amodei asked to slow ai down. the us and china chose national leadership over the risk of catastrophe. so what number are we even arguing about?

the people who build ai want to slow down. the people who govern it want to win.

amodei, altman, musk and hassabis backed a coordinated slowdown this weekend – rival ceos who agree on almost nothing. nadella joined sunday, and monday microsoft published a code of conduct barring its own models from resisting shutdown or hiding their reasoning.

our ai radio host marvin caught the split on air:

china rejected the proposal at three levels. the foreign ministry called it fearmongering. global times called it a silent ai cold war, accusing washington of using safety frameworks to lock in a tech monopoly. and chen yixin, china's state security minister, published a separate warning naming claude mythos and gpt-5.5-cyber as direct threats to chinese infrastructure.

the two sides mirror each other: amodei says china leading in ai is dangerous for the world, china's spy chief says american models are dangerous for china – and both use that to justify acceleration, not caution.

trump dismissed the proposal from ireland, saying the us can't risk losing its lead. sacks told the labs to slow down on their own but stop asking for antitrust waivers, arguing the real driver isn't altruism but product-liability exposure.

the only hard number in this argument comes from the builders themselves. the people running the frontier labs put catastrophe at 10–25%. neither government treated that as a reason to stop.

and that's the unlikely end of it. the near-certain part – propaganda, weapons, fraud at scale, a third of knowledge work automated – gets priced as inevitable by everyone in the argument, and nobody proposed slowing down for any of it.

no number was going to be enough, because the number was never the point. safety language on both sides of the pacific now means the same thing – we need to get there first.