OpenAI has announced three new results in mathematics, computer science, and physics. The company also says it is changing how it shares such work. Instead of dropping everything at once, it now lets handpicked experts decide how each finding reaches the public.
The last few months have been rough between OpenAI and researchers. Mathematicians are still working through hundreds of AI-generated preprints the company released at the start of October. That episode left many in the field exhausted and suspicious of the next announcement.
What OpenAI Says It Found
The mathematics result is the headline. OpenAI claims it resolves the Kakeya conjecture in all dimensions. The problem asks how little space a shape needs to point in every direction. A proof of the three-dimensional case helped NYU mathematician Hong Wang win a Fields Medal this year.
Brown University professor Brendan Hassett said the claim “looks well supported.” He added that he had not yet checked the Lean formalization, which lets computers verify a proof.
The physics work touches the Standard Model and quantum chromodynamics, the theory of the strong nuclear force. It also connects to Yang-Mills theory, a pillar of particle physics. OpenAI’s paper does not address the unresolved “mass gap.”
The third result claims to disprove the exponential time hypothesis. That idea holds that some problems cannot be solved much faster than brute force. It sits close to P versus NP, one of the field’s biggest open questions.
Researchers Welcome the Science, Not the Process
Experts who spoke to The Verge took the findings seriously. They called them “huge,” “major breakthroughs,” and “significant.” The frustration was aimed at how OpenAI handled the release.
“The fact that they handpick the group is a disaster,” said Andreas Karch, a physics professor at the University of Texas at Austin. “The large labs have enough power as it is. They should not be picking winners and losers.” He asked why OpenAI cannot release the proofs now alongside its expert panel.
Constantin Kogler of the Institute for Advanced Study took a softer view. He welcomed a batch of three results over the hundreds released earlier. Still, he questioned the logic of limited early access. “It doesn’t make any sense that first only a few distinguished scientists can look at them,” he said.
The timing is awkward for OpenAI. A day before the announcement, the Advisory Group on Mathematics and Artificial Intelligence urged labs not to adopt this exact practice. The group said prerelease access is unnecessary, borrows the credibility of selected researchers, and creates unequal access. After the news, it pressed OpenAI to publish the natural-language and Lean outputs behind the physics result without delay.
Source: The Verge — Researchers are getting tired of OpenAI gatekeeping its breakthroughs
Related reading on mylistingo: OpenAI now trains its frontier models with no internet access, and GPT-5.6 Sol, Terra and Luna: OpenAI locks down its best AI.
The Netherlands, for internationals
One email a week: the week's news for internationals in the Netherlands, new practical guides and what changed in the cities. No spam, unsubscribe any time.
