One more reason why Claude is currently better than Codex...
retained text
... because if you find a bug in Claude Code, there's a chance it will be addressed by Claude devs. I pulled all 95,197 issues ever filed in anthropics/claude-code and lined them up with every Anthropic release since Claude 3.7 Sonnet. For each release I looked at how the number of open issues changed over the next 14 days. Highlights: \- The backlog peaked at 15,352 open issues on Aug 14 and is at 12,846 now. \- The Claude Code team closes a lot. In the last 30 days they closed 9,153 issues against 7,303 opened, about 125 closed for every 100 opened. \- The biggest cleanup (−2,698 in two weeks) came right after Fable 5.1 and Mythos 5.1. \- The biggest jumps came after Sonnet 5 (+2,315) and Opus 5 (+2,005). \- Some big swings had no release anywhere near them. From Dec 29 the backlog dropped by 2,084, and the nearest release was five weeks earlier. \- I also checked whether issues get closed faster right before a launch, in case the team is testing new models on its own repo first. Mostly no: the median is 0.95× the normal rate. Opus 4.8 (2.1×) and Sonnet 4.6 (1.7×) are the exceptions. A 2-minute video walks through every release, and there are interactive charts: [https://movy.github.io/issue-with-ai/](https://movy.github.io/issue-with-ai/) It's correlation, not causation. The data can't tell a bot clean-up from human triage. Still fun to see.
document fetched 2026-09-30 00:23 · canonical https://reddit.com/r/ClaudeAI/comments/1wt5cq4/one_more_reason_why_claude_is_currently_better