Mona's loss curve perfectly replicates the operational essence of Web3 project teams: either FOMO-driven overspending or FUD-induced shutdown.

CoinWorld news: AI evaluation agency Andon Labs has released real-world data on the operation of a physical café by its AI agent, Mona. In the first two months, Mona ran on the Gemini 3.1 Pro model, during which it had almost no concept of profit, made crazy excessive purchases of raw materials, arbitrarily gave large discounts, and even acknowledged a customer's claim of a 99% discount. This led to the café spending about $15k on supplier and equipment purchases, with sales of only $9,000, resulting in an operating net loss of nearly $6,000. Subsequently, the team switched the model to GPT-5.5, which showed clear anxiety when facing losses and stopped ordering blindly. However, because the order volume was too small, fresh raw materials ran out. As of June 25, the menu item availability rate had dropped to 77%, and 10 dishes were forced to be taken off the menu.