Users praise the agree-to-disagree mechanism in the experiment pitting Claude against ChatGPT because it works really good.
@joshgans