Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I don't think the concept here is blindly taking the output of the LLM and calling it a day. One can test, the test.


I've had about 33% luck with unit tests coming out perfect. Usually the issues are small problems that are easily fixed though.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: