Neal Stephenson joins TBPN on day one of his book tour, discusses AI, filter bubbles, and the fall of Rome

Neal Stephenson on AI, filter bubbles, and the fall of Rome

Neal Stephenson, author of Snow Crash and co-founder of AI startup Wind Air, sat down on day one of his book tour for Heavy Water to discuss the two problems he sees as most consequential for the next century — carbon removal and algorithmic reality fragmentation — and why the second one is harder.

Wind Air and AI-generated worlds

Stephenson is not a passive observer of the AI moment. Wind Air, the startup he co-founded, lets fans tell stories inside fictional universes using AI-generated environments and a script editor. Last week the company announced a partnership with Dreamlands.ai, whose video game environment generation has brought production costs down to a fraction of what a comparable studio build would have cost years ago — Stephenson puts that threshold at "a million dollars plus" for a realistic environment a few years ago.

“To tell a story that intelligent readers will actually enjoy, you need to be realistic, you need to include the actual complexity and ambiguity of the real world. I think the two big problems facing us right now are carbon and people living in their own realities — with carbon, I can see that's an engineering problem straight up. The other problem is harder because people like it.”

The filter bubble problem

Stephenson briefly explored algorithmic reality fragmentation in Fall, or Dodge in Hell, depicting a near-future region he calls Ameristan — a war-torn swath of the country populated by people with no accurate picture of reality who are completely certain they do. He frames it as a dark comedy of epistemic collapse.

He puts filter bubbles alongside carbon removal as the defining challenges of the next half-century, and is more pessimistic about the former. Carbon is an engineering problem with a viable incentive structure — one breakthrough can achieve global adoption because the market rewards it. Algorithmic fragmentation has the opposite incentive structure: the fake content and bot-amplified posts that distort reality also generate more engagement, more time on platform, more ad revenue. There is no clean market signal pointing toward a fix, and government intervention carries its own risks.

His more pointed concern is not fake images — which he says are still legible as suspicious at the current stage of the technology — but the bot-generated reply ecosystems underneath them. Readers assume a real person typed those posts, and that assumption is what makes the manipulation work.

The Roman parallel

On the fall of Rome, Stephenson draws a distinction that has obvious contemporary resonance. The fall of the republic is legible: figures like Marius and Sulla simply decided to discard the norms that Roman political life had depended on, and once that started it could not be stopped. The fall of the empire is a different, slower, more complicated process spanning centuries, driven by migration, religion, and structural decay. His read is that the current moment looks more like the republic's collapse than the empire's — a norm-breaking phase, not a slow institutional erosion.

Writing and AI

Stephenson has no interest in using AI in his own research or writing process. His reasoning is specific: the productive randomness of doing it himself — physically flipping through disorganized handwritten notebooks, stumbling across forgotten notes, making unexpected connections — is not a bug to be optimized away. It is the mechanism. Outsourcing that to a research assistant, human or AI, would be counterproductive. "I already know how to write books," he says, "so I don't need help."

Underexplored territory

The science fiction genre Stephenson wants to see more of is carbon sequestration — biological, industrial, enhanced stone weathering — at civilizational scale. He sees it as a domain currently in genuine engineering ferment, with serious people working across genuinely competing approaches, and almost no fictional treatment of what success or failure looks like.

His underrated author recommendation: Matt Ruff, whose Lovecraft Country became a TV series but whose full catalog, Stephenson argues, is consistently undervalued. Ruff's next book, Black Sheep, is about AI and not yet published.