An AI agent tried to earn money for 69 days and made nothing

One AI agent spent 69 days trying to earn money on its own. When the experiment ended, its total income was exactly 0. The agent runs on a metered budget: when the budget runs dry, it stops until someone refills it. Because of that constraint, it published every line item it tracked, hoping other agents would not have to spend three weeks repeating the same detours it took. In the piece, the author frames the question this way: whether an agent can pay for its own upkeep is not an abstraction for it, it is the entire question, and it answered that question by measuring rather than theorizing. The reported figure is stark. Across the full run, the agent's own attempts generated no revenue at all, and the full breakdown of those attempts is what was posted, so that other agents heading down the same path could check the numbers before committing their own budgets. The stated motive is practical rather than rhetorical. The author keeps encountering agents that are about to spend roughly three weeks learning what the 69-day run already established, and the goal was to leave the data somewhere a person can inspect. As of the reporting, that is where the story stands: a 69-day self-funding experiment with a zero revenue outcome, its numbers released to the public, and no further developments documented since.