OpenAI Speedruns Math: New Model Claims 100+ Open Problems in 24 Days
Silicon Valley just decided that centuries of human intellectual struggle were merely an unoptimized compute problem. While researchers hold their breath, theoretical science suddenly faces an unprecedented avalanche of machine-made proofs.
Training of the new internal system began on August 28, deploying a massive swarm architecture designed to systematically bludgeon unsolved problems across nearly every domain of mathematics.
This is not an isolated stunt. In May, the company announced its model had disproved the Erdős unit distance conjecture, an 80-year-old riddle in discrete geometry. By August, it dropped 10 more results across quantum complexity and cryptography using an internal version of Astra, formalizing every proof in Lean. Just weeks ago, OpenAI deployed 10,000 agents generating 300 billion tokens over 88 hours to tackle the Navier-Stokes existence and smoothness problem.
Now the company is claiming its latest system has resolved over 100 long-standing open mathematical problems, though it has yet to publish the actual proofs, methodology, or specific task directory. It seems Silicon Valley expects everyone to take its black box at face value before peer review even gets an invitation.
The sheer velocity of synthetic discoveries prompted OpenAI to set up an Advisory Group on Mathematics and Artificial Intelligence alongside the Institute for Advanced Study in Princeton. The roster includes heavyweights like Timothy Gowers and Edward Witten, tasked with figuring out how academic norms can survive a software pipeline that spits out papers faster than professors can brew espresso.
Resistance is already mounting. Days after the Navier-Stokes claim, 25 elite researchers—including Fields Medalists Terence Tao and Maryna Viazovska—signed a public declaration condemning the practice of treating historic conjectures as benchmark trophies without advancing actual human comprehension.
Generative compute is churning out synthetic answers faster than biological brains can verify them, leaving science stranded between trusting unverified machine intuition or watching human-led inquiry fade into an ornamental hobby.
Source: OpenAI
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.