I just tend to find LLM code output extremely to read, I guess. It tends to be verbose and do a lot of unnecessary stuff, but I can always get the point easily and edit accordingly.