Prologue

17:22:51 UTC

At 17:18:04 on the afternoon of 9 May 2017, a wallet that had never transacted before sent its first transaction.

The address was 0x3CC44273a97E8fbFbCBd3d60200CC9Fd33d84d66, the nonce was zero, and the transaction deployed a contract to the Ethereum blockchain at block 3678409. The contract was a factory: a small piece of code whose only purpose was to make other contracts. Nobody noticed. Ethereum was four months away from the ICO mania that would clog it for the rest of the year, ether traded around ninety dollars, and the chain processed a few hundred thousand transactions a day from perhaps a few tens of thousands of people. A new contract was not news.

Four minutes and forty-seven seconds later, at 17:22:51, the factory was used for the first time. The call looked like this:

CreateCard(100000, "Curio1", "CRO1", "CurioCard1", "QmZGrL7HrUmkLuUxp7NV4n1H8tHkurmBAdsx8jzJp4v1tn")

A hundred thousand of something. A token called Curio1, ticker CRO1. A name, CurioCard1. And a string beginning Qm, which is what an IPFS content hash looks like — an address in a distributed file system, pointing at a file that was not on Ethereum and never would be, because putting a picture on Ethereum in 2017 was unthinkably expensive.

The picture was of apples.

The transaction cost 0.0426 ether in gas, about four dollars that afternoon. The cards it created would go on sale for twenty-five cents each.

That is the whole of it, and it is worth dwelling on how much of the story is already present in those five arguments. The supply is a round hundred thousand, because the people doing this had not yet decided how many they wanted and planned to burn the rest — a decision that would make the final supply numbers permanently strange. The symbol is CRO1 rather than CURIO, because each card was going to be its own token with its own contract, there being no other way to do it. The name is in the contract, and the artwork is not: what Ethereum holds is a receipt pointing at a file held somewhere else, which is still, nine years later, how almost every NFT works. And the thing being sold is a picture of apples with an edition number, because the person whose idea this was had collected baseball cards as a child and never stopped.

It was a Tuesday. That was not an accident; it would become the project’s only reliable institution.

The chain is the most precise witness in this story, and almost the only one that does not misremember. But it is not the only record of that afternoon, and the others corroborate it within about forty minutes.

At 19:12 forum time, an account called MyCurioCards posted to the announcements board of Bitcointalk, the forum where cryptocurrency projects were announced in 2017 because there was nowhere else. The thread title was flat and declarative:

[ANN] Curio Cards: Welcome to the future of collectibles

The post opened with a sentence that is, depending on your temperament, either charming or insufferable:

“When the grand experiment that is ethereum began, Vitalik Buterin created a modular, stateful, Turing-complete contract scripting system married to a blockchain.”

And then, with a straight face:

“Yet the past two years have painfully demonstrated a missing feature: digital trading cards.”

Underneath was the specification, as posted: “Type: ERC20-compliant token / Supply: 100,000 Limited / Divisibility: 0 decimal places.” Three ways to buy each card, at 0.0025, 0.0050 and 0.01 ether. A Gmail address for enquiries. And a promise — “Plaidpaper: Coming soon!” — which is a joke about whitepapers, and which was never kept.

At 19:52 UTC a link post went up on Reddit, in a subreddit the team had made themselves, titled “Curio Cards - the future of digital collectibles”. It received a score of one and no comments.

At 19:57 UTC, Thomas Hunt tweeted from the @MadBitcoins account, which had an audience:

“Curio Cards - the future of collectibles https://youtu.be/f8In7X2S0uw available now at http://mycuriocards.com”

At 20:30 he posted on Facebook: “Check out my new digital trading cards at https://mycuriocards.com/”. At 23:12, apparently having decided it sounded better, he posted again: “Check out my new Digital Collectibles company at http://MyCurioCards.com”.

And at 22:05 UTC, a little over four and a half hours after the first card went on chain, the project’s own account announced the first thing that had gone right:

“Early Bird Round #1 for all #CurioCards have SOLD OUT”

Here is the strange part, and the reason for a whole book.

Everything above is documented to the second. The contract deployments are on a public ledger that anyone can read today and that cannot be edited. The forum post has a timestamp. The Reddit post has a timestamp accurate to the microsecond. The tweets have snowflake IDs from which the exact moment of posting can be derived arithmetically, which is how the times in this book were established. There is even a capture of mycuriocards.com’s robots.txt in the Internet Archive from the following afternoon, 10 May 2017 at 21:29:23 — a one-day-after-launch fingerprint for the domain.

And yet no page the project published in 2017 states the launch date. The 9 May date, which would eventually be printed by Christie’s, repeated by CNBC and fixed in Wikipedia, rests entirely on the chain, the forum, Reddit and a robots.txt file. The people involved did not think to write it down, because there was no reason to think anyone would ever want it.

For four years, nobody did. The cards did not sell. The company ran out of money and, in the project’s own later words, “went bankrupt”. One of the three founders stopped working on it within months and by January 2018 was describing it in the past tense on his own show — “the idea that we tried to develop” — while noting that the others were still at it. Another took the buying instructions off the website on purpose, because he did not want anyone spending money on something he believed was worthless. The artists went mostly unpaid beyond their share of a few hundred dollars of sales. In August 2017, at a panel event in San Francisco where two of the three founders were among the hosts, a man with a microphone asked the room how many people had heard of Curio Cards, and the answer was evidently so dispiriting that he moved straight on:

“Okay, well two of your hosts here tonight are both working on a project on a blockchain called Curio cards. So now you know.”

Then, in March 2021, a stranger went looking through the git history of an abandoned website, found the addresses of some contracts that had been quietly removed from it, worked out that they still functioned, bought a couple to see what would happen — and then published instructions so everyone else could do the same. Within a day, every remaining Curio Card had been bought. Within seven months, a complete set stood on a screen at Christie’s in Rockefeller Center as lot one of a two-hundred-lot sale, and the auctioneer took live bids in ether for the first time in the firm’s history, and it sold for 393 ETH.

A little over a million and a quarter dollars, for thirty-one pictures that had cost about twenty-five cents each and that nobody had wanted.

This book is about how that happened, and it is about something slightly more awkward than that.

Because the Curio Cards story as it is usually told — by the auction house, by the press that followed the auction house, by the encyclopedia that followed the press — is a story about being first. The first art NFTs on Ethereum. Christie’s went further than anyone, dropping the qualifier entirely: “On May 9th 2017, the first ever NFT-based art project launched, billed as a kind of virtual gallery opening.” That sentence, from the institution whose authority made the claim stick, is bigger than any claim the founders ever made for themselves.

The founders were more careful. Travis Uhrig, asked about it in May 2021, said:

“I never felt comfortable saying that before the first because you know you never know”

And Thomas Hunt, a week later, said the thing that makes an honest book possible:

“What matters is in the Ethereum blockchain, you can see our transactions. It’s completely provable now. It is the thing. So if there’s one earlier than Curio cards and it’s in the blockchain, I’m not going to fight it. I know what the truth is.”

So we will take him at his word, and go and look. There are earlier things on Ethereum. Some of them are plainly not art and some of them arguably are, and one of them — a user-painted billboard called PixelMap, deployed in November 2016 — has never once been addressed by anyone on any side of the argument, including the people who disagree with Curio’s claim. We will get to it in chapter fourteen.

And there is the footnote.

In July 2018, fourteen months after the launch, with the project visibly dying, Hunt told his audience the thing he was proudest of. Curio Cards, he said, had been mentioned by name in ERC-721, the specification that defined what an NFT is.¹

“It’s super funny because in the ERC 31 spec, it actually mentions CurioCards. There’s some certain flags or something that you can do because they saw the way that Rhett wrote our Curio cards and they liked what he did”

And then:

“So it’s pretty cool where we’re a little footnote in the code of Ethereum forever now. Pretty cool for a little startup type project.”

He was right that they were in the spec. He was right that it was a footnote. Then the caller he was speaking to replied: “Well ERC 31 20 doesn’t matter. Anything running on Ethereum is going to be in trouble in the next three years probably.” And the segment ended.

What the footnote actually says, in the Backwards Compatibility section of the standard, where it has sat unchanged since 6 March 2018, is this:

Note: “Limited edition, collectible tokens” like Curio Cards and Rare Pepe are not distinguishable assets.

Not an inspiration. Not a credit. A definitional exclusion, written by someone else, to mark the boundary of the thing being invented — with Curio Cards named as an example of what fell outside it.

And two sections earlier in the same document, under Rationale, the standard lists what it considers the real thing:

“Examples of existing or planned NFTs are LAND in Decentraland, the eponymous punks in CryptoPunks, and in-game items using systems like DMarket or EnjinCoin.”

CryptoPunks in the Rationale, as an example of an NFT. Curio Cards in Backwards Compatibility, as an example of something that is not one. Same file, February 2018, written by people who had plainly looked at both.

It is a better story than the one he thought he was telling. That is roughly the shape of everything that follows.