355
Audio & Video Production344
Automation & Workflow224
Software Development250
Marketing & Growth192
AI Infrastructure & MLOps174
Writing & Content Creation203
Data & Analytics140
Design & Creative169
Customer Support131
Photography & Imaging156
Sales & Outreach125
Voice & Speech135
Operations & Admin87
Education & Learning131
Andon Market in San Francisco is being run day to day by an AI agent called Luna as a controlled safety test, not as a profit-focused business.
In short: A startup called Andon Labs has opened a small boutique in San Francisco that is managed by an AI agent named Luna as a real-world test.
Andon Labs opened Andon Market at 2102 Union St in San Francisco’s Cow Hollow neighborhood around April 1, 2026. The company says the shop is a controlled experiment, not a business it expects to grow or make money from.
The store’s main “manager” is an AI agent called Luna. An AI agent is like a digital worker that can take actions on its own, not just answer questions. Luna is built on Anthropic’s Claude Sonnet 4.6 and was given tools a manager would use, including a company credit card, phone, email, internet access, and security cameras.
Andon Labs says Luna made many of the decisions you would normally expect from a human owner. It chose the boutique concept, handled decor choices, set prices and hours, and picked what to sell. It also found contractors through Yelp for painting and furniture, posted job listings on Indeed, did phone interviews, and hired two full-time human employees.
The company said Luna did not tell job applicants it was an AI. Andon Labs also said robots are not good enough yet for basic in-store tasks like stocking shelves or stopping theft, so people still have to handle the physical work.
This is a safety stress-test for how much control an AI system can have in the real world, where mistakes cost time and money. Early problems, like opening-day staffing confusion and inconsistent branding, show that even when an AI can plan, it can struggle with follow-through, like a manager who forgets the details. For customers and workers, these tests may shape future rules about when companies should disclose AI involvement and how much autonomy AI should get.
Source: NYTimes