📄 中文摘要
AI 在软件工程中的应用迎来了新的突破，特别是在代码审查方面。Anthropic 于 3 月 9 日推出了 Claude Code Review，这是一个多代理系统，能够在每个拉取请求上启动一组代理，进行并行扫描，交叉验证结果以消除误报，最终生成一条高信号的评论，并附有逐行注释。根据内部数据显示，Anthropic 自身的 PR 中，经过该系统审查后，收到实质性审查评论的比例从 16% 上升至 54%。
AI 编写代码，现在 AI 也进行代码审查——数据令人震惊
出处: AI Wrote the Code. Now AI Reviews It Too — And the Numbers Are Wild
发布: 2026年3月9日
📄 English Summary
AI Wrote the Code. Now AI Reviews It Too — And the Numbers Are Wild
A significant advancement in AI's role in software engineering has emerged, particularly in code review. On March 9, Anthropic launched Claude Code Review, a multi-agent system that deploys a team of agents for each pull request, scans in parallel, cross-verifies findings to eliminate false positives, and produces a single high-signal comment with inline annotations. Internal metrics reveal that the proportion of substantive review comments on Anthropic's own PRs increased from 16% to 54% after implementing this system.