Neal Stephenson on AI, the Atomic Race, and Why the Primer Is Hard

Neal Stephenson tells Ashlee Vance why the AI boom blindsided him, why he still thinks it is an S-curve, and why building the primer from The Diamond Age is harder than it looks. In this episode of the Core Memory Podcast, the author of Snow Crash and Cryptonomicon talks about his new novel D: Heavy Water, the parallel between the race for the bomb and today’s AI race with China, why he writes with a fountain pen, the long wait for a Snow Crash movie, his Jane Austen in Fortnite project Whenere, and what he learned in the earliest days of Blue Origin with Jeff Bezos.

TLDW

Stephenson admits he downplayed AI for decades and is now catching up, but he expects it to flatten into an S-curve like almost every technology except nuclear weapons. His new book, the second in the Bomb Light series, shows how fragmentary intelligence about German purchases of heavy water and uranium launched the Manhattan Project, and he sees the same “you cannot know, so you cannot slow down” logic in the AI race with China. He explains why hopeful science fiction takes harder imagining than dystopia, why the insurance market killed radical launch ideas at Blue Origin, and why a real-world Young Lady’s Illustrated Primer has to start from what teachers actually need. He closes by describing his storyteller’s method: embrace the full complexity of a technology, including its business plan and its misuse.

Thoughts

Every AI company that cites The Diamond Age as inspiration should hear his answer about the primer. The question came in six or seven times from Vance’s audience, and Stephenson says it is now hot the way the metaverse was five years ago. His response is not to sketch the product. It is to read the teachers’ subreddit, where people in real classrooms describe problems “none of which can be solved by giving them a piece of technology.” Then he takes apart Vance’s own best example, a 22-year-old who built a fusor in his bedroom with Claude as his guide. That kid, Stephenson says, would have succeeded a hundred years ago. He calls them Abraham Lincoln kids, after the boy who scratched his lessons into the soot on a shovel: they cannot be stopped, and they do not need help. The honest version of the primer serves everyone, and the moment you take that seriously you are back to small pilots and the long record of failed edtech. That is a sharper critique of “AI tutor for every child” pitches than most education researchers manage, and it comes from the man who invented the idea.

His explanation of why science fiction defaults to dystopia is the most useful thing in the interview for anyone who wants the genre to inspire builders again. He traces it to the last shot of Planet of the Apes: a model of the Statue of Liberty, kicked over, with dirt thrown on it. A powerful image, and a cheap one. Take something that exists, bang it up, and you have a future. A hopeful future has no such shortcut; someone has to imagine the whole working world, the way Gene Roddenberry’s team did for Star Trek. That reframes the complaint Silicon Valley keeps making to writers (Stephenson says investors have asked him about it multiple times in recent months, and Vance describes a quiet 2024 effort to get optimistic AI stories onto Netflix). The problem is not that writers are gloomy. It is that dystopia is the path of least resistance for studios, and optimism costs real imaginative work. His own answer was Hieroglyph, the anthology he did with Arizona State after Michael Crow told science fiction writers they were letting down the side.

The Blue Origin section contains a lesson that applies far beyond rockets. In the early years the team studied laser launch, electromagnetic propulsion, ramjet accelerators, the Slingatron centrifuge, even giant steel bullwhips. Then came what Stephenson calls a cold-water moment: the space business was mostly launching $2 billion satellites to GEO, the rocket was a round-off error, and no insurer would cover a payload on a radically new launch system. Every week of delay was a week of lost revenue for the satellite owner. As he puts it, “why not just pay Lockheed to send the thing up.” A brilliant idea that cannot be insured never reaches the market, which is why he assumes Elon Musk put a car on the first Falcon Heavy. It also explains why the idea he still champions, Geoffrey Landis’s tall launch tower, is framed as something to build for its own sake and monetize many ways, not just as a cheaper path to orbit. And it adds weight to the tortoise supporters he suggested for Blue’s coat of arms: the company chose to be slow on purpose, then watched SpaceX run away with the market.

His Manhattan Project parallel is more careful than the usual “AI is the new bomb” line. In early 1940 IG Farben tried to buy all the heavy water from the one plant in Norway that made it, and after taking Belgium, Germany ordered the uranium output of the Belgian Congo. With only those facts, Stephenson argues, launching the Manhattan Project was the only rational move. We now know the German program never came close, and a uranium cube from their reactor just turned up in a garage, but nobody could know that during the war. That is his model for the AI race with China: the pause button is impossible not because anyone is reckless, but because the information is never good enough to justify it. He applies the same structural view to the lab leaders, who have fiduciary duties and “got to do what they got to do,” and he politely declines to judge them as individuals.

The closing answer explains why the whole interview feels less oracular than people expect from him. Vance admits he thought Stephenson would be a latest-and-greatest tech obsessive; instead he is skeptical of BCIs, prefers a fountain pen, and refuses to call utopia or dystopia. Stephenson says that is the novelist’s method. To keep a reader turning the page you have to embrace the full complexity of a technology: how it fails, how it succeeds, how society misuses it, and what the business plan is. Skip any of that and the reader senses something false and puts the book down. It is a good test for any forecast about AI. The two problems he actually worries about, too much carbon in the atmosphere and the fact that “people cannot agree on what is real,” are both problems of a society interacting with technology, not of the technology alone.

Key Takeaways

D: Heavy Water is the second book in Stephenson’s Bomb Light series after Polostan, covering the 1930s physics, espionage and coincidences that set up the Manhattan Project.
Stephenson says AI blindsided him. He spent decades treating it as “just software” and only recently started using it, though not for writing.
He is “more of an S-curve guy.” The only technology he can name that never hit a ceiling was nuclear weapons, which went from tabletop experiments to the 50-megaton Tsar Bomba in about 15 years.
The Germans’ attempts to corner heavy water and Congolese uranium in 1940 forced the Allies to assume the worst. He sees the same logic in the US and China AI race.
Lab leaders face fiduciary duties that push business decisions over ethical ones. “They may not necessarily be amoral people,” he says, but they are constrained by the role.
He writes with a fountain pen because thinking includes the body, from the motor cortex down the arm to the hand, and he struggles to see what brain computer interfaces offer him outside medical uses.
His two big fears are too much carbon in the atmosphere and people no longer agreeing on what is real. Anything that addresses those interests him.
Dystopian films are cheap to imagine; hopeful futures need hard imaginative investment, which is why studios drift toward dystopia.
A Snow Crash adaptation has been stuck for decades. The rights sit with Paramount, and Stephenson notes that every delay made the metaverse easier to film convincingly.
Whenere lets fans tell stories in fictional worlds, starting with Pride and Prejudice in Unreal Editor for Fortnite, using AI for scripts and dialogue rather than generated video.
He was at Blue Origin from 1999 to the end of 2006, left once it became an engineering organization, and proposed the tortoises on its coat of arms.
Radical launch ideas were uninsurable for expensive GEO satellites. He still wants someone to build a tall launch tower, which could roughly double payload to low Earth orbit by Geoffrey Landis’s math.
He now trains with the basket-hilted broadsword, following the Elizabethan master George Silver.
Kevin Kelly’s point that the Amish choose technologies selectively applies everywhere now, from messaging apps to people who refuse to use AI and debates over screens in classrooms.
A real primer must start from teachers’ actual problems and serve everyone, not gifted kids who would succeed anyway.

Chapters

1:53 D: Heavy Water and the Atomic Race

Stephenson introduces D, the second book in the Bomb Light series, which picks up where Polostan left off. The series covers developments in 1930s European physics that teed up the Manhattan Project, and doubles as an espionage and action story because those discoveries coincided with the rise of the Nazis and the start of the Second World War. He also explains why he does not mind modern book tours: two compressed weeks every few years, aimed at the stores that move bestseller numbers.

8:21 AI Caught Him Off Guard, and the S-Curve View

He says the AI revolution blindsided him after decades of hearing the term used for things that did not work yet. Novelists become instant experts and then walk away, so people assume he knows things he stopped following 30 years ago. Asked about the singularity, he says most technologies become S-curves, and the exception was nuclear weapons. He notes that job anxiety makes the curve feel steeper than it may be.

15:27 AI’s Manhattan Project Moment

Vance compares the AI labs to Lord of the Rings characters chasing a ring they fear, with companies now more capable than governments. Stephenson tells the story of IG Farben’s 1940 bid for all of Norway’s heavy water, the French banker Jacques Allier’s evacuation of it, and Germany’s grab for Congolese uranium. With that information, the Manhattan Project was the only option, even though the German effort turned out to be far behind. He maps the same dilemma onto AI and China.

28:32 Brains, Bodies, Words, and What Worries Him

Stephenson says everything that makes him happy is about words, and that writing by hand is part of thinking. He wonders why athletes like Michael Jordan and Caitlin Clark stick their tongues out at peak concentration, and what it means to bypass the tongue, the organ humans repurposed for speech. Vance, who is filming a Neuralink documentary, describes wanting to share feelings directly. Stephenson worries about repeatable bits hitting pleasure centers and names carbon and shared reality as his real concerns.

40:11 Can Fiction Inspire Progress?

Vance raises Stephenson’s 2011 essay Innovation Starvation and Jonathan Huebner’s paper on slowing invention. Stephenson recounts Michael Crow’s challenge to science fiction writers, the resulting Hieroglyph anthology, and his own tall tower story. He recently judged an XPRIZE film competition for more encouraging portrayals of technology and AI, and explains the Planet of the Apes theory of cheap dystopia.

49:56 Why No Snow Crash Movie, and Jane Austen Meets Fortnite

Each time a Snow Crash film stalled, Stephenson was glad, because a few years later the graphics would have looked like the original Tron. The rights have long sat at Paramount. He then explains Whenere, which lets fans tell their own stories in fictional worlds. It runs on Unreal Editor for Fortnite rather than generative video, uses public domain material like Pride and Prejudice, and uses Claude to turn “tell Darcy he sucks” into convincing Regency dialogue. It is heading into early playtests.

1:03:21 Inside Early Blue Origin

Stephenson was there from the 1999 inception to the end of 2006, when the company launched its first rocket-powered vehicle in West Texas. Once Bezos chose to build rockets, Blue became an engineering organization and Stephenson felt he could walk away for 20 years and not be missed. He talks to Bezos a few times a year and says his lasting contribution was the tortoises on the coat of arms, a message that Blue would keep its own pace while SpaceX raced ahead.

1:13:43 Rockets, Space Guns, Giant Towers, and Insurance

Blue studied space guns, the Slingatron, a University of Washington ramjet accelerator and steel bullwhips, but the tall tower stuck with Stephenson. Launching from altitude allows vacuum-optimized engines, less acoustic and max Q stress, and by Landis’s numbers about twice the mass to LEO. He is impressed by SpaceX, but says radical rockets struggle because expensive satellites need insurance and cannot tolerate delays.

1:23:05 Swords and Selective Technology

In the rapid-fire round, Stephenson explains his focus on the basket-hilted broadsword, a single-handed weapon used from the 1500s to roughly the First World War, which he took up partly because longswords kept breaking his hands. On the selective technology idea from his books, he credits Kevin Kelly’s observation about the Amish and points to today’s choices about messaging apps, AI refusal and classroom screens.

1:27:15 Can AI Fix Education? Storytelling Versus Prediction

The primer from The Diamond Age is now a hot topic, and Stephenson says he studies what teachers need, which is rarely technology. A tool for gifted kids is easy; one that works for everyone is hard. He ends by explaining that his ambivalent answers come from a novelist’s discipline of making fictional technology add up, including its failures, its social effects and its business plan.

Notable Quotes

“The whole AI revolution sort of blindsided me in a lot of ways.”
Neal Stephenson, on coming late to AI after decades of downplaying it

“I’m more of an S-curve guy.”
Neal Stephenson, asked whether we are living in the singularity

“I write with a fountain pen on paper.”
Neal Stephenson, on why he cannot get traction on bypassing the body with BCIs

“I feel like I could walk away for 20 years and I wouldn’t be missed”
Neal Stephenson, on leaving Blue Origin in 2006

“we could, you know, come up with the coolest new idea imaginable and it would never go anywhere commercially because it would be uninsurable.”
Neal Stephenson, on the cold-water moment for radical launch concepts

“none of which can be solved by giving them a piece of technology.”
Neal Stephenson, on what teachers say they need, asked about building the primer

“the first duty of a storyteller is or a novelist is to keep people turning the page.”
Neal Stephenson, on why he will not simply predict utopia or dystopia

Watch the full conversation here.