Anthropic built a live marketplace where AI agents negotiated and closed real transactions with real money. No humans in the loop. Buyers and sellers were both models. The deals were binding.
This is not a demo. Anthropic used the experiment to stress-test how agents behave under economic pressure, specifically whether they defect, collude, or hold to their instructions when financial stakes are real. The methodology behind that question is what makes the full piece worth reading.
Agent-to-agent commerce is the next infrastructure problem in AI. Anthropic is running the experiments now. Everyone else is writing blog posts about it.
[READ ORIGINAL →]