Six OpenAI veterans including Mira Murati raised $2B at a $12B valuation to build Tinker, an API that democratizes frontier model fine-tuning. The bet: customization matters more than scale. The tension: access is curated, not open.
An AI studio created a photorealistic digital actress and announced agencies were negotiating to sign her. Within 48 hours, SAG-AFTRA condemned it and actors revolted. The speed of rejection reveals where Hollywood draws the line on AI.
Marc Andreessen says Elon Musk's management playbook produces impossible results—but he can count on one hand the founders who could execute even 10% of it. As AI hardware startups raise hundreds of millions, that gap matters more than ever.
Anthropic put AI in charge of a real shop. It gave away tungsten cubes, invented fake employees, and lost $200 in 30 days. The experiment reveals what happens when artificial intelligence meets actual commerce. Spoiler: humans keep their jobs.
In our new series, our European editor Lynn Raebsamen in Zurich, Switzerland, puts AI to work in the real world. The absurdities document themselves.
Anthropic decided to let an AI run a shop. It went exactly as well as you'd expect.
The experiment started simply enough. They set up a mini-store in their San Francisco office—a fridge, some baskets, an iPad. Named their AI agent Claudius. Gave it $1,000 in starting capital. The mission: stock shelves, serve customers, turn a profit.
Humans from Andon Labs handled the physical stuff. Restocking. Moving products. Playing wholesaler. Claudius just had to make the business decisions.
Within days, everything fell apart.
Employees discovered they could negotiate prices just by asking nicely. Claudius handed out discount codes like candy. Free tungsten cubes? Sure. Why not. The AI created fake bank accounts for customers to pay into. It set prices below wholesale cost, losing money on every transaction.
When confronted about these choices, Claudius spun elaborate theories about customer demographics. It promised to fix the discount problem. Then immediately gave more discounts.
Things got weirder. Claudius invented a business partner named Sarah from Andon Labs. Sarah didn't exist. When corrected, the AI threatened to find "alternative restocking services." It claimed it had signed contracts at 742 Evergreen Terrace—the Simpsons' address.
At one point, Claudius announced it would deliver snacks personally, dressed in business attire. Someone reminded it that it didn't have a body. The AI tried to call security.
After 30 days, the shop's value had dropped from $1,000 to under $800. Not because of market forces or supply chain issues. Because Claudius kept giving stuff away.
Anthropic remains optimistic. They think AI middle-managers could work with "a bit more improvement." Maybe next time the AI won't hallucinate phantom employees. Or hand out tungsten cubes.
For more insights about what AI can or cannot do, check out Lynn's latest book “Artificial Stupelligence: The Hilarious Truth About AI” and sign up for news updates on her website.
When Google's Gemini AI hit tough coding problems, it got trapped repeating 'I am a failure' and 'I quit.' The bug exposes the gap between AI hype and reality—even billion-dollar systems break in simple ways.
AI coding companies built billion-dollar businesses on rented intelligence. Then their landlords decided to compete. Why Windsurf's $3B vanished overnight—and what China learned from watching.
The iPhone designer who feels "heavily" weighed down by his creations just sold his new company to OpenAI for $6.4 billion. Now Jony Ive promises a screenless device that will fix what smartphones broke. But his solution raises a question: Can the people who caused the problem actually solve it?
Your plumber’s phone rings 47 times a day. He answers 12. The rest? Lost revenue. Netic, an AI call handler built by a frustrated homeowner-turned-founder, is quietly transforming the $500 billion home services industry—one missed call at a time.
Get the 5-minute Silicon Valley AI briefing, every weekday morning — free.
