OpenAI Reports Three AI Scientific Breakthroughs

OpenAI announced three claimed scientific advances in mathematics, computer science, and physics on October 26, 2026, saying its AI models had established new mathematical results, challenged a longstanding theory of computational complexity, and produced a rigorous proof related to quantum chromodynamics. Researchers outside the company are examining the findings, while OpenAI has introduced a process intended to involve academic experts in interpreting and publishing the results.

According to OpenAI, its models proved the local smoothing conjecture across every spatial dimension, disproved the Exponential Time Hypothesis (ETH), and developed a mathematically rigorous proof concerning quantum chromodynamics (QCD), the theory describing interactions between quarks and gluons. The company has not established that all three claims have been independently validated.

The mathematics and computer science results are being made available for outside examination, while the physics group has chosen to delay releasing its AI-generated materials as researchers prepare their own account of the findings.

Alex Iosevich, a University of Rochester mathematics professor specializing in harmonic analysis, said the reported solution to the local smoothing conjecture could represent a major development. The conjecture, formulated by his doctoral adviser Chris Sogge, had previously been established only in three dimensions, according to the reports.

"The conjecture was formulated by my PhD adviser, Chris Sogge, so it is near and dear to my heart, and the news was a shock at first," Iosevich told Al Jazeera.

Iosevich said the development could reshape theoretical mathematics, although he acknowledged that the prospect also brought mixed emotions for researchers whose careers have centered on difficult mathematical problems.

"Now I am filled with joy and optimism, because the scientific enterprise as a whole is going to make giant leaps as a result of the AI revolution, and we have the chance to be a part of this historic development," Iosevich said.

He also described the personal and professional uncertainty that could accompany a significant change in how mathematical research is conducted.

"I feel a twinge of sadness as a result," Iosevich said.

He added that he expected those feelings to diminish as researchers recognized the opportunities created by increasingly capable AI systems.

"Many theoretical mathematicians I know feel the same way, but I firmly believe that these feelings will soon give way to the realisation that there is so much we can now accomplish," Iosevich said.

Iosevich said mathematical research has historically drawn on practical scientific questions, creating opportunities for AI to accelerate work across disciplines.

"Problems in pure mathematics did not arise in a vacuum. Many of them came out of concrete physical considerations, and these connections can now be explored rapidly and decisively," he said.

The computer science claim concerns the Exponential Time Hypothesis, an influential but unproven assumption about the computational resources required to solve certain difficult problems. The hypothesis has informed theoretical computer science since the late 1990s, making any valid disproof potentially consequential for researchers studying algorithms and optimization.

Ryan Williams, a computational complexity theory expert at the Massachusetts Institute of Technology, said researchers were assessing whether OpenAI's argument could withstand scrutiny.

"We are working to verify and examine their proof and determine if it holds up," Williams told Al Jazeera.

Williams said the proposed approach appeared to offer a new way to solve several fundamental computational problems more efficiently. However, the significance of the result depends on whether independent examination confirms the proof and its implications.

"The ETH has been widely believed true since the late 1990s and a great deal of theoretical computer science was built on it being true," Williams said.

He cautioned that any practical benefits would not necessarily emerge immediately, even if the claim proves correct.

"It is hard to predict what will be possible next, because it's a shocking development," Williams said.

OpenAI's announcement also addresses concerns about how AI-generated research should be evaluated, distributed, and credited. The company said it had contacted external scientists to help turn its models' outputs into established scientific knowledge, describing the materials as a form of raw research data rather than finished scholarship.

OpenAI said it consulted three groups of experts, known as shepherds, to determine how the findings should be released. The mathematics and computer science groups chose immediate public disclosure, while the physics group opted to prepare its own interpretation before publishing the materials.

"All of these approaches involve tradeoffs," OpenAI said.

The company said collaboration could help researchers explain the results and place them within the wider scientific literature, but acknowledged that selecting the first experts could create unequal access to information.

"On the one hand, the collaborations empower experts to foster unpacking the results and contextualizing them in the broader academic community," OpenAI said.

The company also recognized that its selection process could affect participation in subsequent research.

"On the other hand, choosing initial shepherds necessarily shapes who gets involved first and creates asymmetry of information. We welcome feedback on how to broaden participation and make this process more useful," OpenAI said.

The latest announcement follows earlier disputes over AI-generated mathematical proofs. OpenAI published 377 mathematics results on GitHub earlier in October, with some results formalized or checked using Lean, a proof assistant designed to help verify mathematical arguments. Questions have also been raised about whether an earlier Lean-verified result concerning the Navier-Stokes Millennium Problem established the full claim initially presented.

The developments come amid wider debate about companies announcing AI-assisted scientific discoveries before independent researchers have completed their assessments. Anthropic, the creator of Claude, announced a potential enzyme discovery the previous month that it said resembled CRISPR, the gene-editing mechanism. That announcement also prompted differing reactions among scientists over the significance of the findings.