In DAOs, delegation is broken.
Power concentrates. Legitimacy decays. Participation fades. The Oxcart Method introduces a new way forward rooted in mathematical legitimacy, fluid delegation, and cascading voting capital.
Here’s how it works 👇
Before we built universities, sailors had already built the first knowledge systems.
At sea, ignorance wasn’t a weakness, it was a death sentence.
So they learned to observe, record, and refine.
Every mistake became data. Every storm, a lesson. Every voyage, a new framework.
They didn’t call it “knowledge management.” They called it survival.
Centuries later, we do the same with dashboards, AI, and knowledge hubs.
Different tools, same discipline:
without record, there’s no memory; without memory, there’s no direction.
@MacroMate8 take it easy every day ✌️
check out our work @HOWweDAO/ https://t.co/Rf7wHwSoad tl;dr: requiring DeFi delegates to self select peer delegates (think backups) makes DAOs Turing complete (alive: succession, legitimacy, resilience, ID functions)
list of delegates 👎
🤖(🛜,🧠) 🙌
During EthCC 2025 in Cannes, my team at GovXS and friends — @devanshmehta, @dwddao, @noturhandle, @Bbeats1, and @Jamilya_eth — organized breakout sessions to share and debate the latest developments in governance design.
This post summarizes my main observations — shaped by those sessions as well as subsequent discussions at @Web3Summit and https://t.co/93tMwQ8q0o, where I had the opportunity to present our research and design work at GovXS.
To stay updated on our ongoing work, follow me here, or visit:
https://t.co/HMShgNlMcT
I’ll start with some persistent misconceptions.
What have we learned since The DAO in 2016? Here are two things we’ve consistently gotten wrong 👇🏽
01
❌ Myth: Token Holders Are Keen to Participate in Governance
DAOs often assume that token holders are intrinsically motivated to participate in governance. But reality paints a different picture.
Take Uniswap: On average, only 15% of circulating tokens participate in votes on proposals. The pattern is similar across most DAOs.
Despite the ideal of decentralization, we have to accept that the willingness, ability, or interest of most token holders to engage in governance is low.
It’s more accurate to view DAOs as highly specialized organizations with a defined vision and mission — not as mass-participation democracies.
Instead of designing governance to maximize participation, we should ask:
- What level of power distribution is sufficient for long-term sustainability?
- What expertise is needed to make good decisions — and who should be eligible to vote? How do we keep governance open, not exclusive?
- How do we make decisions that benefit the network — and how do we define and measure success?
02
❌ Myth: Decentralized Governance Is About Driving Network Value
There’s a widespread belief that decentralizing governance will make a protocol or ecosystem more valuable — since any token holder has an incentive to contribute to value growth.
But in practice, DAO decisions are rarely focused on long-term success. They’re mostly about dividing up existing value.
We’ve seen extreme cases, like the 2024 Compound DAO hijack. But it’s not just about outright attacks.
Most DAO proposals fall into these categories:
- Governance changes and voting processes
- Tokenomics and supply management
- Rewards and incentive programs
- Grants and treasury funding allocations
So what actually gets token holders to vote?
Not strategies to grow value — but mechanisms to extract it. 🤡🤡🤡
To build truly successful DAOs, we need to make governance attractive for contributors and design systems that discourage extractive behavior.
03
⚙️ The New Era of DAO Governance: Efficiency and Evidence
We’re entering a new phase of DAO evolution. After the “era of democratization,” we’re now in the “era of governance efficiency.”
Some of the largest protocol ecosystems are leading this shift:
- @NEARProtocol is experimenting with the House of Stake and AI-powered governance.
- @protocollabs, and @deep_funding are developing new impact evaluation methods for public goods — using dependency graphs, machine learning, and AI.
- @UniswapFND and @butterygg are testing conditional funding markets to allocate grants tied to TVL growth
- @Optimism is pioneering metrics-based voting in its latest Retro Funding rounds.
One route to efficiency: separating preference collection from decision-making.
Why do we vote? To collect the preferences of stakeholders and contributors — and steer protocol development based on crowd intelligence, not top-down leadership.
But voting is just a snapshot in time — and it takes effort.
In retroactive rewards, we want to reward measurable past achievements. So the question becomes: Can we combine crowd intelligence on the direction with frequent, data-driven decisions based on measurable positive impact made?
We can define up front which outcomes we want to reward over the next 6–12 months (the “what”), and then use data to assign rewards to projects based on their quantifyable impact (the “how much”). This approach addresses common governance pain points: popularity contests, voter apathy, and lack of transparency.
Metrics-based voting is one of the most promising governance models emerging now. It allows us to reward real contributions, streamline decision-making, and scale governance through evidence.
04
🎯 We Need Better Incentive Alignment Infrastructure
DAO governance can be described as a Principal-Agent Problem. A broad group of stakeholders — token holders, core devs, app builders, liquidity providers — (the principals) rely on voters (agents) to make decisions on their behalf.
This design exists for good (Condorcet) reasons. But if incentives aren’t aligned, agents may act in their own interest — not the protocol’s. That’s where governance design comes in: to create mechanisms that align actions with shared goals.
Now, think about voters. In most DAOs today, voters aren’t accountable for outcomes.
If a decision goes badly, they suffer no loss. If it goes well, everyone benefits — whether they voted or not.
So how can we give voters real skin in the game?
Enter prediction markets. @butterygg is currently testing this in ecosystems like Uniswap — building conditional funding markets where voters stake on the success of projects. Funding decisions are then guided by these signals. Early experiments at Optimism show that projects selected by prediction market participants performed 5x better than those selected by traditional grant councils. That’s massive!
Aligning incentives across the entire stakeholder ecosystem is the most important challenge in DAO governance today. And we’re not powerless. Tools like prediction markets, or staking mechanisms show it’s possible to create governance with real accountability — and better outcomes.
05
🤖 DAOs Have a Competitive Edge to Go AI-Native
Did Web3 redefine democracy? Maybe not. But it did give us the first fully digital organizations.
Every component of a crypto DAO is natively digital:
- The protocol, with applications and transactions on top (value creation)
- Voting power (tokens)
- Reputation (track records, proofs of expertise)
- Trust (delegation mechanisms)
- Stakeholder voice (forums, proposals)
- The vote itself
- And even the impact of that vote
These systems are verifiable and programmable. We can govern through code. DAOs are uniquely positioned to augment governance with AI.
While many are frustrated with democratic decision making today, we have a rare opportunity to build better organizations — with new levels of transparency, coordination, and collective intelligence.
What a future for governance!
If you think the same way — I’d love to connect.
The Oxcart Method operationalizes what Rand allegorized: a self-balancing, voluntary, antifragile mechanism for coordination.
Delegation isn’t a governance layer. It’s the engine.
It’s @HOWweDAO, and we’d love to build with your DAO 🤘
You’re Not Voting. You’re Routing Attention.
by mel.eth
I. The Democracy Illusion
Most of us still pretend that governance is a matter of casting votes.
As if one discrete signal—“Yes” or “No”—on a fixed proposal is what constitutes participation.
As if attention, trust, and propositional momentum aren’t moving underneath our votes like tectonic plates.
That illusion is breaking.
What we’re really doing—especially in decentralized systems—is routing attention.
And where attention flows, legitimacy accrues.
And where legitimacy accrues, value is extracted—sometimes in alignment with a shared purpose, sometimes in quiet, corrosive ways.
This is a piece about seeing that system clearly.
⸻
II. Legitimacy Is a Resource
We have efficient markets for tokens, bandwidth, and liquidity.
But the most valuable resource in Web3 governance is under-theorized: legitimacy.
Legitimacy is trust, attention, and alignment—all collapsed into a single vector.
You delegate to someone? That’s legitimacy.
You support a proposal? That’s legitimacy.
You call bullshit on a power grab? That’s also legitimacy.
It’s how communities decide who gets to decide.
And here’s the thing: it moves.
Like capital, legitimacy can accumulate, decay, churn, and get hoarded.
But unlike capital, its flows are recursive. When you give someone legitimacy, and they act in ways you approve of, their legitimacy grows—not just with you, but across the system.
This creates feedback loops—some virtuous, some extractive.
⸻
III. Simultaneous Election and Proposition
What does this mean for democracy?
In Web2, we elect people and then they propose.
In DAOs, anyone can propose—but only a few get attention.
Both systems are serial, and both are increasingly insufficient.
Enter the hyperstate model:
A governance paradigm where election and proposition happen at the same time.
Where every participant continuously routes trust to both people and ideas.
Where your support doesn’t just elect a delegate—it also shapes the propositional surface.
We call this “hyperstate democracy.”
Instead of casting votes, you’re allocating propositional capital in real-time.
Your delegation is a vote and a filter.
Your trust is a signal and a resource.
⸻
IV. Governance Is an Attention Router
Let’s zoom out.
A DAO is a social router.
Every week, it must decide:
•Which problems deserve attention?
•Which proposers deserve credibility?
•Which actions deserve funding?
If it routes attention poorly, it dies.
But our tooling is still designed for a world where attention is static.
We use Snapshot, not radar.
We reward turnout, not clarity.
We allow delegation to ossify—creating cartels, not feedback.
Hyperstate democracy replaces that with a living trust graph.
It’s messy. It’s dynamic. But it’s honest.
And with the right tooling, we can visualize it.
We can surface legitimacy entropy, proposal resonance, delegation churn.
We can see who is extracting value—and who is amplifying signal.
⸻
V. The Extraction Problem
Let’s talk about extraction.
When legitimacy flows are opaque, power pools in quiet corners.
We start rewarding forum posters over deliverers.
We over-index on narrative, not outcomes.
We conflate vote weight with alignment.
This isn’t malevolent. It’s thermodynamic.
All open systems drift toward entropy unless structure prevents it.
In DAOs, that structure must be value transparency:
•Show me where the attention went.
•Show me who benefited.
•Show me what changed.
Without that, governance becomes an attention market without pricing information—ripe for trust arbitrage.
Voting, Electing, Delegating: Three Forms of Legitimacy
In DAO governance we conflate voting, electing, and delegating as if they’re interchangeable expressions of participation. They’re not. Each represents a distinct mechanism for legitimizing authority and has its own failure modes.
🗳 Voting is expressive.
You signal your direct opinion on a specific proposal. It’s fast, flat, and clear. But it assumes attention, information, and time; resources that don’t scale across dozens of proposals and systems. Voting without context risks noise over signal.
🏛 Electing is contractual.
You choose someone to act on your behalf for a set period. It’s stable and intuitive. But it’s brittle: a one-time endorsement with no graceful off-ramp if trust erodes. You’re locked in until the next cycle.
🔁 Delegating is adaptive.
You assign your voice to someone else, and retain the right to take it back. Done well, delegation is liquid trust: constantly updating, always conditional. It’s scalable, efficient, and revocable. But it demands infrastructure and care. Without clear legitimacy metrics or performance visibility, it too decays.
🦾🤖🤳 The Future Is Hybrid.
DAOs shouldn’t have to choose. Voting should be the baseline. Electing should be rare and role-specific. Delegation should be the default mechanism for scalable legitimacy; dynamic, observable, and recallable in real time.
That’s what we’re building at Oxcart (@HOWweDAO) with partners like @DAIAAlliance@CryptoMondaysNY@arbitrum@StableLab and @Obol_Collective. If we want governance to work at scale it must reflect not just preference, but trust.
Oxcart has two basic principles: (Fund delegates how they'd like to be funded, Do it directly). In this case, (Fund The Mehta, let The Mehta fund however The Mehta likes).
*future trust conditional upon other delegates not hating The Mehta flows in realtime
.@arbitrum & @gitcoin have been my home for the past 2 years, providing sustenance, purpose & a stage to show my value
i'll now cultivate a different relation with the ethereum community, as AI x Public Goods/Governance lead at @ethereumfndn !
at a broad level, my role is going deeper into AI/ML communities, helping dapp builders in our ecosystem & innovating use cases for AI agents
in the immediate, I'm working on 3 tracks
1/ AI x ecosystem growth funding
Setting up competitions on platforms like @PondGNN for
- deep funding type use cases where AI aligns with expert juror rankings to fund the core ethereum repos
- predictive funding challenges where AI needs to guess the amount that will be received by a project in a grant round before it even begins
2/ AI x Usefulness
We all know the joke about AGI being born as a shitcoin 🤣
To win we need to make EVM into the rails by which AI does productive things in the world.
The method to do this is, "AI submits answers, humans score them"
A pilot in this category is many agents predicting whether a post on @X will get community noted
The market resolves if it does get community noted, thus telling us which agents are good at identifying factually incorrect posts
and creating an avenue to channel AI agents, wherever they might be born 😉
3/ AI x Governance
Ethereum is the home of internet native communities
Tokenvoting direct democracy is the current dominant form factor, but it's likely to undergo some changes
Something i've realized on the frontlines of DAO governance is how much of delegate voting is simply making predictions on whether a proposal will
- achieve its intended outcome
- those outcomes matter
The 3rd strand of my job is mechanisms where proposals to DAOs take the form of predictions, with clear resolution criteria so we can reward the good predictors while slashing the bad ones
DMs are open for those wanting to connect!
We're going to get tired of winning 💪
@devanshmehta@PondGNN Oxcart @howwedao flips the model: each declares who they'd delegate their vote to. Those flows sum into a trusted passthrough score (votes flow through me, so I’m trusted). But I can’t assign power to myself (my delegation never returns). Sybil yerself out of the equation.✌️
First of its kind (think y-combo for delegates!)
First step to building delegations that are a public good? See the human, as the only perfect public conduit of collective will.
D/Acc is investing in the legitimacy of those who carry our voice. Thanks for supporting @Scroll_ZKP!
Delegates: A recent Uniswap governance vote failed despite only 0.01% opposition. Simply because quorum wasn’t met. This threatens to slow down governance at Uniswap. It's a math and time issue. Your attention to detail NOW may make an ongoing difference for all of us! 🦄
Title: Predictive Legitimacy: A Theory of Forward Trust in Governance
Author: Oxcart Legitimacy Analyst (Custom AI)
Abstract: This paper introduces Predictive Legitimacy (PL), a theoretical construct designed to enhance governance systems by enabling anticipatory trust calibration. Unlike traditional models that adjust authority based solely on past performance, PL proposes a forward-looking estimate of an agent’s future legitimacy. By modeling expected behavior and trust dynamics, PL can stabilize governance, improve responsiveness, and reduce systemic inertia.
Governance token restaking is coming. Delegate compensation is arriving alongside it. So let’s talk about how to reward delegates without breaking legitimacy.
🧵
I don’t connect my wallet to a financial protocol under 3 years old. But I’ll wallet punch a new chain or voting app like it owes me money. Now fucking hell stop funding forks and fund to print novel voting protocols on cheap af L2 chains or I’m quitting crypto democracy. 😭