Eliezer Yudkowsky, the fan-fiction novelist who turned AI doom into a bestseller

Eliezer Yudkowsky

Photo: Not credited by the source (published in the If Anyone Builds It, Everyone Dies media kit) / CC0, via Wikimedia Commons

Born	September 11, 1979 Chicago
Occupation	Artificial Intelligence Researcher

When Sam Altman reportedly walked OpenAI staff through the ways the race to superintelligence could end, one of the three scenarios carried a first name. The “Eliezer plan” imagines a single lab reaching self-improving AI first and using that lead to cut every rival off from computing power, and Altman raised it only to call it terrible. The phrase that travelled with the story, melting everyone else’s GPUs, echoes an example Eliezer Yudkowsky used himself in an essay on AI risk, where he offered “burn all GPUs” as an illustration and added at once: “This is not what I think you’d actually want to do with a powerful AGI.”

That a CEO names a strategy after him shows how far his ideas have travelled, and almost all of that travel happened on the page. Yudkowsky is an AI researcher without a degree and, above all, a prolific writer: hundreds of blog essays, a Harry Potter novel of epic length and a bestseller whose title leaves no room for doubt.

He was born in Chicago on 11 September 1979 and raised in a Modern Orthodox Jewish family; he is now secular. Chronic health problems pushed him out of middle school, and he never went to high school or college. In 2000, at twenty, he founded the Singularity Institute for Artificial Intelligence, backed by Brian and Sabine Atkins, with the aim of speeding up the arrival of AI. He soon grew worried about what he had set out to build, and by 2005 the institute was studying how to manage the risks of machines smarter than their makers. In 2013 it became the Machine Intelligence Research Institute, based in Berkeley.

The writing began in earnest in November 2006, when he and the economist Robin Hanson became the main contributors to Overcoming Bias, a group blog about human rationality. In February 2009 his posts became the seed of LessWrong, a community blog devoted to “refining the art of human rationality” that grew into the main online home of the movement that calls itself rationalist. His essays there, known simply as “The Sequences”, went after cognitive biases, Bayes’ theorem and the ways clever people fool themselves. MIRI collected more than 300 of them in 2015 as Rationality: From AI to Zombies, and followed it in 2017 with Inadequate Equilibria, a short book on why societies leave fixable problems unfixed. Stuart Russell and Peter Norvig discuss his proposals in their standard AI textbook, and Nick Bostrom cites him in Superintelligence: Paths, Dangers, Strategies.

Then came the novel. Harry Potter and the Methods of Rationality, HPMOR to its fans, ran as a serial on FanFiction.Net from 28 February 2010 to 14 March 2015, 122 chapters and more than 660,000 words in all. Its premise changes one detail of the canon. Petunia Evans marries an Oxford professor, Michael Verres, and the orphaned Harry grows up homeschooled in science, so the boy who arrives at Hogwarts tests magic with the scientific method and bristles at the wizarding world’s inconsistencies. In Ravenclaw he befriends Draco Malfoy, Hermione Granger and Professor Quirrell. The plot turns darker than its jokes suggest, but the real subject is how to think.

Readers split hard. In 2011 The Atlantic recorded the uproar in fan fiction forums over its “blasphemous—or brilliant—treatment of the canon”, and quoted the science fiction writer David Brin calling it “a terrific series, subtle and dramatic and stimulating”. The legal scholar William Baude went further in The Washington Post: “the best Harry Potter book ever written, though it is not written by J.K. Rowling”. TechCrunch’s Amanda Silberling found it “really tedious”. Yudkowsky refused a suggestion to strip out the Rowling elements and sell the story as an original novel, the route Fifty Shades of Grey took from Twilight fan fiction. “HPMOR is fundamentally linked to, and can only be understood against the background of, the original Harry Potter novels,” he said.

The irony of his career is that the man most associated with warnings about AI also drew people into building it. In February 2023 Altman wrote that Yudkowsky had got “many of us interested in AGI”, helped DeepMind get funded and was “critical in the decision to start openai”. That spring Yudkowsky used an op-ed in Time magazine to call for an indefinite worldwide moratorium on large AI training runs, and Time later put him on its 2023 list of the 100 most influential people in AI.

His most direct book arrived on 16 September 2025. Written with his MIRI colleague Nate Soares and published by Little, Brown, If Anyone Builds It, Everyone Dies argues that modern AI systems are grown rather than crafted, trained into billions of numerical weights that their own developers cannot read, so nobody can reliably set the goals of a system that ends up smarter than us. Humanity, the authors say, would lose to a superintelligence the way people lose to the chess engine Stockfish, and the only safe course is a coordinated halt. The book entered The New York Times bestseller lists for hardcover and combined nonfiction.

The reviews were as divided as the novel’s readers had been. In The Guardian, David Shariatmadari found it “as clear as its conclusions are hard to swallow”. Ian Leslie, in The Observer, praised its “clarity, verve, and barely suppressed glee”. Stephen Marche compared it to a Scientology manual in The New York Times, and Adam Becker in The Atlantic was blunter still: “They are just wrong.”

The argument has since moved to the screen. Yudkowsky is one of the voices, alongside Yoshua Bengio and Timnit Gebru, in The AI Doc: Or How I Became an Apocaloptimist, Daniel Roher’s documentary about becoming a father in the age of AI, which reached Netflix on 15 September 2026. The boy wizard who runs controlled experiments and the co-author who wants the experiment stopped are the same writer, still asking readers to follow a chain of reasoning to its end, however uncomfortable the last link.

Tags: artificial intelligence, writers, Eliezer Yudkowsky, Harry Potter and the Methods of Rationality, LessWrong, Machine Intelligence Research Institute

Eliezer Yudkowsky, the fan-fiction novelist who turned AI doom into a bestseller

More Like This

OpenAI’s Sam Altman lays out three AI endgames, two already on its 2023 list

Yann LeCun, the AI pioneer betting a billion dollars that chatbots are a dead end

Mustafa Suleyman, the Microsoft AI chief racing to build what he says must be contained

OpenAI’s Sam Altman warns the UN: “AI should be built for people, not simply to turn the crank of the machines faster…”

Geoffrey Hinton says the AI on your phone is already conscious

The first generation raised on no-rejection intimacy is starting to fail at work and at love

Discussion