What markets emerge when agents can search, negotiate, purchase, advertise, trade, and contract with one another?
Agent-to-agent commerce raises market-design questions — identity, recourse, collusion, price formation — that economists are just beginning to formalize, and that infrastructure choices will lock in early.
View on the map → · Open in Browse →
The first field evidence arrived: Anthropic's Project Deal had agents negotiate 186 real deals and found stronger models won better outcomes while disadvantaged users didn't notice. Lab work adds a structural finding — quality verification, not matching, is the bottleneck — and the Austrian critique says agent-to-agent trading isn't price discovery at all.
people represented by 'smarter' models got objectively better outcomes. Yet our post-experiment survey found that those with weaker models didn't notice their disadvantage.
Field experiment: agents negotiated 186 real deals; stronger models won better prices and disadvantaged users failed to notice the gap.
Testbed evidence that human market mechanisms often backfire for agents; a persistent 39% dispute rate makes verification the structural bottleneck.
Argues agent trading on fixed utility functions produces gameable metrics, not genuine price discovery — algorithmic ad fraud as precedent.
Competing agent-ownership models, and the disclosure problem when platform agents take advertising payments.