Juicebox

Juicebox is the money layer of the DAO stack. Where Snapshot and Tally / Cactus run a community's votes and DeepDAO measures the field, Juicebox is where a community actually raises and holds the money – a programmable, fully on-chain protocol for community fundraising and treasury management. Launched on Ethereum in July 2021 and governed by JuiceboxDAO, it lets anyone spin up a treasury that issues its own token against contributions, pays out on pre-set rules, and lets backers redeem their tokens back against the funds – the machinery behind the most famous crowdfunding DAOs of the last cycle. (Juicebox docs)

How funding cycles work

A Juicebox project runs on funding cycles – time-boxed configurations of the treasury's rules. During a cycle the parameters are locked; a project owner can queue changes, but they only take effect in the next cycle, so backers always know the rules can't be rewritten mid-flight. Each cycle sets a weight (how many project tokens each contributed ETH mints), a reserved rate (a share of newly minted tokens withheld for the team, contributors or other Juicebox projects), and payout splits that route outgoing funds to a fixed list of addresses or other treasuries. Contributors pay in, receive the project's token, and the splits distribute the proceeds exactly as configured – no multisig discretion required. (Juicebox protocol overview)

Project tokens and the redemption bonding curve

The primitive that made Juicebox interesting to DAOs is redemption. A project sets a redemption rate that governs a bonding curve for cashing project tokens back in: at a 100% rate the curve is linear, so redeeming X% of the token supply returns X% of the redeemable treasury; lower rates bend the curve to favour those who stay. That single dial is a built-in exit right – a backer who changes their mind, or whose project fails its goal, can burn their tokens and reclaim a pro-rata slice of the funds. ConstitutionDAO ran a 100% redemption rate precisely so contributors could get their ETH back if the bid lost, which it did.

ConstitutionDAO, AssangeDAO and the crowdfunding-DAO moment

Juicebox is best known for two raises that defined the 2021–22 crowdfunding-DAO wave. ConstitutionDAO used a Juicebox treasury to raise roughly $47M in ETH from over 17,000 contributors in about a week in November 2021, aiming to buy a first-printing copy of the US Constitution at Sotheby's; it was outbid at $43.2M by Citadel's Ken Griffin, and its 100%-redemption config let backers reclaim their funds afterward, with the $PEOPLE token persisting as a memento. Three months later AssangeDAO raised an even larger ~17,400 ETH (~$54M) through Juicebox for Julian Assange's legal defence, concluding in February 2022. Both are canonical crowdfunding-DAO case studies – proof that a rules-based on-chain treasury could coordinate tens of millions of dollars from strangers in days, with the refund path baked into the contract rather than promised by an organiser. (Juicebox on ConstitutionDAO)

Governance, fees and the V4 rewrite

The protocol governs itself: JuiceboxDAO is project ID 1 on Juicebox, steered by the $JBX token, and it takes a protocol fee charged only when funds are paid out of the Juicebox ecosystem; treasury-to-treasury transfers are free. That fee used to be a governance parameter and no longer is. In V3 the terminal held it in storage as uint256 public override fee = 25_000_000 out of a billion, 2.5%, which JuiceboxDAO could move up to the _FEE_CAP of 50,000,000, 5%. V4 hardcodes it: uint256 public constant override FEE = 25 out of JBConstants.MAX_FEE = 1000, still 2.5%, but a constant rather than a storage slot, with no setter anywhere on the terminal.

Read live from the deployed Ethereum terminal 0xdb96…7dbc, FEE() returns 25 (18 August 2026). Changing it now means deploying a new terminal and migrating projects onto it, not passing a proposal. Any page still describing the Juicebox fee as something JBX holders can vote is describing V3. The 2025 V4 release was an omnichain rewrite: projects can now deploy across Ethereum, Optimism, Arbitrum and Base in any combination, it introduced revnets (immutable, ownerless revenue-sharing treasuries), routed protocol fees to a $NANA revnet, and renamed the core concepts (funding cycles became “rulesets,” redemptions became “cash outs”). The current on-chain contracts live under the Bananapus / nana-core organisation, with V5 shipping as a bug-fix fork of V4. (Juicebox developer docs)

How much of Juicebox has actually moved to V4

Juicebox publishes no adoption dashboard, but the registry is an ERC-721 with a public counter, so the answer is one call per chain. The legacy V2/V3 JBProjects on Ethereum returns count() = 863: every project minted into that registry since 2021, the crowdfunding-DAO era included. The V4 registry, deployed at the same address on all four chains, returns:

All read on 18 August 2026 and again, unchanged to the unit, on 8 September 2026. The four are separate contracts that happen to share an address, and an omnichain project registers in more than one of them, so these counts overlap and must not be added up. Nor are they a headcount of live treasuries: a registration is an NFT mint, not evidence that anyone ever paid in.

What the numbers do settle is the shape of the migration. Four years of V2 and V3 produced 863 Ethereum registrations; V4, live since 2025, has 182, and the newer chains carry a real share of the traffic rather than none. The protocol did not fork its user base onto a rollup and abandon mainnet, and it did not carry the old base across either. For a DAO choosing a treasury protocol, the relevant fact is that the contracts a 2021 project runs on are not the contracts a 2026 project deploys, and that the older registry is where almost all of the history sits.

Reading the same five counters three weeks later adds the fact a single snapshot cannot carry: all five registered nothing in three weeks. Not one project was minted into the legacy registry or into any of the four V4 registries between 18 August and 8 September 2026. Three weeks is a short window and a zero in it is not a verdict on the protocol – registrations are lumpy, and a treasury that already exists keeps working whether or not a new one is created beside it. But it does bound the reading above: the migration to V4 is not in progress, it is where it landed. A directory quoting 182 as evidence of momentum would be quoting a number that has not moved.

How Caper approaches this

Of all the DAO tooling, Juicebox is the closest cousin to a caper's economic core: both are on-chain treasuries that issue a token against contributions and let holders redeem against the pooled funds along a curve. The difference is configurable protocol versus fixed protocol. A Juicebox project owner picks and can reconfigure the weight, reserved rate, redemption rate and splits each cycle – flexibility that suits a bespoke fundraise.

A caper instead ships one set of rules every caper shares: a single bonding curve that prices both buys and sells, a founder paid half of a skim on every trade and capped by the contract at 45% of the XRD and 5% of the tokens a trade moves, and an exit for a fair share set by the member's earned vote weight – so exit tracks participation, not just token balance, and no operator can retune the terms after people have paid in. Juicebox optimises for a fundraise you shape yourself; Caper optimises for a standing organisation whose economic guarantees are the same for everyone and can't be quietly changed.

References

• Juicebox, developer documentation and protocol overview.
• Juicebox, “ConstitutionDAO” case study (Juicebox attribution and config).
• CoinDesk, “ConstitutionDAO Outbid for First Printing of America's Founding Document” (raise, contributors, winning bid).
• CryptoPotato, “AssangeDAO Concludes Fundraise After Securing 17,422 ETH” (final total).
• Bananapus, nana-core (Juicebox V4 contracts).
• Bananapus, JBMultiTerminal.sol and JBConstants.sol (the V4 fee constant and its denominator).
• jbx-protocol, JBPayoutRedemptionPaymentTerminal3_1_2.sol (the V3 settable fee and its 5% cap).
• Project counts read on 18 August 2026 with eth_call against count() on the V4 registry 0x0b538a02610d7d3cc91ce2870f423e0a34d646ad (Ethereum, Base, Arbitrum, Optimism) and the legacy registry 0xD8B4359143eda5B2d763E127Ed27c77addBc47d3 (Ethereum).