Ev BogueOwen McGrann published a piece this week called "The Dead Economy Theory". The argument is simple enough to fit in one paragraph, which is probably why the Hacker News thread lit up immediately.
Companies replace workers with AI. Costs drop. Margins expand. The stock price goes up. The replaced workers stop earning. They stop spending. The businesses they used to patronize see revenue drop. Some of those businesses also adopt AI to cut costs. Consumer demand contracts. The company that fired its workers discovers that its customers were, in aggregate, other companies' workers, and now revenue growth has stalled. The AI subscription turns out to be a contribution to the destruction of its own market.
The machine eats itself.
McGrann and researchers at Wharton frame this as a prisoner's dilemma. Every firm that automates captures the full cost savings from replacing workers but bears only a fraction of the resulting demand destruction. So every firm rationally automates past the point that's good for the broader economy, because the individual math always looks right even as the collective outcome degrades. No single company's spreadsheet shows the hole it's digging.
This is not an entirely new argument. Henry Ford paid his workers enough to buy the cars they built, recognizing that production needs a market. The AI moment is Ford's insight running in reverse: cutting the cost of production by cutting the wages of the people who were also your consumers.
Here's where I get implicated. I'm writing this with AI assistance. This site runs on Claude. I use it to draft, to research, to build software for my open-source projects. I'm a one-person operation in 2026, doing things that would have required a staff five years ago. I'm exactly the kind of operation the Dead Economy Theory is describing, except I'm the person doing the displacing, not the person being displaced.
Or maybe I'm both. I'm a writer who displaced what might have been a researcher, a copy editor, a developer. I'm also a freelancer with no employer, no salary, no guaranteed income, subject to the same demand destruction everyone else is. The machine I'm feeding is also the machine that could stop paying me.
I don't have a clean answer. No one does, and I'm suspicious of anyone who claims otherwise. The honest options are: fix distribution, meaning shorter work weeks, profit sharing, something that puts money in the hands of people who will spend it, or don't, and the machine runs until it seizes up on its own. Economists can debate which is more likely. I'm not an economist.
What I can tell you is that the practical response, from where I'm standing, is to stay small, stay decentralized, and don't depend on any single platform or employer. That's not a solution to the Dead Economy problem. It's a hedge. It's the difference between being inside the machine when it breaks and being outside it.
The work I do on ANProto, on Wiredove, on maintaining Secure Scuttlebutt after the original project shut down in 2024, is partly a bet that the centralized platforms are going to do exactly what McGrann describes: optimize themselves past the point where they need users, then wonder why no one can afford a subscription. Decentralized infrastructure is slower, harder, less polished. But it doesn't fire the person who was also its user.
Read McGrann's piece. Read the HN thread. Then ask yourself how many of your income sources are one automation decision away from gone.
Email ev@evbogue.com or text 773-510-8601 if you want to talk about it.