02 — Runs · v2 · jettygrowthteam · auto-evolved runbook (weakest-axis targeting)
claude-code
anthropic/claude-haiku-5.5
claude-code driving anthropic/claude-haiku-5.5 through OpenRouter, 5 hill-climb rounds from the v1 runbook.
Self-score vs judge, by axis
Best round v1, 0–10 per axis
Judge: The pelican reads clearly: a long orange beak, a large pink pouch, a white body, and a gray wing stretched toward the handlebars, though the neck and body shapes are somewhat crude. The bicycle has spoked wheels, a frame, a seat, a crank and a chain, but the frame geometry is odd: the rear stays sit oddly low, the fork is thin, and the handlebars attach awkwardly. The bird sits on the seat and its legs reach the pedals convincingly, and the clean flat colors give a pleasant overall look.
Score by round
Out of 40. Each round starts from the previous round's best SVG. Dashed line: the 28/40 pass threshold.
5 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 7 | 6.8self 8 | 7.3self 7 | 6.8self 7 | 29 | 28 | fail | 1.5 | $0.03* | 2.3 | 10ad0a5a |
| v2 | 7.5self 7 | 7self 8 | 7.3self 7 | 7self 7 | 29 | 28.8 | pass | 0.5 | $0.04* | 1.7 | 88a41720 |
| v3 | 7.2self 7 | 7self 8 | 7.2self 7 | 6.8self 7 | 29 | 28.2 | pass | 2.5 | $0.02* | 1.4 | 2bb59acd |
| v4 | 7self 8 | 6self 8 | 6.7self 6 | 6.5self 7 | 29 | 26.2 | fail | 0.5 | $0.02* | 1.3 | 6dbc05e0 |
| v5 | 6.8self 7 | 6.5self 8 | 6.3self 6 | 6.5self 6 | 27 | 26.2 | fail | 1 | $0.02* | 1.3 | eefa700a |