A year ago, Eliezer Yudkowsky and Nate Soares dropped a book with the subtlety of a brick through a window: *If Anyone Builds It, Everyone Dies*. Not 'might.' Everyone dies. It hit bookshelves as an instant bestseller and landed on the New York Times best-seller lists. The thesis? Building superhuman AI — by anyone — is a species-ending move. Yudkowsky reportedly reached that conclusion back in 2003, and he's been ringing the alarm ever since.
But here's the twist: the one-year anniversary isn't being used to re-litigate the book's claim. It's being used to argue about what to *do*. On March 30, 2026, 80,000 Hours published 'If someone builds it, does everyone die?' — walking through a fictional scenario from the book where an AI company makes a fateful decision and things unspool. That same day, Soares appeared with Greg Kestin in a talk billed as a 'clarion call for humanity.' The authors have moved from publishing a warning to campaigning. That's a big shift.
Meanwhile, the backlash has sharpened. A July 2026 essay on ideatrash.net calls the book 'an unnecessary bait-and-switch.' But read the fine print: the critic concedes the authors might be 'absolutely correct' and that the risk is 'a far sight closer than what the authors present.' That's not denial. That's frustration with rhetoric that overshoots a position the critic half-accepts. And the counter-attack now has a slogan: 'If Nobody Builds It.' The argument has flipped from *can we survive AI?* to *can we survive not building it?*
Why should you care? Because the stakes aren't hypothetical. The Guardian noted in September 2025 that post-ChatGPT investment in the datacenters that would power superintelligent AI is counted in the hundreds of billions. That's not a thought experiment. That's infrastructure. And as OpenAI builds advertising into ChatGPT with 'Sponsored Agents' — AI helpers that recommend and buy products for you — the mundane and the existential are colliding. One year in, the book's real legacy might not be its warning. It might be the fight it started about who gets to decide.
Trends suggest the next year could bring more public-facing events from the authors, more critical essays, and possibly a formal rebuttal. The question isn't whether Yudkowsky and Soares were right. It's whether anyone is listening — or whether we're all just too busy letting our AI agents buy us stuff.