Posted on May 03, 2022Read on Mirror.xyz

learning by daoing: designing for the unique properties of daos

at the most basic level, daos are simple. they facilitate two things: collective ownership and collective decision making. it starts to get more complicated when we consider the many ways to design daos. when it comes to designing daos, we spend too much time talking about how each dao is different and not enough time thinking about the ways in which they are the same. this lack of design framework either forces us to resort to traditional companies as a baseline model (porting over many of the limitations of traditional organizations) or to forgo any framework at all and struggle to manage a dao’s evolution.


through the (currently painful) process of daoing, i’ve identified several inherent properties common across functional daos. by designing daos with these common properties in mind, we can not only improve the experience of daoing but also design systems that accomplish more than traditional corporations ever have (one example here).

these properties are: Autonomy, Emergence, Complexity, Agency, Transparency


  • what it means: the ability to make your own decisions without being controlled by anyone else; the right of an organization, country, or region to be independent and govern itself (Cambridge)
  • what it means in the context of daos: in contrast to traditional corporations, daos cannot be controlled or captured
  • design principle: design for the highest degree of autonomy (weakly linked, autonomous sub-daos) while providing checks and balances on distributed power

the biggest difference between corporations and daos is that daos can be capture-resistant, meaning they incorporate mechanisms to prevent the capture of shared resources by bad actors (more here). in traditional corporate structures, capture resistance is weakly ensured via trust and reputation, since both matter in the repeated game of public business. internal capture is prevented via regulations enforced by higher-level, external agencies. daos require different assurances against capture since trust, reputation, and regulation are mostly absent in large, somewhat anonymous, digitally native networks. daos can be designed to be capture-resistant, but most aren’t. there are three main ways a dao can be captured:

1.) control over decision making

the 2020 wave of daos pioneered “progressive decentralization,” or turning control over to a community once a protocol built by a centralized core team achieved product-market-fit. this approach prevents the core team from being distracted by building a community and a product at the same time, and avoids design by committee, which isn’t usually conducive to shipping products. this method builds community around collective ownership, launching a token once operations are “sufficiently decentralized” and suddenly turning decision making over to a group of new token holders, usually via token-weighted governance. all decision making has been centralized up to this point and a large portion of product-related decisions often remain centralized. since decision making has been centralized from the start, decentralizing it can feel like giving up control, which core teams are often reluctant to do.

community members are drawn in by a token launch, not by co-building or participating in the network over time, which can create communities that are more speculative in nature. given a lack of other criteria, like previous contributions to the dao, tokens are generally distributed based on retroactive usage, which has proven to be a weak contributor acquisition method. contributors quickly become frustrated by a low degree of involvement in decision making. these dynamics can result in lower voter participation, making any decentralized decision making less capture-resistant as less voices are represented.

regardless of the path towards decentralization, token-weighted governance increases the possibility of capture in the decision making process, the risks of which are well outlined here. Optimism’s two house governance system is a positive first step towards more capture-resistant governance (more here.)

for daos to be capture-resistant, we must decouple power (decision making rights) from wealth (token holdings)

2.) control over resources

to avoid some of these challenges, we experimented with a different decentralization method at RabbitHole. our hope was to build a community around collective decision making, by inviting contributors to co-create with us much earlier in the decentralization process (pre-token) rather than building a community around collective ownership of an already finished product and token. We involved our community gradually in non-core business operations, spinning up a series of community-led working groups. however, since there was no mechanism for decentralized decision making, working groups were still reliant on external stakeholders for access to critical resources (in this case a centralized company, without a governance mechanism). as a result, our working groups were vulnerable to capture and we made the decision to pause this experiment until we could empower working groups with more autonomy.

even when decentralized decision making does exist, the party controlling the resources (ie. parent dao) still has lower context as to the value a working group provides. while working groups may be equipped with their own budgets, they are still subject to a budget allocation process that is outside their control. even with a token to align incentives between a parent dao and a working group, the value a working group adds may not be reflected in the token value of the parent (ie. metagovernance, the value of which is hard to tie back to a core business value.)

when one group controls access to resources for another, the relationship becomes adversarial

while working groups help to control and separate scope within a monolithic organization, the higher autonomy design choice is to replace them with multiple sub-daos. autonomous sub-daos allow communities to organize around specific goals with complete autonomy over how to achieve them, to maintain direct ownership over their own resources, and to share directly in any upside generated from their efforts.

these sub-daos can focus on building either the supply side (to increase the scale/scope of product offerings) or the demand side (to explore new use cases or aggregate users). they may also take the form of a subsidiary, providing services across the ecosystem (ie. metagovernance.) all sub-daos should have economically sustainable business models, so they are not dependent on external parties for resources, but a native token quickly becomes an important means of creating “weak strategic alignment” between daos via token swaps (more on this method here; see MetricsDAO and Goblin Sax).

3.) control over execution

the greatest opportunity for capture occurs at the execution phase. governance generally works as follows (see more here):

  • a proposal is put forward
  • the community votes on a proposal
  • the outcome of the proposal is either autonomously executed on-chain or an off-chain action is executed by a select party
  • execution is verified and / or analyzed by the community

execution is the phase most vulnerable to capture since nothing prevents those with execution power (ie. multi-sig signers) from ignoring the will of voters and executing whatever actions they choose. this doesn’t mean we should eliminate all positions of trust within daos. there is a common misconception that in order to avoid capture, decentralized organizations must be flat. but daos are leaderful, not leaderless. daos can still have hierarchy, it’s just that it’s fluid.

positions of trust (like those with execution power) should be elected or legitimized by the broader community with clear mechanisms for removal if/when the responsibilities of those positions are abused or neglected. Yearn’s gov 2.0 proposes such a model, wherein highly autonomous yTeams are empowered by YFI holders to act independently in the best interest of yearn, within a constrained domain of action and discrete decision-making powers, subject to monitoring by YFI holders which have the power to remove contributors if roles are abused (more here).

being a leader in a dao is the most humbling thing I’ve ever done because you cannot rely on static positional authority. you must constantly earn respect with the community. on the good days, it’s meritocracy; on the bad days, trial by jury.

positions of trust don’t need to be open to the entire dao. while power should be distributed, it shouldn’t be permissionless. unauthorized participation can actually threaten a DAO’s resistance to capture as it creates more attack vectors (more here). while participation by as many community members as possible should be encouraged as an input into decision making, the ability to put forward certain proposals, make decisions, and hold positions of trust should be subject to qualifying criteria.

daos (the human layer) do not need to be as decentralized as the protocols they are built around (the tech layer)


  • what it means: the fact of something becoming known or starting to exist (Cambridge)
  • what it means in the context of daos: in corporations, ideas are dictated from the top down. in daos, initiatives surface organically from the community
  • design principle: emergent daos don’t have to be amorphous. a strong mission and culture can guide collective decision making without being overly prescriptive

emergence is vital to maintaining an active community. daos that don’t support emergence become little more than shell communities over time as contributors get frustrated having little ability to influence the dao. the most emergent daos are “day-one daos,” which form organically when groups of people rally around a mission, idea, belief, or event. they’re decentralized from the beginning and decide what to do or build collectively. these daos usually launch tokens immediately as a way to raise funding, which also immediately decentralizes decision making. this provides contributors with full autonomy to shape the dao emergently but day-one daos can struggle to guide decision making given the lack of structure.

let’s define culture as a set of shared beliefs about how decisions should be made (mission) and which behaviors will be rewarded (values). culture is mostly an afterthought in traditional corporations. decision makers are determined via hierarchy and decisions are implemented by force, so there is little need to define a framework around how decisions get made. In contrast, culture is a critical tool for guiding decisions made by a collective.

why are mission statements so important for daos? because they provide a north star for decision making and set a general direction for the dao. directionless daos tend to lose momentum, and i’ve never seen a dao that lost momentum get it back. a strong mission can also create a moat. people think community is a moat but it can’t be since community members are constantly changing. your mission, however, can consistently draw in value-aligned contributors and keep them there when the dream of the dao inevitably fades into the reality of the dao (twoirtter.eth).

contributors can only meaningfully contribute to 3–4 daos, so they need a compelling reason to contribute to one over another. RabbitHole contributors have said they work with us specifically because they understand our mission and share our beliefs. this is important because it creates a sense of belonging, which is among the top things contributors crave.


clear values are the second component of building a strong dao culture. more emergent daos tend to develop around icons, “mimetic filters,” lore, or specific events, which serve as a natural filter to draw in value-aligned contributors. they must then work together to collectively define their mission and values (see Yearn Blue Pill). the best values make clear which behaviors will be rewarded and which actions should be prioritized over others within the dao (see example).

mission statements and values can be suggested by a core team (see CCS, Mirror 1, 2, and 3, and ENS) or collectively created by contributors (see OADAO). regardless of the process for defining them, establishing a clear mission and values provides emergent daos with just enough guidance about how to make decisions without being overly prescriptive.


  • what it means: a complex system emerges organically as a result of the behaviors of the components within it, rather than according to a predetermined plan (Bristol)
  • what it means in the context of daos: most corporations are complicated systems, with any complexity tightly controlled by bureaucracy. daos are complex systems, meaning it’s hard to determine the impact actions in advance (more here)
  • design principle: communities should co-create seasonal charters to guide efforts in a way that accounts for uncertainty while preventing chaos

since well designed daos empower contributors to make decisions that can meaningfully change the trajectory of the organization, outcomes are hard to anticipate. that doesn’t mean that daos have to be chaotic. just because we don’t know the exact destination, doesn’t mean we can’t steer at all.

strategy can exist in daos, the difference relative to corporate strategy is that dao strategy is co-created by contributors rather than set from the top-down. the simplest way to do this is via collectively authored charters (see here) that serve as a mutable guide outlining high-level goals, suggested processes, and guiding principles to inform decision making over a season. these charters maximize contributor scope within a “container,” providing a collectively agreed upon roadmap that defines what needs to be accomplished while allowing contributors to determine how they arrive at the intended destination. allowing for re-evaluation and updates to charters at regular intervals is important since much of how a dao operates is discovered through day-to-day experience (twoirrtter.eth)

while dao visions are collectively discovered over time, dao strategy should be collectively set at regular intervals

while contributors want the ability to shape strategy, i’ve found that they also don’t want to be handed a completely blank slate. at RabbitHole, we found the optimal balance in providing an initial charter outlining ~70% of what a working group will drive towards and empowering the community to provide feedback and fill in the rest.


autonomy is present at the organizational/dao level while agency is present at the individual contributor level

  • what it means: the ability to take action or to choose what action to take (Cambridge)
  • what it means in the context of daos: in traditional corporations, top level management holds most of the agency. in daos, all network participants maximize agency
  • design principle: daos need to be designed with the understanding that individual contributors will optimize for agency via voice and exit

1.) Voice (change the system from within)

governance gives contributors the option to have a voice in decision making. while most governance conversations focus on how to increase voter participation, having the option to vote is actually more important than constantly voting. if contributors have the option to effectuate change, but don’t choose it, they have nobody to blame. if they do exercise the option, they feel empowered by the agency to influence their own destiny. since the option is more important than the action, the expectation should not be that every tokenholder votes on every vote. most everyday users don’t want to participate in democracy for the products they use each day. even if they wanted to, it’s a full time job to keep up with all proposals and so quorum is usually only met on the most controversial proposals. once we accept that option to vote is more important than the action of voting, we can stop applying a one-size-fits-all decision making model to governance. this will allow us to make more effective decisions, which is one of the two primary purposes of daos (along with collective ownership.)

what effective decision making comes down to is empowering high context people to make localized decisions while granting everyone else the option to protest when/if needed. if we design for agency, we can start experimenting more with consent-based governance (see Mirror 0004), optimistic governance (see Aragon and Tribe), and lazy consensus (see Colony) and we can start to think about differing voting rights based on identity, reputation, and contribution (See Element).

Dan Wu: https://twitter.com/itsdanwu/status/1499859953594642439?cxt=HHwWjsC9rcyrydApAAAA

2.) Exit (leave to create start a new system or join a competitor)

part of the beauty of daos is the fluidity of participation and the ability to work, on a part-time basis, for multiple daos at once. it also means that exit is an easier option for contributors than it is at a traditional corporation. core teams should be prepared to fill in any gaps created by a primarily part-time contributor base that may frequently choose exit. maintaining continuity of work when turnover is high increases the need for thorough and transparent process documentation and clearly outlined strategies. you probably don’t want to evaluate your dao on retention. instead, focus on ratifying co-created policies and processes that allow any contributor to plug into well defined, yet flexible, and well documented workflows, reducing dependency on any one contributor without adding top-down bureaucracy.


  • what it means: the quality of being done in an open way without secrets (Cambridge)
  • what it means in the context of daos: daos redefine the meaning of “radical transparency” in a way that would severely rattle private corporations
  • design principle: design for nearly complete transparency and higher levels of accountability relative to traditional organizations

while public corporations make some data transparent, they do so in a highly edited format, released at infrequent intervals. in contrast, all the raw data pertaining to on-chain actions of a dao is publicly viewable, in its entirety, in real time. contributors expect a similar level of transparency for all off-chain data and decision making. providing this level of transparency into off-chain actions is where many daos fall short and run into conflict.

the purpose of transparency in corporations and daos is also different. the purpose of corporate transparency is to inform shareholders about decisions that have already been made. the purpose of dao transparency is to enable contributors to make decisions. this makes transparency all the more important in daos as information asymmetry actually prevents contributors from fulfilling their role (chase chapman). however, this level of transparency also means that dao operators will be held accountable 24/7. contributors will expect rapid responses and open communication at all times. at the same time, anonymity among the contributor base provides less transparency and lowers accountability. since the best collective decisions are the product of disagreement and contest, not consensus or compromise, it will get rough in the Discord. dao operators need mental health benefits and support. burnout is extremely common among community managers, but it’s also preventable. while complete transparency seems like the simplest property to design for, it has the most underestimated impact on dao operators.

it’s completely mindblowing what a group of internet strangers can do when empowered by collective ownership and decision making. we can accomplish even more if we start designing daos in a way that embraces their unique properties: Autonomy, Emergence, Complexity, Agency, and Transparency. If these properties will be disruptive to what you’re trying to achieve, a dao might not be the best structure to accomplish your goal.


we’re still figuring out what great dao design looks like and we have a long way to go. daos are constantly evolving, so the best approach today may not be the best approach tomorrow. i still believe these design principles are useful, even if we disagree about them, because they start a conversation that hopefully moves us all forward, together, and isn’t that the whole point anyway?

If you liked my thoughts on daos, check out my earlier thoughts on DeFi, DeFi 2.0, and Web3 Social

Thanks to pet3rpan for discussions on this topic and for feedback from: Diana Chen, Chase Chapman, Meg Lister, Ben Schecter, and Vishal Kankani