Exclusive: OpenAI’s next flagship model repeatedly processes problems in ways humans can’t read. That makes monitoring its reasoning harder. The company is working to preserve that visibility while fixing safety issues ahead of release. Full story: https://thein.fo/4hXPxSG