There is a point in the LLM usage spectrum where I feel that I am not really understanding the stuff I am creating. I can have the LLM write some code, without starting to feel this. But it is only a small amount of code...