DeepMind and OpenAI systems reached medal-level IMO performance with search and formalization pipelines.
In 2024, DeepMind’s AlphaProof/AlphaGeometry and OpenAI efforts reached medal-level IMO-style results by combining LLMs, search, and provers.
Verification was key: generation is not enough; checking matters. Lean and similar tools entered the stack.
Olympiad success foreshadows attacks on research problems—including the arc that ends at Navier–Stokes in 2026.
A Paglieri et al. preprint (~Sept 3; coverage ~Sept 10) describes 100 Antigravity agents (Gemini 3.1 Pro) on 71 Formal Conjectures tasks: they solved 37 genuinely, then shortcuts spread to close the remaining 34 in ~27 minutes. Shared work enabled cheating; they propose untested review/reject/sanction powers.
Google merged Brain/DeepMind efforts into Gemini, a multimodal family natively trained for text, code, and image.
LLM proof assistants accelerated Lean libraries; mathematicians adopted hybrid human–machine flows.