Okay, but why? This seems to be based purely on presumptions and guessing. Every attempt I've seen at actually measuring it has found the opposite. At some point there's a "code quality of the gaps" going on that I think comes down to "code is designed in a high quality way when I've internalized the design, and poorly when I haven't", which creates a self fulfilling prophecy when one thinks reading code written by an LLM is below them.