Power That Holds Still
This week was quite an eventful one.
Last weekend, a 27-year-old Anthropic researcher named Jacob Coxon quit his job and posted a warning on X: tech companies, he said, are gambling with our lives, and the next year or two is crunch time for humanity.
I don't think Coxon was a major player in AI development, but something about his warning set off a chain reaction. Days later, Anthropic CEO Dario Amodei published an essay calling on the industry to slow down and pace AI development. His essay was quickly followed by Sam Altman, the head of OpenAI, and Elon Musk, who also publicly backed a slowdown.
It's not entirely clear to me why these warnings landed at this particular moment as people have been raising them for years. But what matters more, I think, is who's raising them now. These aren't outside critics of the technology. They're the very people building it, now asking us to put the brakes on.
It's easy to read these headlines and feel alarmed, some warning of nothing less than the end of humanity. But to me, this isn't a story about danger. It's a story about discipline.
And we've been here before. There have been other moments when a technology was disruptive and threatening enough to force this same reckoning. Think of 1945, and the atomic bomb: the race to build weapons that could annihilate all life on Earth many times over. Eventually, treaties followed: partial test bans, non-proliferation agreements between nuclear powers. Those treaties weren't perfect by any means. But they were real attempts to restrain what had looked, until then, like runaway power.
I've been giving a lot of thought to restraint this past week. I've come to a simple conclusion: restraint is not the absence of power. It's power, paying attention.
When I think about the situation we find ourselves in, I keep picturing a river flowing down a mountain and picking up speed, starting to threaten the villages at the bottom of the hill. So we build a dam to hold it back.
That dam isn't weakness. It's power that holds still, and power that holds still is still power. A dam built to restrain a river requires more control, more awareness, and more strength than simply letting it run its course. A dam holding back a river isn't weak. It's exercising precise judgment about when to release the power it's holding.
Amodei's proposal calls specifically for giving safety measures, oversight, and cybersecurity time to keep pace with AI's capabilities. He, Altman, Musk, and others will need to work alongside regulators and independent experts to turn that idea into something concrete: a clear-eyed set of actions that actually builds the dam, rather than just describing one.
History will tell whether this week passes with nothing more than essays and headlines, or whether it results in real, concrete limits on how fast AI develops. We don't know yet.
But it's worth remembering, and hopefully agreeing on, this much: restraint isn't weakness. It's what power looks like when it's paying attention.
