Anna (00:00:01):
Welcome to Zero Knowledge. I'm your host, Anna Rose. In this podcast, we will be exploring the latest in zero knowledge research and the decentralized web, as well as new paradigms that promise to change the way we interact and transact online. This week Tarun and I talk with Martin Köppelmann, CEO of Gnosis. We start off talking about prediction markets, where the project actually started from and how, as a project, they've moved away from that. We then cover the new CowSwap DEX, the success of the Gnosis Safe, its use as the foundation of new generation of DAOs and SafeSnap, a way to automatically execute governance proposals on SnapShot in DeFi governance.

Anna (00:00:48):
Now, before we start in, I just want to share a message to all the folks listening, who might be thinking of jumping into the space. If you're looking for new opportunities to work at top zk-focused companies, we have now a jobs board on the website. I'm adding a link to this in the show notes. It's a simple list of current open positions at some of the best companies in the space. And if you do apply, please mention that you found out about it through us. I also want to thank this week's sponsor, Least Authority. Least Authority, a leader in the security of distributed systems, has created PrivateStorage, which is a private secure, and end-to-end encrypted storage solution. Privacy is at the core of Least Authority's work, and PrivateStorage was created using privacy and security by design principles and includes features like end-to-end encryption, zero knowledge access passes and a countless authorization. This ensures users have greater autonomy and assurance that their data is protected. PrivateStorage cannot see user data when it's being stored on their grid. It's built on a commitment to transparency and open source technology by the Least Authority team. By using technology that is publicly reviewable, they allow the users to know exactly what happens to their data, visit private.storage and register to be notified, once the service has launched, this is a one-time notification email and not a mailing list. So the addresses will be deleted after the notification email is distributed. So thank you once again, Least Authority, for supporting the show. Now here is our interview with Martin.

Anna (00:02:24):
So today Tarun and I are talking with Martin Köppelmann, CEO of Gnosis. Welcome to the show, Martin.

Martin (00:02:30):
Yeah. Thanks. Thank you for having me.

Anna (00:02:32):
Now, we have already had Friederike, your business partner and partner on the show in the past, and she gave a rundown of the origin story of Gnosis, how you guys got started and what had happened up to that point. But that episode was two years ago. So I'm very curious to hear what's new with Gnosis. What have you been up to since then?

Martin (00:02:55):
Yeah, a bunch of things. So, I think, we really morphed into this broader Ethereum infrastructure company. So we started focusing on prediction market, but started just so early within Ethereum that a lot of pieces that were required to build things were missing. And that's why we essentially moved one layer down the stack and infrastructure like wallet, like the Gnosis Safe, like exchanges modules for DAOs and various other elements.

Anna (00:03:26):
I think I actually spoke with Friederike about this idea of Gnosis having so many very, very used projects and very, very cutting-edge ideas and pieces of code or products basically released into the wild. And I had asked her is it a dev shop? Are you doing lots of things? And at the time she actually said something along the lines of like "our current focus is on focus". And I think it was that idea of you expand out with ideas and then you have to narrow down into very specific ideas. Would you say right now, are you at a phase where you are more outward, trying to experiment a lot, or are you actually focusing in on a few specific projects?

Martin (00:04:11):
No, we are finally definitely focusing on the things that turned out to work and that are mainly the wallet, the Gnosis Safe and exchanges or what you'd call Gnosis protocol and, surprisingly, other things that we initially put a lot of folks on, for example, prediction markets, we, to some extent, have put aside currently, because we really focus on the things that are simply working.

Anna (00:04:36):
So what has changed for you on the prediction market front?

Martin (00:04:40):
Yeah, a handful of things. So, I mean, first of all, to build prediction markets, it was always necessary to have some underlying mechanisms. So obviously to have a prediction market, you have to have a market. So DEX was always required. Of course other projects are and have been building DEXes, DEX space is insanely competitive, but still when we started there, there wasn't a DEX, so we naturally needed to build DEXes. But that being said, we still feel building a prediction market, as an application on Ethereum right now, is probably not a good idea, because while transaction fees are one aspect, currently a simple trade can cost a $100 on mainnet, and that already kills a lot of use cases.

Anna (00:05:26):
Yeah, especially when it's prediction market, it's not even like a trade necessarily, it's a bet.

Martin (00:05:33):
Yeah. So it's very clear that the size of the bet would need to be quite large to justify a $100 transaction fee. And again, that already killed a lot of use cases. So of course you could move to layer two, but we decided for now to do more foundational stuff, like again, the Gnosis Safe or work on DEXes on L1.

Anna (00:05:55):
Tarun, there's a question I'm going to throw to you. What are your thoughts on prediction markets? Did you follow that space?

Tarun (00:06:03):
Yeah, I mean, I think probably in a similar vein, I really thought they would be the killer application a long time ago, but I think over time, it's just proven that they need much more care and love than a lot of other applications. And I think we're just not at the point where that will happen, but does this feel like Webvan in 1999, like that later became someone else, 20 years later figured out how to do it, for sure.

Anna (00:06:31):
What is Webvan?

Tarun (00:06:32):
Sorry. So Webvan is like DoorDash or Uber Eats or whatever, but this was in the first dotcom bubble and it didn't work. They literally were the same thing, but they needed the mobile web and cell phones to make it work.

Anna (00:06:48):
Makes sense.

Tarun (00:06:49):
And so I feel like, to Martin's point, the infrastructure is still just not at the level. We don't know what the cell phone is for prediction markets, but it's clear there's something that's like that. I don't think anyone can tell you exactly what the thing that's missing is, because there's just so many things that need to get built. But yeah, I kind of feel similarly, although I think these semi-centralized alternatives, which might have legal issues over time, will be a crutch until we build a good enough infrastructure.

Anna (00:07:24):
Is that the general feeling in that part of the ecosystem? Cause there was other prediction market projects out there, have a lot of them moved away from those original models?

Martin (00:07:35):
I would say there's still kind of both, always new players coming and trying, and that's obviously a good thing, but I have also become more skeptical about other aspects. Prediction markets, one big problem or challenge, at least, is the zero sum nature with the token system, you have this non zero sum games, where everyone goes into a token and everyone gets rich, ideally value is greater and so on. With prediction markets, fundamentally you have markets that are zero sum markets, where someone needs to lose, and the question is who is losing. And the problem was always: you have information and you use this information and you make money, but you can only make money if someone else is losing. There are some answers to that and you have a political betting, where it's quite clear. And we saw in the last election that there was probably a decent amount of irrationality involved, which is, in a way, good for prediction markets, because it answers the question who's losing. It gives in a way, it funds those who potentially provide the valuable information. But in other markets, when you ask, when will Eth2 launch, and the question is who will be losing, who's the fish to finance the revealing of this useful information. And then some answers might be, or what people tried, is someone is willing to pay for this information, but that's not easy. So, I mean, who's paying a hundred thousand dollars to get this information, when then it's public. So those were just a few thoughts.

Anna (00:09:21):
That's interesting. In that election example, I felt like that was one of the times where I saw the topic come up again, in a big way. I think people tweeting like, "Look, look, they're predicting it correct", while the results were still being counted, even though kind of predicted in a wonky way at the time, it was also swinging wildly back and forth, depending on what a newscast said. But what did you make of that experience, watching that in action? Was it promising for the concept in the future? Maybe once the tooling is there?

Martin (00:09:52):
Yeah. Well, I mean, I would say it definitely was a moment, where actually some market activity happened and then definitely the surprising thing was how long after even the election happened. The odds for Trump still being elected or still whatever, winning, where at the surprising 10, 15, sometimes even up to 20%. And I think to some extent that can be explained by inefficiencies of the market set-up. So you would need to lock your capital specifically for market, denominated in stablecoin, then you might get best case, you might get a payout of 20% in timeframe of a few months, when at the same time DeFi craziness is happening and you can get 50% a year without any risk, why would you take the 20% with risk?

Tarun (00:10:51):
The interesting fact that, and this is purely anecdotally, is that people I know, who were professional sports gamblers, who are, in some ways, the ideal customer for prediction market, most of them just trade crypto now, instead of sports betting. When the market die, when the bull market dies and we go back to the bear market, I'm sure they'll go back to sports betting, but a lot of sports betters, especially in the pandemics, were like, "Well, I need to find a new thing". And I think that prediction markets unfortunately have the return structure of sports betting, less of the return structure of these bull and bear cycles in crypto.

Martin (00:11:37):
Yeah, I mean the same thing. So specifically in a bull market, when the opportunity is so large everywhere else, then sports betting or prediction markets are actually quite a tight business and you might expect a few percent on something, if you do it well, but...

Tarun (00:11:54):
With a lot of work.

Martin (00:11:55):
Yeah, exactly. It's a lot of work and it's not that appealing if there are other opportunities, better opportunities.

Anna (00:12:01):
So I think most of this interview, we're actually going to be talking about some newer ideas that have come out of the Gnosis team: CowSwap, SafeSnap. But before we do that, I actually wanted to throw back to an episode we did pretty recently, where Tarun had mentioned that you potentially were at least part of the group that came up with the initial AMM model. And since we have you on the show, I thought maybe we could talk a little bit about what timeframe are we talking? And how were these ideas first developing?

Martin (00:12:34):
Yeah, I mean, as I said earlier, for prediction markets to work, you need to have markets. So we actually started to work on prediction markets, and therefore for markets, even before Ethereum launched. And we wanted to be ready at launch. And actually two weeks after Ethereum launch, we had a first market running with back then thousand ETH in funding — that wasn't a lot of money back then, but anyhow. And from very early on, we used, in the prediction market world, automated market maker, and in the prediction market world, the market maker that's quite famous is LMSR market maker, invented by Robin Hanson. So we were coming from that world or that was kind of the standard for prediction markets then. And then there was this movement, and we're talking about year 2015, 2016, to say, "Okay, well, we might use automated market makers also outside of prediction markets". And I think there are many, many people involved in a very open exchange of ideas. I mean, I think that's really one of the core things that makes Ethereum so great, in my view, is this open marketplace of ideas, sharing ideas all over the place. So discussion happened on Reddit and on forums, and really different people were involved. But yeah, at some point, while we wrote a blog post describing how to build a DEX on Ethereum, that was early 2017, and there we concretely proposed this formula, this simple XY=K formula, that then became very famous through Uniswap.

Anna (00:14:15):
But at that time did you realize what this was? Or was it fully developed? Or what had to happen since then?

Martin (00:14:25):
Well, I mean, we had code for this, so we implemented that actually, but I think it was probably too early. In those days there were more urging problems to be solved. And I think, in the end, Uniswap definitely was super nice and super clean implementation of that, definitely also a cleaner and better execution of it, and then rightfully became a famous exchange for this.

Tarun (00:14:57):
I have to say, it's more amazing to me that these discussions on Reddit posts in 2015, around the "Can you stop trading" incident, I remember there were a lot of Reddit posts between you and Vitalik and a few other people that were really more or less pantomiming, like, "Hey, is there some invariant formulation? We don't need this, we don't need all the properties of an LMSR, it's overkill". And I think it's just interesting, how this set of Reddit posts is more powerful, than 20 years of academic research on this stuff, because most people in academia were focused on this concept of high dimensional markets, because it's kind of sexier, but in reality, how many people are really trading thousand asset portfolios for another thousand asset portfolio? Very few, right? And I think the academics definitely lost, they got caught in the thing that sounded sexy versus the thing that would be useful. And I thought that this scenario is one of the more interesting stories that I hope society at large one day appreciates. But obviously I think maybe it might be our little world. But I think it's a very cool instance of everyone believing the academics' version of the world, it's just like prima facie true or something. And somehow it just turned out that market forces created something a little bit better, which is cool. To me that's somehow the narrative which is very unique in that way.

Martin (00:16:34):
Yeah. Although one reason we actually dropped this effort was our, probably back at the time, overestimation of the problem of MEV, so of Miner Extractable Value, but obviously for the last 2 or 3 years, that hasn't been a big topic, but now it finally hit reality. And so I would say we spent the time already in anticipation of this problem and, hopefully, have now solutions that contribute to this problem.

Tarun (00:17:09):
Well, I think that's a great place to dive into how Gnosis is thinking about MEV, both from the stance of an exchange, but also from the stance of things like Flashbots. Maybe we can start with the etymology of CowSwap, because I think starting with the name might get us into how that's really aimed at MEV.

Martin (00:17:32):
Right. Exactly. So coming from what we see right now is that, if you place your order against a market maker or basically you're sending your order, your transaction to a miner, miners will extract value from you. So basically you're handing them over a transaction and they will use that against you. And they very explicitly do that. They are very public about it, that they will, for example, sandwich your transaction. So you send your transaction trying to buy a token, they will, right before this transaction gets into the block, buy the token, just before you move the price up, then let you buy at a higher price. And immediately after your transaction was mined, they will sell the token again and really just extract this profit. And that's a problem, I would say, in multiple dimensions. So of course, first of all, it's not great for the trader, but it is actually, in my view, a deeper problem, and even challenges the concept of decentralization, because what we are seeing now is the next step, that to some extent, users are starting, or some protocols are starting to defend by, for example, not sending their transactions at all to the public mempool, but instead just doing it privately and having private arrangements with miners, which threatens the concept of censorship resistance and permissionless blockchain. But maybe once they are back, what our high level, our attempt or approach is to say, "Well, instead of sending your transaction to the miner, who already signaled that they will not treat it well for you as a user, we basically ask users in CowSwap or in Gnosis protocol, hat's the underlying protocol, CowSwap is a product on top of it, to send their transactions as a so-called meta transactions or a specific order format. And that cannot directly be used by the miner, so the miner cannot just take this transaction and try to extract value from it. But instead we have this layer in-between, which we call "solvers", and the solvers, they try to execute the transactions or the orders on behalf of the user. But the big difference is that they are bonded and they are governed. So the idea is they need to give us a promise that they will execute the orders in some sense of fairness for the user or in some form of favoring the user's interest. And if they don't do that, they get slashed. Or simply the users would stop sending the transactions to them. But high level, the idea is don't give the transaction to the miner, give it to someone who cares about your interest.

Anna (00:20:26):
What does that intermediate agent look like? Yous said it's built on the Gnosis protocol. So is it the Gnosis protocol that's doing that somehow?

Martin (00:20:33):
Yeah. So, I mean, high level it's, you can say, as you would broadcast your Ethereum transaction to a network, you can instead broadcast your order to the solver network. And the solvers, they can, like miners, pick up those transactions and try to include them, but they are restricted in some forms. So they cannot do arbitrary stuff. They, for example, there are some basic guarantees of fairness, that we can guarantee through the contract. So for example, if there are two trades on the same token, in the same block on Ethereum today, very likely those two traders would get a different price. In the Gnosis protocol it's guaranteed by the contract, the two trades trading on the same token in the same block needs to get the same price. So there are some levels, some requirements guaranteed. And then there are additional requirements that are just softly guaranteed. So the idea is, let's say there's Uniswap, and there's Balancer, and there are different places of liquidity, SushiSwap, then the soft promise is you should execute it wherever the trader gets the best price. And that is something we cannot fully enforce by the protocol, because that would increase gas costs enormously, but that is something we can enforce on the governance layer. So if a solver consistently doesn't act in the best interest of the users, they can get slashed and that then would be done by a governance layer.

Anna (00:22:02):
In this model, is there an AMM? Kind of going back to that topic. I'm trying to map what you're saying, and then also bring it back to that.

Martin (00:22:11):
So there is the whole current Ethereum DEX space with Uniswap, with Balancer, with SushiSwap and all those AMMs, currently, Ethereum trading is dominated by AMMs, but the solver is sort of a layer in-between the AMM and the mining pools. So the solvers decide how to ideally execute a bunch of transactions against AMMs. And this is also where CowSwap comes from, CoW is coincidence of wants. So let's say you have two traders, one selling token A and the other buying token A, then if they would execute today their transactions directly on Ethereum, the first transaction to hit the automated market maker would get a specific price. Then probably some sandwiching would happen, but let's put that aside. But then the second transaction would hit the automated market maker and get, again, a different price. And they would both pay fees. In our model, they are executed in a batch, or basically each block is one batch, and we try to execute all trades happening in one block, or in one batch, at the same time. So those two trades can already be matched against each other and they will probably not be exactly matchable, so they will probably not exactly have the same size, but so only the remaining part then needs be traded against the automated market maker. So we saved the user already fees from the automated market maker and we give those pairs guarantees that they get the same price.

Anna (00:23:58):
Can you say the acronym again for CoW?

Martin (00:24:01):
Coincidence of wants.

Anna (00:24:05):
Coincidence of wants. Okay.

Martin (00:24:05):
So two people, by coincidence, someone wants to trade in that direction and someone wants to trade in that direction.

Anna (00:24:15):
Question about the naming. Did it have anything to do with there being a cow emoji?

Martin (00:24:21):
No, for sure. I mean, with the design, obviously we play with this cow meme and people seem to like that.

Tarun (00:24:29):
I think, one thing that's important to remember is there has been no decentralized exchange that has been successful without actually being an emoji already. I guess, other than Balancer, I don't really know what the emoji is for Balancer...

Anna (00:24:44):
Could have the balance, isn't there a balance?

Tarun (00:24:46):
Yeah. I think you're right, there is. But if you think about UniSwap and SushiSwap, the single emoji meme was one of the driving characteristics of both of those launches.

Anna (00:24:59):
So in this case though, CowSwap, you just described it, it sits between the user and the AMM. Is it taking a fee of its own?

Martin (00:25:06):
Yeah. So it is, and one part is extremely natural, because on CowSwap, as a user, you don't pay the Ethereum gas fee. Usually, if you trade on other market makers, you pay the simple transaction fee. And here you just signed this order that basically gives the solvers the allowance to take your tokens and give you other tokens. But eventually the solver will need to execute the transaction and therefore pay the gas fee. So how we do that is that part of the order is a fee that you then can already pay with the token, either you're selling or you want to buy, and that would go to the protocol.

Anna (00:25:51):
Would you call this? Is this a dApp that lives on top of other dApps?

Martin (00:25:55):
So we see it, to some extent, as a meta-DEX, probably, middleware. CowSwap itself, of course, is a product, but ideally it will not be the only product. And actually we are doing this together with Balancer, we are using their new vault contracts for some gas optimizations. To some extent, you can use internal Balancers or save some gas on ERC-20 token transfers, specifically if you do dispatch settlement and you settle a bunch of trades. But long story short, also the Balancer interface with soon route their trades, or trades coming from that interface should be routed through this Balancer-Gnosis protocol, where again, then on the middle layer, the solvers collect all those orders and try to figure out what's the optimal way to take all this trading volume and distribute it and execute it at fair prices against all the on-chain liquidity, again, coming from Balancer, coming from SushiSwap, Uniswap and so on.

Anna (00:27:07):
Interesting. So in this case, what you described as the benefits to using this, it's reduced MEV and reduced cost. Do you see this project then also coming into competition with the L2s? It's not an L2, but is it solving the same similar problem to what they're solving?

Martin (00:27:26):
Yeah, it is, to some extent. I mean, of course there can be DEXes [that are] completely separate on L2s, but as soon as they want to tap into the L1 liquidity, it can actually be very useful, because we have this concept of batch settlements.So we have a list of, I dont know, 15 trades that might happen on a higher level, and then they are just settled in one larger transaction against the L1 on-chain liquidity.

Anna (00:27:56):
Cool.

Tarun (00:27:58):
Yeah. How did you end up deciding on the Balancer vault and Balancer, working with them versus integrating your existing setup? And then also a little bit about, maybe a slight deviation, but not a super deviation, is talking about Gnosis Auction and the version of world, because I think the dual to the CowSwap is these OTC large-size transactions, allowing DAOs to sell their treasuries more efficiently. And I think it's interesting to understand how you thought about building both of those projects at the same time, because one is very left-brain, one is very right-brain, in some weird way.

Martin (00:28:40):
Right. So maybe first answering the partnership with Balancer. I would say those things are perfectly complimentary. Balancer is developing in a direction. They are trying to address the problems of automated market makers in a different way. So there's this problem of capital inefficiency and UniSwap is solving it currently, or trying to solve it with UniSwap V3, by something, I would say, that brings it closer back to an order book mechanism, in a very elegant way. But Balancer is, I would say, going into their direction of still trying to keep this very retail friendliness, so that someone who has not too much idea about market making can still put their capital in there and hopefully get a good return. And the ways how they achieve that and how they improve the market maker experience or the AMM experience's dynamic fees, so that someone else is being able to set the fees for those who provide the liquidity, and this concept of managed pools or pool manager and asset manager that the assets that sit in the market maker, or in the AMM, can still be used somewhere else. So I think the bottom line is Uniswap potentially wants to even reduce TVL and say, "Okay, with less TVL we still can have a good amount of trading volume", while Balancer really tries to say, "Okay, what can we do with as much capital as possible and how can we use it as good as possible?" And therefore, it makes a lot of sense to say, "Well, we built this layer on top of Balancer. We're trading, to some extent, can already be netted or can already be matched against each other, but then we can match it against this large pool". Although I have to say the protocol is slightly preferring Balancer, or there are slight gas optimizations for Balancer, but it's possible to settle against any on-chain liquidity. So it's not just bound to Balancer.

Tarun (00:30:55):
If we think about the CowSwap, it's really focused on small size traders, because if you're trading large enough size, MEV probably doesn't matter that much to you, to be honest. I mean, it does a little bit, but it's not.

Martin (00:31:07):
I mean, why wouldn't it?

Tarun (00:31:09):
Well, it's not that it doesn't matter. It's just that, I feel like people are computing this as percentage of trading size. And there are certainly times in duress, where it will be large, when you're large size, but when the chain is in the median gas price, it's on.

Martin (00:31:28):
Yeah. I mean, from the meme and so on, we definitely targeted, with CowSwap in the beginning, small traders, but I would actually argue as a small trader, the only thing you should right now care about is gas costs. Because I mean it's insane that sometimes people would pay 10% or even more of the traded value in gas cost. Then you really don't have to care about MEV, because there we are talking about, maybe it makes 1%. So if you're wasting 10%, you don't need to care about 1%, but for large trades, this 1%, that can be a lot. And there the gas costs then play less of a role. So I would definitely say, and we have seen over a $1,000,000+ trades on CowSwap, and we definitely want to see more of those. So we are still targeting that market.

Tarun (00:32:20):
Yeah, for sure. I wasn't necessarily saying that it's impossible. I guess, the notion of what large and small changes over time too. I was just thinking about Gnosis auction that's more interesting mechanism for DAOs that have to sell very large size potentially. And so I guess at least to me, they seemed like opposite, product servicing the opposite sides of the market, even though small might not mean your mom buying whatever the new dog coin is. So maybe actually it would be great if you talked about the Gnosis Auction, describe what the genesis is, what it is, how you guys got there.

Martin (00:32:59):
Right. So Gnosis Auction is a fairly straightforward auction protocol where anyone can at any time set up auction and simply define a bunch of tokens they want to sell. And of course they need to specify for what they want to sell it. And then over a specified period of time, could be days could be 15 minutes, bidders can submit their bids at different prices. And once the auction clears, or basically once the time ends, it's actually possible to do on chain the calculation of this one single clearing price. So you have all the bids and it will calculate the exact bid that is required to clear, basically to sell all the tokens, and it will be then executed at this price. And the one thing where that became most used is IDOs. So the idea of your completely new project, the token is not yet traded on the market, on any market, and you want to have an initial price finding mechanism or trading mechanism for the token. And this auction mechanism is a straightforward way to do it. And it combats a lot of the problems, the MEV problem gets much worse if you're still in price finding mode. So the price might, within minutes, go up by 10 or 5x from starting prices, then the MEV problem on automated market makers is really getting severe. So that is Gnosis Auction and it can definitely also be used for larger trades in general. So we have seen this use case more and more, the DAOs sit on million dollar treasuries, and as DAO, they what to sell, let's say 1 million worth of 5 million worth of Ether, let's say, for stablecoin. And it's not so easy for them to do that. Of course, they could try to trade on, let's say UniSwap, but that would first cause more slippage than necessary. And also, again, you have this problem of MEV and specifically you have the problem that you need to initiate the trade, probably with a long... There needs to be a proposal and then there needs to be voting and eventually the transaction is executed. So at the time you initiated, you don't know what the price would be, so it's hard to do it. So here Gnosis Auction has become an option. And the way we are thinking about it in relationship to Gnosis protocol is that eventually we are going to merge those things. So how that would look like is that you can optionally have this auction and have the bidding going on, but the final clearing, or once the time's over, then you still have the solver that will do the remaining settlement against on-chain liquidity. And it's required to serve all the bids first. But if there's something remaining or there's a better price right now, available on chain, then some of the bids, they can use this settlement against on-chain liquidity.

Anna (00:36:21):
So are those two things being used in tandem? Or is it more like theoretically they could be used?

Martin (00:36:26):
So right now they are still separate.

Anna (00:36:28):
Okay. So this is an idea then.

Tarun (00:36:30):
Yeah. Well, I think a natural transition is basically talking about how we've seen some DAOs use Gnosis Auction. And how does that interact with a lot of the governance tools you guys are working on? I think, especially in the vacuum left by Aragon, it seems like there's a two-horse race of very different philosophies of Compound governance, forks and Gnosis. And so it'd be interesting to think about how you guys have the suite of products that, in theory, actually could all be tied together. So where do you see that going? And also how do you guys think through governance?

Martin (00:37:14):
Right. So the background of our governance activities was the very simple fact that we developed Gnosis Safe coming back from our own needs. So really a long, long time ago, we did an ICO and back in the days there wasn't even a multisig available that could hold tokens. So we developed one and it became the standard for teams to manage funds and out of that effort the Gnosis Safe developed. The Gnosis Safe is a simple smart contract, or a multisig wallet, that allows you to set permissions of who can access the funds, under which rules. Simplest rules are simple, let's say, 2 out of of N signatures that are required to move the funds, but you can set more and more complex rules, kind of say, "Under these conditions, or some out of this group and some out of this group and whatever", you can really try to define rules for who's allowed to do when, what transaction. And what we have seen is that more and more teams/DAOs are using it, or that that became basically the standard solution for teams to manage their funds, they put them in a safe. And the step we then have been doing, or as the second product that emerged, was this idea of SnapShot voting, or off-chain voting. So on-chain voting is simply very complicated and, frankly, very expensive. So an Ethereum transaction can easily cost $50-$100 and I mean, who would want to pay a hundred dollars just to submit the vote, even if, let's say, you're just one out of 10,000 token holders, you would certainly not pay a hundred dollars for your small vote. So what became really the standard was this concept of off-chain voting: you just sign a message with your wallet that contains a token. And then you can verify that whether you have the token and so on. But that unfortunately is all off chain and cannot be easily verified on chain. So that's the bridge we were trying to build with SafeSnap.

Anna (00:39:22):
So you just mentioned Snapshot, let's define what Snapshot is and then how that relates to SafeSnap.

Martin (00:39:28):
Yeah. So Snapshot is this governance tool that allows very easy for communities to say, "Okay, we have a token and you can, within minutes, spin up a simple voting system, where anyone can make a proposal and the token holders vote". But all of that is done off-chain. But it provides tools to read the balances on chain. And that's also where the name Snapshot comes from. So you define a specific block, where a snapshot is taken. So on this block, you say, "Okay, we had 5,000 token holders and token holder 1 had this balance and token holder 2 had this balance and those are your voting rights". And all of that logic, that's what SnapShot provides.

Anna (00:40:14):
Who put that together? Is that a product that you built or is that someone else's?

Martin (00:40:18):
No, that is set as something that came out of Balancer. So Balancer use that for their own governance, but at some point decided, "Okay, this is a useful tool in general". And then the specific teams spun out. And just in general, providing this tooling. And I think currently definitely the majority of DAOs, probably a few hundred may be active, maybe like 50, but it's heavily used by many DAOs.

Anna (00:40:47):
This is still off-chain voting though. So this is, you say it's a Snapshot that gives you the value of, I guess, each user's voting rights, but then where are they actually voting? If it's not on chain, couldn't you just sort of fake it?

Martin (00:41:04):
Yes and no. So when you vote, what you're doing is you are actually using your account to provide signature and therefore that is not fakable. So that can be verified. There are problems around censorship. So it's not clear, you need to trust their server that your vote is actually... They could censor votes basically.

Anna (00:41:33):
I see. So is that a bit of a weak point of Snapshot?

Martin (00:41:36):
Oh, for sure. And that's definitely something they want to address, there are various solutions to that. The easiest one is to take your favorite sidechain or L2 or whatever, and publish the votes there.

Anna (00:41:52):
So let's move on to SafeSnap. This is the project that interacts. Is this also something that lives on top of Snapshot?

Martin (00:42:00):
Exactly. So previously the status was very often... It's a very, very common setup for DAOs effectively have this Snapshot voting where decisions are made, but then still just having multisig. So, I don't know, 5 members from the community effectively control the funds and control sometimes even just rights in the protocol. So rights to upgrade, rights to set fees, rights to make changes. So the idea is, the voting is happening, and then you hope that the multi-stakeholders, they promise that they will do what the voting says. Of course, that's not necessarily what we want. Of course, ideally we would want to make the voting directly enforceable. And that is what SafeSnap tries to do. So what we are saying is, "Okay, you have your safe, anyhow, that has this right, or that might have the funds, or might have the right to do some change". And we built this tool that really tries to connect those two things tightly. So the first thing that is required is that when you write a proposal, you not just write a text saying, "Well, we should set the fee to something", but in addition, you need to provide extra byte code or a specific Ethereum transaction. So you say "that's a proposal" and basically that is what we want to see. And this transaction could be changing a fee parameter, that could be making a payout out of the treasury to someone, whatever, it can be anything, it can be setting an ENS record to another content hash. So that's the first step to be more specific, to be machine readable, to make the machine-readable proposal. The second step is after the voting has happened, then you somehow need to have a mechanism to decide or to let the Ethereum blockchain know that this voting was successful. And there we are using an oracle solution. And that's another project we are building on top, which is called Reality.eth. And the very simple concept here is that it is a staking game. So someone is making the claim saying "Well, that vote happened. And it happened in favor of the proposal, that proposal was accepted, and this vote had this specific payload". And then you're basically making this claim and you're staking a small amount of, let's say, Ether behind it. Let's say, half an Ether. And if no one challenges that, fine, then after a timeout period of, let's say, 24 hours, it passes and the proposal can be executed. If someone is challenging it, then it has escalation game, where you need to double the amount and you could challenge against it. And then again, if it runs out of time, 24 hours, then it's rejected. And actually the one who first did the potentially malicious claim, they will lose their Ether. And you have staking game. There are a few parameters how you can set up the staking game, one is just using Ether and relying on the idea that the majority of Ether holders might be honest. If you don't want to do that, you can also use the token for the staking game. And then you basically say, "Well, we anyhow assume that the majority of token holders can do whatever they want". So then we can also use that security in the staking game.

Anna (00:45:38):
I see now where the link is to DAOs, how it provides a structure to execute those choices that are made. I want to go back to Snapshot. If it's happening off chain, is it just after the vote would happen, a new smart contract would be deployed to do it? Or something would be triggered in a smart contract by some central entity? I'm trying to figure out this comparison a little bit more clearly.

Martin (00:46:07):
Yeah. So usually the idea is that there is already a safe or already a contract that holds, let's say whatever the DAO represents, so that that contract can have some power in other contracts or hold some assets. So that's in a safe and that already exists. And then the proposal is just a transaction that is being executed from that safe.

Anna (00:46:31):
But going back to what it was without SafeSnap, that's what I'm trying to compare.

Martin (00:46:35):
Well, it was exactly the same in that sense. The only difference was, without SafeSnap you just had to rely on the safe owners on the 3 out of 4 signers to actually do this proposal. With SafeSnap you don't need to have those 3-4 authorities. Actually, there's a middle ground that we even currently recommend. The middle ground would be to still have those signers as a backstop, so it can be executed completely automatically through the SafeSnap mechanism, but they could still have some veto rights, but eventually you can get rid of them. And then you have the safe that doesn't have any owners, or the only owner is this complicated governance mechanism with the oracle that says, "Well, that was voted on."

Anna (00:47:22):
But the SafeSnap does not help in the creation of the governance proposals in any way, does it?

Martin (00:47:28):
Right. So the whole process of how you create the proposal, I mean, to some extent, we have a small plug-in that allows you to define an Ethereum transaction, so target and the payload. But the whole process of how you make the proposal and how you come to agreement about the proposal, that is in principle agnostic. So we have built it around, as a SnapShot tool, but actually it's already being used for other mechanisms. So there are, for example, DAOs on xDai, let's say, Colony and other frameworks. So you could also say, the decision is happening there, through that framework, or whatever rules you want to have, but eventually it's executed the same way through this Oracle mechanism that then would not ask, did the SnapShot vote happen, but did this DAO decision happened on xDai or on Polygon or wherever, and it's basically just the way to execute a decision that was somewhere objectively made somewhere else.

Anna (00:48:35):
When I first read about this, I thought for a moment, it was related to what Gauntlet is building, what Tarun is working on, coming up with these governance proposals. But, Tarun, to you, for your tool, does this have to be on-chain governance to be able to do it? Or are you also helping to determine proposals for off-chain or for something like Snapshot?

Tarun (00:48:55):
Yeah. So actually we definitely work with protocols that are using things like Snapshot for adjusting parameters. I think the interesting thing that we're starting to see is really that there's this fast path and slow path in governance of like really big changes will take a whole vote, like executing the byte code vote. And then really small changes might not, up to some bounds, be allowed to be voted on by a smaller subset or a sub-group of people. But I basically think that either way, regardless of whether you do slow governance or fast governance, there's still this sense in which you both need, you need some sense of what recommendations look like, as well as a way for people to really have tools for generating proposals. Because at the end of the day, one of the reasons SafeSnap has a huge advantage over, say, Compound governance from a UX standpoint, this is way easier to make a proposal. The proposed function in Compound actually really does require you to understand what byte code you're sending, quite intimately, whereas I do think in the Snapshot type of thing, you can deal with slightly more qualitative votes.

Martin (00:50:18):
Yeah. I mean, that's an ongoing discussion for both the Safe, in general, even if in a simple multisig, if you want to do more complex or many different actions, for example, that require 3-4 transactions. And if you do that with your hardware wallet, that's not too bad. But if that always requires a round of people in different time zones and they need to sign, that gets complicated. So you want to have a way to bundle or have something that's meaningful on a human-understandable, on the proposal level, and not on the bytecode level, and try to define that fairly easy. And one thing we are working on, as a Safe, is that you will just be able to simulate transactions or use normal depths in a simulated environment and say, "I want to do this. And then I want to do that." And basically then all those transactions are recorded and out of those, a proposal or the byte code to execute all those steps is created.

Anna (00:51:26):
You mentioned a few projects that are using Snapshot and then some are not using it, and they're doing more like this on chain, although I guess it's on chain, on their own dApp chain. I'm trying to picture what the alternative is to some of those examples. Maybe you can give actual project examples, because I'm not as familiar with the landscape of DeFi DAOs, as both of you are.

Martin (00:51:48):
I mean, for example, there's the Compound governance framework. Ans also Maker. So there are some governance frameworks where voting and the execution is happening on chain. First of all, it's great because that really is the ultimate level of guarantees or decentralization. But of course it comes at a very, very high cost. So the participation, just the voting, again, could easily cost a 50 to a 100 dollars. And that's why more and more projects are using at least hybrid approaches or this complete off-chain voting.

Anna (00:52:23):
Got it. So where is SafeSnap at? Is this already rolled out? At what stage is it?

Martin (00:52:27):
Right. So it is rolled out and we have a few DAOs using it. It's still clearly in the space where somewhat still complicated to use, as Tarun hinted, at creating this byte code for the proposal. Right now, the tooling is not as good, that it would be easy for everyone to create a proposal. So that is definitely the most pressing thing that needs to be improved.

Tarun (00:52:57):
I think the interesting thing that we've learned is 2017 had this whole meme, maybe not totally just a meme, but it happened like, "Hey, do tons of on-chain governance, maybe do futarchy", and futarchy was basically a prediction market on governance votes. And we had this very amorphous like, "Hey, here's some idea, but here's no technical spec for it or implementation details, so that's an exercise to the reader". And 3 or 4 years later, we actually have, instead of having kind of "Hey, there's a one size fits all Slayer governance model.", we actually have this hierarchical model of different fits for different types of needs. And I think that inevitably we're going to have this like Hydra of different options for different people. But I think it's kinda cool that we've evolved from trying to have this theory of everything for governance to like, "Hey, well, maybe just use the right tool for your current scenario". And I think DeFi forced that into this space. I think layer ones are inherently conservative and won't do big changes. And I think to some extent that's why things like governance and like Tezos has had extremely few changes. It's A) extremely hard to correctly do, B) it's hard to convince anyone to even vote, because there's a high fraction of non-active asset holders. And C) there are day-to-day changes you do wanna make frequently. And I think only in DeFi is the clear monetary incentive to changing things, where people are willing to go and build this infrastructure. And at least, I think, it's interesting to see Compound governance and the SafeSnap version of the world as two non-competing things, because they both serve a very clear purpose for different use cases. And I think that's the thing I'm always left struck by in non-Ethereum layer ones. It's that somehow there's just too many things left to governance and that just is really hard to do. I don't know. I mean, Anna, you've dealt with a lot of these hard forks in other layer ones. Think about how hard it is.

Anna (00:55:24):
I've dealt with the hard forks?

Tarun (00:55:26):
Well, I mean also trust ceremonies. Upgrades, sorry.

Anna (00:55:31):
Yeah. Actually, recently with the Zero Knowledge Validator, we went for a hub proposal on the Cosmos hub and it was fascinating. That's on-chain governance, it's voting by the people and validators. It was amazing because this is a proposal for events. We're doing 4 events with the ZKV on the Cosmos ecosystem. And to try to explain what an event costs to every single voter there was very challenging. And maybe that says something about marketing costs not going through a hub or a funding like that. But yes, I think my experience so far with on-chain governance is the systems aren't quite in place to make that efficient and to make most efficient use of those funds, because not everyone who's voting has domain expertise. And I think that's maybe where DeFi wins, because it's a very narrow level of expertise needed, in order to do that type of governance. You're not talking about funding the marketing, you're not talking about the larger funding vision. You're talking about one very narrow thing, where there can be a lot of people with that expertise.

Martin (00:56:45):
But even in DeFi... So probably the most complex, Safe, whatever set up is from the Yearn team already there, they are working on a setup with seven different saves basically that would have different authorities about different things. So already there, you will then have this separation of if it's a proposal of the concerning funding, external efforts, it would go through different pipelines.

Tarun (00:57:13):
I think the main thing to understand is that to your point, very scoped, easy to understand and scoped in terms of both the actual thing that's being voted on and the framework to understand why anyone should care, is very important to governance. And I think that's where the layer one like, "Hey, we're going to upgrade the whole network and force everyone to change all their node code. And yeah, we want to do it next week" attitude is very hard to reconcile versus like, "Hey, we want to change the interest rate. Here's why the current thing is wrong" is much more palatable. And I think it's kind of interesting to see that in 2017 we have all these people being like "Ethereum governance sucks", which on the All Core Dev side, you could argue it does, but it's kind of interesting to see the Flippening of the actual governance innovation seems to be in ETH land versus off ETH, which is, if you were in 2017, you would have been saying the opposite thing if you watched ICO's. Everyone was like, "We're going to build better governance into our thing". And I don't think there's a good example, outside of DeFi, of governance actually being effective in a decentralized manner, which is funny. It's just hard to have predicted that 4 years ago, when Gnosis ICOed, it was a different universe, right? Everyone was just huge vision, but no technical details. And 4 years later we aree like, "Okay, these are the things that work. And these are the things that don't", and it's cool to see that it's not necessarily one size fits all, but we found a nice equilibrium.

Anna (00:58:50):
The PleasrDAO that we talked about two episodes ago, is that using any of these things, by the way?

Tarun (00:58:55):
We're actually using Compound governance for the token. I'm officially, as a US citizen, not on the multisig. I just happened to accidentally some of the initial minted tokens. But we do use Gnosis for the multisig.

Anna (00:59:09):
Got it.

Tarun (00:59:11):
All of our bids are done by Gnosis Safe. It's just for minting we had a Compound governance contract, but it's definitely, I think, the idea of building these tools, where I think it's going to be way more useful, in spite of the fact that I'm the first person to tell you, I think DeFi is awesome, I love it. But I actually think it's going to be way more useful for these art DAO-type things, or a collective ownership of some data in some weird way, than it is for DeFi. But the problem is those people don't want to do any programming. So we have to get to the point that they can do that on their own, without needing to understand very low level details about the assets they hold and stuff. And so I think that's the end game, it's like if PleasrDAO didn't have a bunch of developers in it, I think it would have been much harder. But it's pretty easy because of that. And so I think who is the ideal consumer seems to be like people who want to have these collective ownership groups, but are not technical. And that's the next million users or whatever.

Anna (01:00:18):
But, I mean, that was what those original DAO projects were meant to do, right? They were supposed to create the UX of DAOs so that anyone could do it. Would you say that we're back to square one? We have to rebuild on top of the new learnings? Or can we use some of that stuff?

Tarun (01:00:33):
When you say "old DAOs" you mean Aragon? I honestly just think the lightweightness of Gnosis is actually the key. It's just less effort. Aragon, you have to learn their whole vernacular and be ingratiated into the cult, in some ways, to use it.

Martin (01:00:55):
Yeah. I would say, fundamentally it's just a very hard problem, so it will take more time. So the safe is the minimum form of governance, the simplest form is you have a multisig, I don't know, 5 key holders and if 3 of them agree, then you can do something in a way that's at least a starting point. And then the whole topic around, how you describe a transaction, how you make sure that the transaction does what it says to do, that is simply a hard problem we are trying to tackle. And then on top of that, you have the addition to the hard problem that we won't tackle, and that others hopefully will tackle, say, you have a thousand people and somehow they together need to come to an agreement and what is this process? But I think the really cool thing about Ethereum is the possibilities, as Tarun mentioned, like already today, Gnosis Safe can be combined with Snapshot, with Compound, actually also with Aragon. So Aragon DAOs could exist on a layer two and the decision that's being made there is then executed from a Safe on layer one. So there's almost infinite ways to combine things.

Anna (01:02:22):
So I want to bring it back to the point we made right at the beginning, this idea of "Gnosis is focused on focus". So we've heard now of the Gnosis Safe, is it the most used, the largest success from your stack of products?

Martin (01:02:40):
Yeah, probably. It currently holds almost 30 billion worth of assets.

Anna (01:02:46):
Wild. It's done well. So from there you're building sort of the DAO front. I don't know, does the CowSwap still relate to that part of the stack as well?

Martin (01:02:58):
Yes and no. So I would say the two big focuses for Gnosis are currently the wallet, Gnosis Safe and Gnosis protocol exchange mechanisms. There are of course connections. So ideally if you have your funds in the Gnosis DAO, or specifically teams, or also individually you want to trade, we already have the tool. And for safe app, we have this concept of apps within the safe, you can start the Gnosis auction. So there is some overlap, but I would still say those two projects are somewhat separate.

Anna (01:03:33):
Cool. So I want to say thank you so much for coming on the show and sharing with us this update about Gnosis and about these exciting new developments that have happened in the last 2 years.

Martin (01:03:46):
Yeah. Thanks for having me. It was a pleasure to be here.

Anna (01:03:49):
Cool. Thanks again, Tarun.

Tarun (01:03:50):
Thanks.

Anna (01:03:50):
And I want to say thank you to the podcast producer, Andrey, to the podcast editor Henrik, and to our listeners.

Anna (01:03:59):
Thanks for listening.