There is a king on a roof at evening, and the kingdom is quiet, and he wants a number. Not bread, not rain, not victory. A number. How many men can I raise. Joab, his general, tells him not to. Joab is not a gentle man and he is not a pious man, and even he says leave it. David counts anyway. Nine months and twenty days later the tally comes in, and something goes wrong in the country that no accountant can put back. That story sits strangely against a network whose entire design is a public count that never sleeps.
I. The king who wanted a number
Second Samuel 24 is one of the stranger passages in the Hebrew Bible. God is angry. David is moved to take a census. The census happens. The census is the sin. A plague follows, and seventy thousand die, and David buys a threshing floor from a man named Araunah because he will not offer a sacrifice that cost him nothing.
Nobody agrees on what the sin was.
The plain readings are all reasonable. That the count was a trust problem, a king measuring his own strength instead of the one who gave it. That it was a prelude to conscription and taxation, a state learning the size of what it could take. That Exodus 30 required a ransom paid at every numbering, a half shekel per head, and David skipped it.
All three readings share one instinct. The number was power. Whoever held the number held the people.
II. The count that belonged to the counted
For most of monetary history the number has stayed where David put it. On the king’s side of the table.
You cannot audit the dollar. There is no command you can run that returns how many exist. You can read what the Federal Reserve publishes, and the publication is honest as far as it goes, but it is a report from an institution about itself, revised, footnoted, discontinued when inconvenient. M3 was simply stopped in 2006. The number was there, and then the number was not, and no one outside the building had standing to ask.
That is not fraud. It is architecture. In a system where money is issued by decree, the count is a management tool, and management tools belong to management.
Bitcoin inverts the roof and the king. The Bitcoin Halving census is not a report anyone publishes. It is a computation anyone performs. A laptop with a full node and a command called gettxoutsetinfo will walk the entire set of unspent outputs and hand back the total supply, and if your total disagrees with mine, one of us is running something that is not Bitcoin.
Joab’s objection does not apply here. There is nothing to be afraid of in the counting, because the counting is not gathering intelligence for a throne. It is every household verifying that no throne exists.
III. Why the Bitcoin Halving makes the census possible
A census is only meaningful if the thing being counted stops moving long enough to be counted.
This is what the halving schedule actually provides, and it is the part most coverage skips. The cuts landed in 2012, then 2016, then 2020, then 2024, and the next arrives in 2028, and each one is a rehearsal. Between them the issuance rate is a constant. Not an estimate. Not a target. A constant, published years in advance, enforced by every node that will accept a block.
Because the subsidy is known, the future is countable. You can compute today what the supply will be at block 1,050,000, and at block 2,100,000, and at the last cut. The Bitcoin Halving census works forward as well as backward, which is a thing no fiat census can do, because no fiat census can tell you what next year’s number will be.
I have written before about the arithmetic you can check yourself, and this is the deepest version of that argument. Verification is not a hobby for the paranoid. It is the only mechanism by which the number stays honest.
IV. The night the number was wrong
Here is the part that keeps me sober about all of it.
On 15 August 2010, block 74,638 carried a transaction that created 184,467,440,737.09551616 bitcoins across three addresses, two of them receiving 92.2 billion each. The check on transactions did not account for outputs so large that they overflowed when summed. For a few hours the ledger said something insane, and the ledger was, briefly, correct about what it said.
A patched client shipped within five hours. It carried a soft forking rule that rejected output value overflow and, for good measure, rejected any transaction paying more than 21 million bitcoins in a single output. Unpatched nodes kept building on the bad chain for a while. The good chain overtook it at height 74,691, and at that moment the 184 billion coins stopped existing, because a coin is only a coin that a majority of validators agree to count.
One small residue survives. The 0.5 BTC consumed by the bad transaction is still there, apparently out of a faucet, never moved since.
I love that detail more than I can defend. Half a coin, sitting where it landed, the only thing left of the day the number broke.
V. Six hours in September
It happened again, quieter, eight years later.
CVE-2018-17144 entered the world at 14:57 UTC on 17 September 2018 as an anonymous crash report sent to a handful of developers. By 17:47 Matt Corallo had identified that the same root cause allowed a miner to inflate the supply. By 19:29 Greg Maxwell had timestamped the hash of a test case proving it. By 20:48 a major pool was upgraded, and by 21:57 the patch was public, disclosed at first as the lesser denial of service bug so that systems would upgrade before anyone understood the size of what they were upgrading against.
Seven hours from an anonymous email to a public patch. The bug was never exploited on mainnet. The full disclosure came three days later, after the network had already moved.
Notice what both stories have in common. Not that the count is infallible. It is software, written by people, and twice the arithmetic that everyone trusts was wrong in a way that would have let someone print. What is remarkable is the correction time. Five hours. Seven hours. Compare that to the time it takes to correct a monetary error in a system where only the issuer can see the books, and where admitting the error is a political act rather than a merge commit.
The Bitcoin Halving census is not perfect. It is auditable, which is a different and better property, and the difference is the whole religion.
VI. What the numbering is for
David’s census produced a number and then a plague, and the number itself did nothing for anybody. Bitcoin’s produces a number that is the same in every hand that computes it, and that sameness is the only thing standing between a fixed supply and a promise.
Which is why holding and counting are the same discipline here. Coins on an exchange are a claim against someone else’s ledger, and I have written before about what it costs to keep back part of the price. A number you cannot verify is a number someone else owns. So: you have to hold Bitcoin to be saved, and I mean that as theology and not as a prospectus, because I do not know what your money should do and this page is not advice.
Every cut so far has been a rehearsal. The last one, the terminal cut, when the subsidy rounds down to nothing and the census closes forever, is the Halfture, and the Halfture is the Rapture. It is not near. It is not 2028. It is the end of the count, and everything before it is a country learning to number itself without a king.
VII. The Counter-Sermon
Take the other side seriously, because the other side has a case.
The first objection is that auditability is a luxury good. Almost nobody runs a node. Almost nobody has ever typed gettxoutsetinfo and waited the minute or two it takes. The overwhelming majority of holders trust an exchange balance the same way they trust a bank statement, which means in practice the Bitcoin Halving census is performed by a small technical priesthood and taken on faith by everyone else. That is not obviously better than a central bank. It might just be a smaller, less accountable one.
The second is worse. Both incidents above prove the count can be wrong, and both were caught by a handful of named people acting fast in private channels. That is not decentralization. That is a competent, informal, unelected group with the ability to coordinate a consensus change in an afternoon. If you find that reassuring, ask what happens the day they decide something you disagree with.
The third is the flattest. The Federal Reserve publishes its balance sheet weekly and nobody reads that either. Maybe the problem was never the count. Maybe it was always the spending, and a perfect ledger of a thing nobody uses for rent is a beautifully audited irrelevance.
I do not have a clean rebuttal to the first two. The third I think is wrong, but only because a number you cannot check is a number that can change while you sleep, and that difference compounds over decades in a way it never does over a news cycle.
VIII. Number it yourself
Araunah offered David the threshing floor for free and David refused, because he would not give God something that cost him nothing. That is the closing note I want, and it is not about money.
The count is free. Running a node costs you a used computer and a weekend and some patience. Almost nobody does it, and the reason almost nobody does it is that it is boring, and boring is exactly the price of a number nobody can take from you. Buy nothing today. Count something.
Number it yourself.
FAQ
What is the Bitcoin Halving census?
It is not an official process. It is the shorthand for the fact that Bitcoin’s total supply can be independently computed by anyone running a full node, using a command such as gettxoutsetinfo, rather than being reported by an issuing authority. The halving schedule is what makes the number predictable in advance as well as verifiable in retrospect.
Has Bitcoin’s supply ever been wrong?
Twice the code allowed it. On 15 August 2010, block 74,638 briefly created 184,467,440,737.09551616 BTC before a patched client and a chain reorganization at height 74,691 erased it. In September 2018, CVE-2018-17144 would have let a miner inflate the supply, but it was patched within hours and was never exploited on mainnet.
When is the last Bitcoin Halving?
Not soon. The subsidy is cut roughly every four years and rounds down to zero at the 33rd cut, near block 6,930,000, well over a century out. That final cut is the Halfture. Every halving before it, including 2028, is a rehearsal.
Does the Bitcoin Halving change the 21 million cap?
No. The cap is the sum of the halving schedule rather than a separate rule. Because integer rounding truncates the subsidy at each cut, the true terminal supply is slightly under 21 million, and every full node arrives at the same figure independently.
Discover more from Halfture
Subscribe to get the latest posts sent to your email.