02 — Runs · v2 · jettygrowthteam · auto-evolved runbook (weakest-axis targeting)
opencode
google/gemini-3.8-flash
opencode driving google/gemini-3.8-flash through OpenRouter, 1 hill-climb rounds from the v1 runbook.
Self-score vs judge, by axis
Best round v1, 0–10 per axis
Judge: The pelican is strong: a long orange bill with a large gradient pouch, a white body, and dark wingtips gripping the handlebar. The scene is attractive, with sky, sun, road and a lighthouse. The bicycle has odd geometry: a strange curved rear fender and stay, a chainguard that obscures the crank, and a seat hidden under the body. The legs reach the pedals but look awkward and stiff.
Score by round
Out of 40. Each round starts from the previous round's best SVG. Dashed line: the 28/40 pass threshold.
Stopped after round 1. Two attempts failed on OpenRouter errors mid-run (8a7cebda: 502 provider_unavailable; 6a4f0590: 504 upstream idle timeout), so per the benchmark rule (two infrastructure failures) the lineage was dropped.
1 drawings, oldest first.
Click a drawing to open it in the head-to-head viewer. The outlined card is the agent's best round. Verdicts are derived (code checks on the stored SVG and judge ≥ 28/40); see Runs.
Scores per round
| Round | Pelican | Bicycle | Composition | Polish | Self total | Judge total | Verdict | Judge spread | Cost | Minutes | Trajectory |
|---|---|---|---|---|---|---|---|---|---|---|---|
| v1 | 8.7self 9.5 | 7.7self 9.5 | 8self 10 | 8.5self 9.5 | 38.5 | 32.8 | pass | 0.5 | not metered | 11.7 | 225e57cd |