The Latest Math Problems that AI Has Solved in the Last Month (Google Gemini)

What are the latest math problems that AI has solved in the last month?

In the past month, the landscape of AI-driven mathematics experienced a landmark shift. On October 6, 2026, OpenAI publicly released a massive collection of 722 manuscripts—totaling over 700 pages of proof attempts, solutions, and partial progress for 372 open problems.

Rather than tackling benchmark textbook problems, an unreleased advanced model was tasked with attacking long-standing open conjectures across pure and applied mathematics. The mathematical community has spent the past few weeks analyzing, formalizing, and verifying these generated papers.

The notable mathematical problems, conjectures, and breakthroughs produced or verified by AI systems over the past month include:

1. The Quadratic Sensitivity Conjecture (Disproved)

• Field: Theoretical Computer Science / Combinatorics
• The Breakthrough: For decades, a major open question regarding Boolean functions was whether block sensitivity was bounded tightly by the square of sensitivity , specifically . The model constructed an explicit counterexample establishing , settling the strong polynomial relationship conjecture that had stood unresolved since 1994.

2. Free Group Factor Isomorphism Problem (Positive Solution)

• Field: Operator Algebras / Functional Analysis
• The Breakthrough: First explicitly posed by Richard Kadison, this was one of the premier open problems in the theory of von Neumann algebras. The AI generated a proof establishing the isomorphism conditions for free group factors, a result that researchers are actively translating into formal proof languages like Lean.

Wikipedia

3. Open Progress Across 372 Open Problems

• Field: Multi-Disciplinary (Number Theory, Topology, Graph Theory)
• The Scope: The October 6 release covered over 4,000 targeted open problems. Beyond headline proofs, mathematicians verified scores of preliminary results, including counterexamples in algebraic geometry, explicit bounds in spectral graph theory, and new combinatorial structures.

Wikipedia

The Verification Dynamic

The rapid influx of AI-generated mathematics has fundamentally altered academic workflow:

• Reversed Process: Traditionally, mathematicians develop intuition, construct a proof, and publish. Now, AI systems generate hundreds of pages of candidates first, leaving human mathematicians to act as peer-reviewers and formalizers.
• Formal Verification via Lean: To rule out subtle hallucinations, consortia of top mathematicians (such as the First Proof project) are running the AI outputs through interactive theorem provers to verify every single step mechanically.