The Rare Book Thesis for an AI World

I recently read about the sale of a first edition Newton's Principia for $3.7 million, which sounds like a lot of money until you realize that someone paid $23.7 million for CryptoPunk #5822—a 24×24 pixel image of a blue alien wearing a bandana. This is not a judgment about NFTs, but rather an observation about revealed preferences.

Let's do the math: The total value of all ~400 first edition Principias in existence, if we generously assume an average value of $2 million each (most are worth much less), is about $800 million. The total market cap of all 10,000 CryptoPunks at their peak was about $4.5 billion. That means the market valued a collection of CryptoPunks at ~5× the collective value of all surviving copies of the book that fundamentally restructured our understanding of physical reality.

But it gets weirder. The entire rare book market—every first edition of every important work in human history, from Gutenberg Bibles to Shakespeare Folios to Darwin's Origin—is worth maybe $20 billion total. That's only about four times what CryptoPunks alone were worth at their peak. We live in a world where a single NFT collection created five years ago came within striking distance of the collective value of all physically preserved first editions of human knowledge.

To be clear: when I say rare books, I'm referring to first editions of seminal works that permanently altered the trajectory of human civilization—books that became the source code of modernity, like Newton's Principia, Darwin's Origin of Species, Smith's Wealth of Nations, and Galileo's Dialogue. The mispricing feels starker when you zoom out. The global art market commands $2 trillion, while the entire rare book market is optimistically valued at about $20 billion. That's a 100× differential.

The Economics of Rare Books

Books are absurdly easier to store and transport than any comparable alternative asset. You know what happens when you need to move a Basquiat? You hire a team of art handlers who charge six figures, arrange climate-controlled trucks, secure export permits, and pay insurance that costs more than a Lambo. With a first edition Principia, I can just put it in my carry-on and walk through customs like I'm carrying a notebook.

The Netflix series You popularized the notion that rare books require bulletproof cases in climate-controlled underground vaults, but the reality is more mundane: just turn on the A/C and keep them out of direct sunlight. This means that rare books are by far the cheapest to store per dollar of value, and insurance premiums are 2–3× cheaper compared to art or classic cars.

The tax arbitrage is even more delicious. Most jurisdictions treat books like... well, books. There are fascinating quirks in tax laws across different countries. In the UK, for example, rare books have no VAT and are capital gains-exempt. In most European countries, rare books carry substantially less VAT than art, and many offer CGT exemptions with minimal conditions (Germany: hold for more than one year).

The supply dynamics are equally compelling. Every year, more first editions get:

  • Permanently absorbed by institutions (Harvard's library is basically a black hole for rare books)
  • Damaged by people who don't realize their grandfather's old book is worth more than their house
  • Lost to floods, fires, and millennials who throw out "old junk" when cleaning estates

These factors have long been recognized, which is why rare books have been averaging 3–4% real returns, comparable to stamps and art. But what's about to change, and how will these returns look in the future?

The Coming Revaluation

As we contemplate an AI future based heavily on human corpuses, followed by autonomous AIs bootstrapping and recursively self-improving, we have to ask: will that preserve human values? Or will important things be lost or omitted—becoming a photocopy of a photocopy of something that once mattered to us? In a post-scarcity, post-AGI world, meaning and identity become paramount. An aligned AGI would likely recognize that humans derive profound meaning from tangible connections to their intellectual heritage, and that first editions serve as pinnacles of human achievement from before AI's own paradigm dominated. The first Wealth of Nations exists from a pre-capitalist world. The first Communist Manifesto from a pre-communist world. These are messages in bottles from worlds that no longer exist, that can't exist again. An aligned AGI will likely recognize these as the only uncontaminated samples of human thought—ideas from before the ideas themselves reshaped reality—and revalue them accordingly.

Also consider the "culture" of AI itself. An aligned AGI that inherits human values would likely revere its own origins, the same way we preserve fossils and visit historical sites. These books serve as documentation of its own intellectual ancestry. An AGI might value them more than humans do, recognizing them as the primary sources of its own consciousness. If this sounds far-fetched, remember that humans already spend millions preserving cave paintings and stone tools—the artifacts of our own cognitive evolution.

And here's the most poignant point: post-AGI, death might become optional. These books will be the only remaining artifacts from the Age of Death, when all human thought occurred under ultimate constraint. Every great work was written by someone racing against their own mortality.

The Semantic Singularity

We're about to enter what we might call a "semantic singularity"—a point where AI can generate unlimited amounts of meaningful text indistinguishable from human writing. GPT-5 already writes better prose than most humans. When this happens, the value of "content" as an abstract category collapses to approximately zero. Cultural objects will split into two distinct categories:

  1. Pre-Singularity Artifacts: Created under scarcity, bearing the scars of limitation, impossibly precious
  2. Post-Scarcity Productions: Infinite, perfect, meaningless, worth close to zero

This is where first editions become uniquely important. They're the only asset that exists simultaneously as:

  1. Physical artifact (the object)
  2. Semantic content (the text)
  3. Cultural catalyst (what it did to human consciousness)
  4. Temporal proof (when humanity chose to preserve this thought)

In post-scarcity, this unique ontological position makes books impossibly valuable. First editions of seminal works represent the peak moments of purely human discovery. They're physical evidence of when humans were the sole creators of transformative knowledge. In an AGI world, this becomes historically significant in a way that is beyond our current frame of reference. Once AGI surpasses human capability, no human will ever again be the first to discover something truly revolutionary. These books become archaeological artifacts of independent human cognition—the last proofs that our species once led intellectual progress rather than followed it.

Caveats and Metacognition

I might be completely wrong about all of this. Maybe AGI never happens, or happens so differently that none of this matters. But the asymmetry is interesting: the downside is that you own some very nice old books that appreciate at their historical 3–4% real return. The upside is that you own the only authentic remnants of human consciousness from before the phase change—artifacts whose value could explode by orders of magnitude.

In an era of fiat debasement and AI transformation, that seems like a bet worth considering.