Honestly I think LLMs really shine best when your first getting into a language.

I just recently got into JavaScript and typescript and being able to ask the llm how to do something and get some sources and link examples is really nice.

However using it in a language I'm much more familiar with really decreases the usefulness. Even more so when your code base is mid to large sized

I have scaffolded projects using LLMs in languages I don't know and I agree that it can be a great way to learn as it gives you something to iterate on. But that is only if you review/rewrite the code and read documentation alongside it. Many times LLMs will generate code that is just plain bad and confusing even if it works.

I find that LLM coding requires more in-depth understanding, because rather than just coming up with a solution you need to understand the LLMs solution and answer if the complexity is necessary, because it will add structures, defensive code and more that you wouldn't add if you coded it yourself. It's way harder to answer if some code is necessary or the correct way to do something.

This is the one place where I find real value in LLMs. I still wouldn't trust them as teachers because many details are bound to be wrong and potentially dangerous, but they're great initial points of contact for self-directed learning in all kinds of fields.

Yeah this is where I find a lot of value. Typescript is my main language, but I often use C++ and Python where my knowledge is very surface level. Being able to ask it "how do I do ____ in ____" and getting a half decent explanation is awesome.

The best usage is to ask LLM to explain existing code, to search in the legacy codebase.