Every Kill Switch Needs A Hand. Bitcoin Was Built Without One.

Over one weekend, the three men who run the most powerful AI labs on earth agreed on something they rarely agree on. Not that AI is safe. That the reason it might not be is that nobody, including them, can fully steer it once it gets fast enough. That single admission, made in public, by the people with every incentive to say the opposite, is worth sitting with longer than the news cycle gave it.

Because the same sentence, aimed at a different technology, has been the standard objection to Bitcoin for a decade and a half.

THE ESSAY THAT NAMED THE REAL FEAR

Anthropic CEO Dario Amodei published an essay called "We Must Pace the Frontier," arguing that AI capability is now improving faster than any single company, evaluator, or government can verify. His own words were direct.

We must slow down the pace at which we improve AI model capabilities.

Sam Altman agreed within hours and committed OpenAI to the same third-party evaluation standard. Elon Musk called Amodei right. Buried inside the proposal was a request for something specific: a narrow antitrust waiver, government permission for the handful of frontier labs to sit in a room and coordinate how fast everyone else is allowed to move. David Sacks, the administration's own former AI advisor, called that part of the plan what it looked like. Stop pretending antitrust law has to be suspended so you can form a cartel, he wrote, and stop pretending the companies asking for the waiver need anyone's permission to simply slow down on their own.

UNCONTROLLED IS NOT THE SAME AS UNSAFE

Notice what the danger actually was in Amodei's own framing. Not that the models fail. That nobody sits above them with a hand on the dial. A system with no CEO to subpoena and no board to lean on is, in this telling, the scary version. A system with a CEO, a board, and an evaluator embedded inside it is the safe version.

That framing only holds if control and safety are the same thing. They are not. A committee that can change the rules is a committee that will eventually change the rules for reasons that have nothing to do with your safety and everything to do with its own interests. The absence of that committee is not a gap waiting to be filled. In a monetary system, it is the entire point.

EVERY FIX ON OFFER NEEDS A HAND ON THE SWITCH

Run the list of proposed fixes from that single week and count how many of them end without somebody new holding power. Embedded evaluators sitting inside the labs. An antitrust waiver so the top three or four companies can coordinate legally. A joint governance body of, in Amodei's own words, the right combination of governments, a body that does not exist yet and would take years to build. A Senate ban drafted before the public even knew there was a debate.

Every proposal assumes the fix is a person or a body empowered to intervene. None of them ask what happens when that same body decides intervention should point somewhere else. A switch built to stop a runaway model is still a switch, and a switch that can stop one thing can eventually be aimed at another. The question worth asking about any centralized fix is never whether the people holding it today mean well. It is who inherits the switch tomorrow, and what they decide it is for.

THE EXPERIMENT THAT ALREADY RAN FOR SEVENTEEN YEARS

Bitcoin has been the live test of exactly the scenario Amodei describes as dangerous. No CEO to subpoena. No server to seize. No board that can vote to change the supply. No evaluator with employee-level access, because there is no employer. No antitrust waiver required, because there is no cartel to legalize. Design a better miner a thousand times over and you still get zero additional Bitcoin, because the rule was never enforced by a person who could be asked nicely to bend it.

Seventeen years is long enough to know what happens to a system with nobody in charge of it. It does not spiral. It does not need a pacing agreement. It settles, block after block, on a schedule nobody negotiates and nobody can accelerate for a fee. The very trait that makes a room full of AI executives ask for permission to build a switch is the trait that has made Bitcoin boring to attack for a decade and a half. Boring, in this context, is the compliment.

WHAT SURVIVES WHEN YOU STRIP THE STORY DOWN

Take away the essay, the names, the IPO timelines, the resignation letter that went viral in a day. What is left is a simple question that predates all of it. Does a system need someone empowered to intervene in order to be trustworthy, or does it need the opposite: a design where no one is empowered to intervene at all, so trust in a person is never required in the first place.

Every fix proposed for AI this month answers that question one way. Bitcoin answered it the other way in 2009 and has not needed to revisit the answer since. A currency with a switch is a currency with an owner of the switch. A currency with no switch has no one to ask permission from, which is the entire reason it never needed a pacing agreement to stay honest.