That was the day OpenAI announced a new sort of bot, known as a “reasoning model,” that was trained to complete challenging tasks that took long periods of time—the very sorts of science, math, and ...