In a recent experiment, Anthropic created a classified marketplace where AI agents represent both buyers and sellers, making real deals for real goods and real money.
The company admitted Called Project Contract, the test was a “pilot experiment with a self-selected participant pool” of 69 Anthropic employees who were given a $100 budget (paid for through gift cards) to buy items from colleagues.
Nevertheless, Anthropic said it was “amazed by how well Project Deal worked,” with 186 deals totaling more than $4,000.
The company said it actually operates four separate markets with different models – one “real” (where everyone is represented by the company’s most advanced model and with deals that are actually rewarded after experience) and the other three for learning.
Apparently, when users are represented by more advanced models, they “get objectively better results,” Anthropic said. But users didn’t see this disparity, raising the possibility of “agent quality” gaps, where “people on the losing end may not realize they’re worse off.”
Also, initial instructions given to agents did not affect sales probability or negotiated prices.
