Compound DAO

The Compound DAO governs Compound, one of DeFi's original algorithmic money markets, where users supply assets to earn interest and borrow against overcollateralized positions. Its lasting influence on the industry is less about the protocol than about how it is run: Compound authored the Governor + Timelock contract pattern that most on-chain DAOs – Uniswap, ENS, and hundreds more – still copy today. Control over the protocol's markets, its risk parameters, and its treasury rests entirely with holders and delegates of the COMP token, deciding by token-weighted on-chain vote.

COMP and the birth of liquidity mining

Compound distributed its COMP governance token beginning in June 2020, streaming it automatically to anyone supplying or borrowing on the protocol rather than selling it in a sale. The mechanism – later universally called liquidity mining or "yield farming" – kicked off the "DeFi summer" capital rush and became the default playbook for bootstrapping a DAO's token distribution. COMP has a fixed supply of 10,000,000 tokens and, like most governance tokens, originally conferred votes but no direct claim on protocol revenue. (Compound: the COMP token; CoinDesk, 2020)

Community governance went live on 16 April 2020, two months before COMP began streaming to users. Within days the first proposals arrived: one listed Tether's USDT with no collateral factor, and one from the Dharma wallet, the first outside application to propose a change, replaced the DAI interest-rate model because Maker had set its stability fee and savings rate to zero (Boardroom, April 2020).

Governor Bravo – the on-chain governance template

Compound governance is three contracts working together: the COMP token (which tracks voting weight and supports delegation), the Governor Bravo module (which runs proposals and vote tallies), and a Timelock (which holds the admin keys and enforces a delay before any approved change executes). The canonical parameters are baked in:

• Proposal threshold – 25,000 COMP. An address must hold or be delegated at least 25,000 COMP to submit a proposal, a spam filter that also concentrates proposing power.
• 2-day review, then a 3-day vote. After a short review delay, voting is open for roughly three days; voting weight is snapshotted at the block the proposal starts, so tokens bought afterwards do not count.
• 400,000-vote quorum. A proposal passes only with a majority and at least 400,000 votes cast in favour.
• 2-day timelock. A passed proposal is queued in the Timelock and can be executed two days later, giving the community a window to react before code changes land.

"Bravo" was the 2021 upgrade to the original Governor Alpha, adding an upgradeable delegator pattern and vote-reason strings. Because the contracts were audited, permissionless, and copy-pasteable, they became the reference implementation for the whole propose → vote → timelock → execute lifecycle – the same skeleton most protocol DAOs run on. Forks are everywhere: Ondo Finance’s Flux Finance lending market runs a Governor Bravo fork with a 100,000,000-ONDO proposal threshold, one of many protocols that inherited these exact contracts wholesale. (Compound v2 docs: Governance; GovernorBravoDelegate.sol) In January 2025 the DAO replaced the Bravo contracts with the Compound Governor, an OpenZeppelin-based upgrade built by ScopeLift that keeps Bravo's incrementing proposal numbers and guardian roles (design post). Read from the contract on 24 September 2026, it still requires 25,000 COMP to propose and 400,000 votes for quorum.

Governance in practice – and under attack

Compound's model has been stress-tested in public. In 2024 a group nicknamed the "Golden Boys," led by an activist known as Humpy, accumulated and delegated enough COMP to pass Proposal 289 – a bid to move roughly 499,000 COMP (~$24M) of DAO reserves into a "goldCOMP" yield vault the group would influence. It cleared the on-chain vote before the community mobilized; after intense backlash the proposers agreed to rescind it.

The episode is now a textbook case of a governance attack executed entirely within the rules – no exploit, just concentrated voting power aimed at the treasury. Its timing is a case in its own right: 82% of the support arrived in the vote's final 34 minutes, which is the textbook shape of a late-vote swing and the reason Governor Bravo's fixed deadline is now widely paired with an extension or an objection period. (Unchained, 2024)

Proposal 608: who controls Institutional Comet (September 2026)

On 24 September 2026 COMP holders passed Proposal 608, a resolution with no executable action. It requires the operators of Institutional Comet, a Compound market for institutional borrowers built under the DAO-funded V4 programme, to place its admin and upgrade permissions behind a control structure the DAO can revoke without their consent, within 60 days. Its author, forum user ugurmersin, argued that the market's permissions sat with Safes that Compound Governance could not remove. The Compound Governance Working Group (CGWG) and the Compound Foundation said they shared that end state and that work on it had started, but objected to the resolution's first clause, which declares that the DAO's earlier approvals never authorised an irrevocable delegation of control (forum thread).

Most of the For vote came from one address. It cast 1,773,839 COMP For (vote transaction); the 15 addresses voting Against held 984,290 between them, and delegates who explained their votes on the forum, among them Tané, Curia, DAOplomats and Avantgarde, voted Against. The proposal closed with 1,799,026 For (Tally). Without that one address the For side would have held about 25,000 votes, short of the 400,000-vote quorum.

During the vote the delegate Blockful reported that the proposer's own voting power had been delegated on 14 September from an address it linked to Humpy, the organiser of Proposal 289; who controls the 1.77m-COMP address has not been established. The Foundation said it would publish a transition plan whatever the result, and the author asked for a permissions map within 10 business days of passage. The Governor queued the proposal on 27 September and executed it on 29 September 2026, after the two-day Timelock delay. The transition plan it asked for arrived on 8 October, and the transition itself is due around 23 November.

After 608: an optimistic council and two cancelled proposals (October 2026)

The Compound Foundation, Woof and the CGWG published the plan Proposal 608 asked for on 8 October 2026 (forum post). Institutional Comet is a USDC market on Ethereum that opened on 8 September; on 8 October it held about $20.6M supplied against $36.2M of collateral, with $16.3M borrowed. Today a 3-of-4 multisig sets its parameters, a 4-of-7 multisig deploys and upgrades it, and the 5-of-9 Pause Guardian that covers Compound's other markets can pause it. The plan moves administration to an optimistic governance council drawn from the parties who run the market now: the council proposes changes, COMP holders can veto any of them, and governance sets the council's configuration. The full specification and timeline are promised before the end of October.

The Foundation's case against handing the market to ordinary votes is about speed. Comet compiles its supply caps, collateral factors and rate curves into the market's code, so every parameter change is a redeploy, and a governance proposal takes about seven days from submission to execution. Its evidence is April 2026: after unbacked rsETH from the Kelp DAO bridge exploit was posted as collateral on 18 April, the Pause Guardian could only pause whole markets, and withdrawals stayed frozen for about 7 days on the USDC Comet and 16 to 20 days on the WETH and wstETH Comets while fixes waited on the governance cycle. The council targets three days or less for an uncontested change.

Proposals 612 and 613 show the other side of that bargain. Both passed with 1,887,953 COMP For to 1,041,003 Against (proposalVotes on the Governor), and as with 608 one delegate's vote decided them. Proposal 612 would have raised the Treasury Management Committee's escrow cooldown from 2 days to 10 and given the Governor executor and canceller roles over the TMC's timelock, so that a vote could stop a withdrawal before it cleared (author's account). Proposal 613 would have closed the Ecosystem Protection and Continuity Fund created by Proposal 535 and sent its funds and the mainnet SVR revenue to the Governor Timelock (forum).

On 7 October 2026 the Proposal Guardian, held by the Community Multisig, cancelled both before execution, and the Governor's state() for each now reads Canceled. The signers said both met two of the criteria delegates set for the Guardian in Proposals 304 and 538: a single entity controlled the vote, since without the 1.75m COMP cast by one delegate both would have failed, and more than 400,000 COMP voted against (Community Multisig). The author called the cancellation of 612 unjust and offered a narrower alternative to 613 that keeps the fund open and moves its keys from Foundation-controlled Safes to the 5-of-9 Community Multisig.

On 22 October 2026 the Guardian cancelled Proposal 618 before voting opened. It was a signal proposal with no on-chain actions, from the same proposer, asking that the fund move from the wallet the Foundation controls to the Community Multisig. The signers wrote that the proposer had since been banned from the forum and that putting each new proposal from the same source to a full vote prolonged the disruption (Community Multisig). Humpy replied on the forum the same day that none of the Guardian's criteria under Proposal 304 had been met, since the proposal had not been voted on (forum).

So in one fortnight the DAO used both remedies for concentrated voting power that the wider industry argues about: a small council that can veto a passed vote, and a council whose changes stand unless token holders veto them. The governance-attack question is the same in both: who may overrule a majority that a single holder can manufacture.

The DAO also now puts idle reserves to work through outside managers, and that programme has run its full course in public. Its Treasury Management Committee (TMC) opened a request for proposals on 3 June 2026 for an initial ~$20–25M of reserves. Submissions closed on 24 June with public bids from managers including Coinbase Asset Management, Invesco, karpatkey, Steakhouse Financial, Yearn, Delphi and RockawayX; on 17 July 2026 the TMC shortlisted six – Coinbase Asset Management, Invesco, kpk (karpatkey), Steakhouse Financial, Upshift/Flowdesk and Yearn.

The resulting programme splits the DAO's USDC into two sleeves. The RWA sleeve puts $4M into tokenised real-world assets held directly on the DAO's own balance sheet – $1M in Invesco's USTB and $3M in Coinbase Asset Management's CUSHY – with no manager fee layered on top of the underlying fund fees.

The primary allocation sleeve is up to ~$12M curated by Steakhouse Financial in a dedicated single-chain Ethereum vault: the DAO transfers no custody, holds the vault shares, keeps a guardian veto over any change to the vault's structure and may terminate on 30 days' notice, for 0.5% of NAV annually with no performance fee, targeting SOFR + 200bp. Delegates ratified the combined $16M programme in a Snapshot signalling vote that closed 21 August 2026: 512,405 of 631,382 COMP (81.2%) chose “Support as Proposed” across 15 votes. The DAO's WBTC and ETH stay in Treasury Escrow for now. (RWA sleeve memo; primary allocation memo)

The venue that primary sleeve runs on became the programme's sharpest point of contention, because it belongs to a competitor. The TMC's own memo places the mandate in a Morpho Vault V2 that routes through a purpose-written BoxAdapter into Steakhouse's ERC-4626 “Box”, whose funding modules are Aave v3 and Morpho Blue – so the majority of the non-RWA book is lent through Morpho markets. Compound's request for proposals had said it was “not looking for proposals that simply route funds into a competing lending protocol as a primary strategy”, and a delegate raised exactly that on the thread on 15 August 2026.

The memo answers the objection head-on rather than around it: the TMC acknowledges that Morpho is a competing lending protocol and says it is willing to meet its chosen curator where that curator is most active, arguing that the DAO is underwriting the curator rather than the venue – on Morpho a curator selects the markets, where on Compound v3 governance lists them. It is an unusually explicit case of a DAO treasury being deployed into a rival protocol by design, with the trade-off argued in public rather than buried. (primary allocation memo and thread)

Compound is not alone in this: the collective on-chain DAO treasury crossed $26B in Q1 2026 per DeepDAO, and how to steward it has become a governance question in its own right – see DAO treasury management.

Service-provider renewals go to Snapshot, and one fails (September to October 2026)

Compound pays outside firms for security and risk management, and in autumn 2026 the DAO put those renewals to Snapshot signalling votes rather than on-chain proposals, with the Treasury Management Committee paying from the funds it administers. The Compound Foundation posted the security renewal on 16 September 2026: ChainSecurity and Certora for audits and formal verification, zeroShadow for monitoring, incident response and proposal review, and ChainPatrol for phishing takedowns, $1,695,000 in all against $2,036,000 the year before. The Snapshot vote closed on 26 September with 627,011 COMP For and 1,799,026 Against across 16 votes. The TMC wrote on 8 October that two addresses had cast the Against votes, the dCOMP delegate address it attributes to Humpy, with about 1.77m COMP, and the delegate ugurmersin, who receives his delegation from Humpy, that every other delegate had voted For, and that it would fund the renewal anyway through four payment streams, because delegate direction was clear and the only opposition was the concentrated position the Proposal Guardian had already acted against. ugurmersin replied on 9 October that the Snapshot had failed, so nothing had been approved, and that the treasury-management mandate from Proposal 580 lets the TMC administer and recommend spending, not approve it.

Two risk-management proposals followed on the forum. Gauntlet proposed a sixth year on 15 October at a fixed $450,000, paid quarterly in arrears in USDC and backdated to 30 September, with treasury support moved to the TMC. A delegate, Tane, noted that 30% of the Year 5 fee had been held as a conditionally refundable insolvency fund, of which the Foundation reports about $280,000 applied to liquidations after the deUSD collapse and $410,000 returned to the DAO, and asked whether Year 6 ends that arrangement; Arana Digital supported the renewal on 19 October, and a forum member, cmrn, opposed renewing without a competitive request for proposals, citing the Morpho–Polygon growth programme Gauntlet had recommended winding down. Yearn Curation proposed a 12-month risk engagement on 20 October at a fixed $300,000, and Arana Digital answered on 21 October that it saw no need for a second risk vendor while Compound v3 settles after its consolidation. Neither proposal had reached a vote by 23 October.

How Caper approaches this

Compound's Governor Bravo showed the industry how to make a vote binding on-chain – but Prop 289 also showed its soft spot: enough delegated tokens can point the whole treasury at an exit, and dissenters have no way out but to sell COMP on the open market. A caper keeps the binding-execution property – every approved PAYOUT or INVEST proposal settles on-chain – but pairs it with a standing exit: a member who dislikes where a vote is heading can redeem a treasury share sized by their own vote weight directly rather than dumping into a market. And Caper's voting weight deliberately does not reduce to "tokens delegated at a snapshot block," which is the exact lever Prop 289 pulled. The mechanics are described neutrally on those Caper pages.

References

• Compound documentation – protocol and governance reference.
• Compound v2 docs: Governance – Governor Bravo parameters.
• Compound Community Forum (comp.xyz) – where proposals originate.
• GovernorBravoDelegate.sol – the on-chain governor contract.
• Unchained: Golden Boys agree to rescind Proposal 289 (2024).
• Compound RFP: Treasury Management Program (2026) – the RFP, the 17 July shortlist and the sleeve announcements.
• Compound forum: Transfer Ultimate Control of Institutional Comet to Compound Governance (2026) – Proposal 608, its debate and the post announcing its passage.
• ScopeLift: New Compound Governor design approach (2024) – the OpenZeppelin-based successor to Governor Bravo.