lateinteraction: a tool that solves Millennium Prize Problems yet fails 1000 small things reshapes the map

stochasticchasm · x · 2026-11-21

Reposting his earlier take as Riemann Hypothesis news breaks, lateinteraction argues: (1) a tool that can solve Millennium Prize Problems while being dumb at 1000 other things reshapes his understanding of the territory, not his judgment of any single vehicle — compute-vs-human capabilities are mutually jagged; (2) the brute-force vs. generalization ratio in the current paradigm has massively undershot even his low expectations — "my AI that solved Navier-Stokes needs help writing English after seeing the whole Web 10x over"; (3) a third thread about a different philosophy of value/work. A prominent researcher's take on AI's jagged capability frontier.

Original post →

More from AGI Musings

AGI Musings channel →