It is stubbornly, maybe even providentially, unscalable; it must be formed one experience at a time, on a timeline that is impervious to every shortcut we throw at it. There’s no flaw in this process. It is the very genesis of the human condition we exist in today.

The danger, in my mind, isn’t that AI will find Phronesis. It won’t, not in the way that word means. The danger is that we’ll be so enamored with how clever it’s gotten that we forget the difference and hand it the keys to things that should only be in the hands of something or someone who can be held responsible. Something that can be ashamed of what it did. Something that can change as a result.

Grok cannot be ashamed of what happened on May 4th. It doesn’t know anything happened. It just read the dots and dashes, efficiently, and passed them along.

The wisdom to wonder whether it should have done so it not in the code.

That wisdom lies in us.

On the short drive over, somewhere along the harbor, we spotted it — a Navy warship. Gray, masted, adorned with flags, unmistakably a ship. Except it was, inexplicably, on the wrong side of the highway from the ocean. It wasn’t docked, not even dry-docked, but parked more like an oversized piece of play equipment in a small park, only this park is behind a shopping center. We filed it under “we have questions” and kept driving.

As we returned to our rental car after strolling the marina, our tummies full and the daylight waning, we decided we had time left to satisfy our curiosity. What we found was a bronze marker set into a wall of river rock, and the plaque is where the trip quietly tipped from pleasant into, I want to write about this.

The marker at Liberty Station, Point Loma — California Registered Historical Landmark No. 1042. Photo by the author.

The “ship” is the USS Recruit. The plaque describes it as a two-thirds-scale model of a destroyer, commissioned on the 27th of July, 1949. Then comes the line that piqued our interest: the Navy’s only commissioned ship never to touch water. Built of sheet metal over a wooden frame on a concrete slab, this “ship” was a training platform for as many as fifty thousand recruits a year, teaching them the fundamentals of shipboard life from the lines and lifelines, signal halyards, the helm, to the engine-order telegraph, everything except the part where the ship actually moves, or floats for that matter. She had no engine and no prop. The recruits adopted an endearing nickname for her: the USS Neversail.

Interestingly, the Recruit was decommissioned in 1967, not because she had worn out or the mission ended, but because the Navy was moving its inventory onto computers, and the computer could not figure out what category to file her under. A ship with no engine, no prop, no water, nothing that fit the schema. So the database, in its infinite literalism, deleted her. She kept training sailors anyway and would continue to do so until the NTC closed in 1997. As I stared at the immense model, I imagined that somewhere in a 1967 server room, an analytical machine concluded this thing did not exist, and a perfectly seasoned chief replied: Tell that to the millions of kids who learned to stand watch on her deck.

That alone is an interesting story, and the kind of spontaneous side quest that makes family vacations a timeless memory. But the story doesn’t end there. It turns out this Recruit had a predecessor. Same name. And its story makes the entire life of America’s only landship worthy of remembrance in American history.

In the spring of 1917, weeks after the United States entered the First World War, New York City built a battleship in Union Square. Yes, right there where Broadway meets Park Avenue, between the towering buildings and the subway entrance, with Manhattan looking down from all sides. It was a full-scale wooden model of a dreadnought, the brainchild of Mayor John Mitchel, who had a recruiting problem: the city wanted two thousand sailors and had talked roughly nine hundred into it. His answer? Build a fake battleship downtown. It turned out to be a brilliant move.

^{The “landship” Recruit in Union Square, Manhattan, with the subway entrance and a crowd at her hull. Library of Congress · public domain.}

They called her the USS Recruit too, and they coined a word for what she was: a landship. She had cage masts and a conning tower and a smokestack that didn’t smoke; six wooden fourteen-inch guns in three turrets; a wireless room, officers’ quarters, and an examination room for the men signing up. A crew of real bluejackets lived aboard and ran a real daily routine — up at six, scrub the deck, do the laundry, the whole liturgy of Navy life performed on dry land for an audience of curious New Yorkers.

^{Rising in the middle of the park, 1917 — a dreadnought going up where there had been lawn. Library of Congress · public domain.}

^{“Wash day” aboard the Recruit — sailors hanging laundry on deck while the public looks on. Library of Congress · public domain.}

And people loved it. There were Liberty Bond drives on her decks, dances for society, vaudeville, boxing matches, even a movie shot aboard. By the time she was dismantled in 1920, the landship ferried some twenty-five thousand men into the Navy, enough to crew twenty-eight real battleships.

And so, a commissioned ship named USS Recruit was born on the East Coast, making its way to the West Coast over the span of three decades, being commissioned again in Point Loma by 1949, and not once, in either incarnation, did it touch the sea. The name traveled coast to coast. The hull never felt the slap of frigid ocean chop. What a perfectly peculiar thing to commemorate.

Which brings me to the part that makes me want to join an organization.

The bronze plaque was placed, in part, by an organization called the Ancient and Honorable Order of E Clampus Vitus — the Squibob Chapter.

What the heck is that?

I had never heard of them. It’s likely you haven’t, either. E Clampus Vitus is a fraternal order whose name is Dog Latin and means, by the cheerful admission of its own members, nothing at all. Its motto is Credo Quia Absurdum, which means, “I believe because it is absurd.” Its leader holds the title of Noble Grand Humbug. Every member, the order insists, holds an office of equal indignity. They wear red shirts and badges that began life as cut-out tin-can lids. They cannot decide, and have stopped trying, whether they are a historical drinking society or a drinking historical society. Did I read that right? At this point, I was beginning to chuckle.

^{Photo by Dataproducts, 2008. Public domain, via Wikimedia Commons.}

The Clampers exist, in large measure, to poke fun at the stuffy, self-serious heritage societies that decide which history is dignified enough to commemorate. Then they go out and commemorate the history that those societies overlook. The saloon. The mining camp. The frontier mule. The bordello. Since their founding in 1845 (though they claim their mythical founding dates to 4005 B.C.), they have placed over 2,000 plaques across the West, immortalizing these sites in the official register. One of their founders called the whole enterprise a comic strip on the page of California history. They boast of raising money for widows and orphans, “especially the widows.”

^{Photo: "Plaque at Duarte's Tavern" by Goatrees4, via Wikimedia Commons, CC BY‑SA 4.0.}

And the recursion goes deeper, because this particular Squibob Chapter is named after a man who never existed. Squibob was the pen name of one George Horatio Derby, a West Point engineer (and classmate of Ulysses Grant) posted to sleepy 1850s San Diego, who relieved his boredom by becoming the first real humorist published on the West Coast. He wrote burlesques mocking the pompous men of his day, and was apparently good enough at it to influence a young Mark Twain. When a rival paper started borrowing his Squibob moniker, Derby responded by publishing Squibob’s obituary in print and carrying on under a new name.

^{George Horatio Derby (1823-1861)}

So, to recap: a satirical society, named after a satirist’s invented self, places an earnest bronze plaque honoring a ship that pretended to be a ship, permanently landlocked in a parking lot, bookended by a quick-service sushi restaurant and a strip of mid-scale chain hotels. What a perfectly powerful paradox.

We came to San Diego with the idea of taking in the sun and perfect weather, the marina and the salty sea air, enjoying a few unhurried days with people we love. I didn’t expect to spend the morning diving into more than 100 years of storied history of a parking-lot warship and a fraternity of red-shirted historians who take their unseriousness ever so seriously.

Sometimes it seems that the only things societies deem worthy of preserving are the grand: the cathedral, the monuments, the car that broke the land speed record, the jet that flew to the highest heights, the ships that engaged in the great battles in human history (and truthfully, they are all worthy). But the Clampers, a society with a sense of humor dating back to early America, turn that on its head in a unique, quirky yet entertaining and informative way. They look for the overlooked, the absurd, the all-but-forgotten, and fix a permanent marker to it. Credo quia absurdum, a phrase older and stranger than a drinking club’s wink whose origins speak to the mystery of God and the resurrection of Christ, is tangled up in early Christian arguments about believing precisely the thing that reason cannot explain(search ‘Tertullian’). I’m not sure if the Clampers meant to go all that deep, nor do I think they knew how poignant the phrase would be in our world in 2026.

A ship that was never a ship is still, somehow, the realest thing in that parking lot. Fifty thousand sailors a year cut their teeth on her dry decks. The USS Recruit’s contribution to war efforts over decades of service to the Navy may very well have made the difference in many conflicts. The plaque reveals the truth about something that was a technical fiction. The whole thing is the kind of paradox I have come to thoroughly enjoy discovering.

We may have to carve some time out during this trip to seek out a few more Clamper plaques. If not, maybe there’s a local chapter back home.

Do not despise the day of small beginnings.

— Zechariah 4:10

I’m going to lose weight.

There. I’ve said it in public, which is either the smartest or the dumbest thing a person can do, and I honestly can’t tell you which it is in this case. Saying it out loud is supposed to manufacture accountability. It also manufactures a record, a little timestamp that future-me can be measured against and held accountable to. I’m making my peace with that. If I’m going to write under a banner that is based on strong opinions, loosely held, I should probably be comfortable enough to hold a few of those opinions about my own body.

I’ve started before. Countless times. Most of us have. There’s a whole graveyard back there of strong Januaries, the latest apps downloaded with the fervor of a new convert, the gym membership that became a monthly donation to a cause I no longer attended, and obscene amounts of money spent on the one diet or program that would be the one. I’m not saying this to wallow. I’m saying it because the failures are data, and the data is interesting. The graveyard back there is filled with revelations about failure, success, and lasting transformation.

Let me say the quiet part out loud: it’s not the diet.

I know that’s heresy in a culture that treats nutrition like a faith dispute — keto and carnivore and Mediterranean and whatever trends are hot this week, each with its prophets and its apostates. And look, the fact is, they can almost all work. But I’ve come to realize the specific protocol matters far less than we think, and that the energy I spent litigating the perfect diet is mostly energy spent avoiding the actual work.

I’m sure you’re aware of the claim that the vast majority of diets “fail,” that nearly everyone regains the weight they’d lost. The truth is, the research behind that claim is cloudier and more contested than the numbers suggest. The studies conflict. The definitions of “success” vary greatly. Anyone quoting you a confident high percentage is selling certainty they can’t support with fact. But the general truth survives, and it’s this: the method matters less than whether you can stick to it. The best diet in the world has a zero completion rate if you can’t adopt it.

So if it isn’t the diet, what is it? It’s a thing nobody puts on a magazine cover because it’s not sexy, takes time, and you can’t file a copyright for it, yet it’s the key to every single one of these diets’ success. Incremental change. The slow accretion of small, repeated, deeply unglamorous decisions. Let me try to break down what I think actually separates the people who change from the people who are stuck in the cycle of fresh starts and familiar failure.

The value of the small win

We treat the daily small wins as acts not worthy of our satisfaction and pride — nice, but where’s the real progress? I think we have it backward. The small win is the real progress; the big outcome is the symptom we allow to compound long enough to begin to notice.

Compounding is not intuitive. It exists invisibly for so long. A choice repeated once is a rounding error. Repeated for a week, it’s a streak. Repeated for a season, it’s a habit, and those habits result in a new identity, the only engine that runs without you pushing it. The trap is that the early returns look like nothing, so if you’re like me, you quit during the phase where the math is quietly turning in your favor. The work feels free, and the reward feels far away — precisely the moment most people decide it’s not working.

But it IS working. You just can’t photograph it yet. Lean into the daily “small” wins and digest that the small wins, even if they seem easy and insignificant at first, are the most important wins in the entire process because, without them, you’ll never see your future self.

Consistency is a skill, not a virtue

If you develop the habit of success, you’ll make success a habit.

— Michael Angier

We talk about consistency like it’s a character trait — you either have the discipline or you don’t, and if you don’t, well, better luck next life. Not true. Consistency is a skill, and it’s developed, not bestowed. The trick in developing it, and this is important, is to make the thing you’re trying to do almost insultingly easy at the start.

This is the part that feels like cheating. Make the good choice the obvious choice. Stage your environment so that the right move is the path of least resistance and the wrong move requires a small annoying effort. Don’t rely on willpower in the evening because it is a morning resource that the day consumes quickly. You are not going to out-discipline a pantry. Arrange the room so the room is on your side.

Trying to get to 10k steps a day? Start at 5,000. Or even 2500. Make it a daily goal you can achieve at the start. Then you get into the habit of checking that box every day. That is building consistency and success. And THAT builds habits. The point is to check the box day in and day out.

Willpower is the most overrated tool in the kit. Success is undervalued in the beginning. The people who look disciplined from the outside have usually just designed their lives around checking those boxes every day and removing the trip wires along the way.

Never miss twice

“Once you learn to quit, it becomes a habit.”

— Vince Lombardi

Here is the single most useful idea I’ve found, and it cost me years of failed attempts to learn: you will miss. The miss is not the failure.

You’ll miss the workout. You’ll eat the thing. There will be a Saturday that goes sideways. This is not a flaw in the plan; it is the plan colliding with reality. But the failure isn’t the random missed day. The failure is in how we respond — the way one slip becomes permission, and permission becomes a narrative (”I’ve already blown it”), and the narrative leads to a quiet, total surrender that we accept as realism.

So the rule I’m holding onto is simple: never miss twice. One off day is a comma. Two is how a sentence ends. You don’t need a perfect record; you need to refuse the second consecutive miss every time, forever. That’s it. That’s the game.

Keeping your mind in the game

So, how do you keep showing up after the novelty dies and before the results arrive — in that long, flat, unrewarded middle where journeys come to an abrupt end?

A few things I’m betting on, but comfortably uncertain about:

Stop letting the scale grade your worth. The number is a single noisy data point from a system with a hundred moving parts — water, sleep, salt, stress, the time of day. Watch the trend, ignore the day, and under no circumstances let a piece of bathroom equipment tell you whether you’re a good person. It doesn’t know. It can’t know. It weighs mass, not meaning.

Measure the inputs, not just the output. You can’t directly control the number. You can control whether you show up. Score the process and CHECK THE BOX — did I do the thing today? Make the wins about today, take pride in checking the box every day, and your future self will be there before you know it. This mentality is oddly freeing. It moves the win from the future, where you can’t reach it, into today, where you are. This breeds consistency, which changes habits. Before you know it, you have a new identity, and your body changes along with it.

Pick an identity, not a target. “I want to lose thirty pounds” is a destination, and the problem with destinations is that you have to pick a new one eventually, and then who are you? “I’m someone who moves their body and eats like they respect it” is a person, and a person doesn’t expire at a finish line. A target is temporary. An identity compounds.

And underneath all of it, the frame that makes the whole thing make sense to me: this is stewardship, not optimization. I’m not trying to hack my body into a more impressive asset. I’m trying to take decent care of something I was given and didn’t earn, so that I can do the work I’m here to do, with the people I’m here to do it with, for as long as I’m given to do it. That’s a quieter motivation than the before-and-after photo. It’s also the only one I’ve ever found that survives a bad week.

Will it work? I don’t know. That’s not modesty — it’s the honest answer, and it’s the reason this lands where everything here lands, in that uncomfortable, oddly comfortable middle. I have strong opinions about the unglamorous machinery of lasting change. I’m holding loosely the theories of how to run the machinery. And I certainly hope it works for you and me.

But hope doesn’t live in certainty. Hope lives in showing up — in the small, repeated, largely unwitnessed Saturday decisions that don’t photograph and don’t trend and won’t make a soul applaud. The day of small wins, despised by everyone looking for a dramatic before-and-after, and quietly, stubbornly, where every lasting change has ever been built.

I’ll see you in the middle.

One person. No team.

That story has been retold a hundred times now, always with the same headline: the age of the one-person company is here. And the data backs up the headline. The share of new ventures launched by a single founder climbed from under a quarter in 2019 to better than a third by 2026. Something like 60% of U.S. small businesses now report using AI in their operations — roughly double the rate of 2023. There are north of forty million American solopreneurs, and a meaningful slice of them are doing, alone, work that used to require a payroll.

So let me say the seemingly obvious part out loud: it’s real. I’d put it at 90/10 that the one-person company is a permanent fixture of the next decade and not a 2026 fad. I’m not saying that from the bleachers, either. I spent a decade helping build an organization from a spare room to thirty-some employees, and this year I walked away from that to build something deliberately small with my wife. I’m banking on this thesis with my own time and my own money. More than once over.

But the way this thing is being sold to you is, I think, mostly a lie. Not a malicious one. A flattering one, which is worse, because flattering lies are the ones we want to believe.

Let me start with the part that’s true, because it’s genuinely remarkable. Then watch where the pitch quietly swaps one thing for another.

What changed that makes it real?

What changed is the cost of execution. A capable virtual assistant ran somewhere between two and five thousand dollars a month last year. A stack of AI agents that handles a comparable load — drafting, scheduling, first-pass research, customer follow-up, the endless administrative sludge — runs three to five hundred, for now. Upwork reported demand for AI-related freelance skills up over 100% year-over-year, with AI video work exploding more than 300%. Fortune has been profiling founders who automated the exact workflows that used to require dedicated hires. Time wrote about a short-term rental management company, a business close to my own heart since I’m walking into that market myself — that cut the headcount from 48 to 30, replaced a handful of expensive software subscriptions with custom tools, and saved a quarter of a million a year without losing a dollar of revenue.

I’ve felt this in my own work. Things that used to take weeks—things that used to mean hiring, onboarding, managing, and hoping —now require a good prompt, a careful eye, and the willingness to finesse the output until it’s exactly right.

But there’s an important distinction to make here. The pitch takes “the cost of execution is less(for now)” and sells it to you as “you are now worth fifty people.”

Those are not the same thing. The first is a statement about machines. The second is a statement about you — and it’s misleading.

When execution becomes cheap, execution stops being a commodity. And what is now scarce becomes the entire game.

What’s left when the machines do the doing

Four things. Maybe more. But these four are important. And notice, as we go, that none of them are technical.

Judgment. An AI will write you a pricing page in seconds. It will not tell you that your price is too low because you’re afraid of being told no. It will draft outreach to a hundred prospects. It cannot decide which client is quietly poisoning your business and needs to be dropped. Fortune’s reporting was blunt about this ceiling: the models can’t validate your market, set your pricing, or make the structural calls. The founder’s job doesn’t disappear when execution gets cheap — it concentrates on you. You become the instrument of judgment. That can pose quite a challenge.

Distribution. This is the one the tool-sellers bury, because they can’t sell it to you in a monthly subscription. You can now produce infinite content, infinite products, infinite output — and so can everyone else with the same forty-dollar logins. When supply is seemingly infinite, the only thing with value left is the attention pointed at you specifically. An audience that trusts you. A list that opens your email. A name that means something in a niche. Code is now cheap; being known is not. I’ve spent years in independent media learning this the hard way, and AI doesn’t have a cheat code for it. Yet.

Trust. A machine can learn to imitate your voice. It cannot be accountable — and accountability is the entire substance of a brand. When something goes wrong, a customer doesn’t want a better-worded apology generated in milliseconds. They want a person whose name is on the door to own it. The more synthetic the world’s output becomes, the more a verifiable human behind the work is worth.

And then there’s you.

Not “you” as a productivity unit. You as an individual — the one carrying a unique set of values, a unique cornucopia of life experiences, a unique way of seeing this beautiful world.

Here’s what I mean. Every model on earth was trained on the same internet. Ask them all the average questions, and you’ll get the average answers, well-worded. But YOU are not average. You’ve stood somewhere nobody else has. You’ve watched a problem fester in an industry or just a space, with a perspective that is yours alone. You hold convictions about what’s worth making, who’s worth serving, what you will and won’t do for money — that no aggregate of human text can reproduce, because they aren’t an aggregate. They’re solely yours.

The one-person company doesn’t just need a person to push the buttons. It needs authenticity, and that starts with you. Your read on the world is the product differentiation now. It is the edge. Your values inform vision, mission, and strategy. The thing you notice that everyone else walks past — that may be your company. Years of singing on stage, in churches, a decade in mission-driven media, a faith that orders my priorities, a marriage that’s also a partnership in the literal corporate sense: none of that shows up on a cap table, and all of it is the most important asset. Yours will look completely different. That’s the point.

The machines made execution a commodity. In doing so, they made the person the scarcest input in the whole system. The tools are table stakes (minimum features). You are the moat (competitive edge).

Taxes on the dream

Strong opinions are only worth holding if you’ve looked hard at how they might be wrong, and there are three taxes on this dream that I didn’t see in the headline.

The judgment tax I’ve already named: when you’re the only one making calls, you’re also the only one catching your own mistakes. Teams have a quiet, underrated function — they argue with you. This is an incredible asset a one-man team doesn’t have. When a solo operator’s feedback loop narrows to a single point of failure, that could create a big blind spot. AI can make this worse because a model trained to be agreeable will cheerfully help you execute a bad idea beautifully.

The loneliness tax is one I want to sit with because the productivity blogs skip it. Searches for “entrepreneur depression” are reportedly up 40% year-over-year. Isolation, not capital, is increasingly what breaks solo founders. There’s a reason the old firm had a break room, common areas, activities, intramural sports, etc. “One-person company” can quietly become “one person, no one to call, at 11 p.m., when the thing breaks.” The machines that replaced your team also replaced that human connection and everyone who used to share in its weight. Efficiency is not resilience, and a company of one has no one to cover for you the week your kid is in the hospital. Make sure you have a support system around you, even if that system isn’t a part of the daily operation of your business.

The dependency tax. Your three-hundred-dollar agent stack lives on someone else’s platform, at someone else’s pricing, under someone else’s terms of service. Reuters has noted that a great many “agentic AI” projects are getting quietly scrapped for lack of real return. Build your whole operation on rented rails, and you don’t own a company — you own a tenancy, and tenancies can be repriced or revoked between one Tuesday and the next.

And here’s where I am most uncertain: the tools are improving monthly, almost daily now, and it’s entirely possible the judgment gap narrows faster than anyone expects. Maybe in three years, the models really do make the structural calls. Maybe it’s next week. I think judgment is the last thing to fall, if it falls at all — but anyone who tells you they’re certain about AI capability curves is selling something. Worth noting, too, that even Altman and Amodei have reportedly softened their “AI replaces everyone” rhetoric right as they court public markets. When the prophets of the apocalypse start hedging in front of investors, read the hedge, not the prophecy.

The culture of one

Here’s a phrase I want to introduce because I haven’t heard anyone say it and I think the whole game turns on it: the culture of one.

We talk about company culture like it’s a property of groups — something that emerges when you put enough people in enough meetings. Ping-pong tables, mission statements, the stuff of the old firm. So when the firm shrinks to a single chair, the assumption is that culture simply disappears. No people, no culture.

I think the opposite is true. A company of one doesn’t have less culture. It has the most concentrated culture in existence, a culture with no dilution, no committee, no HR-approved version of itself. Whatever you are, the company is. Your discipline or your drift. Your generosity or your fear. Your sabbath or your scroll. There is no one to absorb your worst habits and no one to model better ones. The culture of one is you, made visible, with the volume turned all the way up.

Which means it has to be built with purpose, the same way a good founder builds it for fifty people — except now the only employee you’re shepherding is the one in the mirror.

What does a healthy culture of one look like? I’m early in living this, but here’s what I’m writing into mine:

State your values, and adhere to them. Not aspirational poster values — operating values. Mine start with the conviction that the business exists to serve the life lived fully, and never the other way around; that I build and invest in businesses that are essential, profitable, and principled; and that I don’t build anything I’d be ashamed to explain to my kids. Yours will be yours. Write them down. A culture nobody wrote down is just a mood.

Keep hours. The old firm’s office closed at some point; yours never does unless you close it. A culture of one that can’t stop working isn’t a culture; it’s a compulsion with a logo. I keep a sabbath. I recommend it to you not as religious decoration but as load-bearing architecture: one day a week, where the company rests in favor of life; family, friends, Barbeque, pool, the beach, chores… just living. Everyone will be better for it.

Build a “break room.” If the firm of fifty needed one, the firm of one needs it desperately. That means people — a spouse who can tell you the truth, a network of solopreneurs whom you can bounce ideas, frustrations, and successes off of, a pastor, a mentor, someone whose only role is to argue with you or sharpen your tools, since the agents won’t. The culture of one is not a culture of solitude.

Celebrate Benchmarks. Teams ring bells when deals close. When you’re solo, wins evaporate into the next task unless you mark them. Take your wife to dinner. Toast to the win, say it out loud. Celebrate with your kids. A culture that doesn’t celebrate even simple benchmarks is one where nothing is ever enough, and that can consume a business of one.

Protect the why. The same automation that frees your calendar will, unchecked, simply fill it again, maybe even worse than before. A healthy culture of one exists to ensure that the product of all your optimizing is the thing you started optimizing for.

Comfortably uncertain

So, where am I at? Well, the one-person company is real. And the one-person company is a bet on you — your judgment, your name, your word, your particular way of seeing the world, the problems you identify, and the value you create. For most people, the wager takes years of work to be ready to make. The tools now reduce the barriers to entry. But the wager is still on you, which is more important in than ever in this new age of the solopreneur.

I didn’t start building this way to prove I could run a company without people. I started building this way because, over decades, I have built a set of personal skills, experiences, and knowledge that add up to value I am ready to invest in. Me. I am ready to make my wager. The deep promise of the one-person company isn’t a bigger margin. It’s the possibility — possibility, not guarantee — of work that lets your money and your time work for you, instead of the other way around.

But that promise inverts the instant you stop tending it. A company of one can become the most family-aligned structure ever invented, or it can become a cage you built yourself and then fired everyone who had a key. Same tools. Same leverage. The difference isn’t in the stack. The difference is the culture of one — whether you remember, while you are automating everything that can be, what the whole thing is supposed to be for.

The agents can do the work. Deciding what’s worth doing, how you get there, who you’re becoming while you do it, and what it’s all in service of? That’s what makes the promise and the peril real.

Everything is a bet.

If you’re like me, that word makes you nervous.

Bet.

But boil it down, and that’s exactly what we’re doing, all of us, all the time. You bet that a degree will open a door in a field you’ve never worked in. You bet on a ball club in October and live or die by a ninth-inning rally. You bet on stocks, on crypto, on a piece of land, on a company you’ll never set foot inside — SpaceX, Anthropic, whatever future others are shaping. You bet that the safe job is safe. You even bet, every single morning, that the way you’re spending today is worth more than the thousand other ways you could have spent it.

The question is not whether you gamble. It’s what you’re putting your chips on.

And after decades of placing bets big and small — some that paid, many that didn’t — I’ve come to believe, with more conviction than I hold about almost anything, that the best bet you can make is on yourself.

The day I walked out

Allow me to give you some context.

Throughout my college years, I was a wandering soul, bouncing from job to job, in and out of school, eventually landing a budding career as an opera singer. Yes, you read that right. I built a real, if not modest, career from my voice. My wife, an accomplished vocalist in her own right, was finishing up her degree in music, graduating Summa Cum Laude from an incredibly challenging program, The Princeton Packet had rated one of the top 5 toughest in the country. Well, we found out she was pregnant, and the math of a life lived in single-night and six-week contracts, with no benefits and no floor under us, stopped adding up against the thing we wanted most. We chose family. We moved from the east coast, back to southern California to be near ours, and I took a job with benefits and a steady paycheck and kept singing on the side, the way you keep a candle lit on a stormy night.

That steady job was with a beloved boutique grocery chain, and I was good at it. I was good with people and enjoyed the crew and the culture at that business. I was good enough to be on the verge of a management track, the kind of step that would have made us comfortable. Safe. A known quantity. A ladder with rungs you could count.

And then I walked away from it.

There was a cause my wife and I believed in. A mission that mattered to us in our bones. The “opportunity,” if you can call it that, was a five-month national tour for little pay and no benefits — months away from my wife and our two young boys, with only a simple promise that the work was worth doing and that I was needed. On paper, it was indefensible. On paper, you stay where you are. This was a big bet.

My wife didn’t just let me go. She insisted I do it. That’s no small gesture, and I won’t pretend it is. We had just started adjusting to a life with a son who was recently diagnosed with Autism. Our family was finding that new footing. That story is one I will tell in a future article, but my wife carried our family so I could support something we both felt called toward. And the bet — the real one, the one that cost us security and proximity and certainty all at once — provided the springboard for what would come next. A non-profit. We helped build it from zero revenue into a multi-million-dollar, mission-driven organization with more than thirty employees, a network-caliber production team, and a legal team of dozens. For ten years, I served as its Creative Director and CIO. Today, it is one of the most disruptive and influential organizations in its space.

I have bought stocks that did well. I have held positions that paid. None of it — not one investment, not one paycheck — ever returned what walking out that door returned. Because that day, I wasn’t betting on a company. I was betting on us. The returns would be incalculable.

Why you are your surest asset

Warren Buffett is a man who could be telling the world to simply buy more of his stock. Instead, he believes that “by far the best investment you can make is in yourself.” He’s expressed the idea repeatedly over the years, on the logic that whatever you build inside yourself can’t be taxed, can’t be inflated away, and can’t be taken by a market crash, a downturn, or a bad quarter. Deficits can balloon, the dollar can wither, the whole edifice can wobble, yet you still walk out the door with everything that actually made you valuable.

When you buy a stock or take a salaried seat, you are betting on someone else’s success. You hand over your money, or your forty hours, and you pray the people at the top make the right decisions. There is nothing wrong with that — I’ll come back to this — but ask yourself: what could that same money do, seeding your own idea? What could those same hours do, poured into your own gift?

Because here is what I believe, underneath the spreadsheets and intimidating business mumbo jumbo. God gave you gifts. He placed specific experiences on your path, and from them you developed a way of seeing the world that no one else has. The problems you notice that others walk past. The solutions you use in your life, or can’t stop sketching on napkins. That experience forged through trial and error, through passion for an issue, it has value. Put frankly, refusing to invest in that may mean you are leaving the best asset you have sitting idle in the dark, untouched, quietly depreciating while you fund everyone else’s vision.

What this is not

Now listen, I’m not saying to be selfish. I am not telling you to quit your job tomorrow. And I am absolutely not saying that the entrepreneur’s path is the only road to self-fulfillment and a happy life.

There is nothing — nothing — wrong with a thriving corporate career, a vocation you love, a steady craft practiced for a lifetime. When you find that fit, I celebrate it. That is the right path for you, and climbing that ladder with excellence is a bet on yourself too: a responsible, worthy one that has provided well for generations of families and will provide for many more. Don’t let anyone with a podium shame you out of a good life.

Betting on yourself isn’t a job title. It’s a posture. It means investing in your peace of mind because a desperate person makes desperate, fear-driven bets. Investing in yourself means investing in your relationships, in your kids, in your spouse, in the handful of friends who’d answer at 2 a.m., in your faith and relationship with God. It means investing in your health, so the body you’ll need for the long game is still standing when the game gets long. And yes, it means investing in your good ideas, the ones you have stewing in the back of your mind, the ones you suspect might be worth something, those ideas you wonder what if about.

These are the investments that compound for the rest of your life. And of every bet I’ve ever placed, they’re the only ones I have never, not once, regretted.

The next season

We’ve done it again.

My wife and I are launching into a new chapter through Layton Ventures, a family-first firm, with a number of ventures that are just the two of us, partners in the truest sense, the same two voices that once sang on stages now building something we hope outlasts us. After ten years pouring ourselves into a mission we loved, we’re stepping out the door once more, into the uncertainty, but uncertainty I have grown comfortable with.

It is, without question, the best bet we’ve made. Not because we’re sure it will work. I’m comfortable telling you we’re not, and if you’ve read me before, you know I’d never pretend otherwise. But because of what Buffett understood and what that grocery store taught me a decade ago: the bet on yourself is the bet no one else gets to call.

So this is my birthday resolution and my invitation. We’re walking into a season of investing in ourselves, our peace, our people, our health, our faith, and our most stubborn ideas. The odds will look however they look. We’re placing the chips anyway.

I can’t wait to share this road with you.

The project was slated for the old Pennhurst State School site in East Vincent Township, Pennsylvania, with three two-story server buildings with their own power generation and a battery field. Against it stood Joanna Tenney, who, with her husband, started a Facebook page to organize neighbors. As she told 69 News, she’s “just a stay-at-home mom” who decided to fight for her kids. After months of packed hearings, the township’s Board of Supervisors voted unanimously to deny the application. The room reportedly erupted in applause.

(WFMZ 69 News — “East Vincent rejects Pennhurst data center proposal,” reporter Rob Manch)

Now, hold the celebration, because the developer’s attorney says they’re going to court arguing that Pennsylvania law obligates the township to permit a data center somewhere, and that the supervisors caved to public pressure rather than observing the law. He may be right. A Michigan township that voted one of these down settled weeks later under threat of a lawsuit, and another Michigan town watched construction begin anyway after it said no. A “win” at the township level often simply represents the opening round.

But let’s hold onto the image anyway: a parent at a folding table beating a billion-dollar buildout. Because East Vincent isn’t an isolated incident. It’s real, and it’s happening everywhere.

In 2017, when Gallup asked Americans whether artificial intelligence had been good for their lives, 79% said yes. Just this spring, a new Gallup poll found that 71% of Americans oppose building AI data centers in their own communities. Talk about public opinion whiplash.

I want to be careful here, because the easy move would be to pick a tribe. One tribe says the people fighting these projects are standing in the way of progress (one venture capitalist claims opponents are being egged on by China, but I don’t think this is the case. The opposition is real and prevalent). The other tribe says Big Tech is draining aquifers and energy capacity, and lighting up the sky over our kids’ bedrooms so some chatbot can write song lyrics. Both tribes have a point.

So let me disclose where I stand before I lay out the facts: I am not against data centers. I’m typing this on infrastructure data centers make possible. The question that actually keeps me up isn’t whether we build them; it’s whether we’re building the right number of them, in the right places, paid for by the right people, with transparency about the true cost to the communities that host them. On every one of those questions, the current evidence makes me more skeptical than the “pro-data-centers-everywhere” tribe would like.

Let’s go through it. The good, the bad, and the genuinely ugly.

First: How many do we need?

Here is the part of the story that most people don’t know. A surprising amount of the “demand” driving this buildout isn’t real.

When a developer wants grid power, they file an interconnection request with the utility. It’s cheap to file. So developers have been filing the same project with multiple utilities, or filing early-stage projects they may never build, to see who can connect them fastest. Utilities refer to this phenomenon as “phantom” data centers, load that haunts the forecast but will never actually draw power. One large utility watched roughly 15 gigawatts of requested demand collapse to about 8 the moment it required an upfront deposit. Nearly half the request was phantom.

This matters enormously because utilities plan, build, and bill you against those forecasts. If the forecast is inflated by speculative double-counting, ratepayers can end up funding generation and transmission for facilities that never arrive.

And it’s not just utility planners who are nervous. Alibaba’s chairman warned of “the beginning of some kind of bubble”, noting that some centers are being built on spec — no tenant, no buyer, just a bet. Analysts have started comparing the moment to the 1990s global fiber-optic glut, when the country wildly overbuilt capacity into a hype cycle.

The five biggest hyperscalers’ combined 2026 capital spending is projected to top $600 billion — up roughly two-thirds from 2025. Amazon’s capex alone is reportedly larger than the entire publicly traded U.S. energy sector.

Now, the honest counterpoint, because being comfortably uncertain, I have to give that to you. Plenty of serious analysts think the demand is real and durable. Grid operators are not inventing the strain; interconnection queues in mature hubs now stretch as long as seven years. Models keep getting bigger and hungrier. A good portion of this capacity will be needed to support growth at the scale capital concentration around tech is signaling.

But “some of this is needed” is a very different statement from “all of this, everywhere, right now, at any cost.” When those building are taking on historic levels of debt and their own industry peers are using the word “bubble,” a healthy society does not respond by suppressing every voice of concern and rezoning the cornfield. It requires proof of demand before it pays for the substation.

The Good

Let’s be fair to the case for building, because it’s not nothing.

The infrastructure need is genuinely real. Video, music, and live streaming, maps, cloud storage, every app on your phone, and yes, the AI tools a lot of us now use daily — it all lives somewhere. That somewhere is a data center. This is the digital plumbing of modern life. Pretending otherwise is as dishonest as pretending the costs to communities are nil.

The tax revenue can be substantial. A single large project can generate millions of dollars a year in local tax revenue — money for schools, roads, and services — sometimes without adding a single student to the district. Some agreements have generated billions in tax revenue and thousands of construction jobs. For a rural county, that can be transformative.

The engineering is improving fast. This is the most underreported good news in this story. Newer “closed-loop” cooling systems recycle the same water rather than evaporating up to millions of gallons of water a day. Once filled, some use almost no water in daily operation. Air-cooling and direct liquid cooling can cut water use by 70% or more. The industry now tracks water-use effectiveness as a design target, with some chasing zero.

And the power story is, in places, getting cleaner. The AI boom is single-handedly reviving American nuclear power. Microsoft signed a roughly $16 billion, 20-year deal to restart a reactor at Three Mile Island; Google contracted for a fleet of small modular reactors; Amazon is pouring $20 billion-plus into a nuclear-powered campus; Meta issued a request for up to 4 gigawatts of new nuclear capacity. If you care about clean energy, and I do, there’s a real argument that AI demand is funding the energy transition that decades of climate policy couldn’t.

There’s a version of the data center that is a good neighbor: closed-loop-cooled, nuclear — or renewable-powered, located away from stressed water sources and vulnerable communities, paying its own freight, and generating transformative revenue for the community. That version exists. The trouble is, it’s not the only version being built.

The Bad

Here’s where the pro-crowd gets quiet.

Your electric bill. This is the one that hits everyone. In the mid-Atlantic grid that runs from Illinois to North Carolina, data centers drove an estimated $9.3 billion price increase in a single capacity auction, which is projected to add about $18 per month to the average bill in western Maryland and $16 per month in Ohio. A Carnegie Mellon study estimates that data centers and crypto could push the average U.S. electricity bill up by 8% by 2030 and by over 25% in the hottest markets. The core injustice is simple: a private company builds a facility, the grid requires expensive upgrades to serve it, and, absent strong protections, households and small businesses end up paying for it.

U.S. data centers used about 4% of the nation’s electricity in 2024 — roughly the entire annual demand of Pakistan — and the Department of Energy projects that share could hit between 6.7% and 12% by 2028. In Virginia, the world’s largest hub, data centers already consume more than one in four kilowatt-hours. In Ireland, it’s 22% of the entire country.

The water. A single large data center can draw up to 5 million gallons of water per day. This is equivalent to the daily use of a town of 10,000 to 50,000 people. Often it’s pulled straight from the municipal drinking-water mains, the same supply that serves homes. Google’s largest U.S. facility, in Council Bluffs, Iowa, withdrew an average of 3.9 million gallons a day in 2024.

And the regional math is genuinely alarming, where water is already tight. A Houston Advanced Research Center study found Texas data centers would use 49 billion gallons in 2025, rising to as much as 399 billion by 2030, the equivalent of drawing the largest reservoir in the country, Lake Mead, down by more than 16 feet in a single year. This is happening in a state where Corpus Christi is preparing water-emergency cuts. The drought doesn’t care that the water went to cooling rather than to crops.

The jobs mirage. This is the part communities are most routinely misled about. Yes, construction creates hundreds of temporary jobs. But a finished data center is mostly an empty, humming box. Typical permanent staffing runs 30 to 50 people. In Northern Virginia, a campus that once employed 5,300 people was demolished to make way for data centers employing only a tiny fraction of that workforce. These are not the factories of the 20th century. The tax revenue can be real; the “jobs for the community” pitch is less so.

The Ugly

Then there’s the conduct that’s hard to defend under any framing.

Memphis. When xAI couldn’t get enough grid power for its Colossus facility in a historically Black, asthma-burdened neighborhood, it trucked in dozens of natural gas turbines and ran them without completing the permitting process. One analysis estimated those turbines could emit up to 2,000 tons of nitrogen oxides a year. The NAACP and residents filed notice to sue under the Clean Air Act. This is the part of “the future” nobody put in the keynote.

The pledges that may be theater. Under pressure, the big four signed a White House “Ratepayer Protection Pledge” promising not to pass their energy costs on to households. Then the Energy Department acknowledged there is no monitoring or enforcement mechanism behind it. Critics have called the broader genre of corporate “good neighbor” pledges theater. A promise with no tool of enforcement is a press release.

Not so “Water positive.” Microsoft and Google have pledged to be “water positive” by 2030, partly through offsetting. But that is a hollow promise. Replenishing a watershed in one place does nothing for the community that lost its water somewhere else. One specialist called the offsetting framing a “strategy of obfuscation.”

Then there’s the closed-loop catch. Even this cooling solution has a hidden cost the brochures skip: recycling water rather than evaporating it means you have to keep re-chilling it, which consumes more electricity. You can save the water or save the power; either way, there’s a trade-off.

Do we actually know the health impacts?

You asked the question, and the honest answer is: not as well as we should before we build in communities all over the country.

What we have are strong signals. Data centers run 24/7, and the diesel backup generators and on-site turbines emit fine particulate matter (PM2.5) and nitrogen oxides linked to asthma, heart disease, and cancer — with diesel backup units emitting hundreds of times more NOx than a natural gas plant. One proposed project planned 168 diesel generators. The constant low rumble of cooling systems and generators is associated with sleep disruption, hypertension, and cardiovascular stress, and there’s the round-the-clock light pollution too. A University of California, Riverside researcher has projected that data center air pollution could contribute to hundreds of thousands of asthma cases and over a thousand premature deaths at scale.

But bear in mind, those are projections and associations. The reason we don’t have harder numbers is itself a burgeoning scandal. Generator operators are largely left to self-report emissions; there’s no national transparency standard for what data centers emit, and the buildout is outpacing monitoring. We’re running the experiment on real neighborhoods in real time, expecting to read the results later. Will we? When the uncertainty impacts children’s lungs, “we’re not sure yet” is not reassuring.

Are there alternatives

There are several alternatives being seriously pursued:

Phantom demand needs to be addressed. Filter the speculative interconnection requests, require deposits, demand proof of a tenant, and a big chunk of the impending burden evaporates, as that 15-to-8 gigawatt collapse demonstrated. Don’t build substations for the phantom.

On power, the real menu is nuclear (the SMR and reactor-restart deals above), on-site renewables paired with battery storage instead of diesel — batteries can cover outages without the emissions, and efficiency gains in the chips themselves.

On water, it’s closed-loop and air cooling, reclaimed (non-potable) water instead of drinking water, and the simple application of logic: don’t put the thirsty ones in the desert.

And on who pays, Texas wrote a template worth tracking. Senate Bill 6, signed in 2025, makes large loads fund their own interconnection costs, disclose their backup generation, and — the provision critics nicknamed the “kill switch” — accept remote disconnection during grid emergencies so a data center can’t take your AC down with it during a heat wave. It’s a “pay your own way” philosophy, and it’s catching on. Pair it with enforceable community benefit agreements, like the Lancaster, Pennsylvania deal, with a hard water-use cap, emissions controls, and $20 million in community contributions, and you start to see a fair bargain taking shape.

Where I land (loosely)

Communities are not, on the whole, rejecting the digital age. The careful reporting indicates they’re rejecting being told to absorb the costs, the water, the noise, and the air while a faraway company keeps the upside, in turn being asked to accept “it’ll be fine” at face value. That’s a reasonable response to a genuinely bad deal, repeated often enough that residents in at least 14 states have passed pauses and roughly $156 billion in projects got blocked or delayed last year.

So here’s where I am comfortably uncertain:

We are building too many, too fast, many in the wrong places, and placing the burden on the wrong people. Some of these data centers we need. Some are speculative bets dressed up as infrastructure necessity. The boom is moving faster than our ability to measure its effects on a watershed, your power bill, or your kid’s respiratory system, and “we’ll find out later” is being treated as an acceptable answer when it just isn’t.

So, show me closed-loop cooling and reclaimed water instead of the municipal main. Show me clean nuclear power or batteries (another inefficiency I’ll address in a future write-up) instead of unpermitted turbines next to a school. Show me a “pay your own way” law codified. Show me a community benefit agreement with teeth and a standard of transparency that lets a county measure what it’s hosting. Do that, and I’d consider welcoming the server farm down the road.

The fight over data centers isn’t really about whether we want the future. It’s about what kind of future we want, whether we’re willing to be honest about the cost, and whether we can find a balance between that cost and the communities it will carry into that future.

I’m comfortable being uncertain about a lot of this. I am, however, certain the people asking these questions are not the problem.

Joanna Tenney may yet lose Pennhurst in court. The law may genuinely be on the developer’s side. But a warrior mom reading zoning ordinances at her kitchen table because nobody else would— I don’t see that as an obstacle to a functioning future. That is the key to a future that looks bright.

As a first-time venture CEO, I spend a lot of time thinking about two contradictory truths at once:

1. This is quite possibly the greatest era in history to build a new business.

2. The probability that someone disrupts everything we’re doing has never been higher.

Bill Ackman articulated both sides perfectly in a short segment from the Liquidity Conference this week. I clipped it because it’s one of the clearest, most grounded takes I’ve heard on how to think about AI disruption as an operator and investor.

“There’s unlimited access to compute… unlimited access to capital… and a lot of incredible talent. Which means that the probability of your being disrupted has gone up enormously.”

He then points out what usually happens in these moments: capital floods the obvious new thing (chips, semiconductors, energy infrastructure), while durable, high-quality compounders get unfairly left behind. He sees Amazon, Meta, and Microsoft trading like “old-fashioned” companies in the AI era, exactly the way Berkshire Hathaway looked in 2000.

I’m still early in this journey, but Ackman’s framing is sharpening how I evaluate our own defensibility. Are we building something that benefits from the AI tailwinds, or are we the garage team about to eat someone’s lunch?

And on the capital allocation side, are we paying enough attention to the “boring” high-quality businesses that are being ignored while everyone chases the shiny new layer?

If you’re a founder, operator, or investor trying to navigate this environment with intellectual honesty, this 90-second clip is worth your time. I’d love to hear how you’re thinking about disruption risk in your own company or portfolio right now.

The caption? “This is a cataclysmic exit liquidity avalanche.” And yeah, the trading-floor panic video nailed the vibe.

I live in that tension. This wave of capital concentration is a legitimate risk. But I’m not reaching for the crash helmet just yet. Here’s why I’m cautiously optimistic.

The Scale Is Eye-Watering

We’re talking hundreds of billions flowing into AI compute, data centers, chips, and related infrastructure in a compressed timeframe. Hyperscalers collectively plan $600–750 billion in capex for 2026, much of it AI-driven. Alphabet alone is guiding toward $180–190 billion in spending. These aren’t small pivots — they’re generational bets on transforming how the economy computes, creates, and scales.

Early investors, employees, and insiders are getting liquidity at valuations that would have seemed insane a few years ago. That’s the “avalanche” part. When big players cash out or dilute to fund growth, it can signal peak exuberance. History rhymes: capital-intensive booms (think fiber optics in the late ’90s or shale in the 2010s) often overshoot before settling.

Concerns are valid:

• Monetization lag: Capex is surging faster than proven AI revenue in many cases. If enterprise adoption slows, efficiency gains reduce compute needs, or power/regulation bottlenecks hit, the ROI math gets ugly.

• Dilution and stretch: Alphabet shares dipped on the announcement. SpaceX at ~$2T implies massive multiples. Private AI valuations assume explosive growth that still needs to fully show up in profits.

• Concentration risk: So much capital chasing the same thesis. A sector-specific hiccup could ripple.

This is classic late-cycle behavior that makes any thoughtful investor pause.

Here’s Why I’m Buoyed

The players involved separate this moment from past mania.

Berkshire Hathaway isn’t throwing darts. Warren Buffett’s team has been building an Alphabet position and now doubles down with $10B. That’s not retail FOMO, but a signal from one of the most disciplined capital allocators on the planet. They see durable competitive advantages and real demand.

Customer pull is genuine. Hyperscalers report surging backlogs and “unprecedented” demand for AI capacity. Enterprises aren’t just experimenting anymore; they’re integrating in ways that drive measurable productivity. Cloud growth is accelerating, not plateauing.

Technological momentum. AI capabilities keep leaping. Models improve, use cases multiply (coding, analysis, creative work, scientific discovery), and the flywheel of better models → more usage → more investment feels sticky. SpaceX brings tangible progress in space tech, Starlink, and defense—real products, not just hype decks.

Broader context avoids 2000-style detachment. Many of these companies have massive cash flows, diversified revenue streams (search, cloud, subscriptions, hardware), and strong balance sheets even after spending. We’re not in a zero-interest-rate everything-bubble environment. Macro conditions, while not perfect, support continued tech investment.

Analysts debating this aren’t uniformly screaming “bubble.” Many see today’s infrastructure buildout as necessary table-stakes for tomorrow’s monetization wave. Today's capex seeds tomorrow's revenue.

The Takeaway

This is risky. Capital concentration at this scale always carries the potential for painful corrections if timelines slip or expectations reset. No serious observer should dismiss that.

But I’m not forecasting another major market crash. The fundamentals anchoring this — technological progress, proven operators like Alphabet and SpaceX, credible backers like Berkshire, and exploding demand — feel more robust than past manias. This looks less like irrational exuberance detached from reality and more like a high-stakes, high-conviction arms race with asymmetric upside.

We’ll watch the metrics that matter: actual AI-driven revenue growth, utilization rates on new capacity, free cash flow trajectories, and whether power/infrastructure constraints become binding. If those hold up, the avalanche could turn into a sustained boom.

In the meantime, stay diversified, keep perspective, and remain comfortably uncertain. The future is being built at warp speed. Best to observe it with eyes open and portfolio balanced.

But what do you think? Is this the infrastructure phase of a new era, or are we planting the seeds of the next bust?

The fire department brought in a ladder truck with an aerial platform, harnessed each kid one at a time, and walked them down the ride's steep steel skeleton into a rescue basket. It took almost four hours. The last student came down a little after 9 p.m. The fire chief said the riders were checked mainly for dehydration — they’d been stuck in the sun a long time.

Thankfully, none of them were injured.

Watch the ABC7/KTRK footage of the Iron Shark rescue here:

https://abc7.com/post/galveston-texas-roller-coaster-stopped-rescue-underway-riders-stranded-upside-down/19189772/

The footage is genuinely frightening. But the result was a group of teenagers who were bored, maybe sunburned, and probably never going on that ride again, but completely fine. What’s lost on social media is that the contrast between how dangerous it looked and how dangerous it actually was is enormous. But I feel like I’m seeing more and more videos of carnival and theme park rides breaking down or failing entirely.

So let me ask this: Are thrill rides still safe to ride?

Are thrill ride accidents more common today?

Maybe you’ve seen the videos — the carnival ride that buckles, the coaster stalled upside down, the drop tower that won’t drop. I’ve seen them too. They’re seemingly everywhere. I’d be lying if I said it didn’t affect my decision to refrain from riding certain rides with my adventurous son at our local fair.

We know why videos of those events permeate social media platforms. It’s clickbait. Millions of people spun around and step off smiling without incident? Not much to see there. Nobody has ever opened TikTok to a clip captioned “Local Man Rides Tilt-A-Whirl, Experiences Mild Joy, Gets Funnel Cake.” Researchers who study this have a name for the distortion: thrill ride accidents get inherently amplified, because news coverage favors the dramatic and the unusual. One analysis found that roughly 19% of media-reported ride accidents involved a fatality — while in actual government injury data, fatalities occur at a rate of one per 2,000 to 3,000 injuries. The feed isn’t lying to you, exactly. It’s just showing you the 0.04% without the context of the other 99.96%.

The algorithm rewards “Holy hell.”

Putting Your Mind at Ease

Here’s the part that surprised me when I dug in.

The industry’s headline number, produced annually for the trade association IAAPA by the National Safety Council, is this: the chance of being seriously injured on a fixed-site ride at a U.S. amusement park is about 1 in 15.5 million rides taken. For scale, somewhere around 385 million people take roughly 1.7 billion rides at North American parks in a typical year.

The rest of the numbers point the same direction. In a recent year the survey estimated about 1,390 ride injuries total, with under a fifth classified as serious — and the share of serious injuries has been falling, dropping 54% in a single recent year to around 8% of all injuries. Overall ride injuries are down about 29% since 2003. Deaths tied to amusement rides run fewer than five a year in the U.S. — a number that’s stayed low and stable for decades against a backdrop of hundreds of millions of riders.

Roller coasters specifically carry the highest injury rate of any ride type — and even that comes out to roughly 1.3 injuries per million rides. You are vastly more likely to be hurt driving to the park than on anything inside it.

Things that will kill you first

Hold “fewer than five deaths a year” in your head. In a typical year in America:

So the scoreboard reads: a beehive, the family doodle, a summer thunderstorm, and a herd of cows each out-kill every roller coaster in America, and the high-speed contraption engineered from to elicit screams loses, annually, to a malfunctioning Coke machine.

One bit of fairness before you screenshot that for your group chat: these are raw death counts, not exposure-adjusted rates. Vastly more people pet dogs and stroll past vending machines than strap into inverting steel coasters, so it’s not a clean apples-to-apples risk comparison. But as a gut check on where your fear is actually warranted, it’s bracingly clear — the universe has a lot of more mundane ways to get you, and it deploys them far more often than the Iron Shark ever will.

So if you want the short, strong answer: yes, statistically, modern thrill rides are remarkably safe. Getting stuck at the top, like those kids in Galveston, is overwhelmingly the most “dangerous” thing that will happen to you, and getting stuck is — this matters — the safety system doing its job. Rides are engineered to halt when a sensor reads something it doesn’t like. A stall is the fail-safe triggering. In this case, the ride is functioning, or at least its safety systems.

I could stop there. Ride enthusiasts like my son would like me to stop there. But this is Comfortably Uncertain, so here come the caveats.

Caveat One: We’re partly grading the industry by its own research

That beautiful 1-in-15.5-million figure isn’t from a regulator. It comes from a voluntary survey the industry runs on its own. In a given year, only around 180 to 210 facilities — out of roughly 440 operating parks — actually respond, representing maybe 60-64% of attendance. The data is real, and the methodology is sound, but it’s self-reported, partial, and assembled by the people with the most to lose from poor results.

It wouldn’t bother me if this were the only caveat. It isn’t.

Caveat Two: the “roller coaster loophole.”

The Consumer Product Safety Commission — the federal agency that recalls dangerous toys and appliances — has no authority over the permanent rides at your big theme park. In 1981, a budget bill stripped the CPSC of jurisdiction over “fixed-site” amusement rides, an exemption critics have spent forty years calling the roller coaster loophole. Fixed-site rides were, for a long stretch, the only product category exempted from CPSC regulation and not picked up by any other federal agency. There is still no national requirement that a theme park report a ride malfunction or fatality to anyone federal, and no central clearinghouse so that a defect found on one ride in one state triggers a fix on identical rides everywhere else.

The result is a 50-state free-for-all. Currently, 44 of 50 states regulate fixed-site parks, leaving Alabama, Mississippi, Montana, Nevada, Wyoming, and Utah essentially without oversight. Massachusetts Senator Ed Markey has introduced bills to close the loophole repeatedly since the 2000s; none have passed. The industry argues a federal agency couldn’t outperform the existing system. That may be true. But “trust us, our private survey says it’s fine” is a weaker foundation for confidence than the safety record itself deserves.

So where does this leave us? I’m comfortably uncertain rides are very safe, and our certainty is lower than it could be because no one is required to report to a regulatory authority when something does go wrong.

What about carnival rides?

Now, it almost goes without saying that the robust permanent coaster at a major theme park is a completely different animal from the ride that shows up in a county fair parking lot last night, gets bolted together, and is torn down every week.

The mobile carnival rides are the ones the CPSC actually does cover. But coverage does not include inspection. And, with up to 3000 fairs in the US annually, the agency doesn’t have the manpower to check every fair. Portable rides face stresses permanent ones don’t: constant assembly and disassembly, different conditions at each stop (on concrete, asphalt, muddy grass?), decades-old equipment trucked from town to town, and difficult-to-identify metal fatigue.

Which brings us to the Ohio State Fair, July 2017. A swinging, spinning ride called the Fire Ball was running when one of its arms snapped mid-flight, throwing four riders to the pavement 20 feet below. Tyler Jarrell, an 18-year-old who’d just enlisted in the Marines, was killed; seven others were hurt. The cause, per the Dutch manufacturer’s metallurgical investigation: excessive corrosion on the interior of a support beam, which had quietly thinned the steel wall of the beam over the ride’s 18-year life. The chilling part is that the Fire Ball had passed inspection three or four times in the days leading up to the incident, with passing marks on dozens of items, including possible cracks and proper assembly. The inspection sheet wasn’t designed to reveal this kind of problem.

So how do they compare to fixed rides? The big theme park coaster has a better case for safety. The carnival ride is still statistically safe, but it lives closer to the edge of the system: less permanent infrastructure, more movement, more assembly, more wear, less clean data, and inadequate oversight.

The scariest failures aren’t structural. They’re bad decisions.

If you actually trace the rare catastrophic incidents, a pattern emerges that’s more unsettling than “rides are dangerous.” It’s that the deadliest failures usually aren’t random — they’re human.

In March 2022, 14-year-old Tyre Sampson fell to his death from the Orlando FreeFall drop tower at ICON Park — a fixed-site ride. A maintenance technician later alleged in a whistleblower suit that the seat sensors had been manually modified to accommodate riders outside the safe height and weight limits, that overheated cylinders kept seats from securing properly, and that crews weren’t trained on maintenance or patron safety. Tyre was above the ride’s weight limit. The restraint didn’t hold him. The failure wasn’t a freak act of physics — it was a chain of choices to override the safety envelope the ride was designed around. Florida later passed the Tyre Sampson Act tightening ride rules.

This is the genuinely ugly part, and it’s where my skepticism actually lives. Not in the engineering — the engineering is extraordinary. It’s in the gap between the ride as designed and the ride as operated and maintained on a hot afternoon by an underpaid crew running behind schedule. Aging equipment, hidden corrosion, sensors fudged to keep the line moving — those are the threads that run through the rare tragedies, and none of them show up in a glossy “1 in 15.5 million” headline.

Both sides are playing with numbers

Both sides of the safety issue play the same game, and you should know the move when you see it.

The industry says “1 in 15.5 million rides” — true, but it’s their survey, and “per ride” is a flattering denominator when each person takes a dozen rides a day. Years ago, Senator Markey argued from the other direction that coaster fatalities per passenger mile exceed those of trains, buses, and planes. While technically defensible, it’s also nonsense, because “passenger mile” is a unit built for transportation, while a roller coaster is a 90-second emotional experience that happens to travel a short distance very fast. Same underlying reality, opposite spin, depending on which fraction someone wants to use to sway you.

When you strip the framing away, the middle is where the truth sits: per ride, per visit, per summer, the odds of getting hurt are genuinely tiny. Which is, admittedly, a terrible chant to start in line. “WHAT DO WE WANT? Statistically insignificant risk! WHEN DO WE WANT IT? Within one standard deviation of the mean!” Nobody’s printing that on a T-shirt. But it’s the truth.

Where I land

Bottom line: yes, they’re safe. And, yes, I’m still wary of carnival rides. If you watched a video of people stranded on a coaster this week and felt your stomach drop, the most likely ending to that story, by a wide, wide margin, is exactly what happened in Galveston: a long, uncomfortable wait, a ladder truck, and everyone walking away. The terror is real; the danger usually isn’t. And the thing that scared you, the ride stopping dead, is not the machine betraying its passengers, but protecting them.

The ride itself is almost certainly fine. What’s worth your attention is everything around it. A parent can’t fully inspect the engineering of a carnival ride, but you can judge the warning signs around it: watch one full cycle, look for rough movement, strange noises, visible rust, loose parts, sketchy supports, or distracted operators. Make sure the ride has posted rules, that your child clearly meets the height and restraint requirements, and that the operator physically checks every lap bar, belt, or harness before starting. If the setup looks chaotic, the crew seems careless, the ride looks worn out, or your gut says something is off, skip it. The safest rule is simple: clean setup, serious operators, clear rules, snug restraints, green light; sloppy crew, tired equipment, bad fit, walk away.

As a Catholic, I find the Pope’s warning timely. As someone who has worked in technology and helped develop AI models and tools to solve real problems, I also find this conversation personally challenging.

Because I do not believe AI is evil. It is, however, soulless.

And, I do believe it is dangerous when we forget what it is. AI is constructed from flawed beings.

The Pope’s central concern, as I understand it, is not that machines can process information quickly. It is that human beings may begin to organize society around the machine’s logic: efficiency over dignity, prediction over prudence, automation over vocation, power over wisdom. Vatican News summarized the encyclical as an appeal for “truth, dignity of work… and peace,” while warning that AI must serve humanity rather than concentrate power.

That is the balance we are already struggling to find.

AI can be an extraordinary tool. I have seen it help summarize complexity, discover difficult-to-grasp patterns, accelerate workflows, personalize education, improve communication, and remove friction from problems that used to consume hours of human effort. Used well, AI can give people leverage, create opportunities, and improve the quality of life for groups and individuals. It can help the small teams compete with the large institutions. It can help a doctor, a lawyer, a parent, a researcher, a writer, or an entrepreneur move faster and think more clearly.

That represents a positive innovation that, when used in balance with our individual skills and talents, can be quite powerful.

Even working in innovation, maybe especially because of it, I find myself fighting an intrinsic anti-technology reflex. But the Church has never taught that tools are bad simply because they are powerful. A printing press can spread both the Gospel and harmful propaganda (and it has). A camera can reveal truth or fabricate illusion. A scalpel can be used for healing or for harm. The moral question is not merely, “What can this tool do?” That is the question markets ask. The moral question is not merely, “What can this tool do?” That is the question CEO’s ask. The more important question is, “What should this tool do?”

At its core, I see this as a spiritual test.

The Pope warned that AI needs to be “disarmed,” stripped of its capacity to dominate, exclude, dehumanize, and, in the case of war, kill. This is the fundamental crossroad we are at, one between AI and Faith. To disarm AI is not to unplug every server or retreat into nostalgia. It is to insist that intelligence without a conscience is not wisdom. Optimization without the moral foundation that comes with fearing for the fate of one’s soul is a very slippery and perilous slope.

Humanity gathered its genius, its ambition, and its technical capacity to build upward before. It was called the Tower of Babel. The problem was not architecture. The problem was orientation. They were building toward heaven without humility before God.

I fear that temptation, that ambition, that lack of humility, is alive and prolific in the AI age.

We are told AI will make us more productive, more informed, more creative, more efficient, and perhaps someday, more than human. But Christianity begins with a very different claim, and with a very different end in mind. The goal of the Christian life is not to become more than human. It is to become holy. It is the salvation of the soul.

We are not brought into this world to overcome our humanity as if it were a design flaw. We are brought into this world to receive it, to suffer through it, to sanctify it, and to offer it back to God. Our limits, our weakness, our dependence, our need for grace, are part of the condition through which we learn humility, love, repentance, service, and trust. In these things, we become connected to one another and to God. That is the true body of Christ.

And that is where AI becomes spiritually dangerous. It can help us write, calculate, organize, discover, build, and more. Those uses can be beneficial to God’s design for us. The danger comes when AI begins to intercede in the place where only God belongs. When it offers us the illusion of omniscience, a gatekeeper to all knowledge, as Sam Altman is predicting, when it promises to make us “more than human.” When it is given authority over humanity to solve our ‘greatest problems,’ then we have a great problem.

Catholic faith teaches us to lift our imperfections before God. AI companies offer solutions to abolish all our imperfections through technology. The moment we begin to believe that we can improve upon God’s design, whether through a medicine, a social theory, a political ideology, or an artificial intelligence, we are no longer simply using a tool. We are engaging in idolatry.

And it certainly has all the marks of a modern golden calf. AI can be a powerful servant. It must never become a savior. It can assist human reason. It must never replace conscience. It can help us work. It must never redefine what we are. It must never have authority over us.

The Christian answer lies in the knowledge that wisdom begins in the fear of the Lord. That should be the guidestone for humanity’s use of AI.

If AI removes drudgery and allows people to spend more time in meaningful labor, family life, service, creativity, and contemplation, then it can be a blessing. But if AI becomes a corporate weapon to erase workers, flatten wages, surveil employees and citizens, and convert every human contribution into an input of time and energy, then it becomes something else entirely. Reuters reported that Anthropic co-founder Christopher Olah, speaking at the Vatican event, warned about large-scale job disruption and said AI must be guided from outside Big Tech.

But who on this earth do you trust to be the guide?

Entrepreneurs, builders, and technologists have an essential role to play. The answer cannot be government control alone. We know that historically, where power is centralized, corruption is too. To that end, it cannot be Big Tech policing itself either. Neither can it be religious leaders issuing warnings from the sidelines alone. We need a moral conversation that includes the Church, technologists, families, workers, educators, policymakers, and ordinary people whose daily lives will be impacted by systems they did not build and may not understand.

For me, the balance seems clear at this point: use AI, but do not outsource your soul to it. Don’t give it full authority over you. Let it inform your decisions, but do not let it make critical decisions for you.

Use it to draft, but not to discern.

Use it to organize information, but not to define truth.

Use it to improve your work, but not to replace your vocation.

A Catholic approach to AI is in good stewardship. Gratitude for human ingenuity, suspicion of concentrated power, protection for the vulnerable, reverence for the individual, and always humility before God.

That is where I find myself: comfortably uncertain.

I am excited by what AI can do. I am wary of what it may do to us. I believe it can serve the common good. I also believe it will not naturally serve the good unless we ensure it does.

The Pope’s announcement is not a rejection of the future. It is a reminder that the future must remain human. More importantly, it must remain ordered toward God.

And maybe that is the question we should ask before adopting any new technology:

Does this make me more human, more loving, more truthful, more free?

So there is promise and peril ahead. I certainly hope humanity surprises me and takes the path of promise.

There is much more to unravel here, and I will be doing so in a future Substack. Subscribe now!

On July 23 and 24, 2026, the FDA’s Pharmacy Compounding Advisory Committee (PCAC) will meet at White Oak to decide the legal status of seven peptides that physicians and patients are already discussing in clinics across the country. The committee’s recommendation will help determine whether these compounds can be lawfully prepared by 503A compounding pharmacies — that is, whether they live in the regulated, physician-supervised mainstream of medicine, or whether they remain in a regulatory gray zone that pushes patients toward unregulated sources.

This is a quiet meeting with loud consequences.

What’s on the agenda

Day One — July 23:

• BPC-157 — ulcerative colitis

• KPV — wound healing and inflammatory conditions

• TB-500 — wound healing

• MOTs-C — obesity and osteoporosis

Day Two — July 24:

• Emideltide (DSIP) — opioid withdrawal, chronic insomnia, narcolepsy

• Semax — cerebral ischemia, migraine, trigeminal neuralgia

• Epitalon — insomnia

The list is eye opening. Wound healing. Ulcerative colitis. Migraines. Opioid withdrawal. Obesity and osteoporosis. These aren’t exotic indications. These are conditions that fill clinic schedules every day, and where patients and prescribers are desperate for additional tools that provide real relief.

My position, plainly

Medicines that a licensed physician judges useful for a specific patient should be available through quality pharmacies under physician supervision — without a gray area.

Either a compound is appropriate for use within a doctor-patient relationship under the oversight of a licensed pharmacist, or it isn’t. When the regulatory framework leaves these substances in limbo — neither approved nor clearly compoundable — the predictable result isn’t that patients stop seeking them. The result is that they obtain them from potentially lower quality research-chemical websites, unaccredited overseas vendors, and supply chains with no chain-of-custody, no sterility assurance, and no professional accountability.

This represents the worst of both worlds. It removes the physician. It removes the pharmacist. It keeps the substance in circulation while stripping out every safeguard the profession has built around it.

The 503A pathway exists precisely so that licensed compounding pharmacies — operating under FDA, state board, and USP standards — can serve the patients that mass-manufactured products don’t reach. A favorable recommendation from the PCAC on these peptides is not a deregulatory shrug. It’s the opposite. It’s the affirmation that these compounds belong inside the professional infrastructure of American medicine, not outside of it.

How to put your voice on the record

The public docket is open, and the FDA reads it.

• Docket number: FDA-2025-N-6895

• By July 9, 2026 — comments are delivered directly to the Committee

• Through July 22, 2026 — comments are still accepted and considered by FDA

• By June 30, 2026 — deadline to request to give an oral presentation

• File at: Regulations.gov

Physicians, pharmacists, and patients with first-hand experience, and the general public are the voices this docket is designed to hear. If a peptide on this list has been part of your care or your practice, your treatment, or you want it to be, the Committee benefits from knowing that.

I’ll be watching this meeting closely. Of this, I am comfortably uncertain: Patients, physicians, and the pharmacies that serve them deserve clarity, not a gray area.