Futarchy

Futarchy is a governance system in which decisions are settled by prediction markets rather than by counting votes. Proposed in 2000 by the economist Robin Hanson, its one-line design is "vote on values, but bet on beliefs": a community decides what it wants to maximise – a measurable success metric – and speculative markets decide which proposal is most likely to achieve it. For each decision, traders bet on the value of that metric in two conditional worlds, one where the proposal passes and one where it fails; whichever world the market prices higher becomes policy. It is the most market-native alternative to the ballot-based mechanisms most DAOs use, and after two decades as a mostly theoretical curiosity it now runs live on-chain.

Vote on values, bet on beliefs

Hanson's insight was to separate two questions a normal vote fuses together. A token-weighted vote asks holders to simultaneously express what they want and predict whether a given action will deliver it – and it rewards neither honesty nor accuracy, because a single vote almost never changes the outcome. Futarchy splits the two. The electorate (in a DAO, its token holders) democratically fixes a welfare metric – Hanson's original example was GDP-per-capita; on-chain it is almost always the DAO's own token price, a rough but liquid proxy for organisational value.

Then, for every proposal, a decision market forecasts that metric conditional on the proposal passing versus failing. Because participants stand to win or lose real money on their forecasts, the mechanism is designed to pull information out of whoever holds it – an expert with a strong view can move the price by trading on it, rather than being outvoted by a larger but less-informed crowd. Values stay democratic; beliefs get priced.

Decision markets on-chain

Turning Hanson's proposal into working software is mostly a market-design problem: the markets must be liquid enough to price accurately and hard enough to manipulate that the verdict can be trusted. MetaDAO, the reference implementation on Solana, does it with conditional tokens. When a proposal goes live, the DAO takes roughly half the liquidity from its spot pool and moves it into two conditional markets – a pass market and a fail market – so neither starts empty. Spot tokens are split into conditional versions, and a conditional trade reverts if its world does not happen: buy in the pass market and the proposal fails, and your trade simply unwinds as though it never occurred. That structure lets traders take real positions on the outcome without betting on the wrong branch.

After a fixed window – three days in MetaDAO's current configuration – the protocol compares a manipulation-resistant lagging TWAP (a time-weighted average price, which caps how far the recorded price can jump per update) between the two markets. If the pass market's TWAP beats the fail market's by more than a set threshold, the proposal executes automatically; otherwise it dies. That threshold is not symmetric: MetaDAO's documented figures are −3% for team-sponsored proposals and +3% for everything else, with the docs adding that the parameters are being actively tuned and that the authority is the DAO account on-chain – its own token-mechanics page still states the plain inequality. The threshold and window are governance-tunable parameters, not fixed law.

The gate, the queue and the lag

Descriptions of futarchy usually stop at the mechanism. What decides whether a DAO can actually run on it is the cost of putting one decision through, and on MetaDAO's own documentation that cost is three separate gates rather than one.

Anyone may propose; almost nobody may start a market. Creating a proposal is permissionless, but it does not go live until holders stake tokens on it – by default between 200,000 and 1,500,000, which the docs put at 1–15% of the 10M tokens an ICO issues, varying with the DAO's version and parameters. The stake is a spam gate rather than a bond: no lockup, and no risk of slashing. The practical effect is that the right to open a decision market is held by whoever can assemble a low-eight-figure fraction of the float, not by the proposer.

One at a time. A project can have only one proposal live at once. Add the three-day trading window and the 24-hour delay before TWAP recording even begins – a deliberate pause so traders can price a proposal before the measurement that decides it starts running – and a single decision occupies its DAO for the better part of a week. That is a hard throughput ceiling of roughly fifty to eighty decisions a year per project, where a Snapshot space runs any number of ballots in parallel and closes them over a weekend. It is the structural reason a futarchy DAO's docket is short, and it sharpens the delegate objection recorded below: the mechanism is reserved, by construction, for decisions worth a week of the organisation's entire governance capacity.

The lag has a magnitude. The manipulation defence is not just "a TWAP" but a lagging one, and MetaDAO's docs give the arithmetic: an observation starting at $500 that may move at most $5 per minute takes ten minutes to reflect a market trading at $550. The threat it is sized against is named explicitly – a Solana validator controlling 1% of slots could otherwise drive the pass price up during its own slots and force a proposal through. Manipulation resistance here is a speed limit on how fast truth may enter the record, which is also a limit on how fast genuine news can.

The second venue running futarchy in production prices the same trade-off differently. Seer carries 2,607 markets across four chains, of which 395 are futarchy markets – and every one of those 395 sits on Gnosis Chain, with none on the Ethereum, Optimism or Base deployments. Futarchy is not yet a thing a DAO adopts wherever it lives; it is a thing that happens where the liquidity and the host community already are.

Where it is being tried

Futarchy spent most of its life as a thought experiment – Hanson floated it, Vitalik Buterin sketched a crypto version in 2014, and a handful of projects toyed with it – but until recently no organisation ran on it. MetaDAO changed that: launched in November 2023, it governs itself by futarchy and has since run 96 proposals for 14 organisations, making it the first DAO where market prices, not token votes, are the final word on every material decision. It now also runs a Solana launchpad on the same programs, so the design can be checked against real balances – 20 tokens holding $18.0m in market-governed treasuries as of 15 August 2026.

In 2026 the model reached a blue-chip DAO in advisory form: GnosisDAO's GIP-145, a nine-month "advisory futarchy" pilot with Futarchy Labs, put a widget on Snapshot proposals showing the estimated percentage impact of a proposal on the GNO price, backed by $100,000 in temporary GNO/WETH liquidity to bootstrap the markets. It runs alongside the existing Snapshot vote – a price signal voters may consult but are not bound to follow – a telling middle path for a DAO not yet ready to hand policy entirely to a market.

Gnosis was not the first DAO to wire a market to its ballot. Futarchy Labs, which built the pilot, states in GIP-145 itself that GnosisDAO would be joining Kleros and Velora, and the Kleros Cooperative confirmed in the same thread that it had already been running futarchy.fi for several months, as an advisory tool and as a safeguard on contentious proposals. Gnosis's claim is to be the first large DAO to adopt one.

What the pilot buys is a signal, not a rule, and its specification is unusually precise about how weak a signal it is willing to call. The Snapshot widget shows the spot conditional prices in the pass and fail markets and the percentage difference between them; a five-day TWAP is computed when there is enough liquidity to support one, and a recommendation is issued only when the absolute impact exceeds 1%. The $100,000 is a loan, not a grant – GNO and WETH sent to a 2-of-3 multisig, usable only for providing liquidity and returnable within the nine months. The conditional markets themselves run on Gnosis Chain through Swapr, built on the Gnosis conditional-tokens contracts that also underpins Polymarket, and the resulting positions are composable with CowSwap.

The pilot also priced its own adoption. Before the Snapshot vote opened, Futarchy Labs ran a conditional market on GNO under approval versus rejection; over five days the pass market's TWAP sat about 2.5% above the fail market's. The vote itself is the sharper artefact. With two days left the proposer was publicly asking people to turn out because quorum had not been met, and it closed at 117,291 GNO for and 5.99 against – 117,562 in total against a 75,000 quorum, on 67 votes. A mechanism pitched partly as an answer to voter apathy needed a turnout appeal to authorise itself, against essentially no opposition at all.

Strengths, and where it breaks

Futarchy's appeal is that it attacks two chronic DAO failures at once: voter apathy (you no longer need a quorum of informed voters, only a few informed traders) and the fact that ordinary voting rewards neither conviction nor being right. Its failure modes are equally sharp. Thin markets price badly and are cheap to push, which is why liquidity-bootstrapping is the central engineering problem and why the model works best where the token is already liquid.

Metric gaming is the deeper hazard: whatever proxy you choose, you get Goodhart's law – optimise hard for token price and you can reward decisions that pump the price while hollowing out the thing the price was meant to measure. Choosing whose welfare the metric captures is itself an unresolved political question, not a market one. And a rich actor can, in principle, subsidise a market to steer a decision, betting that the value of controlling the outcome exceeds the trading loss – a live concern the governance-attack literature takes seriously. Futarchy does not remove politics from a DAO; it relocates it to the choice of metric and the design of the market.

Hanson's answer to manipulation was restated in 2017 by Gnosis, the team behind the conditional-token contracts the GnosisDAO pilot runs on, when it announced experiments to test futarchy (How manipulation-resistant are Prediction Markets?). A trader who moves a price to distort a decision, rather than to profit from what they know, is in market terms a noise trader, and informed traders profit by trading against noise. The theory Gnosis cited holds that when traders are only mildly risk-averse and more effort yields more information, extra noise trading raises price accuracy rather than lowering it, because it pays informed traders to trade more and research harder. Gnosis set out to test that on Ethereum, with support from the Ethereum Foundation, through experiments Vitalik Buterin had proposed. The argument assumes informed traders with capital are present to take the other side.

The incentive objection is sharper than the mechanism-design one, and GnosisDAO's own delegates raised it before voting the pilot through. If a proposal's estimated price impact is well under a percent, a trader is asked to take liquidity risk, smart-contract risk and lock-up uncertainty for a fraction of a percent of upside – so who supplies the informed capital the model depends on? Futarchy Labs' answer is that a thin margin is itself information – it means the proposal barely matters – and that on Kleros the absolute impact on contested proposals frequently traded above 4%. That answer concedes the shape of the problem: futarchy is informative exactly where a decision is already divisive, and quietest on the routine business that makes up most of a DAO's docket.

How Caper approaches this

Caper and futarchy start from the same complaint – that one-block, capital-only voting aggregates the wrong signal – but answer it in opposite registers. Futarchy externalises the judgement to a market and lets price stand in for collective belief; a caper keeps the judgement with its members and instead changes what a vote weighs. Voting power on a Caper proposal grows with participation, not with a token balance alone: each ballot a member casts earns them soulbound vote tokens, which nothing else mints and which no other holder can sell them.

That makes a large but disengaged holder structurally weaker than a long-standing contributor, without asking anyone to price the DAO's future in a conditional AMM. It also sidesteps futarchy's metric problem: there is no single proxy to game, because each proposal resolves to a definite executed outcome rather than to a market's verdict on a number. And because that same weight sets a member's pro-rata exit share, the incentive to accumulate participation honestly is built in – a different bet than futarchy's, that who decides is easier to get right than what to measure.

References

• Robin Hanson, "Shall We Vote on Values, But Bet on Beliefs?" (2000). The original proposal.
• Vitalik Buterin, "An Introduction to Futarchy" (Ethereum Foundation blog, 2014). The first sketch of an on-chain version.
• MetaDAO documentation – decision markets and TWAPs. Conditional tokens, the liquidity split, and the lagging TWAP that settles a proposal.
• GIP-145 forum thread (17 November 2025). The full pilot specification, the Kleros and Velora precedents, the delegate incentive objection, and the pre-vote conditional market on GNO.
• GIP-145 Snapshot record. Closed 14 February 2026 at 117,291 for / 5.99 against / 265 abstain against a 75,000 quorum.
• Gnosis conditional-tokens contracts. The settlement primitive the pilot's markets are built on.
• Nadja Beneš, "How manipulation-resistant are Prediction Markets?" (Gnosis, 2017). The noise-trader argument against manipulation, and the announcement of Gnosis's futarchy experiments.