OpenAI Claims a Faster Way to Solve Logic Puzzles, Pending Expert Review

OpenAI says it has a method for counting solutions to certain logic puzzles far faster than trying every option, using little memory. Its 3-SAT case, a standard puzzle type, would disprove a widely held belief about their difficulty. Independent researchers led by Impagliazzo, Paturi and Williams will check the proof; it is under review, not accepted. Announced October 26, 2026.