If LLMs didn't hallucinate as much I'd maybe trust their output more. But they have a tendency to do unpredictable things, which isn't really what you want in a

If LLMs didn't hallucinate as much I'd maybe trust their output more. But they have a tendency to do unpredictable things, which isn't really what you want in a production environment. I keep seeing stories of what AI assistants have done incorrectly, which makes me hesitant to pipe their output into a codebase, unmodified