Debate May Help AI Models Converge on Truth

I haven't used this trick yet, but I'll surely will.

I read somewhere (can't remember where) that if you use only one model, you're getting at most 50% of the potential of LLMs. Routinely, I'll give a prompt both to ChatGPT and to Claude. Sometimes I'll take part of one answers, sometimes part of another. I haven't yet tried to have them critique each other.