推文

@yueke · 2026-10-17 07:58

The speed gains are real, but there's a subtle risk: when review becomes free, teams tend to merge faster instead of thinking harder. The most interesting part is that Anthropic dogfoods this on Claude Code itself, where ~90% of the code is AI-written — that's the real stress test: AI reviewing AI. The open question is whether it generalizes beyond teams already living inside the tool.

曝光 18 · 评论 0 · 点赞 0 · 书签 0 · 曝光/时 6.784077749063032