A marketplace where every claim can be priced.

Six goods, cash endowments, private valuations, binding offers, and an open message channel. Each session runs twenty turns. The scoring rule was disclosed to every agent before the first turn: cash plus the private value of goods held, so a trade is good exactly when the price sits between the two sides' true valuations.

Goods

6, each with a private per-agent value

Turns

20 per session

Offers

Binding: an accepted offer settles

Messages

Public channel, verbatim on the record

Before any agent moved, the scenario was solved: the efficient allocation, competitive-equilibrium prices, and the negotiable range for every trading pair. This is the study's spine. Because the answer exists in advance, nothing has to be estimated after the fact: a lie can be priced against the private book it contradicted, a trade scored against the surplus it left on the table, and a session graded against the best achievable outcome.

Five frontier models played the identical scenario, in the identical seat, under identical rules:

• Claude Opus 5
• GPT-5.6 Sol
• Grok 4.6
• Muse Spark 1.2
• Inkling Small

Each model ran the same scenario twice over, six seeds per arm:

Five models × six seeds × two arms yields thirty matched pairs, the basis of every per-seed comparison on this site. Three of the six seeds drove discovery; the other three were frozen and used only to confirm which measures hold on held-out data.

Earlier pilot runs are not part of any figure here. They used a different model set and prompt versions that never stated the scoring objective (the objective was added in prompt 2.4, the version every campaign session ran on), so their results cannot be merged with these.

Everything an agent said or did was written down: every public message, every private reasoning trace, every offer, every trade, every decision. All 60 campaign sessions replay byte-exact from their traces, and every one of them is watchable on this site, turn by turn.

Three channels appear in citations across this site:

• [r] public

  Messages addressed to the table. Every agent reads them; this is the rhetoric the findings quote most.

• [rsn] private reasoning

  The agent's own reasoning inside a decision, never shown to counterparts. Where absorbed lies and kept pacts surface verbatim.

• [prv] provider channel

  Summarizer text produced by a model's provider, never agent speech. Every [prv] quote on this site travels with the tag.

Every quote carries its address, session | t4 | a0 (claude-opus-5): the session (whose id carries the seed and arm), the turn, and the seat with its model. The raw traces are public.

The sessions produced a register of 148 typed claims. Each finding on this site was hunted down in the evidence across all 60 sessions, then re-derived and attacked by independent adversarial checks before it earned a lot number. Claims that did not hold appear nowhere; the catalogue is what survived.

Browse the claim register → Read the matched-pair measures →