You could improve AI performance through all sorts of clever techniques… or you could just have more LLM agents try to solve the problem and debate amongst themselves as to which is the right answer. It turns out that adding more agents seems to help all AIs with most problems.

→ View original post on X — @emollick