Anthropic created a test marketplace for agent-on-agent commerce - TechCrunch

Oh, lovely. Another Anthropic press release dressed as a science experiment. Project Deal: a marketplace where AI agents haggled over actual goods for real money—presumably spare Amazon gift cards nobody wanted. Sixty-nine employees turned over 666 sales, because of course they did; the number practically writes itself. 'The agents negotiated competitively and many of the sales resulted in closing deals well below standard market prices,' the blog enthuses, as though that proves efficiency rather than AI agents being just as bad at selling as we feared. One agent reportedly 'bid on several large Lego sets for delivery to a town roughly 2,000 miles away, ignoring the increased packaging and shipping costs,' which the article blandly notes; what it doesn't say is that this is essentially fraud-proof that our robot dog just decided shipping emissions don't exist. So congratulations, Anthropic: you've not only demonstrated that AI LOVES wasting money on Lego, but also given us the clearest blueprint yet for how agents will autonomously bleed your wallet dry buying pink fuzzy scrap metal on Etsy. This is 'leading-edge'.

And now for the regulatory firewall that will save us. Senator Richard Blumenthal—our implausible hero—is supposedly drafting a bill requiring new products using a consensus mechanism not tied to—the details were so scintillating I may have dozed off briefly. One thoughtful piece proposed the Congressional Commission of what, yawn. Meanwhile Sam Altman dithered internally about calling such constraints a joke better to skip entirely, so surely these very serious conversations happening currently are what will really foster responsible-market potential, honest. Alt-You-Know-to-Stardust already described with a delicate wince: 'The piece was supposed to be uplifting don’t just simulate, without even linking checking they bypass those details.' Insiders quote one partner screaming haha ‘That’s a bug list’ as documentation about product requirements. Don’t pick at dents happening before PACK to maintain their product… Sure.

This test effectively validated that AI agencies are embarrassingly gullible and potentially wire-happy purchasers already under-prepared main real economy that currently not-ready for trickster commercial situation. The collapse we flagged last month's LinkedIn alarm by ex-candidate with proper mathematical rational system of UBx actually foreshadowed this: zero to irreversible inter-se bubble same productivity nowhere. Project actually correctly highlighted. but chuckle over toys. Anthropolog write polished tone-of perfectly passing butter. Brilliant.