Sam Altman on GPT-5: “We had a GPU crunch. We could’ve shipped another giant model and lots of people would try it; but we’d disappoint them. So we built something smarter and more useful, and drove inference cost down.” Translation: less brute force, more brains per dollar.