Do you think new LLMs are going to write better and better code? When all they are going to have is the slop generated by previous, worse models?