I was thinking the same, although I don't think declarative is the best term. Purely functional may be better. Would getting an llm to use a language like Hask

I was thinking the same, although I don't think declarative is the best term. Purely functional may be better. Would getting an llm to use a language like Haskell allow it to make less errors? I've had chat gpt make errors to do with state changes that maybe haskell or caml code can't have.