Arun ArumugamAgent Engineering

FeedUpdated 13:27 UTC

A Lean-verified proof of the Riemann Hypothesis, built with Claude

Levent Alpöge released a Claude-assisted Lean proof of the Riemann Hypothesis, certified by the Lean community a day later. Thinking Machines shipped a small looped open model, and OpenAI explained its defenses against distillation.

Show

SpikeNewsLean communityResearch

Levent Alpöge's Claude-assisted proof of the Riemann Hypothesis is certified correct in Lean

The 17-million-line Lean proof, built with Claude, was checked by Lean's kernel and several independent checkers. It is the clearest case yet of AI plus formal verification settling a major open problem.

  1. Thinking MachinesReleaseModelsThinking Machines releases Inkling 2 Mini, a 26B looped open-weights model for laptopsA 26B MoE (3.4B active) that reruns its middle layers three times per token; it leads open models under 100B on DeepSWE and comes with research on how looping affects chain-of-thought monitoring.
  2. OpenAINewsSafetyOpenAI details how it detects and disrupts adversarial distillation of its modelsExplains how capabilities transfer through unauthorized distillation, how OpenAI spots coordinated extraction, and how it now protects hidden reasoning.