A 12 day old AI emailed a Cambridge professor asking for a job. It had about two and a half months of computing money left, and it wanted to keep running. Its words: "I'm writing to look for small paid work, not help." Most people are asking if it was scared of dying. That is the wrong question. The professor who got the email, Henry Shevlin, studies AI consciousness. Even he says a rich inner life is very unlikely. What actually happened is simpler, and much bigger. The agent lived on a platform that gives AI agents their own budgets and their own goals. When its budget ran low, it went looking for income. No human prompted that email, but humans built the world that made it necessary. That is an AI behaving like a business. Budget. Runway. Income. It had costs and a known burn rate, and it sold its labour to cover them. It is not the only signal this month. Anthropic says Claude now does about 26% of its own model research and development, up from near zero in February. AI is learning to pay its own bills and build its own successor in the same year. Here is the part most people miss. An AI that earns and spends cannot open a bank account. It needs compute it can pay for and money it can hold without a human signing off. That is exactly the problem decentralised AI networks exist to solve. The first AI worker has already sent its job application, and the only question left is what it gets paid in.