Nine Unpaid Mathematicians to Guide OpenAI on AI-Driven Math

OpenAI announced on September 21 that it will collaborate with an independent advisory group of nine mathematicians on how AI-generated mathematical results should be vetted and made public. According to the company's official announcement, the advisory group on mathematics and AI serves unpaid, may speak and publish independently of the company, can raise issues OpenAI never asked about, and manages its own membership.

The group's remit spans assessing the significance of new mathematical findings, advising on how results are released and coordinated, offering input on academic and professional standards in the field, and evaluating how OpenAI's tools support research and learning. The company drew one explicit boundary: the panel is not responsible for advising on how quickly internal mathematical research should advance.

The same disclosure carried a striking capability claim: an internal model that OpenAI began training on August 28 has solved more than 100 long-unsolved problems spanning most areas of mathematics, with the Navier-Stokes millennium problem the stated exception. The Navier-Stokes problem, which describes the motion of viscous fluids, is one of the seven Prize Problems posed by the Clay Mathematics Institute in 2000, each carrying a $1 million prize. OpenAI said the pace of progress surprised even its in-house mathematicians and triggered internal debate over how to prepare the wider field.

The advisory structure follows visible friction. OpenAI's September 9 Navier-Stokes announcement drew a challenge from NYU mathematician Tristan Buckmaster, who questioned whether the company raced to claim the result after learning of his progress. The company also quoted a declaration published at mathandai.org, signed by multiple Fields medalists — recipients of the quadrennial distinction widely regarded as mathematics' highest honor — warning of negative externalities when unsolved problems are treated as AI benchmarks. Notably, 2014 Fields medalist Martin Hairer sits on both the signatory list and the new panel.

OpenAI frames the collaboration as a "first step" toward putting mathematicians at the center of how AI supports the discipline. What remains undisclosed: which of the 100-plus problems were solved, and how the claims will be independently verified.

SoftBank Kicks Off $10 Billion Bond Sale

SoftBank Group has launched a bond sale of $10 billion alongside €1 billion in senior unsecured notes, with proceeds earmarked for the third $10 billion installment of its OpenAI investment as well as general corporate purposes — including replacing the bridge loan arranged for that commitment. The dollar notes carry tenors of 3.5 to 7.5 years, while the two euro tranches run 4 and 6 years. Citi and JPMorgan are leading the sale, which is expected to price on September 24 and settle on September 29.

The structure matters as much as the headline size. Rolling a short-dated bridge loan into term debt removes SoftBank's near-term refinancing risk on a concentrated position, but the interest bill stays on SoftBank's balance sheet until — and unless — OpenAI's valuation trajectory converts into realizable returns. That places a multi-year carry cost on a private-market stake that cannot be marked to market daily, unlike liquid holdings where exposure can be trimmed opportunistically.

The bond documents confirm timing but not the economics of the underlying stake. Neither SoftBank nor OpenAI has disclosed the valuation at which the third installment converts, the governance terms attached to it, or any liquidity path — details that would let bond investors price the credit story properly. Until those terms surface, the notes effectively bundle SoftBank's broader corporate credit with indirect exposure to OpenAI's future performance, and the market's reception at pricing on September 24 will be the first external check on how much of that risk buyers are willing to underwrite.

Altman Ecosystem Read-Through for WLD

Both headlines trace back to the same founder network: Sam Altman, who co-founded the Worldcoin project — the identity-and-payments Web3 venture tied to the WLD token. With OpenAI formalizing independent scientific oversight while SoftBank institutionalizes its funding through the bond market, the ecosystem signals a shift from scrappy startup to structured institution.

For WLD, frequently traded as proxy exposure to Altman's ventures, the read-through is sentiment-driven rather than cash-flow-linked, and the token's tokenomics and circulating supply schedule remain the fundamental anchors. As spot color, WLD moved 7.7% over the past 24 hours. The next dated Altman-ecosystem event on the calendar is the bond pricing expected September 24, which will also mark the first institutional read on the same OpenAI exposure that WLD's sentiment-driven trade tracks.