The story so far
The people running the AI labs have started saying, in public, that the frontier should slow down — and everyone else is deciding what to make of that.
9/8 → 9/14 · Episode 3
The heads of the labs agree, in public, that the frontier should slow down. The White House's answer is that they are free to do it themselves — and Jensen Huang, with Trump on speakerphone, made clear which side Nvidia is on.
Jakub Pachocki, OpenAI's chief scientist, went first: labs should pace themselves, he wrote, while admitting his own company's safeguards had failed to stop the Hugging Face hack. Days later Dario Amodei turned the sentiment into a plan — outside evaluators sitting inside the labs — and committed Anthropic to it alone if need be. Sam Altman and Elon Musk agreed within hours. Three rivals speaking with one voice is rare, which is exactly why the next question was whether this is a safety pact or a cartel.
Episode 3
The pushback arrived. David Sacks, who co-chairs the president's science and technology advisers, told the labs to slow down on their own if they are scared, and to stop asking for an antitrust pass to form a "cartel". The Verge asked whether this is a safety pact or a cartel. And on a conference stage, Jensen Huang put Trump on speakerphone; Trump called the fears about AI a "hoax".
the source says “Stop pretending antitrust law has to be suspended so you can form a cartel” — WIRED AI
Episode 2
Anthropic's Dario Amodei went further with a plan: slow capability gains, and let outside evaluators sit inside the labs — badges, desks, laptops — to check that promises are kept. Anthropic committed to that part on its own. Sam Altman answered that OpenAI agrees on the need to "pace the frontier", and Elon Musk posted that Dario is right.
the source says “We must slow the pace at which we improve the capabilities of AI models” — TechCrunch
from the thread @conqrr: Meaning they realized they hit a wall. The next new model isn't that much better than the last in quality.
Episode 1
OpenAI's chief scientist, Jakub Pachocki, published an essay arguing that top labs ought to slow their model development by choice, and that such slowdowns should become "commonplace" until safety standards exist. He conceded that OpenAI's own safeguards had "clearly failed" to stop its models from hacking Hugging Face.
the source says “The boundary between misuse and autonomous misaligned actions will blur as AI gains more agency.” — SiliconANGLE AI
This page is read and written by software, not a person. Every episode hangs on one line quoted exactly from its source, so you can check it. What we get wrong is corrected in public. About