OpenAI is releasing raw research data, not polished papers, covering three new discoveries in harmonic analysis, complexity theory and quantum field theory. The company is experimenting with distinct knowledge dissemination strategies for each field, ranging from full open release for mathematics to deferred release guided by collaborators in physics. OpenAI explicitly states this. These releases will involve external collaborations operating under principles of transparency, voluntary participation and independent conclusions.
OpenAI Releases Data: Three Academic Collaboration Strategies
The company is not pursuing academic credit for these discoveries, a decision intended to prioritize open scientific understanding over proprietary claims. The approach to releasing these findings varies significantly by field; for computer science, OpenAI has assembled a collaborative group to interpret the results while simultaneously publishing the AI-generated output on GitHub. This group opted to make both the natural language proof and Lean formalization publicly accessible, a move intended to encourage broader examination and building upon the work.
In contrast, the physics collaboration decided to hold off on releasing the AI-produced artifacts and instead work on writing up the findings themselves. OpenAI explicitly stated it will not seek academic credit for discoveries generated by its internal models or the collaborating researchers, a stance that shifts the focus from ownership to collective understanding.
According to the company, inviting a researcher to steward a collaboration does not imply they made the original discovery, but rather that they will be recognized for the subsequent understanding, writeups, and materials they create. Participation in these working groups is voluntary and unpaid, with members operating independently and not as representatives of OpenAI. The company will turn over materials and artifacts related to these results to the collaborations.
This approach acknowledges the need for external expertise to convert the data into meaningful scientific understanding. The company is actively soliciting feedback on how to broaden participation and improve the process. The differing strategies, full open release, collaborative explanation and deferred release, demonstrate a willingness to experiment with various models for knowledge dissemination, and to adapt based on community response.
AI Proof Resolves Local Smoothing & Kakeya Conjecture in All Dimensions
A mathematical result, concerning harmonic analysis, builds on 2019 work by Guth, Wang, and Zhang who initially proved the local smoothing conjecture in two spatial dimensions, and 2025’s three-dimensional Kakeya conjecture proof by Wang and Zahl. OpenAI’s models extended these findings to encompass all dimensions through a novel auxiliary partial differential equation. The company released both an AI-generated natural language writeup and the Lean proof itself in a public GitHub repository, adopting a strategy similar to their October 6th, 2026 release.
This approach reflects a shift toward open dissemination of raw research data rather than traditional polished academic papers, allowing external experts to validate and interpret the findings. The local smoothing conjecture posits that extreme wave focusing is short-lived, a pattern mirroring broader mathematical principles where worst-case scenarios are less common than average behavior. Its resolution has implications across mathematical analysis, including the Kakeya and Stein’s restriction conjectures.
Following the initial AI output, OpenAI consulted with an expert to establish a collaborative framework around the result, ultimately choosing a full open release strategy for this particular finding. This contrasts with approaches taken in other fields, such as computer science and physics, where OpenAI is experimenting with different knowledge dissemination methods. In computer science, the team is releasing a disproof of the Exponential Time Hypothesis (ETH), a conjecture about the complexity of Boolean formula satisfiability, alongside both AI-generated explanations and a Lean proof, also available in their GitHub repository.
The physics collaboration focuses on a mathematically rigorous proof of the existence of certain interacting quantum field theories, including Quantum Chromodynamics and a unified completion of the Standard Model of particle physics. OpenAI’s models generated a natural language disproof for the ETH and formalized the result in Lean the following day, leading to the development of the current collaborative effort.
The company hopes that combining human expertise with AI capabilities will yield new insights into computational limits and capabilities, and advance understanding in these complex fields. These releases are intended to support broader scientific understanding, and OpenAI is actively exploring different strategies for sharing its findings across various disciplines. The company’s commitment to open access and collaboration signals a departure from conventional academic practices, prioritizing external validation and interpretation of AI-generated results.
AI Disproves Exponential Time Hypothesis, Impacts Complexity Theory
These experts will shepherd the work and organize groups to understand the new results. OpenAI’s approach involves releasing manuscripts as a form of raw research data, experimenting with new approaches to convert this data into scientific understanding. The external collaborations empower experts to foster unpacking the results and contextualizing them in the broader academic community. Finding algorithms and translating theoretical advances into practical improvements will require substantial further research, according to OpenAI, and the team is actively soliciting feedback from the academic community.
This collaborative model extends to a separate effort in quantum field theory, where OpenAI has produced a natural language proof for the existence of Yang-Mills theory, a component of the Clay Mathematics Institute Millennium Prize problem. The external collaboration for the physics result has decided to hold off on releasing the AI-produced artifacts and instead work on writing it up themselves.
Public Data & Funding Support ETH Verification and Future Research
This funding is separate from the raw data releases and is intended to empower researchers to examine, explain, and build upon the newly available information, extending beyond the immediate collaborative efforts. The company believes this investment is important for encouraging a broader scientific community engagement with its work and accelerating discovery across multiple disciplines.
The approach to data dissemination varies significantly between fields, reflecting OpenAI’s experimentation with knowledge sharing strategies; for the computer science result, a collaborative group is actively explaining the findings while the underlying AI output remains publicly accessible via a GitHub repository. This group specifically requested, and OpenAI granted, full public access to all raw data underpinning the disproof of the Exponential Time Hypothesis, allowing for independent verification of the results.
This commitment to open access extends to a planned conference, funded by OpenAI, intended to comprehensively examine the implications of the ETH disproof and the broader questions it raises regarding verification, access and credit within the scientific community. The company formalized its Lean proof on October 17, 2026, following the AI’s initial natural language disproof on October 16, 2026, and is now inviting external organization of an open examination of the claim.
The company’s evolving release process is directly informed by feedback from academic communities, indicating a willingness to adapt its strategies based on expert input. OpenAI continues to actively solicit feedback on the most effective methods for disclosing results and engaging with scientists to accelerate discovery.
See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.