I agree. See also my comment in a sibling thread about using LLMs to write tests: https://news.ycombinator.com/item?id=41961214