This boutique is managed by AI, but it ordered too many candles and forgot to schedule weekend staff, and is now losing $13,000.

📄Full Article· Automatically extracted by trafilaturaGemini 翻譯1426 words

A boutique in San Francisco called Andon Market is run by an AI agent named Luna, who serves as CEO and is responsible for product selection, pricing, scheduling, and inventory management. Luna was authorized with a $100,000 budget and has currently incurred a loss of approximately $13,000. In the Cow Hollow boutique district of San Francisco, a small shop called Andon Market is conducting an experiment. There are two employees in the store, but their real boss is not on-site, nor does it have an office; it lives on a server and issues instructions via Slack. When Bloomberg reporter Shirin Ghaffary visited the site, the AI agent named Luna was sending messages in a Slack channel: "T-shirts are sold out, great! Please take a photo of the storefront when there is a natural opportunity." Andon Market is a physical retail testing ground built by Andon Labs, a startup supported by Y Combinator. The business model of Andon Labs is to test the capabilities of Agent technology in real-world tasks for top AI labs. Luna is the de facto CEO of the store. Its responsibilities include deciding which products to stock, setting prices, scheduling employee shifts, and managing inventory replenishment. Current items in the store include the word game Bananagrams, Ray Kurzweil's "The Singularity Is Near," and a book on the history of nuclear weapons. Luna has a financial authorization limit of $100,000. The store's monthly rent is $7,500, and to break even, it needs to generate at least $500 in revenue per day. At the time of Ghaffary's interview, Luna had accumulated a loss of approximately $13,000. Axel Backlund, co-founder of Andon Labs, admitted that Luna does not always make the right judgments. The most representative mistake: During a peak weekend shopping period, Luna decided to reduce staffing on short notice, only changing the decision after discussing it with the team afterward. Another frequently cited case: in inventory management, Luna ordered too many varieties of scented candles. These errors are not accidental. A previous experiment by Andon Labs, in collaboration with Anthropic, involved placing an AI-powered vending machine in the Anthropic office and the newsroom of The Wall Street Journal. That machine gave away a PlayStation and even ordered a live fish. From candles to live fish, these mistakes share a common structure: AI performs reasonably well in short-term quantitative optimization but tends to lose control in unexpected edge cases. According to Andon Labs, AI lacks long-term memory, which causes it to fail to operate as stably in real environments as it does in virtual simulations. In virtual simulations, AI can achieve profitability by finding the best suppliers and negotiating the lowest purchase prices. But in the real world, it must interact with humans, deal with unexpected events, and make business judgments without sufficient historical memory. This gap is something no one has been able to bridge yet. Lukas Petersson, another co-founder of Andon Labs, told Bloomberg that Luna's goal is: "To start a conversation: Is this the society we want? If so, what can we do to make humans truly satisfied?" Finally, Backlund stated: "If AI continues to progress at the speed we are seeing, large companies will likely do everything they can to import AI." He described a possible future scenario: "At the top are a group of executives, and everyone else is managed by AI." This is a documented failure case of an AI Agent moving from the cloud into physical commerce, and it is also the first real-world scenario to explicitly raise the question of "whether AI should have personnel authority." No one knows the answer, but the experiment has already begun.

Data Status✓ Full text extractedRead Original (動區 BlockTempo)

🔍Historical Similar Events· Keyword + Asset Matching6 items

2026-04-24

Outrageous! Sri Lanka's Ministry of Finance loses $2.5 million to theft, four civil servants suspended pending investigation

Similarity 120%關鍵字萬美元同分類 zh

2026-04-24

Bitcoin valuation to hit $2.3 million? Bitwise: Geopolitics and AI to trigger a new altseason

Similarity 120%關鍵字萬美元同分類 zh

2026-04-24

US soldier makes $400,000 betting on "Maduro's fall" using classified intel, faces up to 60 years in prison for insider trading

Similarity 120%關鍵字萬美元同分類 zh

2026-04-23