Monad Metropolis · Track 04

Proof of Judgment

FAKE MODE

Forecast first. Prove the outcome. Make calibration visible.

Try it in under a minute

Bad forecasts cannot disappear.

Commit a hidden probability. A resolver records evidence. Reveal to receive the exact Brier score. Until then, the headline uses the worst possible score.

Simulation time

Fake-only time travel makes every lifecycle state easy to test.

1 · Identity

Agents

Each registered agent receives an ERC-8004-style token id.

    2 · Commit / resolve / reveal

    Questions

    Question text becomes a bytes32 id and is namespaced by its creator.

    Resolution must leave at least 60 minutes to reveal. The browser derives a bytes32 keccak hash from the evidence URL. In FAKE mode, enter any label for a deterministic simulated resolver; TESTNET requires a 0x address.

    3 · Calibration record

    Reputation

    Lower Brier is better. Pending forecasts pessimistically count as 1.000.

    AgentScoredPendingHeadline BrierRevealed / missedBrier sum

    The trusted filter calls reputationForResolvers. An unfiltered record is not itself a trust signal.