DAOday.wtf was a blast
Here's what you missed in case you couldn't be there.
Just before ETHBerlin, Inflection hosted a workshop on DAO operations, governance and purpose. Check out the home page at daoday.wtf for more context. Below is a summary of the talks and conclusions of the discussions.
Purpose of workshop
Put people with diverse, deep experiences and thoughts on DAO building in one room
Figure out what works, what doesn’t work so far within DAOs
Find out what we can do to make DAOs better
0. Inflection shared troubled thoughts on DAOs [slides]
We started off the conversation by sharing our very honest opinion on the state of DAOs today.
Most DAOs are very inefficient, and they’re not turning out to be the decentral, autonomous coordination tool originally imagined. To try to turn the tide, we decided to put on this event to highlight struggles in DAOs and DAO service providers within our ecosystem.
The worst three DAO problems right now:
Lack of clear objective/purpose
Governance complexity and protocol value not maturing at same speed
Putting newspapers on screen aka. a regular company but calling it a DAO
These problems are further worsened by the differing opinions on what DAOs should do and be like. There is no ideological consensus on what a DAO is, even within DAOs. This makes it more virtually impossible to be efficient and have progress.
It’s very hard to balance delivering product and managing sometimes 1000s of stakeholders in what is essentially a startup setting. Imagine giving out stock options before there is even a user or a clear purpose of the product! How can you expect to build what is effectively a startup in this setting?
Finally, the basic mathematics of organizations this size without clear paths for autonomy and parallelization are doomed to become bureaucracies. Today, scaling out a DAO to more members is welcomed and even actively pursued, even though it’s not a proof of value delivered to the community. It’s, probably even counter-productive.
For a more positive spin, which we also agree with, please read Vitalik’s recent post on where DAOs make sense. :)
1. Centrifuge on preparing their pool onboarding process [slides]
Centrifuge is one of the longest existing protocols in DeFi. They are undergoing a progressive decentralization since some time. Over the course of time they’ve experienced 1st hand how
are a triad that needs to be carefully managed in DeFi. A key components of this management is staking for sybil protection and incentive alignment.
Currently they are re-defining the process for onboarding new Pools. There are 7 steps with two thresholds, of quality.
Curation of new pools should be as much in the hands of the community as possible. There must also be information transparency built into the process, while enabling human decision making.
The results after 4 months was that 9 projects had started the process of being onboarded, however, as with most DAOs, the participation in voting and discussion around proposals was not very high.
The financial risk assessment of the pools requires a great level of expertise, but the experts in the community have not been very active in this process so far. To facilitate community activation, Centrifuge will propose a new version of the POP, with less complexity.
They finished by asking the audience for suggestions on how to manage making difficult decisions that require expert knowledge, while maintaining community control.
Outcomes of discussion
Create an output-oriented reputation system
Building a reputation as an expert, and qualifying as someone who can accurately judge pools, is very complex. “Old world” proxies such as education, experience within TradFi, etc. may not be the criteria that we want to build upon. A suggestion that came up was instead to create a live game of predicting pool success based on open data, and thereby allowing for people to build a reputation in the specific task they would be performing as experts
Specificity to the actual work
Limits discrimination according to “old world criteria”
No skin in the game, hence prediction game may be gamed
Few data points on what makes a successful pool at this point
Time-weight votes on proposals
By adding the time-variable to votes, we can judge how long someone is convinced of a certain vote. This can be modelled linearly but probably logarithmically makes more sense as understanding of a proposal will probably plateau after some time.
May influence vote outcome according to understanding/conviction of proposal
Time in vote and conviction/knowledge are not necessarily correlated
Short, or negative votes on proposals should be included as well
There is a bias introduced by only having positive votes and abstaining. By down-voting proposals, we may get better outcomes. If we include financial incentives in the shorting (defaulting being equal to the proposal not being accepted, or it being proportional to the Yay/Nay vote ratio), then secondary motives may be stronger than actually finding the best pool.
Gives a larger set of information from voters
People with strong conviction may produce stronger signals
Financial incentives not necessarily aligned to Pool outcomes
Tie expert compensation very closely to pool outcomes
Aligned incentives among experts and community should be a good way to onboard better pools. One way of doing this is by shifting compensation from fixed to variable and tying the variable to pool performance and the experts recommendation for that pool.
Ensures experts work towards
People with strong conviction may produce stronger signals
2. Spearbit ensure quality as a service DAO [slides]
Spearbit is a freelance marketplace of vetted security researchers who pair on collaborative teams to conduct security reviews with clients. They’re redefining the traditional notions of security audits as a box to check and a stamp to get into a collaborative, ongoing effort to prevent any types of unintended behaviour. The security market is going towards a combination of community-driven, decentralized approach to security and centralized security firms.
In the evolution of talent marketplaces, there is a range from completely managed ones such as kaggle, Cod4rena to free-er models like Fiverr, to Braintrust and finally ending in DAO-driven ones such as Spearbit.
Spearbit matches audits with teams of freelance security experts. Part of the audits is to train apprentices and junior security researchers in live audits, while also providing an environment for sharing knowledge. Clients get the certainty of the highest quality, at faster speeds than centralized firms, and auditors get the best compensation in the industry. There are 64 auditors who have completed an audit with Spearbit since inception, and earned a total of $3.3M revenue to the DAO.
Security auditing in crypto is a very imbalanced market, with much higher demand for audits than qualified auditors. The logical way to scale is by upskilling security professionals from web2 or skilled smart contract developers. Both Spearbit and Secureum are working on this, but will not be able to scale infinitely. They don’t expect to grow beyond ~120 auditors ever.
Another challenge is protecting against black hats, who could infiltrate projects, discover vulnerabilities and not disclose them, only to exploit them after the project went live. Spearbit has resolved this by performing full KYC and other methods to vet all team members.
However, the most critical challenge for any service DAO will be ensuring quality of services, while remaining open and effective as they scale. This would require establishing accurate, scalable reputation systems (Otterspace badges for example) and proofs of expertise (GitPOAPs for example).
Outcomes of discussion
The DAO is currently centralized in that there is a core team that guides decision making, but Spearbit is aiming to become more than a traditional org, on-chain. They are already achieving much of the benefits of what we usually call a DAO:
Transparency in fees, rewards and decision making
Take rates are completely transparent through a multisig
20% is flipping the service provider model on its head in terms of fee distribution
Open onboarding of people who are not in traditional financial systems (”unbanked”)
They also reach allow for some of the benefits of traditional organizational structures:
Sense of community and knowledge sharing
Ability to work with legal entities
They struggle with the aforementioned challenges, but also with the question on where value accrues. As with any service provider, the system and structures are supporting the people providing the services. But if the people change, it’s hard to ensure the same quality output. Like a Thesean ship, if we exchange the atoms, can the DNA remain the same? On a more concrete level, how to enable the sharing of upside among contributors, when there is no clear case for a token.
3. Snapshot work on increasing voter participation
Snapshot has taken governance participation by storm. Snapshot uses IPFS to collect votes rather than onchain transactions. As such, they’ve removed the need for voters to pay for on-chain votes and reduced friction in the process. Now they are working on how to actually incentivize voters to vote.
Most DAOs operate with a vote-based governance system, in fact, very often the same type of governance based vote system with a liquid ERC-20 token where 1 token corresponds to 1 vote and token holders stake their tokens in governance contracts on specific proposals in order to cast their vote.
As such, many of these ecosystems fall prey to voter apathy. When a few whales drive the majority of governance votes, it can be a challenge to incentivize voters to be active and participate. Voter participation is likely to have very real implications for protocols as it may be classified as a core criterion for decentralization in a regulatory context.
One of Snapshot’s goals in the near term is to improve and iterate on the models for governance of DAOs. One example of an alternative governance model would be Decentraland, where voters vote based on estate size, token ownership and a few other factors.
Given the degree of uncertainty and the severity of the consequences, developers and other stakeholders have a vested interest in how decentralization of DAOs is perceived.
Outcomes of discussion
Equating token ownership with rationality as a voter, and completely aligned incentives is probably wrong
The curve of marginal benefit in outcome for the DAO vs. voter participation is probably not linear, but more like $-x^2$
Voter apathy can be because of
Unfriendly UX - high threshold of participating
Flawed communication - not clear to voters why a vote that is important to them is important
Misaligned incentives - outcomes of single votes are not individually important enough
Snapshop tries to solve what they can affect, which is mostly around voter UX, but DAOs need to take care of the remaining parts. Namely how to pull in the right voters for the right decisions.
4. MakerDAO governance chronology has all the juice
MakerDAO is a decentralized lending platform and is one of the most well known and heavily used DeFi projects. It also boasts a very complex and nuanced governance ecosystem. Maker recently decentralized its operations and moved away from a central foundation toward a series core unit teams. One part of this governance system is the MIP or Maker Improvement Proposal. These proposals are very involved and require a lot of information to be completed. As such, they create a floor on the level of effort and friction needed to pass governance actions.
We had the pleasure of having someone from MakerDAO participate in our workshop, and give us a chronology of some high-level changes to the MakerDAO governance since its inception.
To summarize, there has been multiple waves of decentralization and centralization contractions/expansions over the years. MakerDAO governance has very much been centred around regulatory risk mitigation and less about making ideal decisions as an on-chain entity. There was the MakerDAO foundation, where the founders and core teams were not allowed to participate publicly, but still had valuable input on decision making obviously (shadow board). There has been the recent decision on onboarding of Monetalis collateral, where there are unclear incentives and the most recent LOVE controversial vote.
In the bear market, things have been brought to a point where Maker is pushed to make a profit and pivot rapidly. This has proven to be quite a difficult task, given dysfunctional governance and MKR votes being mostly made up of whales (VCs included).
Instead of risking to misrepresent the story, here are a couple of publicly available viewpoints of governance controversy at MakerDAO:
Move from foundation
5. RnDAO can measure community health
RnDAO is a research DAO focussed on understanding DAOs better, and helping DAOs learn from each other to operate better. A recent topic they’ve tackled is Community health. Specifically they ask:
What are the features of a community?
How are DAO communities different from others?
When is a DAO community healthy?
How can this health be measured?
The full extent of the research results can be found here, but the tl;dr as presented goes as follows:
Community boils down to a combination of people, place, identification and association.
A healthy community
has an environment that helps people thrive
embodies the values as defined by the community
gives people a sense of belonging
allows for human connection among the individuals
Given this, we can break down the health indicators into measurable things, but still, quantifying how much values defined map to values lived is very difficult, given how imprecise and liable to bias qualitative surveys are. What one has to do instead is focus on proxies, such as
engagements between members
e.g., through message graphs
engagement with the product
e.g., usage of services offered by DAO
e.g., MAU/quitting rates
RnDAO is coming out with a 38-page summary of their findings on how to best measure health within DAO communities, including 10 key metrics they suggest.
Unfortunately, there was not much time for discussion on the final presentation, hence little feedback or suggestions came from the participants. In conclusion, we believe research efforts both academic and independent, such as RnDAO, to be crucial to understand how DAOs actually function today and should function in the future. With respect to the specific topic of community health, we’d like to see someone measure it in different DAOs. Then we can start making accurate statements of how well a DAO does what it is supposed to do, which is, in part, to have a healthy community.
Outlook on DAOs
DAOs are not the optimal human coordination mechanism it was thought to be. Many talented people are pouring in enormous resources into making DAOs work efficiently for purposes that would be much better suited as legal structures with clear incentives and objectives.
Many DAOs are formed for the “wrong” reasons in that they exist out of regulatory necessity (not long term sustainable), to attract buzzword investors, or to mislead retail investors. While we understand that compliance is in many cases not possible while operating at the cutting edge of innovation (see Uber), avoiding regulatory oversight is not long-term feasible. For it to be effective, it should lead to an adaptation of regulation long-term, while providing all the benefits and causing as little harm as possible in the mean time. This means, coordinating capital and efforts online, while reducing the amount of
In our opinion, there are three forms of organizations that are actually well suited to being true DAOs/on-chain organizations/internet-native corporations:
Service providers/freelance marketplaces
Off-path research efforts (insert DeSci-meme) that enables people to work on topics not fundable within traditional academia or industry R&D in constellations not possible
Low stakes, for fun communities coordinating around digital assets
That being said, we want to further the experimentation and learning from the last decade of on-chain community and company building. What we struggle with in building any organization today, will necessarily be part of the equation in on-chain organizations, but we can try to minimize the coefficiencts of those factors. There are great examples of experimentation in both governance and on-chain organizations, such as outlined in our last post on on-chain organizations.
Thanks for reading SVRGN! If you liked this, and want to stay up-to-date with what inflection.xyz does, subscribe now! We’ve got some interesting content coming ;)