> A big wrinkle in AI evangelism is that proponents don’t understand the concept of human judgment as a “learned” skill

Which is ironic given how much variation in output quality there is based on the judgement of the person using the LLM (work scope, domain, prompt quality, etc.)