DAOs in 2026 feel like Schrödinger’s cat: they are dead AND alive at the same time.
That’s the best comparison I can think of to capture the whole sentiment about all the conversations I have had at EthCC.
The part that died
It is the naive vision we had about DAOs that died and is now buried. The naive vision was that DAOs would function like anarchy, with a group of uncoordinated actors managing to cooperate without any imposed authority.
The problem is that among these actors, at least one eventually asserts its authority over the others.
A major sign confirming this observation was the centralisation of governance within MakerDAO/Sky: the distribution of active votes suggests that voting power is spread across several delegates, except that it is actually the founder, Rune Christensen, who has delegated his voting power to these entities.
Then it was Aave that brought everyone round to the same view. Before December 2025, Aave had a reputation for being “the only DAO that actually worked” in the sense that the voting power of delegates came mainly from other AAVE token holders who weren’t eager to take over the protocol, and all changes to Aave could be anticipated on the governance forum.
However, following the dispute between Aave Labs and the Aave Chan Initiative, Aave Labs decided to accumulate enough voting power to ensure that the proposals they wanted would either be accepted or rejected👇
It is now clear that Labs has imposed its authority over the entire Aave ecosystem. And the message clearly got through to BGD Labs and ACI, as both have decided to leave Aave.
The most prominent examples of DAOs that have been successful are no longer decentralized or autonomous.
Schrödinger’s DAO
Now, if DAOs are dead and buried, why are the governance forums still open, and above all, why do the actors who imposed their authority continue to vote?
Thanks to this tweet, I found the answer👇
Leaving aside the issue of participation rates (which is a problem in its own right within DAOs), the point made in the tweet perfectly sums up the situation: if DAOs are dead, actors wouldn’t bother voting for the proposals and start to change things on their own, which is not the case.
At Maker/Sky and Aave, there have already been delegates like PaperImperium and ACI who have attempted to accumulate voting power to prevent authoritarian takeovers, but Rune and Aave Labs responded by acquiring even more voting power to retain control.
It is the onchain vote that counts. If DAOs were truly dead, these votes would never have mattered in the slightest.
On top of that, Maker/sky and Aave continue to emphasise the governance approval and the “community” aspect (even though this isn’t really the case), because the DAO has always been part of their identity and renouncing it would cause them far more problems than it would solve.
Long story short, this is Schrödinger’s DAO: DAOs are captured so they are dead, but onchain votes still matter, so they are alive at the same time.
How to capture a DAO
At present, we don’t know how to create a capture-resistant DAO, assuming that such resistance is even feasible.
But given how long DeFi has been trying to get DAOs to work, we know certain things shouldn’t be done.
1 token = 1 vote
This is the model used in most DAOs, including the largest ones, and its recurring issues are well known:
- Low participation. The ratio of tokens actually used for voting to the total number of tokens is only a few per cent for the most active DAOs, even though holders have the option to delegate their voting power (Yes, we have a 30% participation ratio on Sky governance, but considering what was mentioned above, we know that ratio can’t be taken seriously)
- Plutocratic tendencies. In this system, anyone can purchase tokens within the limits of available liquidity, and there is no distinction between legitimate governance participants and hostile participants. Ultimately, the vote belongs to whoever has the money, not who is legit.
But the biggest concern is the lack of mechanisms to prevent abuse in this type of governance. According to anticapture, most of the DAOs have the following traits:
And because of all these oversights, numerous DAOs were exploited (Beanstalk, Build Finance, Compound GoldenBoyz…), whereas the mere presence of a timelock or restrictions might have prevented certain accidents from occurring.
The 1 token = 1 vote model was introduced to replicate shareholder ownership, but it turns out that it was not the most suitable for that purpose.
Complete control through governance
In the largest DAOs, it is often said that “those who hold the token own the protocol,” which is true in the sense that token holders control the treasury, future updates, and everything else.
Token holders’ participation is essential, but this role is demanding: it involves in-depth knowledge of the protocol and monitoring its activity. Not everyone has the time or energy to do so, hence there is a recentralization through the delegation platforms.
T****he more utility a token has, the more voting power is concentrated among a small number of participants, so all it takes for those participants is to collude to take over the DAO. The brute-force solution is to increase the number of participants and ensure that there is no obvious leader, but in the end, the decision process takes much more time for everything.
Apart from this problem, the ability to control the DAO increases the incentive to capture. Controlling the treasury, or modifying the system to suit our needs (including stealing users’ funds), are all additional incentives to undermine governance.
How to build different
Hardcoded specific utilities
If protocols have been exploited due to governance attacks, it is also because those protocols were designed in a way that made such attacks possible. But what if there are hard-coded limitations?
By limiting governance to specific utilities, attackers are unable to go beyond them:
Aero: The voting power of AERO only allows it to direct AERO emissions and get swap fees from the voted pools. Disrupting the governance would change the yields for liquidity providers, but the DEX still works the same.
Liquity: 25% of interests paid by borrowers are distributed weekly to support Liquity’s growth. Disrupting it would just mean less money for growth, meanwhile, 75% of interests paid are still directed to stability pools, and the way the stablecoin issuer operates remains unchanged.
Polaris: Polaris introduces the Stewardship, where stewards have power over quantitative parameters like interest rates, bounded within minimum and maximum values. If it is disrupted, the stablecoin issuer, the bonding curve, and the liquidation system can’t be changed.
Specific utilities also have the advantage of requiring less expertise to make decisions.
Different voting systems
As mentioned earlier, there is no distinction between legitimate and malicious participants among token holders. However, we can implement mechanisms that allow us to make that distinction.
Time-boosted voting. We adjust the voting power of tokens over time, using several different methods:
- veTokens and veNFTs imply locking the tokens for a given time period. The longer it is, the more we have voting power. The only problem is that we still can purchase massive voting power in one go.
- Now, something more appropriate is staking tokens with zero voting power at the beginning, but getting increased voting power over time, noting that voting power is reset to zero when unstaked. So attackers must wait before their voting power gets significant, and everyone can anticipate their commitment.
Quadratic voting. This is something already in production with the “quadratic funding”, where crowdfunding depends not only on the amounts donated, but also on the number of donors who have contributed to a project.
We can do the same with governance, where voting power depends on the number of tokens, but also on another statistic, which can be the number of voters or something completely different, like bonus voting power based on KPIs
That said, the quadratic system introduces further weaknesses into the voting process. Including the number of voters encourages Sybil attacks, whilst choosing the wrong KPIs may hurt the protocol more than anything else.
Speaking of Sybil attacks, there are already mechanisms to prove legit voters (Human Passport), and technologies like zkTLS will certainly help to identify voters without KYC.
Some teams are trying to build DAOs differently, and we must build differently anyway.
DAOs are a tool of power
When people say DAOs are dead, it is just the innocent vision of DAOs that died. Ultimately, DAOs should be seen as tools of power:
- Even when the votes are rigged, governance works as intended
- DAOs fail because there is a greater rational incentive to seize control of governance for one’s own benefit than there is to build something for the DAO
They allow control over a protocol while being marketed as community-driven.
So just like any tool of power, we must expect actors to try to capture it at all times for their own purposes, and must build a DAO accordingly.
Even Fluid has decided to transfer all brand assets to the Fluid Foundation, which is owned by the DAO, following everything that happened with Aave, proof that governance issues are still being taken seriously.
Blockchain is a technology that allows us to experiment with governance systems far more than any other technology would have allowed us to do. In doing so, I hope all the drama about DAOs will help to gain maturity, and get out of the Schrödinger’s state they are right now.