DAOstack was the first DAO framework built around the claim that a DAO's binding constraint is attention rather than capital. Its three layers – the Arc contracts, the Infra voting machines, and the Alchemy web client – shipped a governance design called holographic consensus, in which an open prediction market decides which proposals the membership should even look at. Where Moloch made leaving the primitive and Aragon made permissions the primitive, DAOstack made agenda-setting the primitive. The framework is now closed, and because its own domain has since changed hands, its record is easier to read on-chain than on the web.

The three layers

Arc held the organization: an Avatar contract that owned the assets, a Reputation contract that tracked non-transferable voting power, and Schemes that were the only things allowed to act on the Avatar. Infra held the voting machines – the Genesis Protocol above all – as swappable modules, so a DAO could change how it decided without redeploying what it owned. Alchemy was the user-facing client where members read proposals, voted their reputation, and staked GEN.

Two of those choices outlived the framework. Splitting the treasury-owning contract from the modules permitted to touch it is the same separation Zodiac later marketed as modularity, and non-transferable membership is now a category of its own under soulbound tokens. What did not outlive it was the assumption that members would hold reputation they could never sell while a separate crowd traded the token that set the agenda.

GEN: a fixed supply that outlived its purpose

GEN was minted once and never again. Read directly from the token contract on Ethereum mainnet at block 25,753,484 (14 August 2026, 13:10 UTC), 0x543Ff227…DCAddf returns name() “DAOstack”, symbol() “GEN”, 18 decimals and a totalSupply() of exactly 60,000,000 tokens across roughly 7,250 holding addresses. There is no minter, no emission schedule and no burn – the entire monetary design was that a scarce, transferable token would be worth holding because staking it on proposal outcomes paid.

The token still trades. Blockscout's transfer index for GEN in 2026 is almost entirely DEX router traffic – swap and swapExactAmountInOnUniswapV2 calls a few times a month. Nothing in it touches a proposal. That gap between a live ticker and a dead mechanism is the honest state of most wound-down governance tokens, and it is why a metrics page built on price or transfer count will mis-read a framework as alive years after its last vote.

The Genesis Protocol, dated on-chain

Blockscout's mainnet index carries 13 verified GenesisProtocol contracts. Each one is a separate voting machine wired to one or more Avatars, and each one records its whole life in events: NewProposal, Stake, VoteProposal, StateChange, ExecuteProposal, Redeem. Reading the last of each across all 13 gives a death certificate that no write-up has published.

• The busiest deployment, 0x917a2C44…0A31, took its last stake on 25 July 2021 and its last vote two days later.
• 0x1C18bAd5…3055, whose events all name DXdao's Avatar, ran a complete proposal cycle in March 2023 and then stopped; its last event of any kind is a RedeemReputation on 11 November 2023.
• 0x332B8C97…5B84, also DXdao's, is the last machine standing. Everything after it is silence.

So the common summary that DAOstack “wound down around 2021” is off by nearly three years on the only measure that counts – whether the prediction market was still pricing anything.

The last holographic-consensus proposal

Proposal 0xa090fb30…017d, on 0x332B8C97…5B84 for the DXdao Avatar, is the final one the mechanism ever processed. Its whole life is four days of events:

• 17 Jan 2024, 17:02:59 UTC – NewProposal by 0x091F982b…4861.
• 17 Jan 2024, 17:27:35 UTC – Stake of 666 GEN for the proposal passing, by the same address that filed it. (A first attempt 16 blocks earlier staked 666 wei, a rounding error of a token.) The stake carried the proposal to state 4, PreBoosted.
• 18 Jan 2024, 17:29:23 UTC – StateChange to state 5, Boosted. The enum is read from the deployed, verified source: enum ProposalState { None, ExpiredInQueue, Executed, Queued, PreBoosted, Boosted, QuietEndingPeriod }.
• 19 and 21 Jan 2024 – two votes in favour, carrying 500 and 75,621.64 units of reputation.
• Then nothing. The last ExecuteProposal on that contract is from April 2023, so the final boosted proposal was never executed at all.

It is a fair epitaph for the design. The prediction market was supposed to let outsiders earn by surfacing proposals a distracted membership would ratify; by the end the only staker was the proposer, betting on himself to get his own item looked at, and the item that cleared the boost bar still went nowhere. An attention layer stops working when the attention it was rationing has already left, which is the failure how DAOs fail treats as the general case.

Where the artefacts are now

The code is frozen but readable. None of the four principal repositories is archived, and all four stopped moving in the first quarter of 2023: arc last pushed 24 January 2023, infra the same day, the Hackers Kit on 13 January 2023, and alchemy on 1 March 2023 with its default branch still set to dev. An unarchived repository whose default branch was never merged to master is the ordinary shape of a project that stopped rather than concluded.

The hosted surfaces are gone in two different ways, and the difference matters to anyone citing this framework. alchemy.daostack.io, the client itself, has no DNS record at all – an unambiguous death. daostack.io, by contrast, answers HTTP 200 today and is not DAOstack: the Internet Archive's 11 October 2023 capture still shows the DAOstack homepage, while the domain now resolves to a WordPress site of crypto-casino guides. It is a live illustration of the citation that dies at status 200, and the reason the surviving links worth keeping are the GitHub-hosted Hackers Kit, the archived GitBook, and the chain itself.

A third surface is worse than dead, because it is the one a researcher actually needs. DAOstack's DAOs deliberated on daotalk.org, and that domain answers HTTP 200 today while serving Indonesian online-gambling affiliate marketing under the title "DOLLY4D # Menjual Akun Mahjong Cepat Maxwin Hanya 10K Saja." Every Discourse endpoint returns an identical 1,251-byte 404, as does a nonsense path, so it is a catch-all rather than an outage. The Internet Archive dates the handover: live forum captures of 13–14 KB through 23 July 2024, a 1,785-byte husk on 2 December 2024, and from 16 December 2025 the affiliate site, byte-identical by digest through 6 June 2026. Two of the three domains in this framework's history now belong to gambling operators. The record survives on-chain, which is how DXdao, the deployment whose Avatar appears throughout this page, can be reconstructed at all.

How Caper approaches this

Caper starts from the same premise DAOstack did – that a vote nobody attends is not legitimacy – but it prices attention inside the vote instead of alongside it. A member's vote weight is the canonical product w = (t · v) / (V · T): tokens held multiplied by votes actually cast, normalised by the totals. There is no second token to stake and no outside crowd to pay, because the thing that would have been bought – a claim on the agenda – is only earned by participating.

The sharper contrast is what happens to someone who never participates. DAOstack let reputation sit idle indefinitely; a caper makes non-participation cost the one thing it protects. Exit requires handing back both a governance-token bucket and a soulbound vote-token bucket – assert!(vote_tokens.amount() > 0, "Exit requires vote tokens") in the contract source – so a holder who never voted has no exit to take, and the share of the treasury an exiting member receives is that same w. Voice and exit are one number, verified against contracts/logic/src/lib.rs.