Skip to content

Demo

Watch the agent work, step by step

Replay costs nothing and needs no account: these are real recorded episodes, scored against ground truth. To run a live model, bring your own provider key, which is used for that request and never stored.

45 recorded episodes · bench v1.0

Measured 2026-08-20 on bench v1 (8 perturbations, 9-type flag vocabulary) with claude-sonnet-4-6. Case inputs rebuild byte-identically under current code; aggregate metrics from this archive must NOT be pooled with bench v2 results, whose prompt vocabulary differs.