It's an interesting concept, but inapplicable here because I don't trust the media reporting on LLMs and I personally believe expert programmers are never going to be replaced. My concept of the value of LLMs is that they are good for generating throwaway test code to assess the use of a library or to prototype a feature.