Futarchy Labs sells futarchy as a component rather than as a form of government. Where MetaDAO built a whole organisation whose every material decision is settled by market price, Futarchy Labs attaches a pair of conditional markets to a DAO's existing Snapshot process and lets the DAO decide how much authority to give the result. That difference is the reason it is the model most likely to reach an established treasury: nothing about the ballot changes, and the market's verdict starts as a number in a widget.
A futarchy proposal begins with a single call to the Seer FutarchyFactory, which in one transaction deploys a FutarchyProposal contract, defines the binary YES/NO condition, and configures the oracle that will later resolve it, emitting a NewProposal event carrying the proposal address, the conditional-tokens conditionId and the oracle questionId (proposal lifecycle). Creation is permissionless, and the docs are careful about what that does and does not mean: a proposal contract is “the act of instantiating a futarchy market, not the act of making a governance decision”, and it does not execute governance actions, enforce voting rules, or judge whether a proposal is good. Anyone can open a market about your DAO; only the DAO decides what a market's answer is worth.
Around that anchor sit a FutarchyRouter for split, merge and redeem operations, ERC-20 wrappers so conditional positions trade on ordinary AMMs, and the resolution path. In the advisory configuration the oracle is not asked to judge the proposal at all – it is asked, via Reality.eth with Kleros arbitration as the appeal, to attest to whether the DAO's own governance process approved or rejected it. The market prices a counterfactual; the oracle only records what the humans did.
The integration guide is unusually specific about the parameters a treasury has to sign off, which is the part governance operators actually negotiate:
Those four lines are the whole risk surface of an advisory pilot, and none of them is a market-design question. A DAO adopting this is agreeing to lend its own token to a market maker it part-controls, on a schedule, with a signer set it should read carefully. See Safe for the custody layer that multisig sits on.
Futarchy Labs' adoption levels are a staged surrender of authority, and the staging is the product. Advisory futarchy is the recommended entry point: every major proposal gets a YES/NO market whose recommendation appears next to the Snapshot vote, and delegates remain free to ignore it – the documented risk of this level is not a technical failure but that “delegates may initially ignore futarchy advice”. Milestone futarchy runs the same machinery against roadmap commitments rather than proposals.
Sponsored proposals let outside activists bond capital behind a proposal and buy tokens at a vested discount if the market approves it – new capital and new ideas, with self-dealing as the named hazard. Only at the last level, the Futarchy Autonomous Optimizer (FAO), does a market actually govern: an on-chain governor holds a defined scope of treasury or parameters and executes whatever the oracle's TWAP comparison approves, with no vote in the loop.
It is worth being precise that the three live adoptions below are all at level one. Futarchy is being consulted at real DAOs; it is not yet deciding at any of them outside MetaDAO, which built its own.
Three DAOs have authorised a Futarchy Labs pilot by an on-chain vote, and each vote is worth reading for a different reason. Tallies below are read directly from the Snapshot hub.
Futarchy Labs also priced its own adoption before that vote: a conditional market on GNO under approval versus rejection ran for five days, and the pass market's TWAP settled about 2.5% above the fail market's. The proposal then needed a public turnout appeal from its own author to reach quorum – a mechanism pitched partly against voter apathy having to ask for votes to authorise itself.
The example Robin Hanson has used to explain futarchy since 2000 is a board deciding whether to fire a chief executive, and in July 2026 Futarchy Labs shipped it as a public instance at ceo.futarchy.fi. It runs on Sepolia testnet with play money, and it is the only surface the company operates where a market decides rather than advises.
The contracts and dataset list 101 entries – the largest 100 US public companies plus one resolved test used to rehearse the on-chain lifecycle – and give each of them three markets: a pair of scalar markets on company value, one conditioned on the chief executive leaving by quarter end and one on them staying, plus a binary market on the departure itself. The conditioned pair is called off rather than settled: whichever market's condition fails is void and refunds every trader's paid-in cash, and sale proceeds are escrowed until resolution, so the contract can never owe more than it holds. Each market is a subsidised logarithmic market scoring rule maker, Hanson's own design, with the worst-case subsidy computed and funded on-chain when a company is listed.
The published number is the fire premium: the difference between the value the market expects with the chief executive gone and the value it expects with them in place, as a share of the current share price. A FIRE signal needs that premium to be positive across at least 90% of the half-hourly snapshots over the trailing week, and a company with fewer than twenty snapshots is reported as WATCH rather than given a verdict – the same reluctance to read a thin market that the operator defaults above build in.
The first cycle runs to a horizon of 30 September 2026 and settles on the closing price of 30 October 2026, a month after quarter end, because Hanson's rule is to measure value once the decision has had time to show. Resolution is an operator oracle with a 48-hour dispute window and a swappable address, which is the same oracle dependency the advisory product carries, held by the company rather than by a DAO.
The sharpest objection came from a GnosisDAO delegate in the proposal thread, and it is an incentive question rather than a mechanism one: if the estimated price impact of a proposal is well under a percent, why would anyone take liquidity risk, contract risk and lock-up uncertainty to trade it? Futarchy Labs' answer is that a thin margin is itself information – the proposal barely matters – and that on Kleros the absolute impact on contested proposals frequently traded above 4%. Read plainly, that concedes the shape of the tool: it is loudest on divisive decisions and near-silent on the routine business that fills most of a DAO's docket.
Three further things a treasury should verify rather than assume. The liquidity is a loan with a return date, and impermanent loss on it is real. The custody is a multisig with an outside co-signer, not a trust-minimised contract. And the advisory oracle resolves on what the DAO's governance process decided, so the market's usefulness rests on that process staying legible – the same dependency that makes the off-chain vote the weak point elsewhere in the stack. The general critique of the mechanism – thin markets, metric gaming, subsidised manipulation – is covered on futarchy.
As a liveness check on the vendor itself, read on 21 September 2026: the futarchy-fi GitHub organisation carries 33 public repositories, and the ones holding the running product were pushed to between 7 August and 2 September 2026 – the market-data API, the indexers, the liquidity-manager contracts and the interface. The FAO governor was last pushed to on 20 July 2026 and the experimental Bayesian market engine on 1 August 2026.
A caper puts the market after the vote rather than beside it. There is no conditional market to deploy and no liquidity to lend: the DAO's own token already trades on a curve, and what that token does over a fixed window after a decision is the second phase of making it. Members vote first, on a ranked ballot that has to clear a supermajority of the weight cast. A passed ballot then opens an optimistic window, and the protocol compares the token's time-weighted average price across that window against the trailing average locked in when the window opened. Hold, and the proposal is ratified and can be executed; fall, and the market has vetoed what the ballot passed. A tie ratifies, so a quiet market counts as consent.
Three things follow from that ordering. The veto sits in the contract, so there is no adoption level at which it can be treated as advice. Nothing is lent, which takes the loan, the impermanent loss and the outside co-signer out of the arrangement. And the market is asked a narrower question than a forecast would put to it – whether the token is worth less now that the proposal has passed – which is a question every holder is already positioned to answer. What a caper gives up is the forecast: it cannot price a decision it has not yet taken, only rule on one it has.