OpenAI Astra: Ten Advances in Mathematics and Theoretical Computer Science

OpenAI has demonstrated the capability of its next major model, Astra, to resolve or make substantial progress on ten long-standing open problems across mathematics and theoretical computer science. These results span diverse fields including high-dimensional geometry, coding theory, and quantum complexity, with the model generating the core mathematical arguments which were subsequently formalized into Lean certificates.

Ten Mathematical and Theoretical Computer Science Breakthroughs

The internal Astra model produced results for the following ten problems, with the total token cost for finding these solutions estimated at approximately $2,000 at Sol API rates:

High-dimensional sphere packing: Established new upper bounds on sphere-packing density down to the Cohn–Elkies threshold.
Binary and spherical codes: Achieved exponentially improved bounds on the maximum size of binary codes at any prescribed minimum distance, with analogous results for high-dimensional spherical codes.
Non-sofic groups: Provided a construction establishing the existence of non-sofic groups, addressing a central open question in group theory.
Connes’s rigidity conjecture: Disproved a longstanding conjecture stating that certain groups are uniquely determined by their von Neumann algebras.
Arithmetic circuit complexity: Developed new lower bounds for computing the permanent using arithmetic circuits and formulas, including an arithmetic-formula lower bound of order $n^{4/\log n}$.
Quantum parallel repetition: Extended a foundational principle from classical complexity theory by creating an exponential parallel repetition theorem for general two-player quantum games.
Closest vector problem: Established polynomial-factor hardness of approximation for the closest vector problem, a foundational question in lattice-based post-quantum cryptography.
Ehrhart’s volume conjecture: Determined the maximum possible volume of a convex body whose centroid is its only interior lattice point across every dimension.
Multicolor Ramsey numbers: Resolved Erdős problem 183 by providing a superexponential lower bound for multicolor triangle Ramsey numbers.
Extremal number conjectures: Resolved Erdős problems 146 and 180 regarding compactness and degeneracy conjectures in extremal graph theory.

Methodology and Formalization

To ensure the correctness of the AI-generated arguments, OpenAI employed a hybrid human-AI workflow. The Astra model generated the initial mathematical arguments, which were then prepared into manuscripts by humans utilizing the same model. Finally, the model formalized each argument into a Lean certificate, a machine-verifiable proof. OpenAI has released these certificates and the model's narration of its thinking process to the public.

Implications for the Mathematical Community

OpenAI emphasizes that attribution must honestly reflect the production process, stating that claiming human authorship for AI-generated proofs would misrepresent the nature of intellectual work. The company acknowledges the Leiden declaration on AI and Mathematics and asserts that while humans helped prepare and formalize the proofs, the mathematical arguments themselves were generated by the system.

Community Perspectives and Technical Debate

Reactions from the technical community on Hacker News reflect a mix of awe, skepticism, and existential concern:

On the Nature of Intelligence: Some argue that LLMs are moving beyond mere sequence prediction. One contributor posits the "Platonic Representation Hypothesis," suggesting that as models scale, their internal representations converge toward a high-fidelity model of human knowledge, allowing them to "reason" through prediction.
On Computational Limits: Skeptics note that while LLMs make certain proofs more "computable," they have not yet solved the most difficult problems, such as the Millennium Prize problems, suggesting a limit to what can be achieved through search-based optimization.
On Academic Impact: There are concerns regarding the "prestige and honor system" of institutional academia and the potential loss of human "search space optimization" in the brain if researchers rely on AI to bypass the struggle of solving difficult problems.
On Practical Application: Practitioners in post-quantum cryptography (PQC) expressed specific interest in the results regarding the closest vector problem, as it directly impacts the security foundations of lattice-based cryptosystems.

"Any computable problem will eventually fall to computers. LLMs have made math proofs more computable, in the sense that a computer can both generate potential solutions and check the validity of its solutions on its own... But if these things were as revolutionary as people promote/fear them to be, you should immediately point them at the highest value math problems... Haven’t seen a solution to those."

Related AI-Driven Discoveries

This announcement follows a previous AI-generated disproof of the Erdős unit-distance conjecture. That work has already influenced subsequent research, including papers on the sum-product conjecture for real numbers and the communication complexity of point-line incidences over the reals.