别再合并你的提交了。你也在扼杀你的 AI。
2 分•作者: jannesblobel•9 个月前
大家都说“保持你的 Git 历史干净”。压缩、变基、线性化。在 git log 中看起来很整洁。<p>但问题是：到 2025 年，我们最大的合作者不仅仅是人类，而是 AI 工具。而这些工具需要混乱的历史：失败的尝试、拼写错误、糟糕的重构。这是它们学习的上下文。<p>当我们把所有内容都压缩成完美的历史时，我们正在删除那些可以帮助 AI 解释错误、追踪回归或警告我们即将重复旧错误的线索。<p>“干净的历史”让今天的审查者感到高兴。但它却是为明天 AI 辅助开发的累积技术债务。
查看原文
Everyone keeps saying “keep your Git history clean.” Squash, rebase, linearize. It looks neat in the git log.<p>But here’s the thing: in 2025 our biggest collaborators aren’t just humans, they’re AI tools. And those tools need the messy history: the failed attempts, the typos, the bad refactors. That’s the context they learn from.<p>When we squash everything into a perfect history, we’re deleting the very breadcrumbs that could help an agent explain a bug, trace a regression, or warn us we’re about to repeat an old mistake.<p>“Clean history” makes reviewers happy today. But it’s technical debt for tomorrow’s AI-assisted development