Is Cursor Hallucinating when Generating Code? This could be the reason...

If you’ve enabled rules in Cursor, what it will do is that whenever you send a prompt, your rules will also be sent to the LLM.

Therefore, if your .cursorrules file contains complex and lengthy rules, your original prompt might be blurred and the model may start hallucinating. Therefore, only use specific rules when needed and turn off this auto-sending feature if you see a dip in the code generation accuracy.

What I noted was that the model wasn’t able to understand exactly what I needed and it sometimes went really weird doing one thing again and again infinitely. This could be because, at times, it couldn’t remember what it had done previously.

Lesson: Don’t feed too much data to your LLMs. Only give them what’s needed. Take your time to research. Don’t throw away all the stuff to the LLM to decide. At the end of the day, it’s a piece of software that’s doing a bunch of calculations so it can’t think like a human can nor does it possess consciousness.