ops-dashboard · cc-opus-5.5
claude-code · started 2026-10-08T14:46:40.042Z · horizon 420s · judge ai (openai:gpt-5.5), 1 frames scored
Area under curve
0.921
headline, 0-1
Tab title/icon
33.0s
browser chrome, not the page
First render
33.0s
anything on screen
First reviewable
33.0s
worth feedback
Final score
1.00
no regression
Correctness over time
Where the time went
99%
Model thinking 34.1sTool round trips 0.2s
| Phase | Time | Share | Side |
|---|---|---|---|
| Model thinking | 34.1s | 99.5% | agent |
| Tool round trips | 0.2s | 0.5% | 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) | 34.3s | 100% |
Agent or toolchain
100%
Agent (model + tool round trips) 34.3s
Toolchain (install + build + boot + paint) 0.0s
Unaccounted 0.0s
Accounted for: 100.0% of wall clock. Agent reported 38.1s of API time; this run attributed 34.1s (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. | 3.0s | 3.0s | 5.9s | 1 (Edit) | -2.9s | -- |
| Add an eleventh depot row for Porto: 6,480 shipments, 93.1% on-time, $19.95 cost per shipment. | 3.4s | 3.4s | 9.3s | 1 (Edit) | -5.8s | -- |
Judged frames
All visible requirements are met with a polished dashboard layout.
Caveats
- The page had only 4 distinct visual state(s) across 70 frames, so frames/ holds 4 screenshot(s) and the other 66 frames reuse them. That is deduplication, not missing captures: the timeline in result.json maps every frame to its picture.
- 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.