I was 23 when I almost died in battle.

It was September 26, 1983, around 9:30 in the evening. I was hunched over a manual typewriter in a rented room in Cambridge, England, pounding out the first chapter of my Ph.D. thesis in archaeology. I had just come back from four months of fieldwork in the Greek islands. My work was going well. I was in love. Life was good.

I had no idea that 2,000 miles away, Stanislav Petrov was deciding whether to kill me.

Petrov was the deputy chief for combat algorithms at Serpukhov-15, the nerve center of the Soviet Union’s early-warning system. He was a methodical man, an engineer, a writer of computer code—and not, fortunately for me, a man given to panic. But when the siren went off a little after midnight (Moscow time), even Petrov leaped out of his chair. A red bulb blinked into life on the giant map of the Northern Hemisphere that filled one wall of the control room. It signaled that a missile had been launched from Montana.

Above the map, red letters came to life, spelling out the worst word Petrov knew: “LAUNCH.”

Computers checked and double-checked their data. Again the red lights flashed, this time with more certainty: “LAUNCH—HIGH RELIABILITY.”

You may not be very interested in war, Trotsky is supposed to have said, but war is very interested in you. Cambridge was—and still is—a sleepy university town, far from the seats of power. In 1983, though, it was ringed by air-force bases high on Moscow’s list of targets. If the Soviet General Staff had believed Petrov’s algorithms, I would have been dead within 15 minutes, vaporized in a fireball hotter than the surface of the sun. King’s College and its choir, the cows grazing as punts drifted by, the scholars in their gowns passing the port at High Table—all would have been blasted into radioactive dust.

If the Soviets had launched only the missiles that they were pointing at military targets (what strategists called a counterforce attack), and if the United States had responded in kind, I would have been one of roughly a hundred million people blown apart, burned up, and poisoned on the first day of the war. But that is probably not what would have happened. Just three months before Petrov’s moment of truth, the U.S. Strategic Concepts Development Center had run a war game to see how the opening stages of a nuclear exchange might go. They found that no player managed to draw the line at counterforce attacks. In every case, they escalated to countervalue attacks, firing on cities as well as silos. And when that happened, the first few days’ death toll rose to around half a billion, with fallout, starvation, and further fighting killing another half billion in the weeks and months that followed.

Back in the real world, however, Petrov did draw a line. He later admitted to having been so scared that his legs gave way under him, but he still trusted his instincts over his algorithms. Going with his gut, he told the duty officer that this was a false alarm. The missile-attack message was stopped before it worked its way up the chain of command. Twelve thousand Soviet warheads stayed in their silos; a billion of us lived to fight another day.

A world like this—in which Armageddon hung on shoddy engineering and the snap judgments of computer programmers—had surely gone mad. People cried out for answers, and on both sides of the Iron Curtain the young turned away from aging, compromised politicians toward louder voices.