DAO Governance
DAO Governance
On-chain governance is the mechanism by which decentralized protocols make decisions without a central authority. Token holders vote on proposals — parameter changes, protocol upgrades, treasury allocations — and the result executes automatically via smart contract, with no human intermediary in the execution path.
The core tension in DAO governance is the same as in all governance design: the people with the most votes are usually the people with the most capital, not the people with the best judgment or the most legitimate stake in the outcome.
Core Mechanics
Token voting The dominant model. One token = one vote. Holders submit proposals, reach quorum (a minimum participation threshold), and pass by simple or supermajority. Compound and Uniswap pioneered this model.
Vote delegation Token holders can delegate their voting power to another address — a protocol politician, a research organization, or themselves. Compound's delegation model has become the standard: tokens accrue voting power to the delegated address, allowing passive holders to participate without direct involvement. Delegation enables a small number of active participants to carry the governance load, but concentrates power in whoever attracts delegation.
Timelock Winning proposals don't execute immediately. They enter a timelock queue — typically 24–72 hours — during which token holders can observe the outcome and exit the protocol if the change is unacceptable. The timelock is the primary safety mechanism against governance attacks; it gives the community time to react.
Guardian roles Some protocols maintain a multisig or guardian address that can veto or delay timelock execution in extreme situations. This is a trust tradeoff: the guardian provides a last-resort safety valve but introduces centralization. MakerDAO's Emergency Shutdown Module follows this pattern.
Attack Vectors
Governance attacks An attacker accumulates enough tokens — or borrows them via flash loans — to pass a malicious proposal. The 2023 Tornado Cash governance attack is the canonical case: an attacker distributed 1.2 million malicious TORN tokens to existing holders in an airdrop, then used the accumulated voting power to pass a proposal that assigned all governance control to themselves, effectively draining the governance treasury. The attack succeeded because the proposal was disguised as a routine grant and the voting power was real — borrowed, not synthetic.
Vote buying Token voting creates direct markets for governance influence. Platforms like Votium and Hidden Hand let protocols bribe Curve and Convex holders to direct CRV emissions toward their pools. This is legal and transparent, but it separates the voting decision from any judgment about protocol quality — the vote goes to whoever pays the most, not whoever has the best argument.
Low quorum / last-minute votes Most governance systems have low real participation. A determined actor who waits until the last few hours of a vote can tip outcomes with fewer tokens than expected, because passive holders who didn't vote are effectively absent. Snapshot and Tally data consistently show that major governance decisions pass with less than 5% of total supply voting.
Proposal spam Some protocols have no minimum token threshold to submit proposals. A spam attacker can flood the governance interface with malicious proposals, causing voter fatigue and increasing the chance that a bad proposal passes unnoticed.
Governance Models
Compound / Uniswap pattern Governor Bravo (Compound) and its forks are the most widely deployed governance contracts: propose, vote, timelock, execute. Works well for infrequent high-stakes decisions. Poorly suited for continuous parameter management (liquidation thresholds, interest rates) because governance speed is measured in days.
MakerDAO (Sky) model
Continuous approval voting on core parameters via dss-exec-lib spells, with weekly executive votes. The MKR voter base is small and engaged; the risk is that a concentrated MKR position can pass any weekly vote. Sky (the rebrand) adds SubDAOs for more distributed governance.
veToken model (Curve, Velodrome) Vote-escrowed tokens: lock CRV for up to 4 years to receive veCRV, which controls gauge weight voting (directing CRV inflation to specific pools). Creates protocol stickiness — locked tokens can't be sold — and a market for governance influence via bribe protocols. The model has been widely forked.
Optimism Bicameral Governance Two houses: Token House (one token = one vote, handles protocol upgrades and treasury grants) and Citizens' House (non-transferable badges, handles retroactive public goods funding). The design is an explicit attempt to separate capital-weighted decisions from legitimacy-weighted decisions.
Governance Fatigue
Participation in token governance has consistently disappointed protocol teams. Typical patterns:
- Most token holders never delegate or vote
- A small number of delegates or whales drive most decisions
- Contentious votes (protocol changes, treasury allocations) see higher participation; routine parameter changes often fail quorum
- "Governance fatigue" — holders who voted early, got frustrated with the process, and stopped — is endemic
The result: most DAO governance is formally decentralized but operationally controlled by a small group of core contributors, large funds, and active delegates. This isn't necessarily bad — representative democracy has the same structure — but it's different from the "community-governed" framing.
Progressive Decentralization
Most protocols launch with significant centralization (admin keys, multisigs, upgrade authority) and gradually hand off control to governance as the protocol matures. The op-stack governance progression is the best-documented example: Optimism started with a core team multisig and has progressively moved authority to the bicameral governance system over 2022–2025.
Related
eigenlayer · real-world-assets · optimism · ethereum-l2s