Terence Tao one of the greatest mathematicians alive just said something that stuck with me: "We still don't have a complete picture of AI capabilities because companies rarely share negative results." Read that twice. The people building these systems are quietly sitting on the failures. So the picture we're all reacting to is the highlight reel, not the full story. And yet, even with that incomplete picture, here's the part that genuinely moved me: AI is already cracking some extremely hard math problems—not by being smarter than any human, but by combining known techniques at a scale no single mathematician could realistically attempt. It's not magic. It's patience and reach we simply don't have. Both things are true at once, and I think that's the honest place to stand: → We should be humble about what we don't know, because the wins get published and the losses get buried. → We should be curious about what's already real, because the scale advantage is genuine and it's here now. The temptation is to pick a side—hype or doom. The more useful move is to hold both. Celebrate what AI is actually doing, and keep asking for the results nobody's rushing to show us. That tension isn't a problem to resolve. It's the job. Honest question: when you evaluate a new AI capability, do you go looking for the negative results—or do you only see the highlight reel?