I'm more concerned about the "falling behind" we're doing chewing up compute generating, and then running, LLM slop. That is a lot of energy used and heat generated for a less-than-optimal payoff.