/s3/static.nrc.nl/wp-content/uploads/2026/10/15214514/201026OPI_2037475872_wisk.jpg)
OpenAI systems autonomously solved complex theoretical mathematics problems this October, marking a sudden shift that has left academic researchers grappling with the automation of high-level intellectual labor. Sorbonne University mathematics researcher Piet Lammers reported in nrc.nl that his own projects and those of his postdocs and promovendi are vanishing following the automated breakthroughs.
OpenAI Agents Automate Millennium Problems
OpenAI presented hundreds of solutions across multiple mathematical domains by deploying high-performing agents that collaborated in internal organizational roles without human intervention. Lammers stated in nrc.nl that human researchers played no role in asking the questions, directing the research, or checking the proofs. Mathematical accuracy on advanced theoretical problems jumped from two percent to thirty-five percent in data published at the beginning of October, according to SOURCE 2. That same outlet noted that OpenAI released a prototype in September that successfully resolved the Navier-Stokes equations, an enduring fluid dynamics challenge.
Sorbonne Researchers Lose EU Grant Focus
The mathematical physics subfield experienced heavy representation in the OpenAI publications, directly disrupting ongoing academic work supported by external funding. Lammers received a 1.4 million euro grant from Europe that now requires completely revised objectives after machines solved problems at speeds that bypassed human investigators. Gabriel Manso and Claudio Mayrink Verdun joined an academic effort collecting unsolved problems between December 2025 and May 2026, testing models including o1, o3, GPT-5, GPT-5.1, GPT-5.4, and GPT-5.5 on challenges published after training cutoffs, SOURCE 2 reported.
Theoretical Physics Faces Similar Disruption
Computer Science and Artificial Intelligence Laboratory researcher Neil Thompson at the Massachusetts Institute of Technology stated in SOURCE 2 that probability and statistics, economic theory, cosmology, and quantum physics will serve as the next disciplines affected by advanced models. SOURCE 3 reported that critics fear the sudden influx of machine-generated proofs will undermine mathematics before physics experiences a parallel transformation.
Human Oversight Remains Essential for Verification
Sara Hooker, technology expert and cofounder of the startup Adaption, stated in SOURCE 2 that artificial intelligence alters working methods but cannot replace human researchers because machines cannot independently formulate fundamental questions. Lammers observed in nrc.nl that some colleagues still believe machines write proofs poorly, questioning whether that limitation will persist for more than a few weeks or months.
