The oldest problem in governing anything is the gap between deciding and doing – and the person trusted to cross it. On advisory votes, multisig theatre, and a decision that carries itself out.
In 1932, Adolf Berle and Gardiner Means described what had happened to the American corporation: ownership and control had come apart. Thousands of shareholders owned the firm, but a small management controlled it, and the two were no longer the same people with the same interests. The owners could vote; the managers decided what the vote meant and whether to act on it. Nearly a century later, most shareholder votes are still, in law, advice. The say-on-pay vote Congress gave American shareholders in 2010 says so in the statute itself: it “shall not be binding on the issuer or the board of directors”. The board is asked. It is not bound.
This gap – between the body that decides and the body that acts – is one of the oldest problems in the government of anything. A parliament passes a law; an executive has to enforce it, and may not. A referendum returns a verdict; a government has to implement it, and can slow-walk it for years. The decision is one thing, the deed is another, and in between sits someone trusted to carry the first into the second, who can dawdle, dilute, or simply decline. Every constitution ever written is in part an attempt to narrow that gap and to bind the person crossing it.
The first wave of on-chain organizations promised to close the gap and mostly rebuilt it. Votes were cast on an off-chain forum – a signal, cheap and non-binding – and then a small group holding the treasury's keys was meant to go and execute what the vote had decided. The keys were the catch. A proposal could pass and nothing could follow, because moving the money still required the signers to move it, and the signers could hesitate, disagree, or fail to show. The vote was a petition to an executor who happened, this time, to be a handful of people with a multisig. Governance turned into theatre: a great deal of deciding, with the doing left to trust.
A caper closes the gap by refusing to keep an executor at all. A proposal is not a message asking someone to act. The winning option is itself an on-chain action the contract performs, drawn from a closed list of five: pay these funds to this recipient, spend this much of the treasury buying into another caper, sell that holding back out again, rewrite the caper's own token metadata, or – on the platform caper alone – swap the logic that every caper runs on. Since 5 September 2026 there is a sixth kind that is deliberately not an action: a debate option is a position rather than an instruction, and a caper whose members rank one highest has settled a question without authorising anything. The rest of this essay is about the five that move something. For those, the decision does not wait on a trusted signer's willingness. Anyone at all can submit the transactions that carry it out, and the contract acts only on the option that won, exactly as it was written into the proposal. The deed is not entrusted to a person.
Removing the human executor would be reckless if it meant executing carelessly, so the contract is exacting about what it will carry out. Before anything can act, it tallies the ballot on-chain and confirms that the leading option cleared the bar the caper demands. That bar is one and a half divided by the number of options, measured against the weight actually cast: 75 per cent where the choice is between one action and doing nothing, 30 per cent on the five-option maximum the contract allows. Those two numbers look like a bar that softens as the ballot widens, and they are not, because the ceiling falls with them. Every ballot is a full ranking of every option, scored Borda-style, so an option ranked first by every voter who turns out takes two divided by the option count of the total weight – all of it at two options, two fifths at five – and nothing can take more. The bar is three quarters of that ceiling at every width. It is a genuine supermajority; only the denominator moves. The contract also confirms that voters did not instead choose the “do nothing” option that every ballot is required to carry, so a proposal can be defeated rather than only redirected. If the winner falls short, the proposal closes with no action taken.
Underneath that sits a deeper piece of discipline, and it is the part that took the most care to build. Counting the winner and carrying out the winner are held apart. The count tallies the ballot under the caper's rules – anyone can fold ballots into the running total, across as many transactions as the turnout needs – and then freezes the result, exactly once: this option won, by this weight, pass or fail. Everything downstream reads back that frozen result and has no power to recompute it differently. The payoff is subtle and total. The outcome the world sees, the outcome recorded on-chain, and the outcome actually executed are guaranteed to be one and the same number. Nothing can be tallied one way in public and carried out another way off to the side, because the count and the deed read from the same frozen result.
Between the frozen count and the act sits the part this essay originally described as a timelock, and it is not one. Since 29 August 2026 a passed tally does not authorise anything by itself; it earns the right to be tested. Anyone may trigger a passed proposal, which locks the token's trailing time-weighted price as a baseline and opens the market window. When that window closes, anyone may resolve it, and the contract integrates the token's price across the window and compares it against the baseline. At or above, the proposal passes and the action becomes executable. Below, it does not, and the action never binds – the check sits in the state tier, in front of every execution path there is.
A timelock is a delay you can watch go by. This is a test the decision has to survive, and the electorate is a different one: the ballot is weighted by holding and participation, the window by money at risk. Holding, or buying, is consent. Only selling – enough of it, for enough of the window, to hold the price under the baseline – blocks what the vote decided. It is a veto that costs its user real money to exercise, which is what keeps it from being a heckler's veto, and it cannot be faked cheaply either: because the verdict integrates the whole window rather than reading the price at the instant of the crank, an atomic buy, resolve and sell round-trip purchases nothing. A debate option, having nothing to execute, is refused at the trigger and never reaches this phase at all.
When the act runs, it runs whole or not at all. A payout leaves the treasury in one motion. An investment moves through withdraw, buy, and deposit inside the contract's own hands, so the treasury's money never lies exposed on the table for whoever triggered the transaction to skim. A divestment reverses that motion inside the same frame – withdraw the holding, sell it into its own curve, redeposit the proceeds – for the same reason. If any step fails, the whole thing rewinds as though it never began. There is no half-executed decision, and no moment at which a passing executor could take a cut of what moves through their hands.
The result is that a caper's decisions have the one property the corporation's owners and the first DAOs' voters both lacked: nothing discretionary stands between the decision and the deed. To decide is not quite to do – the market still gets its window – but what remains in the gap is a rule, priced and public and open to anyone holding a token to sell, rather than a person who might not sign. A holder voting on a caper proposal is not lodging an opinion with a management that may or may not honour it; they are pulling one lever of a machine that will carry out whatever the machine counts, unless the people who own the thing pay to say otherwise. The distance between the will of the members and the action of the organization – the distance every earlier form left open and asked you to trust someone to cross – is not shortened to nothing. It is shortened to a mechanism.
The rule of law was always meant to be exactly this: a decision that binds the decider, carried out the same for everyone, immune to the discretion of whoever happens to hold the keys. In practice it has always leaked, because a law is only as good as the officials who execute it, and officials are human. A caper is a small, complete instance of what the phrase promises – a body whose decisions reach its treasury on rails, whose treasury moves only when its members move it and only as they moved it, where anyone at all may run the machinery and nobody anywhere in it has the power to alter what was decided – only, at a price, to stop it. It is not that the members are trusted. It is that they no longer have to be.