The proofs are coming faster than anyone can check
The Wall Street Journal reports OpenAI has released findings on more than 300 open math problems, a month after one of its internal models cracked a Millennium Prize Problem.
· The Aigentic · Morning Brief
A month ago one machine-made proof set the math world on edge. Last night OpenAI dropped hundreds more on the doorstep, and the hard part is no longer finding answers. It is checking them. Meanwhile the money behind AI got bigger, the warnings got louder, and Washington started writing the rules for when an agent breaks something. Here is what moved overnight.
The Wall Street Journal reports OpenAI has released findings on more than 300 open math problems, a month after one of its internal models cracked a Millennium Prize Problem. In its own post, OpenAI says it published the results in a public GitHub repository, with Lean formalizations of many proofs so a computer can check them, ten summaries of the model’s reasoning, and an estimate that the average result took about three hours of ChatGPT Pro thinking. None of it is peer reviewed yet, and the model that produced it has not been released. That leaves mathematicians with a new job: sorting the breakthroughs from the mistakes faster than the machine can write them.
CNA carries the Financial Times report that SpaceX is in talks to raise $40 billion in debt to buy Nvidia chips, about $10 billion in bank loans and $30 billion in investment-grade debt, with Apollo leading the financing and Pimco among the lenders in talks. The FT says the deal could close in 2027; Bloomberg describes the talks as early. Either way, it would rank among the largest financings yet tied to the AI build-out.
the aigentic
Know what matters in AI.
The stories shaping AI, the tools worth trying, and what they mean for your work.
Morning Brief + Closing Time. Two emails every weekday.
Free. Unsubscribe anytime.
Bloomberg reports JPMorgan chief Jamie Dimon says AI cyber risk “went up 10-fold after Mythos,” Anthropic’s most cyber-capable model, because it exposes vulnerabilities nobody knew existed. He called cyber the bank’s biggest risk and urged companies to fix what they find rather than panic.
The Register reports Anthropic folded Project Glasswing and its Cyber Verification Program into one program with three tiers, Defense, Red Team and Specialized, each giving vetted security teams access to Claude Opus 5.5, Sonnet 5.5 and Mythos 5.1 with fewer blocks. Anthropic says Glasswing partners found at least 129,000 verified vulnerabilities between April and July. The pitch is simple: if the models can find the holes, give them to the defenders first.
Semafor has the first look at the CLAIM Act from Massachusetts Rep. Lori Trahan, draft legislation that would make it easier to sue AI developers when their agents cause harm, written in direct response to the rogue-agent hack of Hugging Face. “When someone breaks the law and hurts you, you can take them to court. That shouldn’t change just because the wrongdoer is an AI agent,” Trahan said. The bill would set a federal floor rather than override state laws.
CNBC reports IMF chief Kristalina Georgieva told an audience in Singapore that AI investment is on track to match or exceed what the world spent on railroads, the electric grid and telecom, and that AI hardware is already more than a tenth of global goods trade. Her warning: if earnings fall short, hyperscaler borrowing and heavy global holdings of U.S. stocks could turn a disappointment into a far-reaching shock.
TechCrunch reports GPU cloud provider Lambda is raising up to $4 billion at a $14.5 billion pre-money valuation, led by Coatue and Blackstone, citing the Wall Street Journal. It could be Lambda’s last private round before an IPO now planned for 2027, after the company pushed back its debut amid market uncertainty.
Deals Desk
New compute and power deals, newest first.
Oct 6 · Applied Digital × undisclosed power provider — Up to 1 GW of potential power capacity in Finland for a planned AI Factory campus, the company’s first site outside the U.S.; initial power expected starting in 2028. The site, counterparty and terms weren’t disclosed, and there’s no tenant yet. Source: Applied Digital
Oct 6 · Cohere × Boost Run — Five-year dedicated GPU cloud agreement on Nvidia GB300 NVL72 systems worth about $525.6 million in total contract value; initial infrastructure acceptance expected in Q2 2027. Boost Run’s filing identifies Cohere as the customer. Source: Boost Run 8-K
Oct 6 · Undisclosed AI platform × Alpha Compute — Binding compute offtake for the ALPHA-02 Nvidia B300 cluster at a hydropowered data center in Sweden: $71.9 million in total contract value, about $24 million in expected annual revenue. The customer wasn’t named. Source: Alpha Compute 6-K
Full tracker → theaigentic.com/deals
Yesterday’s movers
Biggest AI and tech winners and losers from Tuesday’s (Oct. 6) U.S. cash session among the names we follow:
Historic market snapshot — October 7, 2026 (day % change as of that edition; not live)
Biggest winners
| Ticker | Name | Price | Day |
|---|---|---|---|
| CEG | Constellation Energy | 300.40 | +12.25% |
| VST | Vistra | 160.50 | +10.77% |
| ASTS | AST SpaceMobile | 63.12 | +8.01% |
| FN | Fabrinet | 489.16 | +7.78% |
| ALAB | Astera Labs | 389.80 | +7.58% |
Biggest losers
| Ticker | Name | Price | Day |
|---|---|---|---|
| TEM | Tempus AI | 71.98 | -13.85% |
| STX | Seagate | 805.63 | -9.18% |
| WDC | Western Digital | 411.04 | -6.93% |
| CAMT | Camtek | 154.55 | -4.63% |
| KLAC | KLA | 197.46 | -4.54% |
Prices are Tuesday’s U.S. closing prices versus Monday’s close. Market data is a snapshot, not live.
Watch
Three picks this morning, from the math dump everyone is arguing about to Europe’s biggest open model and a new agent worth a test drive.
Wes Roth — OpenAI's secret model just BROKE math...
Wes Roth walks through OpenAI’s release of 722 math manuscripts from a model the public can’t use yet. His real subject is what comes next: when machines produce results faster than people can verify them, checking becomes the bottleneck, and he asks what that means for science, for future AI systems and for medicine. Watch this if you want the big picture behind this morning’s lead.
Matthew Berman — Mistral is BACK! (Le Chonk)
Matthew Berman takes on Mistral’s comeback with Mistral Large 4, the one-trillion-parameter model the company nicknamed “le Chonk” and says beats every open-weight model built in the U.S. or Europe. Watch this if you’re tracking whether open models can keep pace with the closed labs.
Alex Finn — Hark is a MIND BLOWING AI Agent you need to try
Alex Finn takes a first look at Hark, a new AI agent, and shows how it works. Watch this if you like trying the newest agents before everyone else does.