Google Research said a multi-agent AI system had solved several open problems in mathematics and computer science; the claims have not yet been independently verified.
On Nov. 13, Google Research reported that its Antigravity /teamwork multi-agent system, using Gemini 4 Argon, Carbon and Hydrogen, had helped researchers tackle open problems in mathematics, computer science and other fields. Among the results, teams from Google, CUHK, USC and CMU proved the 2013 CourtadeāKumar conjecture in full generality. Google also said its LEAP framework autonomously generated a machine-checked Lean formalization of the Weil conjecturesāabout 1.8 million lines across 28,800 subagent invocationsāand reproduced a formalization of recent work on finite-time blowup in 3D NavierāStokes equations. The announcement matters because verified AI-assisted proofs could accelerate research and intensify competition among frontier-model labs, while giving mathematicians and software researchers new tools. But these are company-reported achievements, not yet independently validated; formal verification checks encoded reasoning, not whether every original claim or interpretation is sound.