Been trying to figure out a way to use LLMs to better understand code that comes from LLMs at a level of abstraction somewhere between the code itself and the p

Been trying to figure out a way to use LLMs to better understand code that comes from LLMs at a level of abstraction somewhere between the code itself and the prompt, but haven't succeeded yet.