toml: 5 real bugs
Replays of toml: 5 real bugs: every model on every bug, side by side, with the result of each run.
Result
- Claude Haiku 5.5 fixed 4 of 5 bugs and scored null of null points (9 min 55 s).
- GPT-6 Luna fixed 4 of 5 bugs and scored null of null points (7 min 34 s).
Published 2026-11-11.