Physicist @skdh says @OpenAI's GPT-5.6 solved a math problem after GPT-5.5 spent six months calling it unsolvable. @tunguz's takeaway is practical. Save problems that stump today's models, then try them again when the next generation arrives.(Digg)