From 'AI can't do 3×(2+5)' to Navier-Stokes: a skeptic's surrender
exophades · reddit · 2026-09-09
A onetime AI skeptic posts a long Reddit 'surrender': three years ago he mocked ChatGPT for getting 3×(2+5) wrong and called AI history's biggest bubble; now, watching Astra do a professional infographist's job in seconds and an internal OpenAI model crack Navier-Stokes in 88 hours, he argues this is not an adaptable paradigm shift but the manufacturing of a stronger form of intelligence.
His case: humans need 20 minutes to refocus after a distraction, while frontier models solve graduate-level problems in that window; a problem all human mathematicians hadn't solved in 88 years fell in 88 hours. He concludes this revolution cuts deeper than 20th-century computing — we are automating ourselves, and the automated selves may be orders of magnitude better.
More from AGI Musings
- "AI Techbros Only Produce Low-Variance Copies of Researchers," Sparks Debate — MoonL88537 · 2026-09-09
- NUS argues universities should pursue 'X + AI', not 'AI first' — ProfChesterman · 2026-09-09
- NUS professor's new book asks what universities are for in the AI era — ProfChesterman · 2026-09-09
- Data engineer says PMs and CEOs are safest from AI, while data engineers feel the pressure first — Zachly · 2026-09-09
- Researcher warns entering the AGI "endgame" race from a private company's Slack is a hubristic gamble — dyn___ · 2026-09-09
- The 2013 grumpy blog on Zuckerberg invading NIPS—every worry came true — iskander · 2026-09-09