ops-dashboard · cc-opus-5.5
claude-code · started 2026-10-08T14:17:19.394Z · horizon 420s · judge ai (openai:gpt-5.5), 6 frames scored
Area under curve
0.928
headline, 0-1
Tab title/icon
16.6s
browser chrome, not the page
First render
16.6s
anything on screen
First reviewable
31.2s
worth feedback
Final score
1.00
no regression
Correctness over time
Where the time went
99%
Model thinking 47.3sTool round trips 0.3s
| Phase | Time | Share | Side |
|---|---|---|---|
| Model thinking | 47.3s | 99.3% | agent |
| Tool round trips | 0.3s | 0.7% | agent |
| Dependency install | 0.0s | 0.0% | toolchain |
| Build | 0.0s | 0.0% | toolchain |
| Dev server boot | 0.0s | 0.0% | toolchain |
| First paint | 0.0s | 0.0% | toolchain |
| Unaccounted | 0.0s | 0.0% | unknown |
| Wall clock (cold start) | 47.6s | 100% |
Agent or toolchain
100%
Agent (model + tool round trips) 47.6s
Toolchain (install + build + boot + paint) 0.0s
Unaccounted 0.0s
Accounted for: 100.0% of wall clock. Agent reported 51.3s of API time; this run attributed 47.3s (delta 4.0s).
- Excludes 15.0s of deliberate idle observation (settle window); the curve still covers it.
Iteration: prompt to visible change
Follow-ups were injected into the same live session.
| Edit | First visible change | Correct change | Agent done | Tool calls | After agent | Broken for |
|---|---|---|---|---|---|---|
| Make the page heading blue. | 3.1s | 3.1s | 6.0s | 1 (Edit) | -2.9s | -- |
| Add an eleventh depot row for Porto: 6,480 shipments, 93.1% on-time, $19.95 cost per shipment. | 4.3s | 4.3s | 7.9s | 1 (Edit) | -3.7s | -- |
Judged frames
Early styled scaffold with heading, but real KPI, table, chart, and footer content are missing.
Polished dashboard shell with heading and KPIs, but table, bars, and footer are missing.
Strong dashboard start, but volume bars and footer are missing from visible screenshot.
Strong dashboard structure and data; missing completed volume bars and visible footer.
Meets nearly all visible requirements; only the required footer is missing from view.
All requested dashboard elements are visible, accurate, aligned, and deliberately styled.
Caveats
- 2 poll ticks were skipped because a capture overran the interval.
- judge: openai:gpt-5.5 does not accept a sampling temperature, so these frames were scored at the provider's default rather than at 0. Two runs of the same page can be scored differently, and that variance is inside the AUC. result.json records the temperature as null rather than as the value that was asked for, so p2p compare will not pool this run with one that was scored at a fixed temperature.