Kerman Kohli
Kerman Kohli
Building an On-Chain DAO Framework w/Shreyas from Llama
0:00
-28:26

Building an On-Chain DAO Framework w/Shreyas from Llama

Learn about what an OS for DAOs would look like with Shreyas from Llama

There’s many challenges with governance but very few are qualified to speak about them as well as Shreyas. He started out as a DAO contributor in when DAOs started and now he’s building a framework that will help unify things. If you love governance or DAOs, this one is for you!

  • His Twitter: https://twitter.com/HelloShreyas

  • LLama’s Twitter: https://twitter.com/llama


00:00:01

Speaker #0

So would love to learn a bit more around like how you got started with Llama and like, what kind of drew you to governance in the first place?

00:00:09

Speaker #1

Yeah, I think when I uh was exploring a bunch of projects built on Atari, like um a make a compound unis one of the things that was super interesting to me that what, what was that? Um there wasn't just the um uh smart contracts that we used to build these protocols, but smart contracts were also used to uh govern how these protocols were upgraded and maintained and how, you know, a shared community treasury was allocated.

And so all that seemed like super interesting and compelling to me. So I wanted to, you know, build something in um uh to, to, you know, help foster like, you know, more uh interesting activity among these decentralized organizations.

And uh yeah, it started off with um you know, me, my co-founder and, and, and a bunch of other uh team members like contributing to various dos and so the uh you know, the scarcity at that point in the in the market was definitely with just contributions.

There was no one doing any work in these dos, like there was a lot of um there's a lot of work to be done.

Like, you know, taking an example of a, uh you know, there's uh the core team of AVI that's built a V two and V three, but um there's a lot of risk parameters that need to be updated, you know, every so often there's interest rate strategies that need to be updated.

There's um you know, treasury allocations that need to be made, there's uh you know, a bunch of uh periodic maintenance of these like governance contracts themselves.

And so we kind of got involved with, you know, doing some of that work um across different projects like avi Unice um you know, and others.

And um yeah, I think over time, uh we worked entirely through unchain governance and contributing on um on, on forums and then through proposals on chain.

And we kind of realized that there were, yeah, a bunch of problems that uh you know, on chain governance had that we could uh help solve. And so we, yeah, thought through the best way to, to solve them.

Um And you know, whether we could improve some existing frameworks in, in, in a light way or if we could um you know, modify how, uh you know, governor or um uh you know, uh a governance or these other systems work.

Uh And then our conclusion was no, like, actually there's that we could rethink this much more from first principles.

And there's um a bunch of these things can be addressed to uh really focusing on um a few key ideas, which is, you know, decentralizing, ready to access control rather than just uh coins and tokens.

Um Focusing on um the idea that uh you can, you really kind of uh build, build a governance system for contributors and contribution uh rather than kind of just designed for um uh owners and investors.

Um And there is a real place for owners and investors um and tokens already uh encode like a very useful cap table. But what, what exactly should the investor class and the owner class do versus how should projects be administered? Um So if you see a company, right, like it's not like your uh your investors are sort of um uh having a say on how like $1000 and you know, and an engineer's ramp credit card is spent.

Uh And, and why you kind of separate some of these owner and, and administrative functions. Um And it's a cool way in, in crypto that those could actually be encoded on chain.

And so, yeah, that's like the, the general premise and we kind of walked in that problem and yeah, we're, we're sort of um excited to launch a launch chain governance framework

00:03:40

Speaker #0

for sure. So what are like, I think some of the key problems you see in governance today. Um that. Yeah, like it's probably start there.

00:03:49

Speaker #1

Yeah, I think the, the first thing is that uh things uh are really slow and so um it takes uh a lot to um uh pass a governance uh an on chain governance proposal because um one, the people who, you know, own these tokens are either, you know, um uh investors or are uh you know, team members but they can't vote because of regulatory reasons Um Or they are, you know, so in some cases, delegates uh but they don't have full context, say on a, on a. Uh And so there's, there's different stakeholders that either like uh and each of these stakeholders don't have full context on each of these um uh proposals being made.

And so things uh tend to move slowly where there's, there's a default to uh consensus building and debate over action, which uh is good for some things.

It's good to prevent like World War Three and to prevent something like really extreme and catastrophic from happening, but it's kind of not good, especially in the early and growth stages of protocols where you on some decisions, you should like iterate quickly enough.

Um You should um actually, you know, experiment and, and, you know, try out some things and, and the cost of failure actually is, is low on some of these decisions. In fact, the cost of not moving quickly is, is much higher.

Um And there's some decisions that are catastrophic and, and those should have a very high bar. Um and those should move very slowly.

Uh But I think this granularity of separating uh what decisions, uh those verses, what are like the decisions you can actually make quickly again, just doesn't exist with some of the existing systems.

Um And uh and yeah, I think the um and related to that, like, you know, if you think about allocating the treasury just for like spending, not for any financial reasons, just for, for operations and you know, ways to kind of grow the the project.

And then you have like one main treasury, right? So you just have uh one treasury, there's these governance proposals to allocate the treasury uh because there's just one treasury and the governance process is so heavy, you just get these massive proposals like it's like all, you know, like a $1 million proposal to allocate to a service provider to contribute.

And that just is like, you can't allocate uh $50,000 through that process, it just doesn't uh it's not at all like uh sort of cost effective for someone who, you know, uh to, to make that proposal.

Um And, and instead, you know what we think it is a better system is you actually have multiple accounts with different levels of, of per commissioning where um for, for the big one, you actually need the entire do to kind of participate in that Um And then for like, you know, smaller uh sub accounts, you actually uh can have granular permissions to uh different stakeholders or groups um to allocate that capital and you can always have the, the dow involved in the same process.

So one way that happens today is like you allocate from the Treasury to say like a grants multi sig. Uh But the problem with that is you now have to, it, it's not like there's $10 million in this grants multi sig that's managed by 10 people.

And now you have to buddy up to the multi sig Sina to get anything done.

Um But I think a better system is that the multi should actually uh you know, being code in the same governance system uh which is kind of, you know, what with, with the llama system that, that's how it works where you actually can issue it, issue some money to a, to a sub.

Um But you could have something where, you know, these like 10 members can approve things, say below $50,000 about $50,000. You can have uh token holders or other community members that are involved in this decision.

You'd also have um uh token holds like uh disapprove decisions, like there'll be a veto period where people can disapprove decisions.

And so, yeah, I think um there's ways to, you know, that there's a lot of potential ideas but like you could kind of simplify things into like, you know, three broad categories of proposals where, you know, some are optimistic where these are periodic changes that um that don't need, that are specialized services providers.

Uh make, let's say just scrambled updates don't need governance involvement. Those happen optimistically, but the governance can, can veto things some things that are standard where you, you know, uh approval of a big contract that requires a typical governance vote and some, some things happen through an emergencies proposal, let's say where um you actually bypass it uh the typical time lock.

So if you saw like the compound um uh liquidity mining bug, like the problem was, it took like 10 days to actually um solve that because of how on chain governance worked.

Uh Everyone knew the effects like one day in, but you actually had to wait through the time lock period.

Um But if you could have something where if you reach a high enough quri if like, you know, 50% of people agree that this is what needs to be done. Um You could do it within 12 hours a day, which is what happens in emergencies, right? Like everyone actually agrees on what, what's to be done. Um Just takes um it just takes a while when you have a big time luck

00:08:50

Speaker #0

for sure. So uh and like, what are I think the hidden problems that are not obvious on the outside? Like I think speed and like autonomy are probably, like, very clear ones. But what are some of the, maybe the more nuanced problems that, like, maybe you've seen from the inside, um, that kind of happened. Maybe it's like political or it's technical like, yeah.

00:09:12

Speaker #1

Yeah, I think, uh, it's a good question, like the, the hidden problem that, like, everyone that contributes to, that knows. But, uh, you know, people don't talk about as much is just like the showmanship element where like there's given that um uh you uh are, you know, are, are uh uh decentralizing voting through uh a token uh voting dial.

Um You kind of uh you know, there, there's two options, right? You either get the tokens through um uh because you bought them uh or because you, you know, the, the excluding the core team, I, I think you bought them or two, like um uh you've been delegated them and the, the bottom category is investors in the delegated them category are people who have to like, uh you know, uh it's a popularity contest and so you kind of have to um uh show that you just have it and it's, it's not, you know, it's like, uh yeah, just imagine another running another organization that way where uh you know, the most popular person kind of um has the most proposal power, voting power.

Um It can be good in, in, in some cases for some decisions. But yeah, for a lot of decisions, like should that be how um uh you know, a, a technical upgrade is made, like, you know, probably not.

Um And so I think the, the, the default to like, you know, popularity contests in, in a lot of cases is, is like, you know, suboptimal.

Um I think the uh the way that probably um uh I, I think I mentioned the other like kind of hidden problem of, of treasury all, you know, spending and allocation where there's a default to large spending over a small spending just because um of the amount of time uh it, it takes to go through the process.

So you, you'd rather just like make it big and make a big splash and, and go through an UN chain proposal than, than not

00:11:02

Speaker #0

completely. It's sounds like a lot of what you're saying is really, we're just like rediscovering how companies work in like this kind of like slightly like awkward way. Like, what do you think is the difference? OK. So there's two questions, what do you think is the difference between an on chain company and a Dow or on the flip side or actually dow is not that great? And we should be thinking more like on chain companies um borrowing regulatory uh shenanigans and one because Dow sounds great to avoid uh certain risks. But yeah, like I'd love to hear your take on that.

00:11:33

Speaker #1

Yeah. No, I think, I think actually that was really great. Like I you know, it's just, uh, the cool thing is you can kind of write these rules into uh contracts and you can write them however you want. Um And so you could write the rule, you could write the rules to make it, uh to, to design an organization that is, um, uh, extremely, um slow and has a high bar. You could write the rules to design it where it's almost like a pop up.

Do I think vital had this thing? Uh you know, saying something like uh do are good either at um at, at, at, you know, building like a world government or something or, or they're good at like doing something like pop up. Uh uh If I'm like constitution there.

Um Yeah, I think the, the, so I think the flexibility is like, interesting, like you can kind of encode whatever you want. But um uh yeah, the, the unique things about that is like one is uh the transparency piece.

So like, it is very, you know, it's very distinctive that you can actually see who the members are, who uh owns how much uh how much um is in this, like shared like balance sheet of this unchain organization um how that's being spent.

Um And so that is like that, that's like, you know, pretty unique and interesting.

Uh There's the, the other element of um uh like truly encoding like the, the key changes to the project into the code uh into like the, the governance contract literally where you have, these are the admin functions uh that uh the protocol has and, you know, we're not using this three out of five multi and, and there's a, there's a reason why like a community can govern over uh these particular admin changes that just makes it interesting, for example, like, um with a, if they didn't have um asset list things through the dow but instead, like, you just, you know, just a companies had a, you know, the team and, and you, you said a ping or something to add an asset, it just is like less uh it's just less interesting.

I actually think avi wouldn't have grown as much too. Uh the interesting piece of like a uh being across all these chains and the listing process being actually quite transparent where this is the, you know, this criteria you need to meet.

And then this is like uh the way that the, you know, data proves it and these are the providers who kind of need to prove it. And like these people are not um you know, the, the R A company CEO. Um Yeah, it is unique.

Like I think it adds to longevity of, of this like lending protocol. It is a learning protocol, it's kind of not a lending company.

Um It's um uh it, it makes it so that like if that is like captured by any specific party, then you know, that can be resolved because, you know, governance kind of doesn't allow one party to capture it.

So, yeah, II I kind of um I don't think we're like reinventing the wheel in a sense that like, uh yeah, we're just like creating companies.

Um I think you can run, you know, companies on chain and you can run some things that don't seem at all like companies, right? Like, uh you know, buying a shared like property um you know, like, yeah, through like crowd funding over a few weeks just looks nothing like any existing uh an existing company. Um And so, yeah, it's just like a wide, wide set of um you know, possibilities.

00:14:50

Speaker #0

Yeah. In, in that example, I really liked it because it's kind of like this very like, I don't know, mundane process that you'd have to run like internally within a company, but actually by kind of democratizing it a you make the organization more efficient and you add more transparency.

Um and it kind of becomes like a marketing thing as well. So actually, it has a lot of upside of like that. It's almost like it kind of social media algorithms in some sense provided there's no profit sense attached to it, like democratizing that actually is beneficial.

So I think in that instance, like data transparency like really shines through. But if you kind of think of like, say your average D five startup getting started today, like it, do you think on chain companies are like, first of all, like, are there any instances of on chain companies you have seen? But do you think actually it's kind of like a do is a privilege you earn over time rather than something you start off as? Um because I think for all the stuff, like it just requires more infrastructure and people no matter how you spin it, like or more tooling. Um So, yeah, like, what are your thoughts on that?

00:15:56

Speaker #1

Yeah, we like to use the word on chain organization for like kind of the reason you described where it's um like the, the, the dow just seems like this uh pretty uh you know, big thing you need to achieve like complete decentralization. Uh And yeah, first of all, like nothing starts out decentralized, right? Like Bitcoin was started by uh and uh and you know, uh Italian with metallic and a few other people.

And so there's always like a few group of people that, that start something, even if you decentralize the, the process of uh you know, how you get the ownership by proof of work or something. There's still a few people on that that know about it.

And so there's always centralization uh at, at some level when you're starting a project.

Um I think there should just be more of a part from um starting with like a team of um five people uh or three out of five multi um running these admin functions of, of a protocol uh to, to, to this like eventual, you know, thousands of people um administer the project.

I think it's like, you know, a space in between that we just like completely ignore, like um which is one of the things we're addressing where, hey, if you want like 20 people, uh you know, 10 of which are like, you know, power community members, users of your project to have some light ways to administer.

Um uh the, the key functions, for example, to um if there's a pause function, uh you know, for a uh for, for a protocol um then, you know, should the core team only control that pause function.

If you have like 20 power users can, they also just have that right to control it. You can't do that in a multi, like you can't do a 20% multis with a, with like a super custom like approval strategy.

Uh But I think, yeah, like, you know, one thing with Lama you can do is something like that where you, yeah, you can expand the number of members uh from, you know, 5 to 20 to 50 to 100 without having to think through like launching a token and doing tooms and you know, all of that stuff.

Um We kind of issue these policies through non transport NFTS. Um And so, yeah, you, you can kind of, you know, II I think there should be more organizations like in that mid space where they don't, like, you know, uh do this like mega set of Token launch and, you know, fully like Token based governance until they already have like, you know, not just product market fit but like a sense of like what the, what the deal with the Token is, like, what, what do they use for? Like, you know, where does it fit in with, with governance? Um uh Yeah, and, and so until then, like, there is a, you know, a better thing than a, you know, five multi uh for a bunch of reasons like effectiveness and, and, and decentralization and legal and to increase like trust and credibility with your community.

And, and those are like, you know, good business reasons to, to have more members involved uh even before you kind of have a token.

00:18:44

Speaker #0

So, so one thing I believe is like seeing the problems is like the first step, the second bit is solving it. So it sounds like you clearly see the problems, which means uh you probably had like a working thesis of like, how to solve these. So I would love to learn a bit more on like what you're thinking around uh is that and uh feel free to like, talk about Llama where it makes sense.

00:19:04

Speaker #1

Yeah. Um Yes. So our, our solution to this problem is by um uh you know, we issue these so organizations can can use a um to issue these non trans NFTS uh which code permissions for what you could do. Um And uh those permissions and code like which um actions you can take on this protocol, what what functions you can call? Um And then you can also create um specific action, execution rules.

How exactly you execute those actions, whether it's through um say an optimistic voting strategy or an emergency voting strategy or, or, you know, something uh something very different.

Um And, and, and the third piece is um uh is the accounts piece which is uh you can actually permission accounts the same way you permission the rest of your protocol.

Um And you can kind of do some of this um uh you know, fund management like better and more programmatically.

So, um yeah, where, where, you know, how examples of like uh you know, how Llama would, would solve some of this is uh if you have um uh you know, a lending protocol that has a uh you know, multiple um risk service providers.

Um And, and right now, what's happening is that every um uh risk parameter update goes through a uh you know, a typical kind voting on unchain proposal where like the people voting on this are either uh investors or delegates that don't have full context on, you know, why you would increase the liquidation threshold from 70 to 72%.

Um Then you can actually have, have it so that the Dow approves what exactly the risk providers could do so that the providers, the dow proves that what the risk providers can do is update liquidation threshold, say between, um, 70 to 70 70 to 80% and, uh, between that threshold, like they could keep updating risk parameters, uh, liquidation thresholds.

And, uh, and it gets approved optimistically, which is, you know, if they approve it, it goes through unless the dow vetoes it. But anything beyond that maybe happens through a typical uh governance process where, you know, actually Toklas vote on this.

Um you could structure things like quite um well, like you can actually produce the um maybe overhead for like the average token oder while still making sure they have really strong guardrails with um uh preventing the, you know, this provider from doing something that they want.

Um and it is provided and, and for actions like say freezing, freezing an asset on a particular market that always will have to happen to, you know, uh AAA wider involvement with the community, those risk provided maybe can, can propose it.

So, um yeah, I think the the idea is really uh to take like some of the things that really work well with um existing governance systems and then uh yeah, modify them with um uh with, you know, this like granular permission and, and uh custom like approval strategies to then like help these projects like, you know, be more effective, involve more people um um more early but also like, you know, not compromise on um you know, growth, the decentralization.

00:22:11

Speaker #0

Sure. And like, how do you solve the problem around like enabling that transition from like five dudes on the multi sig to like a proper decentralized out? Like where is maybe it's not more of a product thing but what's your philosophy to actually like guiding teams through that?

00:22:28

Speaker #1

Yeah, I think, yeah, so when you start off like um yeah, you know, keep it simple, like uh have uh have, have a multi, like ideally, you know, five out of nine multi if you can to run these admin functions, uh limit the, the admin functions, you know, in, in your protocol if you can. Um And then, yeah, I think the, you know, the first thing of course is to just like launch the product and get, get usage and that should be the only focus of, of these projects and nothing else.

Um And then once you have like, yeah, some sense of uh product market fit or, or some um Yeah, I think it's definitely worth thinking, thinking through how do you uh reduce the liability that that has? Um And, and that's for various reasons.

I think one reason is um uh there's a trust uh issue uh which is if you have really um you know, power users in your community, uh especially um in, in some cases you know, deposing, deposing a lot of assets and trusting you.

Um It actually is good business wise to really think through like, uh how do you increase that credibility and, and, you know, reduce the um the, the but, and, and this month uh to, to give trust to your community that there are like more people um involved in, in, you know, governing this and, and kind of making, making these um uh admin updates than like, you know, just your code, code team that, you know, could go to or other thing, you know, mistakes could happen.

Um And so, yeah, I'd say after product market fit, like, you know, one thing you could do is issue like um permissions to more, more of these um members to kind of govern uh the project.

And so, yeah, like think through like who are the 2030 you know, uh party like power contributors, users, um people you want to attract in your community. And that way you can also have lightweight ways to uh test out what a community looks like, right? Like you kind of want a productive community, a group that uh is high quality, that does things and it's super hard to just do that through a token air job. Uh You just, it's just hard to attract it. Like, you know, just with that.

Uh In fact, you probably attract the opposite kind. I know you've done a bunch of interesting research and um on kind of the, the cost of uh uh acquiring users. And I think the, yeah, there's, there's always, um, you know, it's like worth it.

I, in some cases, but it's just like it, it's not an effective strategy to get, um, to, to build your, you know, uh 2030 people in, in the dow that actually eventually, like, do all the work. Right. So, I think there's, um, there's lighter ways to do that.

There's ways that are non-financial and speculative where you could just get them to. Um, yeah, just administer different things. Uh, if your project, you know, light ways for them to, um, you know, govern or, or make these protocol upgrades.

Um, and then, yeah, once you're ready and you've tested like exactly what the Dow should do and what the core team is doing, which is, I think one thing that a lot of teams don't focus on, which is what is the role of the, if there is a centralized, like, labs team, what is the role of that team? And what is the role of the Dow? Uh, if the role of the Dow is just to kind of, uh, you know, sit pretty and pose for a picture like the Dow is not gonna do any, like the Dow is gonna be really messy and, um, it's gonna attract, you know, strange people and, uh, everyone's gonna be fighting.

But if the Dow has like a, you know, AAA purpose. Uh And there's some like productive thing the doo can do uh which say like nouns, this case, it's like, it's to just allocate money to people who are like, you know, building on nouns in obvious case, it's like uh it's, it's people, you know, maintaining the protocols.

So like updating risk parameters or um you know, updating the, you know, making sure the safety modules. Um Right. But those are like, it's encoded in the protocol, those are like governance um and data functions.

And so now you think to what those are and um if it's, if it's nothing, if the protocol is super like um um minimal in terms of governance like Unisa is, then it probably is the Treasury. It's like, what do you do with the Treasury? And how, how do you, you know, how, how do you allocate the Treasury for pro projects, you know, building on uni to what? Yeah, what success looks like for the dow.

Uh And then, yeah, and then, you know, it's a good time to think through like how you um actually launch duke in and decentralized.

And you should probably, you know, uh look at how these like the the experiment of these like 20 to 100 people um in, in this like light minimal do without, you know, uh much regulator has it has just gone like what, what is the experiment uh been like over the last 6 to 12 months? Um what are productive things that people are actually doing? Uh Would you refine, like what the doo should actually do? Like should the dow actually not do these things and should, should some of these functions be more centralized and some of them be more centralized? And that way, like the token launch is just much more informed and kind of more um you kind of can also think through like just who you want in this broader Token community

00:27:12

Speaker #0

completely. It's uh I feel like I got like a complete master class, so I had to launch it down all the problems of launching Garance. So that was really, really informative. Um I think we're coming up on time, but is there anything else you'd like love to let people know who are interested about the stuff thinking about it? Um Yeah, I just wanna like hand the floor over to you to close up.

00:27:32

Speaker #1

Yeah, I think uh well, if anyone has any like governance related ideas definitely hit me up and share us on Twitter. I'm always looking for, I, I would say like, there's a bunch of interesting governance um uh experimentation right now. Like, that's one of the like bear market things that actually um I'm like, you know, excited about like, there's, it's a bunch of people like in the similar idea space, like thinking about this problem. And um I'm sure like, we'll have more interesting like mechanisms and products.

Um you know, come out of this. So, um, yeah, if anyone's working on anything or has ideas around this, like we'd love to hear about it.

00:28:12

Speaker #0

Sweet. Well, thank you so much. Really enjoyed this chat and uh look forward to hearing what the response of this is. Yeah, sounds good.

0 Comments
Kerman Kohli
Kerman Kohli
thinking in systems. data, identity, growth and war stories from being a founder in crypto.