I agree, this is clearly an indictment against LLMs. If LLMs and agents were capable they'd 100% write it natively but they realize the current limitations.