//
Discussion · 2026-09-13

The three CEOs agreeing is the tell: can the slowdown even be coordinated?

Jared Kubin called the weekend's coordinated slowdown rhetoric “the tell”: if Anthropic stops, OpenAI…

The Debate

For vs against
Essence
Jared Kubin called the weekend's coordinated slowdown rhetoric “the tell”: if Anthropic stops, OpenAI continues; if every US lab stops, China continues — so only government coercion can enforce simultaneous compliance, which he reads as the limit of the CEOs' conviction. Anxur Capital sharpened the game theory — self-regulation fails as it does for nukes and bioweapons — while the counter-camp, led by Nico Deva and mag_pl, argued the calls are perfectly rational strategy: the training arms race is destroying profits, compute costs are too high, and regulatory limits would guarantee profits and smooth the Anthropic IPO. At stake: whether “pace the frontier” is an enforceable safety regime, an unenforceable gesture China will defect from, or cover for capex-cost relief.
For
  • “inference is very profitable but the training arms race is destroying profits, regulatory limits would help guarantee profits and successful IPOs” (Nico Deva, @NicoDeva_)
  • “Nah, costs of compute are just too high, you need some new narration and more $” (mag, @mag_pl)
  • “So far no evidence that suggests anything remotely close to that. It's always easy to play the what if game” (Michael Rosmer, @MichaelRosmer)
Against
  • “If ANT stops, OAI continues. If OAI stops, xAI or GOOG continues. If every American company stops, China continues. THEREFORE, "only govt can force simultaneous compliance…monitor cheating…control"” (Jared Kubin, @JaredKubin)
  • “what good would it do if they stopped and China or other US lab didn't? in that scenario they would have stopped for nothing vs. going ahead and hoping they can somehow have a pretty aligned model. self-regulation wouldn't work as it doesn't for nukes or bioweapons” (Anxur Capital, @anxurcapital)
  • “Also we can't trust china to impose or follow any regulations / laws. Likely they (Dario and Sam) are scared of open source and want the monopoly” (Scalp Trading, @ScalpsTrading)

View the thread on X

The Tweets

4 posts
JKJared Kubin@JaredKubinX Post

Thinking about it more while my kids swim … the 3 ceos agreeing is the "tell" If ANT stops, OAI continues If OAI stops, xAI or GOOG continues If every American company stops, China continues THEREFORE, "only govt can force simultaneous compliance…monitor cheating…control"

17 Reposts · 163 Likes · View on X →
ACAnxur Capital@anxurcapitalX Post

what good would it do if they stopped and China or other US lab didn't? in that scenario they would have stopped for nothing vs. going ahead and hoping they can somehow have a pretty aligned model. self-regulation wouldn't work as it doesn't for nukes or bioweapons

2 Likes · View on X →
NDNico Deva@NicoDeva_X Post

inference is very profitable but the training arms race is destroying profits, regulatory limits would help guarantee profits and successful IPOs

8 Likes · View on X →
MAmag@mag_plX Post

Nah, costs of compute are just too high, you need some new narration and more $

1 Likes · View on X →