OpenAI’s path to building superintelligent machines is looking ever fuzzier.

The latest sign came from an embarrassing blunder by the company’s leading scientists this month. "GPT-5 just found solutions to 10 (!) previously unsolved Erdos problems,” the company’s vice president for science, Kevin Weil, breathlessly posted on X in a post that has since been deleted.

It turned out that was misleading. The company’s latest model had simply scraped answers off the internet and regurgitated them as its own.