Latest experiment in #ArtificialIncompetence:
I asked my go-to LLM to review the code it had produced. It rejected its own code. It specifically rejected a line for which I had asked some clarification and where it had convinced me that its approach was OK. It also rejected it for missing pieces of code that I had explicitly prompted it to skip. It hallucinated some bugs that didn't exist, got some explanations wrong... and missed all the crashes that it had originally created.
Incompetent.