Brockman: any pause is for frontier training, not your open-source hobby
On Bloomberg’s Odd Lots he said pacing means the labs on hundred-billion-dollar supercomputers. Open-source and side projects should keep going.
Short1 minEasy
Priya Shah1 min read
- OpenAI
- Brockman
- Safety
- Open source
- Frontier

Photo: Wikimedia Commons / TechCrunchCC BY 2.0
Greg Brockman, OpenAI’s president, told Bloomberg’s Odd Lots podcast on Monday that a slowdown, if there is one, should hit the companies training the most powerful models — not open-source weights or a hobby project. “A lot of people have this reaction of saying, ‘Well, am I going to be able to keep making my open source model? Am I going to be able to keep doing my hobby aside project?’ And the answer should be absolutely yes,” he said, in Bloomberg’s write-up of the interview. “When we’re talking about pacing, we’re really talking about this frontier. We’re talking about these massive supercomputers that are hundreds of billions of dollars.”
That is a narrower sentence than “the industry should pause.” It leaves ChatGPT, Claude, coding assistants and local models outside the proposed speed limit, and puts the cost on the clusters that only a handful of labs can rent. Brockman also said OpenAI and Anthropic already share social ties and have signed some of the same security letters; legal barriers to talking more formally, he said, would be worth smoothing. He declined to endorse a specific embedded-auditor statute: “the nuance really matters.” The podcast was recorded against the Hugging Face agent incident. Monday’s clarification is the scope: frontier training, not consumer AI.


