I tend to write doc-comments before the functions they document, because it helps me think more clearly about what I want to happen - and sometimes causes me to

I tend to write doc-comments before the functions they document, because it helps me think more clearly about what I want to happen - and sometimes causes me to entirely re-think my approach and abandon the function entirely. I can definitely see such a practice improving LLM output. Meanwhile, there are programmers that think comments are a "code smell".