提问 HN：有没有 GitHub 仓库能展示 AI 辅助开发？
1 分•作者: rhelz•4 天前
团队如何使用 AI 增强流程来开发/维护/扩展一个非平凡的代码库？
目前我所了解到的最好的情况是，GitHub 仓库包含一个关于代码应该做什么的“精选知识库”。
但我不清楚如何确切地保持这些提示与代码库的同步。
它们是否类似于 yacc 源代码文件，每次我们对仓库进行干净构建时，生成的代码都会被清除，然后 AI 使用它们重新生成代码，这次是根据，例如，添加的几个新功能？
或者我们是否定期让 AI 读取提示和代码，并建议从一个到另一个的编辑？
这里有哪些最佳实践？是否有 GitHub 仓库展示了如何以世界一流的方式完成这项工作？
查看原文
How are teams using AI-enhanced flows for developing/maintaining/extending a non-trivial code base?<p>Best I can get is that the GitHub repo contains a "curated knowledge base" on what the code should do.<p>But it is unclear to me exactly how to keep these prompts in sync with the code base.<p>Are they like yacc source files, and every time we do a clean build of the repo, the generated code is blown away and AI uses them to re-generate the code again, this time according to, say, a few new features added?<p>Or do we periodically have AI read both the prompts and the code, and suggests edits from one to the other?<p>What are the best practices here? Are there GitHub repos which illustrate how its done in a world-class way?