That was the topic of my talk at EthCC this year. Year after year, DAOs enjoy more attention from the crypto community and beyond, but we’re still far from prime-time exposure. My goal was to share an update on how we see DAOs today and how they can be used.
Like every year, 2020 was announced as the year of the DAO. Still, most people who heard about the word have a hard job defining the concept. I’ve tried elsewhere to provide such definitions and shed light on the associated narratives. Aragon recently tackled the challenge of defining what DAOs are in a beautiful manner. In the context of this post, let’s just keep in mind Vitalik Buterin’s original proposition that DAOs are organizations with automation at the center and humans at the edges.
For decades, firms have been considering automation as a way to increase the productivity and quality of manufacturing processes. When one thinks of automated factories, the magical ballet of robotic arms on a production line comes to mind－or, more recently, mobile robots in an Amazon warehouse. Natural candidate processes for automation are routine tasks, executed with little to no space for decision-making on the spot.
Tasks associated with design, coordination, or optimization of such processes, on the other hand, are seen as requiring humans with soft skills such as making sense of complex situations and interpersonal communication. Hence, the central functions of organizations remain the indisputable privilege of humans, barely threatened by the distant advent of strong AI.
The typology introduced by Vitalik reverses this view by suggesting that the edges of a productive organization might still rely on human effort, while most of the complex coordination nexus that sits at its center may run autonomously and use its own, internal capital to reward workers. According to this view, software offers a way to replace traditional economic and social coordination mechanisms.
So, what can we say about this vision today, after a few years of buidling and experimentation? Let’s deconstruct ‘Decentralized Autonomous Organizations’ in light of their most recent developments.
DAOs are autonomous in the sense that they are immune to third parties’ actions. Thanks to the public blockchain networks they run on, DAOs control their own capital and allocate it according to self-enforcing, tamper-resistant rules. This is why TheDAO in 2016 was presented as “unstoppable”, meaning that nobody could change its behavior nor pull the plug of the computer it was running on.
As fascinating as the idea appeared at the time, it turned out that running a fully autonomous system comes with a number of undesirable potential consequences. When TheDAO was hacked using a vulnerability in its code, millions of dollars were drawn out of its treasury in plain sight while no one could do anything about it. The system was behaving the way it was programmed, to the point that some people argued that the attack was less a hack than a feature. If “code is law”, then the loophole in TheDAO code that made the exploit possible is also part of the agreement between the parties, who agreed to be bound exclusively by TheDAO’s smart contract.
That’s what autonomy means here. The funds of a DAO are controlled by its code, and only by its code.
What have we learned?
Using blockchain-based, autonomous code artifacts was quickly found to be problematic, and not only for DAOs. As individuals as well as organizations, we expect to have a recourse when something goes awry. If we make a mistake in bank transfer instructions, it is assumed that the bank will fix it and that our money will be returned. When two companies disagree over the execution of a contract, they might plead their cases before a court.
Compare that with what happens when one sends cryptocurrencies to a wrong address, or when a bug locks millions in a wallet such as Parity’s multisig contract. If measures haven’t been planned ahead of time and implemented in the code, there is no way to cancel or reverse these operations. And since it is impossible to guarantee that a complex piece of code is free from any bug or vulnerability, there’s always a possibility that even the most carefully secured and audited code will fail to protect its users.
Immutability and censorship resistance are valued by those who want to protect their assets and their independence from corrupt judges or failed states. The autonomy of the code is a way to set personal and collective arrangements out of reach of any third party, regardless of how powerful or wealthy they are. But it’s worth emphasizing here that the autonomy that we really crave is for the people (as individuals or as collectives), and that the autonomy of the code is a means rather than an end.
Autonomy in 2020? Balancing code vs. community
The focus has shifted from the absolute autonomy of the code praised by cypherpunks to the relative autonomy of a combination of dry code and human judgment.
One way to approach such a combination is to make room for subjectivity in automated processes. Kleros and Aragon pioneered decentralized arbitration services that can be used today to make DAOs more flexible. Aragon Agreements, for instance, enable organizations to define human-readable agreements that can be enforced by a decentralized court of jurors when a dispute is raised.
Another way to protect people from blind application of automated rules is to provide them with a way to escape a system before being harmed. This is the idea behind the ragequit mechanism offered by Moloch DAO: any investor in the DAO’s fund can withdraw their investment when they disagree with a decision voted by a majority of members. To be fair, such protection was already present in TheDAO in 2016, but ironically it was a flaw in the way it was implemented that allowed TheDAO to be hacked…
Lastly, we’re seeing mounting concern regarding legal risks for DAO members, which can be addressed by wrapping a DAO in a legal form such as an LLC. Several jurisdictions now offer such legal vehicles so that DAOs as well as their members are protected from legal uncertainties that may lead to unlimited personal liability. Examples include dOrg, a cooperative of developers that incorporated as a Vermont registered Blockchain-Based Limited Liability Company, and The LAO, a venture fund organized as a Delaware LLC.
In each of these cases, the ideal of a pure cyberspace solely determined by code is balanced with the legitimate concern of protecting the people for whom such space has been built in the first place. Decentralized courts make it possible to protect communities from rogue individuals. Ragequitting prevents majorities from exerting excessive power upon minorities. Legal wrappers offer protection against threats to DAO members via regulations and applicable laws.
Decentralization can be understood as the process of reducing or avoiding the concentration of power within organizations. In decentralized structures, resources and decision-making power cannot be seized by a small number of parties.
In the past, decentralization has been generally achieved through legal arrangements, such as the statutes of a cooperative or antitrust laws. DAOs and crypto-networks inspired by Bitcoin have used a combination of economic incentives and programmable rules to foster decentralization.
For instance, cryptocurrencies are based on a consensual state of information with respect to holdings and transactions between addresses. With public, permissionless networks, anyone can participate in the consensus. No gatekeeper means that there is no rent-seeking mechanism, no subordination between members of the network, and equal rewards for contributing members.
What have we learned?
It took a while to understand and admit that public blockchains are not as decentralized as they seemed to be. Angela Walsh, for instance, reminded us that decentralization exists on a spectrum and that depending on the criterion used to measure decentralization, even the largest permissionless cryptocurrencies tend to have significant concentrations of power.
The same criticism can be leveled against DAOs. The token distribution in many DAOs is so unequal that a very small number of members can block or pass any proposal. In other cases, the personal influence of founders or key members is much more important than the actual voting power that they hold through their tokens.
This is not necessarily a problem when one considers decentralization as a gradual process. While decentralization is lauded as a way to reduce inequalities and guarantee the accessibility of public goods, it’s totally unproven in the context of starting a new project. In the crypto space as anywhere else, nothing seems to beat small teams with strong leaders, not even protocols precisely engineered to enable decentralized coordination such as DAOstack’s Holographic Consensus.
Reality check: Put aside the decentralization razzle-dazzle, and have a look at real projects in the crypto space. All of them are under the control of their founders, and most of the significant ones are organized according to the traditional corporate governance system, with shareholders and accountable executives. Cooperative and meritocratic communities like dOrg, Metagame or LeapDAO are still an exception and have yet to prove themselves.
Decentralization in 2020? More options to effectively decentralize
Decentralization is hard, but there have never been so many ideas to make it real! Here are some of the most interesting approaches we’ve seen recently.
Token-based voting has been the main instrument for collective decision-making in DAOs since it’s the easiest way to prevent Sybil attacks, i.e., voting with multiple identities. Sybil attacks have always been a headache in the crypto space, which is essentially pseudonymous. People do not need to have their identity verified in order to participate in a public crypto-network such as Bitcoin or Ethereum, they just need a pair of cryptographic keys, and anyone can create an arbitrary number of keys at no cost. Token-based voting is an efficient way to be Sybil resistant since tokens are crypto-assets that cannot be created at will.
However, token-based voting is also criticized because of its adverse effects on decentralization. As mentioned previously, most crypto-networks are imperfectly decentralized, precisely because a small number of addresses control a majority of the tokens. Token-based voting favors plutocracy and perpetuates power asymmetries. It also discourages the participation of the majority of people when they know that their vote has no weight.
Conviction voting enables the preferences of a community to be continuously captured over time instead of organizing votes within a narrow time period, and to let stakeholders choose where they put their influence. It makes Sybill attacks and vote-buying harder, and it can also help to fight voter apathy. 1Hive has delivered an implementation of conviction voting, available as an Aragon app.
Quadratic voting is also a way to limit the plutocratic effects of token-based voting, while still giving people with more skin in the game a way to signal their preferences. It has been successfully experimented with by Gitcoin for funding public goods.
Another take at decentralization relies on the allocation of decision-making power to the people who contribute to the network. When the voting power of DAO members is based on their actual contributions and cannot be transferred to others, power asymmetries tend to be much less acute. Various reputation systems have been designed to that end, for instance by DAOstack, Colony, or SourceCred.
The same idea recently appeared in the DeFi space under the term “Simple Agreement for Future Governance” (SAFG). Exemplified by Compound and its COMP token, this approach consists of granting governance tokens with no economic rights to people who participate in a decentralized network or protocol. Later on, when the network is widely used, token holders can opt for an economic model and add economic rights and transferability to the token. Allocating decision power to actual users is an encouraging development towards decentralizing a network to prevent rent-seeking. However, it must be done cautiously as it opens up new attack surfaces for token holders who might be tempted to game the system in their own interest.
Drawing on political philosophy, Lawrence Lundy suggested that the separation of powers at the cornerstone of our modern democracies should be applied to crypto-networks in order to prevent power grabs by a particular category of stakeholders. Legislative power is expressed by token holders through a vote, executive power is delegated to organizations whose duty is to implement policies and operations voted by legislature, and judicial power is exerted by decentralized courts so that executives cannot violate a DAO’s intangible principles.
Up to now, the executive branches of most crypto projects were held by firms or foundations that had no fiduciary duties towards token holders. We’ve seen a trend towards more integration between these institutions and token holders with voting rights. The Aragon Association, for instance, was vetting proposals made and voted by Aragon token holders. Now that a judicial arm is available thanks to decentralized courts, we are seeing DAOs like Pocket Network integrate the three branches so that checks and balances effectively protect decentralized communities.
Another take at decentralization consists of giving a voice to groups of stakeholders whose interests are different. For instance, the governance of the Melon protocol relies on two main bodies, the Melon Technical Council, representing the developers, and a group of Melon Exposed Businesses representatives, representing the users. Multistakeholder governance is indeed a very promising way to make collective decisions legitimate across large and diverse groups. John Light shared a detailed tutorial on how to create a multistakeholder DAO on Aragon.
Finally, it’s worth mentioning that communities can be decentralized not only because of institutional or crypto-economic incentives, but also because they want to be. Decentralization is an ethos. MetaCartel, MetaGame, 1Hive or the Commons Stack all use tokenized instruments for measuring value and weighing voting power, but they also cultivate a participatory culture that is the most authentic and resilient way to embody decentralization.
The “O” of DAO points to the seemingly simplest term: “Organization”. It’s also the most deceitful. When one thinks about an organization, the image of a firm comes to mind. But it’s hard to picture a firm that would be decentralized and autonomous. Firms’ structures are hierarchical — even cooperatives appoint executives. And firms transact between each other thanks to contractual relationships that heavily rely on the rule of law, so they can hardly be considered autonomous.
Therefore, we still need to question the nature of DAOs as organizational structures.
What have we learned?
The creators of TheDAO claim that it was a decentralized fund, controlled by a crowd of investors, without any corporate structure in the traditional sense: no shareholders, no executives, no management. Corporations have sometimes been defined as a collection of contracts between different parties (shareholders, directors, employees, customers, suppliers, etc). TheDAO was precisely an attempt to dismiss the legal fiction of the personhood of a firm and to keep only the contractual relationships between the parties, expressed and enforced as smart contracts running on a public blockchain.
In this perspective, DAOs may replace firms through disintermediation. One could argue that it makes them something like the “ghosts” of firms rather than a new form of corporation. Instead of considering DAOs as automated firms, we may state that automated and self-governed markets eliminate the need for certain intermediaries. Another way to phrase it is to say that DAOs might be seen as organized economies, rather than automated organizations.
Anyway, we are still at the very beginning of the age of DAOs. In the past few years, we have seen DAO experiments ranging from teams of a few individuals to protocols used by hundreds of thousands of people. Some DAOs have even included artifacts (such as works of art) or natural entities (such as forests). Therefore, we should be very open-minded when thinking about the type of organization a DAO can be.
As of today, the two dominant types of DAOs are protocols and collectives. Protocol DAOs enable interested parties to make collective decisions regarding the critical parameters of a crypto-network based on an open-source protocol: blockchain protocols such as Tezos or Decred, financial protocols such as Compound, DeversiFi or Maker.
Collectives are closer to traditional organizations: They have a smaller number of active members and a wider range of activities than protocol DAOs. 1Hive, dOrg, or MetaGame are examples of collectives that use DAOs to be more transparent, more open, more flexible and fairer towards their members than usual corporations. DXDao is worth a special mention, as a hybrid DAO that gathers an open collective to govern multiple DeFi protocols.
DAOs in 2020? Not a thing, but a dynamic process
It is now pretty obvious that there is no such thing as a pure “Decentralized Autonomous Organization”. DAOs are imperfectly decentralized, very moderately autonomous, and hardly comparable to organizations in the traditional sense.
In 2020, it’s time to acknowledge the fact that most DAOs are actually associated with firms or foundations. In most cases, it is not a problem but a necessity. DAOs are less a thing than a gradual, dynamic process. It is possible to “daoify” organizations, firms, markets, and protocols, i.e., to use blockchain technologies to transform these entities into transparent, rent-free networks owned by their members.
There is now ample evidence that decentralization can be introduced progressively. Melon started as a company and dissolved into a DAO once their protocol reached the production stage. DeversiFi started as a spin-off of a centralized exchange and is making great strides towards a DEX fully governed by a DAO. The venture-backed startup Compound introduced its governance token in 2020, once its protocol already had several hundred million USD under management.
It’s worth noting that this path can be walked in both directions. MakerDAO’s story is one of such contradictions: Initiated as one of the first DAOs, with a strong ethos of decentralization, it has been through multiple twists and turns, from control taken back by its founder last year to a new plan for giving the full authority over the protocol to its community. Another telling case is the takeover of Steemit by Justin Sun’s Tron, with the notable use of STEEM tokens stored on centralized exchanges, and the response of the community forking itself out as a new blockchain called Hive.
DAOs are so comparable to living entities that they can even die! Early on in 2020, DigixDAO announced its own dissolution. Its members faced an arbitrage between the value creation potential of the project and its war chest of 366,000 ETH, raised in one of the first token sales in 2016. DAO members decided to burn their tokens in order to receive their share of the treasury, leading to the demise of TheDAO.
So what can we do with a DAO in 2020? It’s not prime time yet. I don’t think that we’re close to seeing mass adoption. DAOs are still a weird name for an obscure concept, only known by a tiny group of people, even compared to the number of people interested in cryptocurrencies and blockchain technologies in general.
It may seem paradoxical. A lot of people today are interested in self-organization, collectives, and cooperatives. Many appreciate the shortcomings of both free markets and political institutions with respect to the mounting social and environmental challenges of our time. One might think that a technology designed to help communities to govern and decide by themselves, to produce and share value together, would raise more interest from the general public.
My personal belief is that DAOs are poised to succeed once “crypto” itself becomes mainstream. DAOs suffer from the same plague as Dapps in general: user experience is simply unbearable to people unfamiliar with crypto. I appreciate the efforts made by Aragon, Colony, Abridged, and others to make things prettier, faster, and simpler. We’re just not on par yet with Web 2 applications. As long as this is the case, there will only be a handful of people willing to use DAOs.
And that is ok. Today is a blessed time for experimenting, testing, learning. It’s time for creation and innovation. If you are already into crypto (and I expect that the majority of people reading this post are), then today’s DAO tech is good enough for you.
If you start a project, consider raising funds with bonding curves, foster initiatives through open proposals, make decisions using token and reputation-weighted voting rights, onboard new members collectively and transparently.
If you are building a software product or a financial service, consider creating an open network of users, developers, and investors around it, and engineer a token system that will serve them all by giving them a voice and a share of the value they contribute to creating.
Bonus points for taking care of externalities as well. The environment and society at large are the invisible contributors that we all benefit from and rely upon. DAOs help to create inclusive economies, through transparency and cooperation incentives.
When the year of the DAOs eventually happens, it will be because they serve the common good, rather than the limited interests of the crypto community.
Many thanks to Luke Duncan, Jordan Ellis, Griff Green, Hammad, Jack Laing, Auryn MacMillan, Kirstin Maulding, and Ori Shimony.