I think about this more than is probably healthy. Drop me into 1958 and I do not think I would have hesitated. I would have spent my life trying to keep the United States and the Soviet Union from ending the world, not out of any special courage, but because that was the fight that decided whether there would be a future to argue about at all. Every other cause sat downstream of it.
I did not get that fight. The bomb was built, aimed, and, so far, never used against a city again after 1945. The people who made that "so far" hold, the ones who spent the Cold War building the test bans, the hotlines, and the verification regimes that kept a terrifying standoff from becoming a final one, did some of the most important work of the last century. Most of them will never be famous for it. Prevention is quiet by nature. When it works, nothing happens, and nothing is a hard thing to put on a monument.
The same shape, a new weapon
When I say we are living through a new Cold War, I do not mean it as a slogan. I mean the situation has the same shape. A small number of rival powers, the United States and China above all, are racing to build a technology that each is convinced it cannot afford to let the other reach first. The technology could end the human story if it goes wrong. And no one doing the racing can honestly claim to know how to control what waits at the finish line. Swap the hydrogen bomb for artificial superintelligence and you have described 1958 and 2026 in a single sentence.
I am not the only one hearing the echo. In June, the director of the CIA told an audience in Washington that it would not be misplaced to call the capabilities of today's frontier AI "digital nuclear weapons."
The language of the last existential arms race is coming back, because the structure that produced it is back.
The differences run the wrong way
There are real differences between then and now, and this is the part that keeps me up. Almost all of them make our situation harder, not easier. A nuclear weapon stays in human hands; a superintelligence would not. Nuclear material is physical, so it can be mined and counted. The inputs to AI are chips and code, harder to see and easier to hide, though not impossible to govern. And the Cold War handed its participants one thing we may not get, which is time. The world had years to absorb Hiroshima, to feel the fear in its stomach, and to build institutions in response. AI capability is compounding faster than any institution we have can keep up with.
What the Cold War actually proved
But the same era demolished the one claim the accelerationists lean on hardest, that rivals who distrust each other can never agree to hold back. Washington and Moscow trusted each other not at all, and they still signed binding, verifiable arms-control agreements, because both could see that an unbounded race left them worse off than a rule that bound them equally. The distrust was the reason for the treaties, not the obstacle to them. If two governments could find that logic with tens of thousands of warheads pointed at each other's cities, the confident claim that the United States and China can never do the same on AI is not hard-nosed realism. It is an alibi for continuing.
Why the Foundation exists
So this is my answer to the question I cannot stop asking myself. I did not get to work on preventing thermonuclear war; that was an older generation's, and they carried it. But its equivalent landed in mine, and this time I am here for it. That is what the Nakada Foundation is. Not a think tank cataloguing an interesting risk from a safe distance. A refusal to be part of the generation that built the weapon capable of aiming itself and told itself, the whole way there, that it had no choice.
The people who prevented the worst of the Cold War did it by insisting, over the objections of everyone who called them naive, that some technologies are too dangerous to be left to a race. They were right then. The only task that matters now is to be right again, and to be right in time.
Common questions.
The parallel is structural, not decorative. A small number of rival powers, chiefly the United States and China, are racing to build a technology each believes it cannot let the other reach first, one that could be catastrophic if it goes wrong and that no one yet knows how to control. That is the shape of the nuclear standoff, with artificial superintelligence in place of the hydrogen bomb. Several of the differences, especially the speed of AI progress and the fact that a superintelligence would not stay in human hands, make the new version more dangerous, not less.
The opposite. The Cold War is the clearest proof we have that rival powers who distrust each other can still agree to hold back. The United States and the Soviet Union signed binding, verifiable arms-control treaties precisely because both sides understood that an unlimited race made them less safe. The same self-interest applies to superintelligence: no country benefits from building a system that could destroy its own people, so the shared interest in prevention is real.