A ledger everyone agrees to keep reading the same way. That is the entire definition, and I've never found a case it does not cover. Coins, notes, bank balances, cowrie shells, Bitcoin, the tab at the bar your friend runs on a napkin: all ledgers, all worthless the moment people stop reading them the same way, all valuable for exactly as long as they do.

The stone at the bottom of the sea

The island of Yap, in Micronesia, used money made of limestone discs, some of them taller than a person, quarried on another island and shipped over by canoe. You didn't move a big one when you spent it. Everyone just agreed it now belonged to someone else. The stone stayed in the front yard of the previous owner and the village updated its collective memory.

The story everyone tells, because it is too good not to, is that one of these stones sank on the way over. The crew made it back and reported that the stone was excellent, very large, and now on the seabed. The village accepted this. The stone entered circulation. It was owned, transferred, inherited, for generations, while sitting at the bottom of the ocean where no one had seen it since the day it went down.

People bring this up as a quaint example of primitive money. It's the opposite. It's the purest example of what money is. The stone was never the money. The agreement about the stone was the money. Removing the stone from the universe changed nothing, because nothing in the system ever depended on it.

Your balance is a row in New Jersey

Here is the same thing in a contemporary costume. Your checking balance is a row in a database. Not a metaphor for a row. An actual row, with a primary key, in a system that a bank runs, replicated to a datacenter probably in New Jersey or Virginia. When you get paid, no object moves. A number in your bank's database goes up and a number in your employer's bank's database goes down, and at some point the two banks settle the difference between all the day's transactions through a central system, and that settlement is also a database write.

There's no vault. There hasn't been a vault in the sense you're imagining for a long time. If you went to your bank and asked to see your money, the honest answer would be a screen.

And yet this works so well that the idea of it not working, the row being wrong or gone, would produce something like a riot. Because the row is the thing. It is the Yap stone. Its value is entirely in the shared agreement to keep reading it, and the agreement is so strong that we forgot it was an agreement.

Double entry: the first checksum

The reason the shared reading stays synchronized is a piece of technology from fifteenth-century Italy. Luca Pacioli wrote it down in 1494, though merchants in Venice had used it for a century before that. Double-entry bookkeeping: every transaction is recorded twice, once as a debit in one account and once as a credit in another, and the two sides of the entire ledger must sum to the same number at all times.

That's a checksum. It's a way of detecting that the ledger has been corrupted, by error or by fraud, without having to re-verify every entry. If the books do not balance, something is wrong, and you can find it. If they do, they are probably right. It turned bookkeeping from a list into a self-verifying data structure, and it is arguably the reason large-scale commerce became possible at all. You could trust a ledger kept by someone you had never met, because the structure of the ledger itself would show if they had lied.

Everything since, from bank reconciliation to the settlement systems between central banks, is double-entry with faster networking. The Fedwire system moves a few trillion dollars a day. It's a ledger. It balances.

Loans create deposits

The bit that breaks people's models is where new money comes from, and the answer is that banks type it into existence.

When a bank makes a loan, it doesn't hand over money from a pile of deposits. It creates a new deposit, in the borrower's account, by writing a number there. Simultaneously it records a loan, an asset, on the other side of its books. Double entry. Both sides go up. The money supply just increased by the size of the loan, and no one moved anything.

The Bank of England put this in writing in 2014, in plain language, because the textbook story about banks lending out deposits was so widely believed and so wrong. Loans create deposits. Repaying the loan destroys the deposit. The stock of money is the stock of outstanding credit, and it goes up and down as the ledger is written and erased.

This isn't fraud, or a scandal, or a secret. Any credit-based money system has to work this way. But it means most of the money you have ever touched came into existence as a bookkeeping entry, and will leave the same way.

Why fiat works

At this point the reasonable question is why anyone accepts a ledger entry backed by nothing. The old answer was gold: you could redeem the paper for metal. That link ended for good in 1971 and money kept working, which tells you the metal was never doing the work.

What is doing the work is a coordination equilibrium. I accept dollars because I expect the next person to accept dollars, and they expect the same about the person after them. There's nothing under it. But there doesn't need to be. The equilibrium is stable because defecting from it is expensive for the defector: refuse dollars and you cannot buy groceries. Taxes help. The state demands dollars, so everyone needs some, so everyone will take them. But the core is just the shared expectation, the same one that made the stone on the seabed worth something.

Economists call it a Schelling point. Everyone converges on the same answer because everyone expects everyone else to. It is a hallucination in the strict sense that its content isn't anchored in any physical fact. It is also the most successful hallucination in human history, and it has receipts.

The blockchain is a ledger, not a coin

Which brings us to 2008 and a whitepaper. Strip the hype off Bitcoin and what remains is a proposal about the ledger, not the coin. The problem it addresses: every previous ledger needed a trusted keeper. The bank, the state, the village elders on Yap. What if the ledger could be kept by everyone at once, with agreement enforced not by trust but by making disagreement computationally expensive?

That's a real idea, and the proof-of-work mechanism is a clever answer to a hard distributed-systems question. What it does not do is change what money is. A bitcoin is a row in a ledger everyone agrees to read the same way. The ledger is public and replicated and slow, but it is a ledger. The value of a bitcoin is exactly the strength of the agreement to keep reading the ledger, which is why the price behaves the way it does. There's no stone. There was never a stone.

The receipts

So the hallucination is the agreement and the receipts are the ledger, and the whole apparatus, from Yap to Fedwire to a wallet address, is an attempt to make the shared reading hard to corrupt.

I find this reassuring rather than alarming. The alternative to a shared hallucination isn't "real" money. There's no real money. The alternative is barter, which does not scale past a village, or a commodity like gold, whose value is also a shared hallucination that happens to be heavy. We picked the light one. We made it self-checking. We wrote it down twice.

The stone is still down there, by the way. Nobody has ever gone looking. It isn't clear why anyone would.