A number theorist conceded a bet four years early: AI can produce Annals-quality number theory for under $100k per paper. Prediction markets price an AI-solved Millennium Prize Problem by 2028 at 52%.
The response from practitioners is not competitive but elegiac. One described a dark night of mathematics, arguing discovery was how humans touched the ineffable, and asking whether foreclosing it for future mathematicians is itself a kind of evil.