ARTIFICE DAILY AI + CYBER SECURITY BRIEFING

Artificial Intelligence

Google Deepmind's AlphaProof Nexus solves open maths problems

A Science study said Google Deepmind's AlphaProof Nexus solved nine open Erdős problems, 44 OEIS conjectures and one algebraic geometry question, and its paper was reviewed before publication.

THE-DECODER

A study in Science said Google Deepmind's AlphaProof Nexus solved nine of 353 open Erdős problems, including two that had remained unsolved for 56 years. The team said it had posted those solutions online before the paper appeared.

The report also said the system proved 44 conjectures from the OEIS integer sequence database, resolved an open question in algebraic geometry and improved a bound in min-max optimisation.

Google Deepmind said AlphaProof Nexus was built around Gemini 3.1 Pro, which drafted and revised formal proofs in Lean. Lean checked each edit and returned errors that guided the next attempt.

For harder sub-problems, the system could use AlphaProof, which Google Deepmind described as a specialised proof tool. The company said the system still solved the nine Erdős problems without it.

The study appeared during a dispute over AI maths proofs. OpenAI had recently posted hundreds of proofs on GitHub rather than sending them to journals, and the Association for Human Mathematics, headed by Terence Tao, called for a boycott.

Named in this story

People

Terence Tao
headed the Association for Human Mathematics

Companies

Google Deepmind
said its system solved multiple open mathematics problems
OpenAI
was criticised for posting AI proofs online without journal submission

Organisations

Science
published the study
Association for Human Mathematics
called for a boycott

Products and systems

AlphaProof Nexus
the system described in the study
OEIS integer sequence database
supplied conjectures that the system proved
Gemini 3.1 Pro
formed the basis of the proof system
Lean
checked each proof draft after every edit
AlphaProof
served as a specialised proof tool for harder sub-problems

How the source tells it

The piece reads like a celebratory product note, with a contrastive jab at OpenAI and a brief controversy hook.

  • hype the headline and lead lean on loaded emphasis about the work's reception and success
  • triumph the article piles up counts of problems, conjectures and bounds solved to amplify the achievement
  • controversy the OpenAI comparison and boycott references frame the story through a live dispute