Bitcoin’s supply is quoted as if every coin ever issued could still move. This site measures how true that is: output by output, from a fully indexed node, with every figure traced to a public rule. Two questions organise the work. Which coins can never move? And which coins can move, but behind a public key that a large enough quantum computer could break?
The whole cap, in one bar
That picture is the first finding. Popular estimates put three to four million bitcoin “lost”. What can actually be demonstrated from the chain is a few thousand. The millions in those estimates are coins that have not moved for a long time, which is a different claim, and one this site keeps separate on purpose.
Two lenses, kept apart
Lost means the coin cannot move, or almost certainly never will. Provably lost coins fail on mathematics or consensus: an output script no signature can satisfy, a key that is not a point on the curve, a coin the software never added to the spendable set. Probably lost coins could in principle move but no one plausibly holds the key, such as burn addresses. The lost-coins research covers both.
Quantum-exposed means the coin can move, but its public key is already visible on-chain, so an attacker with a cryptographically relevant quantum computer could derive the private key and move it first. Some output types publish the key when they are created; the rest publish it on their first spend, and from then on everything at that address is exposed. The quantum research covers where that coin is and how to check your own.
Each output gets exactly one verdict, in a fixed order: provably lost, then probably lost, then key visible in the script. So a lost coin is never also counted as exposed, and the two totals can be added without double counting. Dormancy is a third view, not a verdict: it is how long an output has sat still, and it overlaps everything. The million-odd early coins attributed to Satoshi are dormant, exposed, and deliberately not called lost.
How we try to be right
- Every figure traces to a rule. Each classification is a public proposal with its rationale, scale and edge cases. Nothing on the site is an estimate unless it says so.
- We count outputs, not addresses. An address can hold coins of different kinds and ages; counting per output keeps an exposed coin and an unexposed one at the same address apart, and stops a dust payment from making a dormant fortune look active.
- We check ourselves against independent work. The quantum total is compared every hour with Project Eleven’s, computed by a different method on different infrastructure. After the corrections below, the totals agree to within about a hundredth of a percent.
- We publish our corrections. Our first quantum figure was half the real one. Two hundred and thirty-eight million zero-value outputs once inflated every loss count. A coin we had called provably lost turned out to be spendable. Each is written up in the research, with what changed and why.
- We say what we can’t see. Keys disclosed off-chain, coins lost to forgotten passwords, exchange collapses: none of it is visible in the chain, and none of it is guessed at here.