Answered from Noam Brown – Agent swarms, alignment, & recursive self-improvement · Dwarkesh Podcast · 2026-09-17
Noam Brown says he expected a Millennium Prize problem to fall around 2028, not in 2026. He had tracked a tenfold yearly rise in how long the hardest math problems models could solve would take a human: about five seconds for GSM8K, a minute for the MATH benchmark, ten minutes for AIME and about 100 minutes for International Math Olympiad gold in 2025. Following that line suggested roughly 15 hours next, which he judged too little for a Millennium Prize problem.
“he was willing to bet me $1,000 that it would take past 2027.” Noam Brown · at 35:05 — open in the transcript →
Source: our own transcription of the episode. The full conversation is on Dwarkesh Podcast (2026-09-17). Read the whole episode summary →