Absolutely. The comment I was responding to argued the difficulty was that LLM code wasn't loaded cognitively. I'm arguing the problem is actually the code prod

Absolutely. The comment I was responding to argued the difficulty was that LLM code wasn't loaded cognitively. I'm arguing the problem is actually the code produced by LLMs tends to be messy and hard to follow beyond trivial examples.