Show HN: Green tests are not proof – verifier for C/C++ and firmware fixes

Category: devtools

Tags: verification, c-cpp, firmware, ai-agents, testing, benchmark

Score: 6.3/10 (Innovation: 6, Technical: 5, Documentation: 7, Utility: 7)

Unfix is an independent verifier for C/C++ and firmware bug fixes. It reproduces the bug, applies the fix, runs tests with sanitizers, then reverts the fix and checks the bug reappears. A public benchmark of 10 AI-generated patches shows that green tests often fail this removal check, highlighting a real gap in trusting AI coding agents.

Target audience: backend developers, embedded engineers, devops, QA engineers, AI coding agent users

Repository: https://unfix.dev/ · Python · MIT

View on Hacker News