cryptopath on Nostr: The Drawbridge Hundreds of billions are being poured into AI that earns almost ...

The Drawbridge

Hundreds of billions are being poured into AI that earns almost nothing back. That's not a reason to stop — it's a reason the people who spent it now need you afraid. The oldest move in capitalism is about to run again, and the tell is a fallacy you already know.

You've bought the concert ticket. You get to the venue and you feel sick — genuinely unwell, the kind of night where staying home is obviously the right call. And you go anyway, because you already paid for the ticket.

That's the sunk cost fallacy: throwing good money, time, or effort after bad, because of what you've already spent — which is gone either way and should have no bearing on the next decision. The ticket price is spent whether you go or not. The only sane question is whether tonight, from here, going is worth it. But the human mind can't let the spent thing go. So it doubles down to justify the loss.

Everyone knows this fallacy as a personal foible. Almost no one notices when it's running at the scale of a civilization. Right now, it is — and following it tells you exactly what's coming.

Where the fallacy is running now

Look at the money going into AI. Hundreds of billions of dollars — data centers, chips, energy, the largest capital buildout in the history of private enterprise — against revenue that, for the frontier labs, is a rounding error next to the spend. The numbers don't close. By any ordinary measure, enormous sums have been committed to something that does not yet pay for itself, and may not for a very long time.

Now apply the fallacy. When you have spent that much, you cannot stop. Stopping means admitting the hole is a hole. So the only psychologically and financially bearable move is to spend more — to keep building, keep raising, keep promising the payoff is just over the next ridge. The deeper you are, the more impossible retreat becomes. The sunk cost doesn't warn the players off. It traps them into escalation.

And here's where it stops being a story about reckless companies and becomes a story about you. Because an enterprise that has spent far more than it can earn back has only a few ways out. Grow into the spend — maybe. Collapse — possible. Or find someone else to make the bet pay off. Someone with deep pockets and no choice but to help. Someone like a government. And the surest way to get a government to pour public money into your private hole is to make the public afraid. The drawbridge

Here is the oldest move in capitalism, and once you see its shape you'll see it everywhere.

You're the first to climb the castle. You get inside, you take the high ground — and then you raise the drawbridge behind you, so no one else can follow. You don't stay on top by being better than the next climber. You stay on top by making the climb illegal for everyone after you.

In a free market, the way you raise the drawbridge is regulation — and the way you get the public to pay for raising it is fear. The script runs like this. First, warn loudly that your own product is so powerful it could be catastrophically dangerous. Then, position yourself as the responsible one, the only grown-up who can be trusted to handle the dangerous thing safely. Then let the government respond to the fear you manufactured — with funding for the "serious" players, and with rules, licenses, audits, and compliance regimes so expensive that only a company with billions can clear them.

And notice what that does. The rules don't fall on the giants. The giants helped write them, and can afford them easily. The rules fall on everyone smaller — the startup, the open-source project, the foreign challenger — who now can't afford to enter at all. The safety regime, whatever its intentions, becomes the moat. Two or three incumbents are sealed inside the castle. The competition is locked out, not by being worse, but by arriving after the drawbridge went up. And the sunk-cost hole gets filled with public money, justified by a danger the hole-diggers were the loudest to announce.

That is the thing to watch for. Not "will AI be regulated," but who writes the regulation, and who can afford it. When the leaders in an industry start begging to be regulated, they are rarely asking for a cage. They're usually asking for a wall.

It's an old script — and here's where to be careful

This pattern has a long history, and your own instinct to pattern-match it is correct. "Too big to fail" was a drawbridge: the banks that caused 2008 emerged more concentrated, because the rescue and the rules that followed entrenched exactly the giants that broke it — a theme in an earlier essay. Occupational licensing does it to hairdressers and florists. Defense contractors do it with the fear of the enemy. The mechanism is always the same: manufacture or amplify a fear, route public money and protective rules toward the incumbents, and raise the cost of entry for everyone else.

But here is where the discipline of this whole series matters, because it's easy to take this instinct one step too far. There's a difference between two claims, and collapsing them is how a sharp argument becomes a lazy one: This industry has capture dynamics — public money and protective rules flowing to a few incumbents. Often true, and worth exposing.

Therefore the product is fake, or the danger is pure invention. That does not follow, and asserting it is its own kind of fence.

Take pharmaceuticals, since the comparison tempts everyone. It is entirely fair to scrutinize how public funding, liability shields, and procurement concentrate an industry into a few dominant firms — that's real market-structure analysis. It is a completely separate question from whether a given medicine works, and the two must never be welded together. A vaccine that genuinely prevents disease and an industry with genuine capture dynamics can both be true at the same time; neither fact refutes the other. The moment you let "the industry games the system" slide into "so the product is a scam," you've stopped analyzing and started believing — exactly the move the essay on the surgeon and the quack warned against. Keep the two blades separate. The capture critique is strongest precisely when it refuses to overreach.

So is AI actually dangerous?

Which lands on the real question, and I'll be straight that it's genuinely contested reasonable, informed people land on opposite sides, and anyone who tells you it's obvious is selling something.

Your view — that AI isn't as dangerous as the loudest voices claim, and that the danger is being inflated to serve the incumbents — is a serious position, held by serious people, and the capture argument above gives it real weight. The financial incentive to exaggerate the threat is enormous and undeniable.

But the honest counter deserves its full hearing. It's possible for the danger to be both real and cynically exploited — those don't cancel. Incumbents overstating a risk doesn't prove the risk is zero; it proves they found it useful, which is a different thing. The people most worried about AI include many with no castle to defend. And a civilization that decided a technology was harmless because worrying about it was profitable to someone would be making its own kind of motivated error — the mirror image of the one it's trying to avoid. The capture can be real and the caution can be warranted. The hard part is that you have to weigh the actual danger on its own evidence, separately from the fact that some people profit from your answer either way.

Subscription is the moat's other half

There's a second drawbridge, quieter than the regulatory one, and it's the one the money is actually chasing: the subscription. Watch where every AI company is racing — not to sell you a thing once, but to enrol you in a monthly charge that never ends. That's not an accident of pricing. It is the whole point, and it's the business-model twin of the regulatory moat.

Here's why. A sale that happens once forces you to earn the next one — the customer can leave the moment something better appears, so money keeps flowing to whoever is currently best. A subscription severs that. Once you're enrolled, you keep paying whether or not the product stays the best, because leaving takes effort, and inertia is the cheapest moat there is. Subscription is captured revenue — the recurring, locked-in income that makes a sunk-cost hole survivable, and the reason investors fund the hole in the first place. The regulatory drawbridge keeps competitors out of the castle. The subscription keeps you inside it. Together they are the same instinct: stop the free market from doing the one thing a free market does, which is move money to the best option, continuously, the instant it changes.

Now put the alternative next to it, because this is where it ties back to the rest of this publication. The opposite of a subscription isn't a one-time sale. It's a stream. Money that flows by the second, for exactly the value received, and stops the instant the value does. That's what a native money layer — Bitcoin, Lightning, value-for-value — actually makes possible, and it inverts the whole game. In a streaming market, you don't enrol and forget; you pay while the thing is good and the stream dies the moment a better one appears. There's no lock-in to inherit, no inertia to hide behind. Money distribution tracks live quality — it runs toward the good product and the real invention and abandons the stale incumbent in real time, instead of being trapped in a monthly charge nobody remembers signing up for.

See what that does to the drawbridge. A moat only works if the customer is stuck behind it. Streaming money un-sticks them — it turns every moment into a fresh decision, which is exactly the thing both the subscription and the regulation were built to prevent. The incumbent's entire strategy is to make leaving hard; a stream makes leaving the default and staying the choice. That is why the free-market answer and the sound-money answer turn out to be the same answer: money that moves freely, to whatever is genuinely best, right now, is the one force a moat cannot survive.

The one line to keep

The sunk cost fallacy isn't just why you sat through a concert you didn't enjoy. Scaled up, it's why a mountain of money committed to AI with no revenue behind it cannot be allowed to stop — and why the people who committed it now need you afraid, so that a government will fill the hole and raise a drawbridge behind the two or three left standing. And it's why they need you subscribed — locked into the recurring charge that makes the hole survivable, kept inside the castle while the regulation keeps rivals out. Watch for the move. When an industry's leaders ask to be regulated, ask who can afford the regulation. When they warn you their own product could doom us all, ask what the warning builds — a cage, or a wall. The big eating the small is the oldest story in markets, and fear has always been the cheapest fork.

But keep the blade honest, because that's what makes it cut: an industry can be gaming the system and its product can be real, and a danger can be both exaggerated by incumbents and worth taking seriously. The lazy version of this argument says everything scary is a scam. The sharp version says: follow the money, name the drawbridge, and still judge the danger on its own evidence — because the people raising the wall are counting on you being too frightened, or too cynical, to do both at once.

And the structural answer — the only real one — is the theme under this whole publication. A castle has a drawbridge because a castle has a single gate. The reason decentralized systems matter is that there's no drawbridge to raise, because there was never one gate to hold. You don't beat the moat by capturing the castle. You beat it by building something with no walls to climb.