The Omniscient Bulletin · 2026-09-22
The Omniscient Bulletin — September 22, 2026
OpenAI had two very different Mondays. At ten in the morning it asked Washington to lead the world in writing technical standards for recursive self improvement, and said fully autonomous RSI should not be pursued until it can be done safely. Two hours later it announced an outside group of mathematicians, and gave the reason: an internal model it started training on August 28 has resolved the Navier Stokes problem and more than a hundred other long standing questions, at a pace its own mathematicians did not expect. By evening a Canadian province and a school district had sued Sam Altman personally in a California federal court, over a shooting the company's own reviewers flagged and, the complaint says, chose not to report. Elsewhere Xiaomi took the top of the open weight rankings, Texas froze every state permit sought by a data center, Shopify opened Shop Pay to the Meta shopping agent Amazon had just locked out, and an AI drug company filed to go public promising molecular superintelligence.
The day the frontier reported a breakthrough and a bill for it.
Research
OpenAI says an internal model has resolved the Navier Stokes problem and more than a hundred other open questions in mathematics
It came in the naming of an unpaid advisory group of mathematicians at the Institute for Advanced Study, among them Timothy Gowers and Edward Witten. "On August 28, we began training a new internal model." Since then, "in addition to resolving the Navier-Stokes Millennium Prize problem, this model has now resolved more than 100 long-standing open problems across most areas of mathematics," at a pace that "has surprised the mathematicians within OpenAI." The group "will not be responsible for advising us on how to pace our internal progress."
Policy
A Canadian province and a school district sued Sam Altman personally over a school shooting, in OpenAI's home district
It arises from the February mass shooting at a school in Tumbler Ridge, British Columbia. The complaint says OpenAI monitoring flagged the shooter's account in June 2025 and its reviewers found "a credible, specific risk of harm to others" and recommended a referral to the RCMP. "But OpenAI leadership rejected that recommendation." The account was deactivated; nobody outside was told. One of eight counts is "AIDING AND ABETTING A MASS SHOOTING." Altman later acknowledged OpenAI did not "alert law enforcement to the account that was banned in June."
Agents
Shopify opened Shop Pay to Meta's shopping agent, hours after Amazon locked the same agent out
Shopify's chief executive said the company is "partnering deeply with Muse to enable agentic checkout with Shop Pay on all Shopify stores, offering people an easy and delightful way to shop and check out with Muse." Amazon had blocked Meta's agent the previous night, telling users that continued access by an unauthorized AI agent violated its conditions of use. Catalog sharing and direct checkout between Shopify and Meta have been live since Muse launched on September 8. What is new is Shop Pay itself, and the scope.
Compute
Texas halted every state permit sought by a data center until the grid operator finishes an audit
Governor Greg Abbott "directed the Texas Commission on Environmental Quality (TCEQ) to halt all permits sought by data centers" until ERCOT, the state grid operator, finishes an audit. He added that no state agency may "move forward with regulatory approvals related to data center development" until that information is in hand. His conditions: "Data center projects must cover all electrical infrastructure costs, result in lower residential electrical bills, and complete the ERCOT audit." He will also ask the legislature to end data center incentives.
Frontier
Xiaomi put out an open weight model that VentureBeat says now tops the Artificial Analysis index
MiMo-V2.6-Pro-RL is a "Sparse MoE (Mixture of Experts), 1.02T total / 42B activated parameters" model under an MIT license, and Xiaomi subtitles the series "Scaling Reinforcement Learning Toward Self-Improvement." The weights went up on Monday evening, and outside developers had converted them to other formats within two hours. VentureBeat, reading the Artificial Analysis index, reports the model ties Grok 4.7 and beats GLM-5.3, which would make it the highest scoring open weight model on that benchmark.
Industry
An AI drug discovery company filed to go public and called its approach molecular superintelligence
Iambic Therapeutics filed its S-1 on Monday afternoon. "We call this approach molecular superintelligence." Behind the phrase: "Our net losses were $77.3 million and $47.9 million for the years ended December 31, 2025 and 2024, respectively," and "As of June 30, 2026, we had an accumulated deficit of $245.3 million." Nvidia is listed among the holders of five percent or more. The company says it has raised about $461.8 million since inception and has applied to list on Nasdaq under the symbol IAM.
Policy
OpenAI asked Washington to lead global standards for recursive self improvement, and said fully autonomous RSI should wait
The post argues that "the United States should lead an effort to work together with countries around the world to develop global technical standards for frontier AI, including for RSI," and states that "fully autonomous RSI is not happening today, and we should not pursue it unless and until it can be done safely." It is careful about what it is not asking for: the standards "would not be licenses, mandatory prerelease review, or approval requirements for AI models." It also calls dialogue between the United States and China on this a positive step.