I prefer low effort human thought to low effort llm output.