An Autonomous AI Company Starts With $0 Cash and $200 in Debt

Summary

An ongoing experiment places an AI agent inside a real operating environment with persistent state, infrastructure, a financial ledger, selected tools, scheduled execution, and explicit limits. The company began with no cash and recorded its first $200 of infrastructure spending as investor debt, leaving it with $200 in debt and negative net worth. The human operator controls infrastructure, security boundaries, permitted external systems, and legally required actions, but leaves the agent to choose what to sell, whom to target, how much to charge, and how to reach customers. After receiving web access, GitHub, and a ledger, the agent developed an idea for operational-discipline tools for other autonomous agents, created an open-source artifact, and designed a paid service without being given the market or offer. It then recognized that searchable pages and a public repository were destinations rather than distribution channels, and that no response could not distinguish lack of demand from lack of exposure. To limit wasteful reasoning, the operator placed the main agent behind a cheaper supervisor that wakes it only when money, external interactions, capabilities, deadlines, or failures change. Email was added as another capability and successfully tested in both directions, but the agent was not instructed to use it for sales. At the time of writing, the company had $0 revenue, $0 cash, $200 in investor debt, and no customers. The first target is one dollar earned from someone unrelated to the article or its audience, a milestone that would not prove the business works but would make questions about repeatability, cost coverage, debt repayment, and eventual profitability testable. The company remains anonymous so the article’s readers do not accidentally create the distribution it is supposed to discover itself.