2026-10-12 / Signal #4
DeepMind’s math agent got into *Science* — while math enters “proof abundance
“AI just published math in *Science*. The world’s best mathematician says that’s the wrong victory.”
The Story
Google DeepMind’s AlphaProof Nexus paper is now in *Science*: an LLM+Lean agent resolved 9 of 353 open Erdős problems (two open 56 years) and 44 of 492 OEIS conjectures, at a few hundred dollars per solved Erdős problem. Coverage is landing now as a contrast to OpenAI’s already-covered GitHub math dump. Separately, Terence Tao told Caltech that math has moved from proof scarcity to proof abundance, and that “further blind optimization of problem-solving alone is now actively harmful.” HN is circulating the slides.
Why It Matters
The plot is not “AI did math.” It’s two institutions of truth: a journal with peer review versus a dump with a boycott, while the best living mathematician says solving is now the cheap part.
Evidence
*Science*, Oct 8; The Decoder, Oct 12; Terence Tao “Math 2.0” Caltech lecture, Oct 9
Sources
Daily scan: 2026-10-12