The DAO industry's one working counter-example

MetaDAO is the reference implementation of futarchy on a public chain, and the only organisation of consequence that has removed voting from its governance entirely. Its own documentation puts the claim plainly: “The difference between governance on MetaDAO and elsewhere is that there isn't any voting, only trading.” Every material decision – spending treasury USDC, issuing new tokens, changing token metadata, moving protocol liquidity – is settled by comparing the price of the token in two conditional markets, one priced on the proposal passing and one on it failing.

It matters to this wiki for a reason separate from the mechanism. Most alternatives to token-weighted voting exist as forum posts, pilots or advisory widgets. MetaDAO has run its own treasury this way since November 2023, and has since become the machinery under a working Solana launchpad, so the design is observable against real balances rather than argued from first principles. As of 20 September 2026 the protocol's market feed lists 23 tokens whose treasuries hold $13,999,569 in USDC, every dollar of which can only be spent by passing a market.

How a decision is made

The lifecycle has four stages, all documented and all on-chain.

• Anyone proposes. A proposal can spend treasury USDC, issue new tokens, update token metadata, or change treasury-provided liquidity. There is no proposer whitelist.
• Stake, not quorum. Tokens must be staked on a proposal before it goes live – a default of 200,000 to 1,500,000 tokens, 1–15% of an ICO's 10m supply, tuned per DAO. Staking is anti-spam only: no lockup, no slashing, stakes returned once trading opens. Only one proposal can be live at a time.
• Three days of conditional trading. The project moves half its spot liquidity into the pass and fail markets, so neither starts empty, and traders take positions that revert if their world does not happen – buy in the pass market, watch the proposal fail, and the trade unwinds as though it never occurred. TWAP recording does not start immediately: a 24-hour delay gives traders time to price the proposal first.
• A lagged average decides. Settlement compares a lagging-price TWAP in each market – an observation series that can only move a bounded amount per update, so a validator controlling a slice of slots cannot spike the recorded price. What a system inherits by settling on a price at all – the manipulation cost that becomes its real quorum, and the retention bound that shortens a nominal window – is set out under price oracles in governance. Execution is immediate on resolution; the three-day window is itself the timelock.

The pass condition is not a simple comparison, and the difference is worth reading precisely. MetaDAO's threshold table sets −3% for team-sponsored proposals and +3% for everything else – teams are given the benefit of the doubt, external proposers must clear a margin – and the docs add that “we're actively tuning these parameters. Check the DAO structure onchain to understand exactly what thresholds have been configured for any specific organization.” A second documentation page still describes settlement as the plain inequality “Pass market TWAP > Fail market TWAP”; where the two disagree, the on-chain DAO account is the authority, not either page.

The three-day window is a per-DAO parameter too, and shortening it shortens the only delay there is. BASKET-002 set Basket's proposal duration to 24 hours and its TWAP start delay to 8 hours on 21 August 2026, and ACCRUE-001 and LFOWN-002 did the same for their own DAOs on 15 September. Where that is configured, a proposal is proposed, priced and executed inside a day. Read the duration off the DAO account alongside the thresholds; neither is a platform constant.

The launchpad, and the discretionary cap

Since 2025 the same programs sit under a token launchpad, and it is the launchpad that made MetaDAO visible outside governance circles. Investors get four days to commit USDC; 10m tokens are distributed proportionally; everyone pays the same price. Allocation is by an accumulator – committed_amount × elapsed_seconds – with a fill boost for people who commit while the pool is still sparse, so discovering a raise early is rewarded as well as arriving early.

The unusual clause is the discretionary cap: the founder chooses how much of the committed USDC the project actually takes. MetaDAO's stated purpose is “to allow believers to participate while preventing projects from over-raising”. On success the USDC goes to a market-governed treasury, mint authority transfers to that treasury, and 20% of the USDC plus 2.9m tokens seed the liquidity pools; the team then draws a configured monthly budget and must pass a proposal for anything larger.

Rip Cars is the extreme case, and it is the clearest number on this page. The Hot Wheels gacha project's raise closed with $32.0m committed against a $250,000 minimum – and $250,000 raised. A 128× oversubscription was refused in full. Contemporary coverage reported the sale at $20.9m in commitments mid-flight and correctly cautioned that “commitments are not the same as capital settled” without being able to say what settled; the protocol's own market feed can. As of 17 September 2026 the Rip Cars treasury holds $90,000, down from $110,000 on 4 September and $160,000 on 15 August. For comparison, Avici raised $3.5m against a $2.0m goal and Umbra $3.0m against $750,000 – caps exercised, but nothing like Rip Cars'.

Since September the launchpad also funds ideas that have no team. Two raises settled that way: LFOWN took $10,000 from 28 owners on 8 September and ACCRUE $10,000 from 39 owners on 15 September, each posted as an idea rather than a company – “No team yet: builders compete for the treasury and a decision market picks the winner. No winner? Full refund.” That inverts the usual order. A conventional raise prices a team’s plan; these price the idea first and then point the same mechanism at hiring against the money. Both are small, and whether a decision market can pick a founder is not a question either has answered – but it is the first time futarchy here has been aimed at an appointment rather than a spend.

The Bid Wall – a NAV-priced buyback that burned tokens sold into it – is worth noting mainly as a retirement. MetaDAO's own docs now head the page “The Bid Wall is deprecated. It was used in one raise and has not been used since.” A mechanism designed, shipped, used once and openly marked dead is a better disclosure standard than most of this industry manages.

What the numbers say, read 20 September 2026

Every figure below comes from MetaDAO's own CoinGecko-compatible market API, which is public and needs no key, though since 4 September it refuses a bare script and wants a browser user-agent. treasury_usdc_aum is the USDC currently sitting in each project's market-governed treasury – not what it raised, and, as the first two findings below set out, not a measure of what it has left to spend.

The confound this page warned about has fired, and it moved $680,000. Three days ago this section said that three live decision markets were about to convert treasury USDC into USDv without spending it, and that a reader who had learned to read a fall in this column as a spend-down would be reading it wrong. All three passed. Credible's, shown on the site as CREDIBLE-001, settled at a pass TWAP of $0.5567 against $0.5407 in the fail market on $8,508 of volume; Umbra's, shown as UMBRA-006, at $0.1945 against $0.1926 on $2,039; OMNIPAIR-006 at $0.1589 against $0.1581 on $387. Each of the three treasuries then fell by exactly the amount its proposal authorised – $250,000, $300,000, $130,000 – and, in a table of twenty-three rows, no other treasury moved by as much as a dollar except Laso's, which rose by $3,335. The column fell $676,344 and not one cent of it was spent.

The drain this page reported in August was two thirds the same thing, which is the more useful correction. On 19 August METADAO-040, Convert $2,000,000 of MetaDAO Treasury USDC to USDv, passed on a pass TWAP of $5.4505 against $5.1580 – on $274,223 of volume, the deepest decision market the platform has run. MetaDAO's own row read $5,943,551.02 on 15 August and reads $3,703,551.02 today, a fall of $2,240,000, of which $2,000,000 is that conversion. So of the $4,037,173 this column has lost since 15 August, $2,680,000 is a change of denomination and about $1.36m is everything else – real spending, Paystream leaving the feed on liquidation, and four new launches arriving with money. A metric is safe to build a finding on only while the thing it measures and the thing it is taken to measure stay the same thing, and this one stopped being safe in August rather than this week.

The distribution is unchanged and is still the structural finding. The ten rows above hold 98.0% of the treasury total; the other thirteen hold $275,422 between them, sixteen of the twenty-three hold under $120,000, and the two at the bottom – Loyal ($0.54) and Flash.Trade ($0.00) – have not moved a cent since 4 September. A launchpad's headline is its cumulative raise; its actual state is a short head and a long, empty tail.

The independent cross-check has closed, and the reason it ever opened is mechanical. This page reported an unexplained 11-day divergence between the market feed's liquidity total and DeFiLlama's TVL for the protocol – 11.7% apart on 6 September, 12.5% on 17 September – and concluded the two were counting different things. They are not. They were being read at different points in a proposal cycle. MetaDAO's documentation states that once a proposal has enough stake “the project takes half of its liquidity from the spot … market and moves it into a proposal's conditional markets”, and liquidity_in_usd reads the spot pool. A project with a live decision market therefore reads at roughly half its book. The three projects with live markets on 17 September have each since roughly doubled – Credible $914,573 to $1,912,586, Umbra $486,896 to $973,848, Omnipair $261,645 to $540,956 – while every other row in the table sits within a tenth of its 17 September figure. The case this page had already recorded as inexplicable fits the same shape: Solomon's book halved from $2,901,515 to $1,478,708 between 4 and 6 September and was back at $3,190,480 by 17 September, and SOLOMON-004 was created on 4 September and passed on $96,368 of volume. This page said no proposal had been cast either way; that was wrong. As of this reading the site's decisions page reports no live markets, and the two sources agree to 0.8% – $13,654,643 on the feed against $13,546,870 on DeFiLlama, stamped 20 September 19:16 UTC. The lesson generalises past this protocol: a liquidity figure for a futarchy platform is only meaningful with the state of its proposal queue attached.

Fees moved with it. DeFiLlama puts protocol fees – a 0.25% trade fee on all Futarchy AMM trades – at $3,209,676 all-time, $153,746 over 30 days, $38,134 over the trailing week and $13,697 over 24 hours, read 20 September 2026. All-time grew $24,132 in the three days since this page last read it, about $8,000 a day against roughly $4,200 a day over the thirteen days before, and the trailing week is up 75% on the 17 September figure. Revenue here arrives in bursts shaped by launches and by decisions, so one reading of it is a level and not a trend.

Two counts need care. MetaDAO's documentation still states it has “run 96 proposals for 14 organizations” since November 2023, and that sentence has not moved while the protocol has; the homepage count of launches that used to sit beside it is gone, since metadao.fi was rebuilt between 15 August and 17 September 2026 around per-company runway and a live decision feed. And the feed's twenty-three rows are not twenty-three launches. Every launchpad token carries a vanity mint ending meta – METAwkXcqyXKy1AtsSgJ8JiUHwGCafnZL38n3vYmeta, omfgRBnxHsNJh6YeGbGAmWenNkenzsXyBXm3WDhmeta – and exactly two rows do not: Sanctum's CLOUD (CLoUDKc4Ane7HeQcPpE3YHnznRxhMimJ4MyaUqyHFzAu) and Flash.Trade's FAF (FAFxVxnkzZHMCodkWyoccgUNgVScqMw2mhhQBYDFjFAF), the two governance clients that came for decision markets and never ran an ICO. CLOUD is also the only row whose treasury_usdc_aum is null rather than zero. So the feed carries twenty-one launched tokens and two governance clients, and this page's earlier reading of Cloud as a 15 September launch was wrong.

One citation hazard, because this section leans on the decision pages. A decision's URL slug does not always match the identifier the site prints on it: /decisions/credible-002 renders CREDIBLE-001, /decisions/umbra-008 renders UMBRA-006, and /decisions/lfown-004 renders LFOWN-002. The links above are the working URLs; where the two disagree, the printed identifier is named alongside.

When the mechanism said no

Four cases are on the record, and between them they test the launch, the wind-down and the token.

Hurupay, February 2026. The payments project's ICO opened on 3 February 2026 and closed at roughly $2,003,593 against a $3,000,000 minimum. Under the documented rule – “when a project fails to reach its minimum, everyone is refunded their USDC back” – every participant was refunded and no token listed. It was the launchpad's first failure, and the interesting part is that the failure mode was the specified one: a raise that could not clear its own floor did not get quietly relisted at a lower valuation.

Paystream, September 2026. The first wind-down settled the same way a spend does. PAYSTREAM-003, Liquidate Paystream, passed on a final TWAP of $0.02582 in the pass market against $0.02495 in the fail market, and it is a dissolution rather than a pause: the DAO withdraws all protocol-controlled liquidity, permanently revokes the PAYS mint and freeze authorities, burns the 94.28% of supply under protocol control, moves the recovered USDC into a liquidation vault, terminates the Paystream DAO LLC and open-sources the intellectual property. Holders redeem at a rate fixed once at execution – a reference 0.0267 USDC per PAYS, a 35.62% recovery against the 0.075 ICO price – for three years, with the redeemed tokens held in the vault so the ratio does not move as people claim: “There is no first-mover advantage and no last-holder penalty.” The token left the market feed with it. Publishing a wind-down at the same address, in the same format and through the same mechanism as a raise is a disclosure standard almost nothing else on this page's subject matches.

Kimia, August 2026 – the one the market refused. Two weeks before Paystream's DAO voted itself out of existence, Kimia's was asked to and declined. KIMIA-001, Authorizes the orderly liquidation of the Kimia treasury and the return of IP to the original owners, failed on a pass TWAP of $0.004575 against $0.004636 in the fail market: the market priced the project as worth more alive than wound up. It drew 28 participants and $44,794 of volume – 240 times what settled the Paystream liquidation – and it is the only failed decision in the platform's recent record. KIMIA-001 carried no team-sponsored flag, so under the threshold table above it needed to clear +3% rather than −3%, but the flag did not decide it; the pass market came in below the fail market outright. Kimia still holds $28,800 in treasury a month on. The pair is what earns its place here: the same mechanism, asked the same question by two projects a fortnight apart, returned opposite answers, and the one that drew real trading is the one that said no.

META's supply, continuously. META has no hard cap at the token-program level – mint authority was never burned. What replaces the cap is that the authority is the governance program itself, not a human operator, so every new token must be proposed publicly, trade for its configured window, and clear the threshold. Supply read on-chain on 15 August and again on 4 September 2026 was 22,684,693.458722 against an initial 10m, total and circulating identical, so no issuance passed a market in those twenty days. It was 22,684,691.071575 on 17 September and 22,684,690.908316 on 20 September, still total-equals-circulating and now 2.550406 below the August reading. A decrease is not something an issuance proposal can produce, and nothing readable from here says what produces it; what the four dated readings establish is that the number drifts downward in small amounts, so a no-issuance claim has to be re-checked against a fresh reading rather than inherited from this page. Whether that is reassuring depends entirely on whether you believe the markets are thick enough to reject a bad issuance – which is exactly the bet the protocol is asking you to make, stated honestly rather than hidden behind a cap that mint authority could restore anyway.

The legal shell

MetaDAO is not a pseudonymous protocol wearing a DAO label. It operates as METADAO LLC, and its transparency page publishes an operating agreement, a certificate of formation, and unaudited quarterly reports running from Q1 2025 to Q1 2026, alongside a daily balance breakdown – $9.51m total, $8,219,832.59 of it USDC across named accounts.

In July 2026 it went further than nearly any DAO has. MetaDAO LLC filed a MiCA Title II crypto-asset white paper for META – notification dated 2 July 2026, Digital Token Identifier BQ53DH590, legal entity identifier 254900XHQIYLONV5P484 – published as both a PDF and an Inline XBRL filing. The filing carries the standard Article 6(3) disclaimer that no competent authority has approved it and that the offeror is solely responsible for its content.

The combination is unusual and worth stating for what it is: an organisation that has removed human discretion from its spending decisions has simultaneously accepted the disclosure obligations of a named legal entity in a regulated market. Those are not in tension – a registered offeror still needs somewhere for authority to live, and here it lives in a market rather than a board. Set beside the survey in DAO legal structures, it is a different answer from the foundation and the DUNA: an ordinary LLC whose treasury it cannot unilaterally spend.

What it does not solve

The strongest objections to futarchy apply here undiminished, and the treasury table above sharpens two of them.

Thin markets price badly, and it is not only the tail. Sixteen of MetaDAO's twenty-three tokens hold under $120,000 in treasury and correspondingly little liquidity; a decision market on a project with $3,050 in the bank is a decision market in name. But the volumes that settle real decisions are the sharper version of the objection. The Paystream liquidation carried $186 across both conditional books. The three USDv conversions read above moved $680,000 of treasury on $10,934 between them, Omnipair's $130,000 settling on $387 from three participants. And two DAOs have shortened their own timelock this way: BASKET-002 cut Basket's proposal duration from three days to twenty-four hours and its TWAP start delay to eight, on $97 of volume, and LFOWN-002 did the same on $54 and a margin of one hundredth of a cent. Shortening the window is the one change that makes every later decision cheaper to move, and it was made by markets almost nobody traded. This is the same objection GnosisDAO's delegates raised against the advisory pilot – who supplies informed capital when the upside is a fraction of a percent – arriving from the opposite direction.

Goodhart's law does not care that you removed the ballot. The welfare metric is the token price. Optimising hard for the token price can reward decisions that pump it and hollow out whatever the price was meant to proxy, and no amount of market depth fixes a badly chosen metric. Futarchy relocates politics to the choice of metric; it does not abolish it.

Nothing here is a claim on the entity. Holding META buys governance influence through trading and nothing else. The tokenomics question of what a governance token entitles a holder to is left exactly where every other DAO leaves it.

And a fourth, particular to this protocol: the launchpad and the governance layer are being judged as one thing. Decision markets are a governance claim; the discretionary cap, the four-day sale and the refund floor are a fundraising claim. Rip Cars refusing $31.75m is evidence about the second, not the first.

How Caper approaches this

This section has been rewritten twice in a month, and the dates matter, because a reader who saw it last week is holding a different model. Until 26 August 2026 Caper settled proposals on a ranked ballot alone. Between 26 and 28 August it settled them on a price alone, and this page said so – that MetaDAO and Caper had converged on the same answer. Since the redeploy of 29 August 2026 a caper does both, in sequence, and the convergence claim is retired.

A caper proposal now passes two gates. The legislative gate is a ballot: two to five typed options, one of which must be “do nothing” and at least one of which must be an answer other than it (executable, or since 5 September 2026 a non-executing debate position); a full-permutation ranked vote for a flat 100 XRD fee banked to the caper’s own treasury; a Borda tally; and the leader passes only if its share of the weight actually cast clears 1.5 ÷ the option count – 75% on a two-option ballot, 30% on five, both of which are three quarters of what a unanimous first place could take at that width, since a full-permutation Borda fold caps any single option at 2 / option_count – and is not the do-nothing option. The optimistic gate follows: triggering the winning action locks the caper’s trailing TWAP as a baseline and opens the market window, and the action executes if and only if the TWAP measured across that window is at or above the baseline. Holding is consent; only selling hard enough to hold the price below the baseline for a real share of the window blocks it.

That makes the contrast with MetaDAO sharper than the convergence it replaced, because the two protocols now put the price in different places. MetaDAO puts it instead of a vote: a decision market prices two conditional worlds against each other, and there is no electorate that could overrule the result. Caper puts it after a vote: the ballot decides what the proposal is and whether it has a mandate, and the market is handed a veto over that verdict rather than the decision itself. A caper can therefore fail in a way MetaDAO structurally cannot – a proposal that wins its ballot outright and is then sold down through its own baseline – and that failure is the design, not a defect in it.

The mechanical differences run the other way. MetaDAO’s markets are conditional: two books, one per world, with positions that unwind if their world does not happen, so the comparison is between two simultaneous prices for the same token and the proposal is what differs between them. A caper has one book, and compares the token’s price during the window against its own price from before the window opened. That is far cheaper – no conditional liquidity to seed, no pass and fail AMMs to keep honest, and any passer-by can crank the settlement – but it cannot separate the market likes this proposal from the market moved this week. The 24-hour recording delay and the −3% / +3% thresholds described above are MetaDAO’s answers to that problem inside a single proposal; a caper’s answer is that both sides of the comparison are integrals rather than spot reads, which defeats an atomic buy-resolve-sell round trip but not a genuine trend.

The second difference is that a caper has a membership as well as a market. Only a ballot mints the soulbound token: casting one mints exactly one v (VOTE_MINT, contracts/logic/src/lib.rs), and buying, selling and transferring mint none – the trade-side mint was removed in the genesis redeploy of 11 September 2026. Holding some v is a precondition of exit – the call asserts a non-zero v amount, withdrawn from the member’s own account, and aborts otherwise – and the share it pays out is w = (t · v) / (V · T), the same expression that weighs a ballot. So the exit opens only to members who have voted, and the one formula prices both. MetaDAO has no analogue: a decision market has no members to redeem, and holding META is a position rather than a claim. See how vote tokens are earned.

References

• MetaDAO docs – Introduction to Decision Markets. The “no voting, only trading” framing, and the 96-proposals / 14-organisations count since November 2023.
• MetaDAO docs – Finalizing Proposals. The lagging-price TWAP, the 24-hour recording delay, and the −3% / +3% pass thresholds.
• MetaDAO docs – The ICO. Four-day commitment window, the accumulator and fill boost, the discretionary cap, and the refund rule.
• MetaDAO docs – Token Mechanics. Proposal lifecycle, the absence of a hard cap, and mint authority held by the governance program.
• Futarchy AMM market API – tickers and META supply. Keyless, first-party; the source of every treasury and supply figure on this page, read 4 September 2026 (and previously 15 August 2026, cited above for the twenty-day comparison).
• MetaDAO transparency report. Incorporation documents, quarterly unaudited reports, and the daily account-level balance breakdown.
• META Token MiCA white paper (notified 2 July 2026) – PDF.
• metaDAOproject/programs. The Solana programs, with every deployed version and address listed in the analytics page.
• Kevin Heavey, “Futarchy as Trustless Joint Ownership”, Umbra Research. The essay MetaDAO's own documentation quotes at length.
• Crypto Briefing on the Rip Cars ICO, July 2026 – commitments mid-flight, and the caveat that commitments are not settled capital.