MIT researchers asked a simple question: What if AI could check its own work from multiple angles before giving you an answer? Not just “think step by step.” Actually verify. Score confidence. Flag uncertainty. The results were insane.

→ View original post on X — @godofprompt