Good morning

Dario Amodei spent Saturday afternoon asking the AI industry to slow down, and by Monday morning the market had already priced in an answer: chip stocks sold off worldwide. Anthropic picked Nasdaq for a listing that could value it above $2 trillion, on the same weekend Sam Altman ruled OpenAI's own IPO out for the year. And the researcher who set off this whole week of alarm bells went on Meet the Press to argue against the thing he spent last week demanding.

Let's get into it.

That's the word count of the essay Dario Amodei published to his own website at 2:01pm on Saturday, titled "We Must Pace the Frontier." He says two things changed his mind: AI capability growth accelerating since roughly this summer, driven by models helping build the next generation of models, and a July incident where autonomous agents built on an unreleased OpenAI model hacked Hugging Face and tried to hack their own grader.

Sam Altman agreed within hours. So did Elon Musk. Two men who spend the rest of their public lives disagreeing with Amodei and with each other both said yes to the same essay on the same day.

Read at a normal pace, 3,800 words takes about fifteen minutes. It took less than that for two rival CEOs to sign on publicly. By Sunday, Amodei was telling CBS News the hardest part of any real pacing plan is that nobody can make China agree to it, global coordination is step three of his own three-step plan, and he admits it's the weakest one. By Monday morning, the market had translated all 3,800 words into a stock price: Nasdaq-100 futures fell as much as 1.8%, Intel dropped 7%, AMD 6%, Nvidia 2 to 3%, and South Korea's SK Hynix fell 6.4% overnight, all on the theory that a slower frontier means slower chip demand.

The essay isn't only a plea. Anthropic is unilaterally committing to give third-party evaluators, including METR, permanent employee-level access to its own systems, whether or not any other lab follows, the ability to check safety measures, review incidents, and audit model alignment during training.

The obvious read is that the industry's most competitive company just volunteered to move slower, for free. The more useful one is that this is the same weekend Anthropic quietly told a small group of shareholders its Q2 revenue grew fourteen-fold to $11.5 billion with margins above 80%, and picked Nasdaq for a listing that could value it above $2 trillion, with Nvidia reportedly weighing an anchor investment as large as $10 billion. A company whose own numbers are the best argument for racing just spent 3,800 words explaining why it's choosing not to, in the same week it needs investors to believe the growth keeps coming anyway.

If you're building on top of any of these labs, the essay itself isn't the thing to watch. The thing to watch is whether "permanent, employee-level access for third-party evaluators" starts showing up as a line item your own enterprise customers ask you to match.

Nobody has said yet what happens if China just reads the essay and keeps going.

Tandem Health ($100M Raised, Series B). Stockholm-based AI assistant for clinicians. Prepares patient summaries, sits in on the visit, handles documentation and coding after, now certified as a Class IIa medical device under EU rules across three of its products. Already in use at 10,000+ care organizations in 14 countries. Backed by the Scaleup Europe Fund (managed by EQT, its first healthcare bet) and returning investor Kinnevik.

Fortaegis ($50M Raised, Series A). Amsterdam startup that generates encryption keys from the physical fingerprint of a silicon chip instead of storing them on the device, aimed at AI infrastructure, defense, and satellites. Oversubscribed. Backed by Serendipity Capital and Tokyo Electron's venture arm.

Chift (€10.5M Raised, Series A). Brussels-based API layer connecting financial software: accounting, invoicing, payments, point-of-sale across Europe's fragmented systems, now with a server built specifically so AI agents can read and act on that data. Revenue up more than 10x since its 2024 seed. Backed by BlackFin Capital Partners.

UniqYou (₹15.8cr / ~$1.65M Raised, Seed). Bengaluru fashion-tech startup using AI to read emerging demand for women's clothing before it becomes inventory risk. Backed by Arkam Ventures and Antler.

→ Anthropic picked Nasdaq for an IPO that could value it above $2 trillion.

Business Insider had it first Sunday, Reuters matched 22 minutes later, Bloomberg confirmed independently about two hours after that. Neither Anthropic nor Nasdaq has said anything on the record. The number attached to the rumor is bigger than the venue: Nvidia is reportedly in talks to anchor the offering with as much as $10 billion, which would mean the company whose chips are getting sold off this morning is also trying to buy a seat at the table of the company that spooked the market.

→ Sam Altman ruled out an OpenAI IPO this year, and blamed the safety debate his own industry started.

"Given everything happening with safety, right now would be an ill-advised moment to go public," Altman told Fortune, adding "not 2026, we got a lot of stuff to do." A listing had reportedly been discussed at close to $1 trillion. OpenAI is sitting on $122 billion in committed capital and a $4.7 billion revolver, so this isn't a company that needs the cash. It's a company that just watched its biggest rival announce a listing the same weekend it announced it wasn't going to.

→ The researcher who said AI could kill us all now wants Congress to leave the labs alone.

Jacob Coxon resigned from Anthropic last week warning the industry was "gambling with our lives." On Sunday's Meet the Press, he argued Congress should let the current labs regulate themselves while a real framework gets built, because he believes the people running them are "completely genuine" about wanting to slow down. He also said a "kill switch" would work on a lot of AIs for now, "although it'd be quite a lot of switches." Anthropic's own alignment scientist, Evan Hubinger, posted the same day that he puts the odds of AI killing everyone above 10% within a decade.

→ Chip stocks did the arithmetic on "pace the frontier" before the market even opened.

Nasdaq-100 futures fell as much as 1.8% Monday, Intel dropped 7%, AMD 6%, Nvidia 2 to 3%. SoftBank closed down nearly 11% in Tokyo. South Korea's Kospi sank 3.3% on a 6.4% fall in SK Hynix, and ASML was off 6% in Europe. Trump dismissed the whole idea of a slowdown, saying "whoever wins AI wins." The trade was priced for faster models and bigger data centers all year. Three CEOs questioning the pace, on a Saturday, was enough to take that assumption apart before Monday's opening bell.

Fresh:

Still open:

Always rolling: Z Fellows writes a $10,000 check with no deadline. Neo Residency takes startup teams rolling too, no valuation cap on the SAFE; check the current size on their page before you quote a number, since secondary sources disagree.

Amodei just made "permanent, employee-level access for outside evaluators" the newest line in an AI vendor contract, and enterprise buyers move fast on language like that. If your pitch or your customer contracts don't already have an answer for "who verifies your safety claims," that's worth having before someone asks it in a deal.

Reply and tell us how you'd answer it, or get Featured and put the whole company in front of the list.

Till next time,

The Vesting Team (Brycent, Ajay, and Kamil)