A Note on Sources

This book has an unusual advantage and an unusual problem, and they are the same thing.

The advantage: Thomas Hunt was broadcasting throughout. From April 2013 onward he hosted Mad Bitcoins, World Crypto Network and The Bitcoin Group, several shows a week, for years. When Curio Cards launched in May 2017 he talked about it on air. When it failed he talked about that too. When it came back in 2021 he recorded nine hours of interviews with the people who had built it, and then did it again on every anniversary. The result is something almost no failed startup has: a continuous contemporaneous record, in the founder’s own voice, of a project nobody was paying attention to.

The problem: that record exists as roughly 2,800 machine transcripts. They have no speaker labels and they contain real errors — names especially. This is said once, here, and then the book gets on with it.

The rule that follows is absolute: nothing is silently corrected. Every quotation in this book appears exactly as its source has it, errors included. Where an error would actually mislead — a name rendered unrecognisably, or a word that reverses a sentence’s meaning — there is a footnote. Where the meaning is plain despite the imperfection, it stands unremarked, and the reader is trusted. Tidying quotes into what someone probably meant is how invented quotations get born, so it is not done here at all.

Two consequences, also said once. Speaker attribution is often inferred — in a two-person interview, short questions are the host and long answers the guest — and where it matters and rests on inference, the inference is stated; where eight voices are in a room, this book says so rather than guess. And the transcript is not the recording: every spoken quote carries its episode and date so the audio can be checked, and for anything load-bearing — a price, a date, a number of cards — the chain or a document is cited instead, with the recollection set beside it.

A full glossary of the recurring errors, and the verification method, are in the quote bank at the back for anyone who wants them.

Four kinds of material appear in this book, and the reader should always know which one they are reading.

Thomas Hunt’s words at the time. Tweets with recoverable timestamps, Facebook posts, and show transcripts from 2017 and 2018. This is the most valuable material in the archive because it was not recalled. A man explaining his project in June 2017 to an audience that does not care is a different and better witness than the same man explaining it in 2021 to an audience that suddenly does.

Thomas Hunt’s words later. The 2021 interviews and the anniversary retrospectives. Rich, vivid, and — on numbers especially — frequently wrong. He gets the eclipse date wrong three times across three years. He guesses Robek’s card supplies and is corrected on air, in real time, by Travis Uhrig reading the official page. None of this is a knock on him; it is what memory does over nine years. But it means his recollections are treated as testimony, not as fact, and checked against the record.

Other people’s words, sourced. The artists, Travis Uhrig, Rhett Creighton’s single on-record statement, Adam McBride, Christie’s specialist Noah Davis, journalists. Every one carries its source. No quotation in this book is invented, reconstructed from gist, or attributed to a person who cannot be shown to have said it. Where a widely circulated quote could not be verified — and there is at least one, a line attributed to McBride via Quartz — it is named as unverified and not used.

Documented fact. The Ethereum blockchain, which is the best witness in the story and does not misremember. The ERC-721 standard text and its git history. Archived 2017 web pages. The Bitcointalk announcement thread. The Christie’s lot listing. These settle what can be settled.

And then there is my own prose, which is everything not in quotation marks and not attributed: the connective tissue, the sequencing, the arguments about what things meant. It is distinguishable by not claiming to be anyone’s words. Where it makes an inference, it says so. Where the evidence runs out, it stops.

A word on the disagreements, because there are many and they are not a defect.

Reconstructing Curio Cards from this archive turned up fourteen substantial places where the record and the memory diverge, or where two participants remember differently, or where a published source contradicts a primary one. They are numbered and set out at the back. Smaller disagreements — mostly single figures, and there are a dozen or so more — are recorded at the point where they arise. Christie’s states in print that the artist Marisol Vengas is “an anonymous artist collective”; Max Infeld says on air that Marisol Vengas is him. The project’s own documentation says card 30 celebrates an eclipse on 21 September 2017; the eclipse was 21 August and the contract went up on the 23rd. Four different numbers circulate for how many complete sets of Curio Cards can exist, and all four are defensible, and none are interchangeable.

In every such case both accounts appear here, with their sources, and the book says which is better evidenced and why. Where neither is, it says that instead. Early NFT history is full of genuine gaps — the word “NFT” did not exist in its current sense until the standard that excluded Curio Cards from it — and an honest gap is worth more than a confident guess.

Separately from the divergences, there are fifteen things this research could not establish at all — an unread tweet, a dead auction page, a company registration never pulled. Those are listed at the back too, under what could not be established, so that nobody has to guess at them twice.

All quotations drawn from show transcripts in this book were verified by exact substring match against their source files before use. Quotes that failed that check were discarded rather than repaired.