Hacker News new | past | comments | ask | show | jobs | submit login

Have you tried writing tests with an LLM?

Because I have, and it's not been the experience you're describing. The LLM hallucinated the error message it was testing for (that it had itself written 5 minutes earlier, in the file it had been given as the source to test).

I don't think this can be solved with the current methodology we're using to create these assistants. I remain arguably highly intelligent and definitely convinced that LLMs need to evolve more to surpass humans as coders.




Consider applying for YC's W25 batch! Applications are open till Nov 12.

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

Search: