What Is a DAO? On-Chain Governance Explained by People Who Take Part in It

A protocol upgrade on Vaulta doesn't happen until 15 of the chain's 21 active block producers sign it. RORO is one of those 21, through EOS Support, so we spend real time on the voting side of decentralized governance. That gives us a fairly unromantic view of what decentralized governance looks like day to day, and it's the view this post is written from.

Short version: a DAO is a set of smart contracts that hold money and only move it, or change their own rules, when a defined group votes to do so. The rest is detail, and the detail is where DAOs succeed or fall over.

The moving parts

Strip away the branding and a typical DAO has four pieces.

A membership rule says who gets a vote and how much. Usually that's a governance token, where one token means one vote. Sometimes it's an NFT, a list of approved addresses, or a reputation score that can't be transferred.

A treasury is a contract (or a multisig wallet controlled by the governance contracts) holding the organization's funds. Nobody can spend from it directly. Spending is the output of a vote.

A proposal process defines who can submit a proposal, usually anyone holding a minimum number of tokens, and what a proposal contains. In well-built DAOs a proposal isn't a paragraph of text; it's the exact list of contract calls that will execute if it passes. Voters approve code, not intentions.

And execution rules set the quorum, the approval threshold, the voting period and, ideally, a timelock: a delay between a proposal passing and it executing, so that people who disagree have time to exit or raise the alarm.

On Ethereum, OpenZeppelin's Governor contracts paired with a timelock are the common starting point for all of this. Plenty of DAOs also vote off chain with signed messages (Snapshot is the usual tool) and then have a multisig carry out the result. That's cheaper and more practical, but it means a handful of signers are trusted to honour the vote. It's a fair compromise as long as everyone is honest about it.

The "autonomous" in the name is mostly aspirational. Contracts execute automatically, but people write the proposals, review the code, run the front end and argue on the forum. A DAO automates the final step of a decision. It doesn't automate the deciding.

The DAO, 2016, and what it taught everyone

The first well-known one was literally called The DAO. It launched on Ethereum in 2016 as an investor-directed venture fund and raised around 12.7 million ETH, a large share of all ETH in existence at the time. In June 2016 an attacker used a reentrancy bug in its withdrawal logic, calling back into the contract before it had updated balances, and drained about 3.6 million ETH.

Ethereum's community responded with a hard fork that moved the funds to a recovery contract. Part of the community refused the fork and kept the original chain going as Ethereum Classic.

Two lessons came out of it that still hold. Smart contract bugs in a treasury are permanent unless something outside the contract intervenes, which is why audits and timelocks are standard now. And "code is law" lasts exactly until the losses are big enough for people to change the law. Every DAO is ultimately backed by social consensus, whether or not it admits it. We cover the engineering side of this in our post on smart contracts in production.

Where token voting goes wrong

Token-weighted voting is easy to implement and hard to make work. These are the problems we see repeatedly.

Start with turnout, which is low almost everywhere. Most token holders never vote. Reading proposals takes time, and a single small holder has no realistic effect on the result anyway. The predictable outcome is that a small fraction of supply decides most questions, which makes quorum rules both important and awkward: set quorum too high and nothing passes, set it too low and a few wallets run the place.

Then there's concentration. One token, one vote means the largest holders, often early investors and the founding team, can outvote everyone else combined. The code is doing exactly what the rule says. Whether it's acceptable depends on what the DAO is for.

Votes can also be bought or borrowed. If voting power is a transferable token, it can be rented. The sharpest example is Beanstalk in 2022, where an attacker took a flash loan, used the borrowed tokens to push through a malicious proposal, and repaid the loan within the same transaction. Snapshotting balances at the time a proposal is created, rather than when votes are cast, closes that specific hole. Bribery markets that pay holders to vote a certain way are harder to stop.

Delegation helps, with caveats. Most mature DAOs let holders delegate their votes to someone who follows governance closely. That raises effective turnout and puts decisions in the hands of people who read the proposals. It also concentrates power in a small set of well-known delegates, which starts to look a lot like representative government, with all its usual tradeoffs.

The alternative is one person, one vote. It's fairer in principle, and it requires identity, because otherwise anyone can create a thousand addresses. Our BlockzHub platform runs on-chain polls on Vaulta with enforced one-user-one-vote logic for exactly this reason, using its identity layer (with optional KYC) to stop the same person voting twice. That works for community polls. It's a harder sell for a treasury where large stakeholders expect their stake to count.

Vaulta as a working example

The network we know best is a useful case study, even though it rarely gets called a DAO.

Vaulta, the chain formerly known as EOS, uses delegated proof of stake. Token holders stake and vote for up to 30 block producer candidates, weighted by stake. The 21 candidates with the most votes become active producers: they take turns producing blocks and collectively hold the network's highest permission. Anyone ranked below 21 is a standby producer, ready to step in if votes shift.

Changes to the system itself, such as a protocol upgrade or a change to a system contract, go through an on-chain multisig that needs approval from 15 of the 21 active producers. Someone proposes the transaction, producers review it, and it executes once 15 have signed. That's two thirds plus one, the same threshold the chain uses for finality.

So Vaulta has the same pieces as any DAO. Token holders choose representatives, and nothing changes until a supermajority of those representatives signs an exact transaction. The difference is that voters elect operators rather than voting on every proposal directly. It's representative rather than direct democracy, and in our experience that's a reasonable fit for decisions like protocol upgrades, which need people who will actually test the release and stay on call during the switch.

We've seen this up close. EOS Support was among the producers listed in the March 2025 network transition discussions, the period around EOS becoming Vaulta. Being on the producer side means reading what you're asked to sign before you sign it, and coordinating timing with the other producers so the switch happens cleanly. (We wrote more about the producer role in our post on crypto nodes.)

The weaknesses are the same as token voting anywhere. Large holders carry a lot of weight in producer elections, and turnout depends on people bothering to stake and vote. Vaulta doesn't hide those problems, but it does make the people with power easy to identify. You can look up the 21 producers, see who they are and vote against them. That accountability is underrated.

Legal wrappers, briefly

A DAO with no legal form can leave its members exposed, because in some places courts may treat an unincorporated group as a general partnership, which can make individual members liable for the group's debts. Some states have responded. Wyoming has offered a DAO LLC structure since 2021, which gives a DAO a recognised legal entity with limited liability. This isn't legal advice, and the right structure depends heavily on where members are and what the DAO does, so talk to a lawyer who has done this before.

Should you build one?

Only if there's a real reason for decisions to be made by a group that doesn't trust a single operator. A protocol that wants to outlive its founding team is a good candidate, and so is a treasury pooled by parties who don't fully trust each other. A startup that wants to say "decentralized" in its pitch deck isn't.

If you do build one, our view is fairly firm. Put proposals in code rather than prose. Snapshot voting power at proposal creation. Add a timelock long enough for people to react. Start with a guardian or veto role you plan to remove later, and write down when and how it gets removed. And decide early whether you want token-weighted, delegated or identity-based voting, because changing it afterwards means asking the people who currently hold power to give some of it up.

If you're designing governance for a protocol or community and want people who've been on the voting side of it, talk to us.