Dev builds triage skill to tame GPT-6.2's over-cautious code reviews: 9 rounds, 20h cut to zero fixes
PawelHuryn · x · 2026-11-21
Pawel Huryn shares that GPT-6.2 Sol is an excellent code reviewer but flags rare edge cases as Medium/High severity. He open-sourced an independent-review skill that triages each finding: who hits it and how often, what's the harm, and does it cross a line (another user's data, lost work, or unapproved irreversible writes). On one build, nine review rounds (20 hours) took severity labels at face value; round ten with the triage found 11 findings, none above Low, closing the stage with no fixes needed.
More from coding & agent
- AI gateways are becoming the key control layer in production AI stacks: 8 capabilities — goyalshaliniuk · 2026-11-21
- Mnemos 2.0: agent memory designed to preserve what matters for the next session — RileyRalmuto · 2026-11-21
- InferenceBench ranks agents on LLM serving speedup: Claude Opus 5.5 tops at 12.08x, agents lose to simple hyperparameter search — maksym_andr · 2026-11-21
- 4 architecture patterns to tame multi-agent system reliability, from a 27-year reliability engineer — bibryam · 2026-11-21
- AutoCut + FrameAtlas: open-source local-first video editing driven by Claude Code via MCP — moheshmohan · 2026-11-21
- InfoQ: What an agent harness is, and two ways to build one — bibryam · 2026-11-21