The Wrong Apocalypse

Why we panicked about the robots and forgot about the weather

In September 2025, Eliezer Yudkowsky and Nate Soares published a book titled, with a restraint that has become the genre’s trademark, If Anyone Builds It, Everyone Dies. It became a bestseller. Six months later, in March 2026, the Harvard Book Store hosted Soares for an evening discussion of the book’s argument that artificial superintelligence will, with near-certainty, kill us all. Researchers ran a survey of attendees afterwards. 96 per cent agreed that mitigating the risk of AI extinction should be a global priority. The median estimate of the probability of human extinction from AI was seventy per cent.

That same month, Gallup’s annual environment poll found that forty-four per cent of Americans were worried “a great deal” about climate change, near the highest figure ever recorded. The American March of 2026 was the warmest in the country’s history, 9.4 degrees Fahrenheit above the twentieth-century average. The Ipsos international report, published a few weeks later, confirmed what the Pew data had been suggesting for years: climate concern remains high across thirty-one countries, but the proportion of people who believe their own actions matter has fallen in every single one of them since 2021.

The thing that is already happening, the one with the temperature records and the wildfires and the homes turning to mud, has not lost its grip on public concern. People are worried. They have been worried for some time. What has quietly collapsed is the sense that anything they personally do about it matters. The thing that is hypothetical, contested, and entirely speculative, by contrast, has produced a bestseller, a moral panic, and a Harvard audience prepared to assign it a seventy per cent chance of killing them all.