Timeline extrapolation: from sub-10% AIME to solving a Millennium Prize Problem
scaling01 · x · 2026-09-09
A timeline of AI acceleration: GPT-4o scored under 10% on AIME 2024 in May 2024, o3-preview saturated it by December; Gemini DeepThink took IMO gold in July 2025; internal GPT models reportedly made an Erdős unit-distance problem breakthrough in May 2026. The author predicts an OpenAI internal model solves a Millennium Prize Problem by Sep 2026 and argues that's not far-fetched.
More from Fun
- Touch Rock: an Unnecessarily Delightful Typing-Based Stone-Skimming Roguelike — jh3yy · 2026-09-09
- DeepMind's Samuel Albanie: AI cracking Navier-Stokes is an epic human achievement — SamuelAlbanie · 2026-09-09
- "Navier-Stokes is solved — can we get back to curing cancer now?" — CtrlAltDwayne · 2026-09-09
- Miles Brundage spoofs Anthropic rumors: models found a path to federal AI law — Miles_Brundage · 2026-09-09
- Rumor: Anthropic discovered a new class of antibiotics, plans IPO reveal — generativist · 2026-09-09
- 'Spawning 100,000 sub agents': the meme about agents over-refactoring on demand — NERDDISCO · 2026-09-09