02 — Runs · v1 · jettyio · auto-evolved runbook (weakest-axis targeting)
hermes
openrouter/google/gemini-3.5-flash
Same hermes agent, Sonnet swapped for Flash: +5 self-score points from the model alone.
Self-score vs judge, by axis
Best round v9, 0–10 per axis
Judge: The bicycle is well formed, with spoked wheels, chain, crank, fenders and a fish crate, set against an attractive seaside backdrop. The pelican's anatomy is confused: the huge orange pouch reads as a detached sheet, the head is tiny and tilted back, the body is a jumble of white blobs with an odd grey tail, and the feet only loosely meet the pedals.
Score by round
Out of 40. Each round starts from the previous round's best SVG. Dashed line: the 28/40 pass threshold.
10 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 | 7self 10 | 7.5self 9.5 | 6.8self 10 | 7.2self 9.5 | 39 | 28.5 | pass | 1 | – | – | c118fb80 |
| v2 | 7self 9 | 7.8self 10 | 6.3self 9 | 7self 10 | 38 | 28.2 | pass | 1 | – | – | 2128ffc1 |
| v3 | 6.3self 9.5 | 7.3self 9 | 6.3self 9.5 | 6.5self 9.5 | 37 | 26.5 | fail | 0 | – | – | 2405db86 |
| v4 | 7self 10 | 7.7self 10 | 6.8self 9.5 | 7.2self 9.5 | 39 | 28.7 | pass | 1.5 | – | – | 822a3bca |
| v5 | 7self 9.5 | 8.3self 9.5 | 7.2self 10 | 7.5self 9.8 | 38.8 | 30 | pass | 0 | – | – | 60502506 |
| v6 | 6.3self 9.5 | 7.5self 9.5 | 6.3self 9.5 | 6.7self 9.5 | 38 | 26.8 | fail | 0.5 | – | – | 2c2a74e8 |
| v7 | 5.8self 9.8 | 7.8self 9.5 | 6.7self 9.5 | 6.7self 9.7 | 38.5 | 27 | fail | 2.5 | – | – | c165de92 |
| v8 | 5.5self 9 | 8self 10 | 6.2self 9 | 6.7self 10 | 38 | 26.3 | fail | 1 | – | – | d391e018 |
| v9 | 6self 9.5 | 8self 10 | 6.8self 10 | 7.3self 10 | 39.5 | 28.2 | pass | 2 | – | – | e6d4548b |
| v10 | 5.3self 9.8 | 8self 9.5 | 6.3self 9.5 | 6.8self 9.7 | 38.5 | 26.5 | fail | 1 | – | – | 3e21bf05 |