LLMs might be good enough at coding a solution, but a human might need to understand the problem and apply some of their wisdom to it.