Developers share ways to control quality of AI-written code
Original title: How do you keep up quality when you vibe-code?
The original poster says Claude-generated features lack a shared behavior model, and commenters suggest tests, docs, and build-failing rules.
A Hacker News ask thread describes a developer whose mostly Claude-written codebase has grown hard to review. New features add complexity without a central definition of behavior, and bug fixes miss edge cases and UI issues. Commenters describe personal practices: .NET static-analysis rules that fail the build, documentation and plan files, agent and second-model review before pull requests, and heavy tests including mutation coverage. One reply plugs a local open-source UI called obeya.si.
Unconfirmed: Commenters describe personal workflows; one says they did not run a systematic study.
- The poster says the agent only considers surrounding code in the current session.
- One reply fails the build with .NET static-analysis and editorconfig style rules.
- Another commenter has Codex Sol 6.1 review changes before opening a pull request.
- A third says agent-written codebases are generally more than 70 percent tests.
- The poster says a pre-PMF startup leaves little room for testing and benchmarking.
From the original article
8 min read at news.ycombinator.com[](https://news. ycombinator. com/)Hacker Newsnew past comments ask show jobs submitlogin [](https://news.
By nkcodes · news.ycombinator.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.