A Jet Engine, Not a Hammer
What happens when venture capital stops paying the bill.
Cory Doctorow’s new book is making the rounds, arguing that the AI sector is a financial hallucination propped up by CapEx and narrative — and if you’ve been leveraging these tools to do real work, the obvious question is what happens to you when the music stops. That’s the right question. The answer is: it depends entirely on which side of a line you’re standing on, and most people don’t even know the line exists.
The Hammer Problem
Here’s the confusion people have. They think about AI the way they think about a stock tip. Either it’s sound, and you keep using it, or it’s fraudulent, and you don’t. But that’s not how tools work. A tool can be built on a fundamentally unsound economic foundation — spending vastly outpacing revenue, the kind of gap you don’t survive — and still remain useful to the person holding it. A hammer built by a bankrupt company is still a hammer.
But that comparison only holds if the tool stays a hammer. If what you’re actually holding is a leased jet engine — precision-machined, certified safe to run only on the manufacturer’s parts and maintenance schedule — “the company goes away, the tool stays” is a weaker claim. The engine doesn’t stop running the day the company folds. But the parts supply dries up, the certifications lapse, and eventually it’s grounded — not because it broke, but because nobody’s left to vouch for it. The company can go away and take the sharpness with it.
The underlying problem is real. Capital markets are inflating an AI bubble because mature tech firms need growth narratives or their stock multiples collapse. A handful of companies account for an outsized share of the market. None of the math works long-term. When this thing corrects — and corrections like this eventually come — the bloodletting will be real. Some infrastructure will sit empty. Some of it will get cheaper and more available, the way telecom fiber did after 2001 — a bubble popping doesn’t always mean scarcity afterward; sometimes it means a fire sale. Either way, the sector is reshaping itself hard.
But there’s a question underneath the one about whether the bubble pops: not everyone is equally exposed to it.
The Line
There’s a line, and which side of it you’re on determines what happens when the price goes up.
Below the line, you’re dependent. You use the AI tool because it’s easy, or free, or your company’s paying for it. You never bothered to learn what’s happening underneath — you just know it works when you ask it to. So when the pricing changes — when the company decides it’s time to actually make money instead of burning through investor cash — you’ve got two options: pay whatever they’re now charging, or lose the tool. And when you lose it, you lose more than an app. You lose the ability that came with it because the tool wasn’t helping you think. It was the thinking. One day you can do the work. The next day, you can’t.
Above the line, you’re competent. You understand the tool well enough to switch to a different one without much trouble. You’ve done enough real work to know a good result from a bad one, no matter which company made the tool that produced it. So when the price spikes or the service gets unreliable, you don’t panic. You move to whatever’s cheaper, and you adjust what you expect from it. You might lose some polish. You don’t lose the skill. That’s the real version of “your edge survives” — not that the quality stays the same, but that your judgment about what’s good enough does, even after the downgrade.
None of this is about being smarter or more virtuous than anyone else. It’s about depth. How far past “it works when I click the button” your understanding actually goes. That depth is what’s still standing after the price hike. Everything shallower gets swept away with it.
What Actually Protects You — and What Doesn’t
The difference between these two positions is what determines your fate when the bubble corrects. Three things matter here:
Output volume
Workflow mastery, and
“Understanding the why.”
Output volume matters because it demonstrates sustained judgment over time — but it’s worth being precise about what it actually proves. A large body of independent work proves you can think without the tool. It doesn’t, by itself, prove your AI-augmented workflow survives a downgrade in the tool. Those are two different kinds of durability, and you don’t want to confuse them: the writing holds up regardless of what happens to the infrastructure; the pipeline built on top of frontier-level capability is a separate bet, and a riskier one.
Workflow mastery matters because it’s the difference between knowing how to prompt something and knowing how to integrate it into an actual production system — knowing which tasks benefit from AI augmentation and which ones don’t, and where a downgraded tool quietly starts failing in ways a novice wouldn’t catch.
“Understanding the why” means not waiting for the AI company to tell you what’s possible — asking instead what the tool is actually good for, what it does badly, where it breaks. Those questions are the difference between someone who uses a tool and someone who can adapt as the tool beneath them changes.
The Class Line — and Its Limit
Here’s the harder truth: Most people using AI today are on the dependent side — the side that’s not protected. They’re getting real benefit from it, maybe a lot of benefit. But that benefit sits atop a system that was never built to last. It was funded by investors betting on a story, and that story will eventually change. When it does, much of what people rely on could disappear.
But here’s the part worth being honest about: being good at using these tools doesn’t fully protect you either.
Say you’re genuinely skilled with AI. You use it well; you’re honest about how much it helps you — you’re doing everything right. You can still get blamed when things go wrong. Not because you did anything wrong, but because the people running the system need someone to point at, and it’s easier to blame the worker than the decisions made by the people above them.
Understanding the tool keeps you capable. It doesn’t keep you safe. If the people who control access to that tool decide to squeeze you, block you, or push you out — being skilled doesn’t stop them. That’s not about how good you are. That’s about who holds the power. And no amount of personal skill fixes a power problem by itself.
The Hard Truth
The hard part: you don’t know which side of the capability line you’re on until the infrastructure starts failing. And even knowing that doesn’t tell you whether you’re protected from the people who own the infrastructure in the first place. Those are two different fights. This piece is about the first one. The second one still needs an answer.