The report also highlights bug severity, with Claude Sonnet 4, the highest-scoring model on functional benchmarks, producing nearly double the proportion of BLOCKER bugs compared to its ...

Anthropic's Claude Code tool contained buggy auto-update commands that broke some systems, according to reports.

If something fails, Codex reports the problem and doesn’t hide errors. meet codex: openai’s new ai coding agent that writes code, handles multiple tasks and fixes bugs ...