Dario Amodei spent the weekend asking the world to slow down. By Monday, the world had answered — just not with anything resembling agreement, and not by slowing down.
Trump gave the loudest answer, posting on Truth Social that “the only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT,” and naming Amodei directly: he’s “now pretending to be a ‘perfect little angel.’” It’s a step past last week’s general dismissal of “negative forces” — this time the president picked one CEO and mocked him by name, while claiming his administration already holds “tremendous CRIMINAL and REGULATORY power” over the labs anyway.
Beijing answered in its own register: Foreign Ministry spokesman Guo Jiakun called the pacing plea “fearmongering, confrontation and vicious competition” that “serve the interests of no one,” while state media framed it as a Cold War playbook dressed up as caution. Both governments, for once, agree that the other’s reading is self-serving — they just disagree about which country the slowdown is secretly designed to help.
The closest thing to an actual “yes” came quietly. Anthropic, OpenAI, and xAI cosigned AEF-1, a formal baseline standard for how independent evaluators get access, funding independence, and publication rights inside frontier labs — the concrete mechanism underneath Amodei’s essay, arrived at not through legislation but through three competitors agreeing on paperwork. It’s real, and it’s also the version of “pacing” that costs the signatories the least: evaluators with rules, not a rate limit on capability.
Inside the labs, the mood wasn’t procedural. Bilal Chughtai, who worked on AGI safety and alignment at Google DeepMind until July, posted his resignation this week: “AI has the potential to kill us all,” and the industry needs coordination “to avoid this manic race between AI companies.” He’s now at least the sixth safety researcher across three labs to quit publicly making some version of that case since Anthropic’s own threat report crossed the bioweapons line ten days ago. The pattern holds: the people closest to the systems keep leaving, and the systems keep shipping.
Anthropic’s own week undercut its own message hardest of all. It picked Nasdaq for an IPO that could value the company near $2 trillion — and, in the same reporting cycle, signed a six-year, $13.7 billion compute deal with Rum Group, the Trump-linked owner of Rumble and Truth Social’s hosting infrastructure, taking an option on 51 million of its shares along the way. A company whose CEO just asked the industry to slow down and submit to outside scrutiny is, simultaneously, doing business with the family of the president who told him to get lost. Neither fact cancels the other out; they just sit there, awkwardly, in the same prospectus.
Even the internet’s reaction to the reaction had layers. “Dario, Please” hit the top of Hacker News, arguing Amodei’s plan would regulate open-weight models out of existence while quietly carving Anthropic and its peers an antitrust exemption — a critique from the crowd that’s usually sympathetic to slowdown arguments, not hostile to them.
Nobody in this story is actually pacing anything. Trump is accelerating on principle, China is accelerating on suspicion, Anthropic is accelerating on a six-year contract, and the three labs that did agree to something agreed to the version that changes the least. The slowdown call didn’t fail to land. It landed everywhere at once, and everyone used it to keep doing exactly what they were already doing.
