Claude Opus 5 spots code issues earlier versions and Codex kept missing
PawelHuryn · x · 2026-07-25
The thread says the issues found are not random bugs, but ones previous Claude Opus/Codex sessions had also missed, plus synthetic analogs.
The commenter says this is a strange model: it was the first in the Claude family that did not understand the instruction “analyze this solution” and instead started inspecting uncommitted changes. At the same time, it did actions that initially seemed odd, like analyzing git history, but those steps led to unexpected insights.
The author also claims that in dozens of sessions with Fable 5 and Opus 4.8, essential issues were missed that Opus 5 identified.
Related event: GPT-5.6 Beats Claude Opus 5 in Bug Hunt Benchmark(3 posts)→
More from coding & agent
- AIE Code Summit 2026 opens applications: 800 seats, keynotes streamed free — trq212 · 2026-10-27
- Decade-long cloud architect: enterprise AI architecture is far more than bolting on an LLM — kashifmanzoor · 2026-10-27
- Analysis of thousands of coding sessions: 40-89% of files agents read are never used — rseroter · 2026-10-27
- TRL adds Score Centering, a simple fix for RLHF training instability — lvwerra · 2026-10-27
- Precompute golden answers and match at runtime for exact wording control — colinmcnamara · 2026-10-27
- Dev argues there's still alpha in working with your coding agent interactively — alex_frantic · 2026-10-27