Runs

Every run is stamped with the git SHA of the harness, the full config snapshot, and a content hash of the challenge set. Scores are only comparable between runs with the same challenge-set hash.

Run	Started (UTC)	Git	Challenge set	Graded	Failures	Note
run 2	2026-07-17 15:31	c4dd881	96e86e1711a49f7d	29	8	run 2: 4 local models, 9 challenges incl. 4 harder
run 1	2026-07-17 14:55	de38c28	43d1f79070dee5ce	8	2	first live run: qwen2.5-coder:3b, 5 simple challenges