Key Moments
AI Keeps Pissing Off Mathematicians
ColdFusionKey Moments
OpenAI claims to have solved a $1M math problem, but a mathematician alleges they stole his work from chat logs, sparking a debate about AI ethics and data privacy.
Key Insights
OpenAI announced the solution to the Navier-Stokes problem, one of the seven Millennium Prize Problems, a feat that has eluded mathematicians for nearly a century.
Mathematician Tristan Buckmaster claims OpenAI used his private chat logs with their Codex model to derive the solution, which he and a colleague had been developing using a novel approach.
The core of Buckmaster's argument rests on the timing of OpenAI's sudden intensive work on the problem immediately after learning of his progress and the use of his specialized methodology.
OpenAI's narrative shifted, initially denying and then admitting the possibility that anonymized user data, including potentially Buckmaster's work, could have influenced their models.
Prominent mathematicians have expressed concern that AI is being used for 'marketing events' rather than fostering genuine understanding, likening it to 'using bulldozers to loot an archaeological site.'
The controversy highlights a broader concern that AI might inadvertently learn from and replicate users' unpublished or sensitive work, raising questions about data ownership and intellectual property.
OpenAI's bold claim and initial mathematician skepticism
OpenAI has claimed to have made significant progress on over 300 major unsolved mathematical problems, with some even reportedly solved. This announcement has generated a mix of awe and concern within the mathematical community. While some mathematicians acknowledge the potential for AI to accelerate discovery, others, like Tristan Buckmaster, a mathematician at NYU, feel that companies like OpenAI are overstepping by attempting to solve complex problems without transparently showing their work, potentially even by misusing user data. The controversy ignited when OpenAI announced it had solved the Navier-Stokes problem, a $1 million Millennium Prize Problem that has puzzled experts for decades. This claim, however, is overshadowed by Buckmaster's serious allegations that OpenAI stole his research from his private chat logs.
The Navier-Stokes problem and the million-dollar question
The Navier-Stokes problem is one of seven Millennium Prize Problems identified by the Clay Mathematics Institute, each carrying a $1 million reward for a valid solution. These equations, formulated in the 19th century, describe the motion of fluids like water and air and are crucial for engineering and weather prediction. Despite their practical use, proving their consistent regularity—whether smooth fluids always remain smooth or can develop infinite speeds in finite time—has remained elusive for nearly a century. The problem essentially asks if the equations remain well-behaved at all times or if they allow for singularities where fluid behavior becomes chaotic and unpredictable.
Buckmaster's collaborative approach and the role of AI tools
Tristan Buckmaster, alongside his collaborator Levant Albujey from Anthropic, was exploring a specific, less-traveled path to proving the Navier-Stokes equations could become irregular. Their strategy involved a series of precisely timed pushes designed to drive a fluid towards infinite velocity. Buckmaster describes their collaboration as purely personal and free from institutional agreements. They utilized various machine learning models, including Claude from Anthropic, Codex from OpenAI, and particularly GPT-3.5 and Astra. Crucially, their OpenAI chat log sessions were stored, a detail that later became central to the controversy. The duo made significant progress, with Albujey sending Buckmaster 'the most horrifying proof I've ever read' on August 15, which was officially validated by August 29, 2026, thus paving the way for a potential solution.
The alleged theft and OpenAI's sudden involvement
On September 3, 2026, Buckmaster reached out to a prominent mathematician at OpenAI, clarifying that his team, not Anthropic, was working on the problem. This inquiry followed rumors about Anthropic's supposed breakthrough. It turned out that OpenAI was indeed working on the same problem. A mathematician at OpenAI informed Buckmaster that an internal OpenAI model had produced a lengthy proof related to the Navier-Stokes equations. The method used by OpenAI was described as the same unconventional approach Buckmaster and Albujey had been secretly developing for over a year, which Buckmaster saw as a 'clear warning sign' that they had likely used his methodology.
Conflicting narratives and accusations of coercion
Buckmaster's extensive PDF document details his argument, emphasizing the suspicious timing of OpenAI's intensive work—starting immediately after his progress became known to them—and their completion of the problem in under a week using his specialized method. He claims OpenAI did not initially disclose when they began their work, vaguely referring to 'amazing models.' He also disputes claims by an OpenAI mathematician that human intervention was minimal, asserting a full OpenAI team was involved. A key point of contention is whether OpenAI's models were trained on Buckmaster's private Codex sessions, where all project drafts were stored. When questioned, OpenAI initially stated their models do not search user data, but later failed to provide a clear answer about training data. The situation escalated when OpenAI allegedly offered to let Buckmaster publish first and claim the $1 million prize, but only if his colleague, Albujey, was removed from the contributor list because he works for Anthropic. Buckmaster refused, and the interaction reportedly took a threatening turn, with an OpenAI representative asking, 'If you don't want me to be nice, I don't have to be.'
OpenAI's shifting statements and broader implications
OpenAI initially dismissed Buckmaster's claims as 'false and inflammatory.' However, as the New York Times investigated, OpenAI's narrative evolved. Their initial statement claimed researchers had no access to the duo's work until publication. This was later modified to state it was 'impossible to cut and dry' that Buckmaster's inputs on Codex influenced their system, and that no inputs after July 3 could affect it. Sam Altman also tweeted that the team acted with integrity. This shifting stance fuels suspicion about whether OpenAI's solution was derived, directly or indirectly, from Buckmaster's work stored in chat logs. The controversy extends beyond mathematics, raising profound questions about data privacy and intellectual property in the age of AI, as researchers and professionals input sensitive, unpublished work into AI tools.
Concerns about understanding versus answers and the future of research
Many mathematicians, including Terence Tao, worry that AI might solve complex problems but fail to contribute to genuine understanding. This mirrors concerns about students using AI for answers without grasping the concepts. The incident has also led to a chilling effect in the mathematical community, with researchers hesitant to discuss their work for fear of AI companies appropriating it. The Clay Mathematics Institute is reviewing the situation, but the Navier-Stokes problem remains officially unsolved. OpenAI has stated they will not pursue the $1 million prize. The incident highlights a critical need for legal frameworks to keep pace with technological advancements, ensuring trust and fair play in how AI interacts with human knowledge and creativity.
Mentioned in This Episode
●Software & Apps
●Companies
●Organizations
●Concepts
●People Referenced
Common Questions
OpenAI claimed to have made significant progress on over 300 unsolved mathematical problems, potentially including the Navier-Stokes problem. However, these claims are surrounded by controversy regarding their methodology and accusations of data theft.
Topics
Mentioned in this video
A prominent AI research company that claimed to have solved major mathematical problems, sparking controversy over its methods and data usage.
A company where Levint Albugea works, which was rumored to have solved a major math problem before OpenAI's announcement.
An interactive learning platform for math and programming that helps students develop problem-solving skills without just giving answers.
Leader of OpenAI, mentioned in the context of political maneuvering and potentially reckless decisions within the AI industry.
A Spanish mathematician who, along with Diego Córdoba, developed a new strategy for approaching the Navier-Stokes problem.
A representative from Monash University who highlighted the need for researchers to trust that their unpublished work entered into AI systems won't benefit competing companies.
A mathematician working at Anthropic who collaborated with Tristan Needleman on the Navier-Stokes problem research.
A Spanish mathematician who, along with Luis Martínez Zornoza, developed a new strategy for approaching the Navier-Stokes problem.
A mathematician who proved the Poincaré conjecture, one of the Millennium Prize Problems, but refused the prize money and Fields Medal.
A mathematician at OpenAI who discussed the company's work on the Navier-Stokes problem with Tristan Needleman.
A mathematician from New York University who, with Levint Albugea, advanced a novel approach to solving the Navier-Stokes problem, later accusing OpenAI of data theft.
A renowned mathematician who praised Needleman and Albugea's work but warned against AI companies treating problem-solving as marketing events.
An AI model from Anthropic used by Needleman and Albugea during their research.
A programming language mentioned as an example of AI programming's strong points, currently experiencing a revolution in the gaming industry.
An AI foundational model expected to revolutionize drug discovery, highlighting AI's potential in medicine.
A programming language mentioned in the context of a course taken on Brilliant.org, demonstrating how Koji can assist in learning.
An OpenAI model used by Needleman and Albugea, later central to accusations of data theft.
An AI model from OpenAI used by Needleman and Albugea in their research.
A more capable internal AI model used by OpenAI, reportedly more advanced than GPT-6 Astra, which worked on the Navier-Stokes problem.
A feature within Brilliant.org designed to guide students through problem-solving without directly providing answers.
More from ColdFusion
View all 92 summaries
22 minThe AI Industry is a Complete Mess
23 minHow to lose $35 Billion Dollars Betting on AI
23 minWhy Kids Can't Read Anymore
30 minWhat Does Palantir Actually Want? (Re-upload)
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free