OpenAI announced it solved and Lean-formalized a variant of the Navier-Stokes existence and smoothness problem using approximately 10,000 coordinated AI agents over 88 hours. Research lead Mark Chen noted the compute cost reached millions of dollars using an internal model described as more capable than GPT-6 Astra.
The announcement arrived alongside controversy after mathematician Tristan Buckmaster alleged that OpenAI pressured him to remove his Anthropic-affiliated co-author, Levent Alpöge, from a related paper after learning of their AI-assisted breakthroughs. Buckmaster claimed OpenAI made repeated urgent outreach attempts and threats after information leaked.
OpenAI executives denied using the academic research, with Sébastien Bubeck stating the company did not see the researchers’ work before its public release. Bubeck confirmed OpenAI accelerated its own math model training and resource allocation in late August after hearing rumors of progress at Anthropic.
Why it matters
Multi-agent AI frameworks are solving high-level mathematical proofs, showcasing a milestone for automated scientific discovery.
Intense frontier lab rivalries are escalating pressure on academic collaborators working across competing AI ecosystems.
Reasoning compute costs continue to drop rapidly, turning million-dollar research efforts into affordable API calls over time.
Source: the-decoder.com
