Making DAOs Actually Decentralized and Autonomous - John Shutt | UMA

thanks is this on hey everybody uh I'm John I'm with Uma uh we're a decentralization Maxis so we were trying to use our optimistic Oracle to solve a really big problem the Dallas have which is that they're not particularly decentralized there's kind of a missing step there where you wind up just having a multi-stick do everything at the end of a whole lot of voting on Snapshot so this is live and deployed uh I think the first thing to explain is what Uma is actually doing hey okay so uma's an optimistic Oracle so when a lot of people think about oracles they're just thinking about price feeds and delivering data to D5 which is what chain link and some other oracles do what Uma does is it allows you to assert any kind of Truth on chain and allows anybody to dispute it so instead of having a bunch of nodes try to verify some specific piece of data in advance you allow anybody who knows some facts about the world to make that assertion put some money behind it if nobody disputes it within the challenge window it can be accepted and you can use that data and information to trigger some smart contract actions so we describe it as a decentralized truth machine and we've kind of geared our latest contracts to really fall around that pattern of asserting a truth which can be natural language by the way so it could be like a sentence like you know the weather is not going to go above 80 degrees and like Belgrade today is something that you could assert so it makes it really flexible and open-ended and we'll talk about a couple of the use cases in addition to governance which is the main focus here talks about that already one thing that we didn't talk about is you actually have recourse for bad data so sometimes you'll have some bad data like a manipulated Market that price data arguably is bad you could say that the Oracle works correctly but if it's clearly somebody that's just manipulating one big Market to change the price of some token to attack some protocol you could say that's not the real price and that's something that you can actually dispute then within Uma that you can't really do if it's like an API based thing where it's like well this is the only source of data and that Source says this so therefore that's the price and it allows you to be a little bit more oh I'm not even clicking through sorry about that yeah this one this one yeah we're here okay recourse the contracts I was clicking through on my computer and forgot that it wasn't connected so I'm talking about this last thing so adding recourse to contracts if some data is wrong in some unanticipated way uh maybe like the maybe binance's API got hacked and it's delivering that data now you know a bunch of nodes technically verifying that data from the API can't necessarily catch that but you could say like this is clearly hacked this is clearly wrong none of the other exchanges agree with that Etc Uma lets you dispute things for basically any reason but you also have to put money behind your disabuse you can't just arbitrarily dispute stuff you have to dispute when you expect to actually win so what goes into it we talked about requesting some data proposing some data disputers that are watching it and all of our proposals go to like one interface and a lot of people have Bots set up too because there's a monetary incentive to dispute you get like a big check that the proposer's bond for disputing so we do see these disputes pretty regularly with poly Market when there's some ambiguous wording about an outcome and you have to have some kind of methodology of the truth if you have some kind of complicated data that you need like the across Bridge uses Zuma under the hood and it's about like verifying a bundle of relay or refund data from a whole bunch of different chains and making sure that all the transactions are included and it's highly Technical and that's very bot driven and economic guarantees that bound the cost of manipulating the system which I'm happy to talk more about that after this talk if anybody's interested but you basically want to have some kind of um either use uma's DVM like all the Numa token holders coming in to vote to get the final answer which is what poly Market does but if you were the uniswap treasury or something you might want to off-ramp and go to your on-chain governance if you're using osnap to control your treasury just because the unit swap treasury is so much larger than uma's system you want to make sure that you don't have an economic incentive to try to attack uma's voting system so these are some examples these are all like live examples uh poly Market I talked about prediction markets are an obvious use case you don't know in advance What markets you want to create you want to have maximum flexibility insurance is another obvious one so Sherlock and cozy use us and they do a lot of things like was this protocol hacked which is not something you can technically verify but you can basically know so we did some payouts around the Euler hack for both of those protocols recently actually across I mentioned extremely technical driven it's all Bots with that like humans are not manually proposing or disputing things within a cross it's Bots that are watching all these different chains and seeing what's going on proposing and disputing success tokens this is kind of an interesting one so you can do uh Define native VC fundraising on chain using Uma where you basically lock up your native token in the shapeshift case it was the fox token and you sell it to investors and it basically has an embedded call option and if the fox token price is above some level at the expiry date in 2024 those VCS are getting the payout of that embedded call option if it doesn't go past that level you just get like one to one with the native token and you just had this long lockup period and might have bought it at a bad price and uh then this is a kpi option with chidao where because Uma is extremely flexible and it can verify truths in a natural language way you can describe really custom kpis and use that to incentivize your users instead of just saying here's an airdrop here's a bunch of tokens I hope you want the same things I want you could have your users only get paid out if the thing that you want actually happens and you can see that success tokens are basically the same idea but around price all right so let's talk about governance though so this is a big problem we have Dows that run on form discussions and then snapshot votes and that's great and you might have hundreds of people participating but then it just kind of drops off and you just have a multi-sig that controls everything anyway and it kind of makes the first two things a little bit fake it's also really inefficient like it's not fun to be a multi-six signer so if you're on a multi-sig you're constantly being asked to verify things you're probably just like pushing it through like oh I trust this person that transaction is probably fine massive legal and Regulatory liabilities being a multi-sig signer especially as a project gets bigger so it's inefficient it's insecure it's not decentralized Uma can help so we basically built a tool called the optimistic Governor that allows you to attach some rules to a safe and transactions that follow those rules can go through transactions that don't follow those rules get disputed and then we made it much more narrow and focused on the problem of relaying snapshot votes on chain and executing them so right now once you add osnap your Dow is now actually controlling it like it's not the multi-sig that has to execute everything anyone can initiate some transactions that were approved on Snapshot anyone can dispute that too like you can just look on Snapshot see if it was approved see if the transactions all match and the truth is economically incentivized and it's really easy to set up we tried to make it really beginner friendly it's all UI based you can deploy this it says in an afternoon really like that's with reading all the docs and trying to understand how Uma works and everything if you're just clicking through the buttons it's like 10 minutes uh so it's gotten a lot of adoption and defy in particular um my theory for that is just uh that D5 protocols do tend to be decentralization Maxis and they know that they have this problem and they just feel deep in their heart that they need to change it uh plus maybe like the regulatory thing of financial Protocols are a little bit more hazardous to be a multi-sig designer but really it should also work really well for nft projects where you can use your nfts for voting on Snapshot projects that have a really wide voter base that don't want to pay for gas I think I was reading an article that uniswap governance has cost over three million dollars in gas so far and it also became very concentrated for the same reason because people don't want to pay gas but they want to participate in governance they delegate and so now you've got concentrated voting power and it's still costing a ton of money just for those delegates and like handful of really big voters to vote on chain so osnap not only makes snapshot execute like the combination of snapshot and osnap means more people can participate it's a lot cheaper and it makes it more decentralized especially if you're doing ethereum mainnet governance where gas is super expensive so this is the flow I mean it's like a pretty simple flow um the big part is just like understanding uh you know the step of disputing and voting and so on like uh you can pretty easily verify whether a snapshot vote passed or not whether the transactions match and actually if there is a dispute it's thrown away you can repurpose you don't have to wait for the Uma voting system you can push it through if somebody disputed by accident or something and that's an area that we're iterating on too because we want to maybe have some like off-ramps where after a certain number of disputes you go to on-chain voting as a fallback or something like that but this is where it stands today and we're just trying to go where people actually are so a lot of dows use safes already a lot of dials use Snapshot oh snap just fits in between it's just like a plug-in thing that you add and a small module that you deploy and attach to your safe and then you have uma's Network verifying all of this and by the way since we have a lot of people watching the network for poly market and for a cross and for all these other systems that are really highly active that helps with security and monitoring you get a lot of security and monitoring out of the box a lot of these proposals have long challenge Windows like a day or more poly Market challenge windows are about two hours and usually a dispute comes in within like 30 minutes if there's something obviously wrong and that's just mostly somebody spotting it and just like thinking about it double checking the wording and everything so disputes come really fast so your treasury is not only protected by your dial watching or your multi-six signers watching you have the entire ethereum ecosystem potentially watching and trying to earn some money by blocking bad proposals so this is getting set up um I think that are we able to share the slides at the end there's like links and stuff I mean we have like the docs on this like it's a pretty easy set of steps I'll walk through a couple of them that are maybe more interesting uh you can set a custom Bond so you could say like oh this is uh too wet to propose something or this is 100 of our protocols native token to propose something and you can also set a custom challenge period That's how long you have when a proposal is made to actually dispute it so shorter challenge period means you can execute faster longer challenge period means more time to review so the more conservative approach is a fairly long challenge period especially because a lot of these Protocols are using this to secure their native treasury which would be a really big problem if that got hacked you just attach it to your snapshot space with the safe snap plugin which some of you might be familiar with it's a very similar principle a big difference is what's actually executing it instead of using the reality module you're using the osnap module run with Uma and a big advantage of that is just this monitoring Network you have a lot of validators watching Uma proposals which you don't really get with reality you have to bring your own monitoring with reality and you can contact us then and we'll tweet about it and do Twitter spaces with you and help to promote your integration because we're excited uh so what it looks like just literally circling it you go to the zodiac app in your safe apps so zodiac lets you add all of these additional little modular things to your safe the osnap module is fairly easy to find and then here are some rules actually so you can read through the rules these rules are posted as a string on chain so when a proposal is made to the Oracle you can just go and like read these natural language rules as a human being and see whether or not the transactions follow those rules you know just go over to the snapshot space check the transactions check the transactions on chain we're working to actually show the transaction details in the Oracle interface just make it super super easy we're going to have some Bots too that can automatically dispute obviously wrong ones like if you had a proposal that doesn't link to any set of transactions at all or if the transactions clearly don't match like that's an easy bot dispute that would automate the process and unfortunately out-compete human validators for the proposer's bond so you could run that bot and try to run it more frequently than us and catch the bonds uh just a bunch of extra stuff um it's all audited we get open Zeppelin to audit all our stuff um you can set it up right now on ethereum optimism arbitrum polygon and Gurley but if you're on some other evm compatible chain and you have a dial that wants to use this it's not that hard to deploy it there we can deploy the Uma stack and we can talk with the Zodiac folks and deploy that stack too and safe is probably already there it's really beginner friendly too so you don't really need a lot of technical expertise you can set up a safe and snapshot and osnap without doing any coding at all and you also don't need to know how to code to validate that these things are correct like these are things that you sort of like eyeball you can compare you can like you know do like a diff on like the transaction bundle or something you don't need to write your own Scripts and it lets anybody participate you know human readable rules make it more inclusive you're not like looking at this inscrutable multi-sig transaction to try to understand what it's doing you're just checking if it passed on Snapshot all right so then moving beyond voting systems and multi-cigs so this is like a really easy set of rules to enforce but you could add other rules like this address is allowed to spend 1 000 usdc a month uh because that's like the marketing person or something and you just tack that onto your custom rules and now you have like a self-service kind of payment thing you don't have to have a vote every time that happens and you also don't need to send it to another safe you don't need to give this person potentially multi-sig control a lot of reasons a lot of the people that use multi-cigs do it for efficiency that it's just not efficient to go back to a vote every time it's ridiculous so you just trust the multi-sig signers with everything this allows you to have customized limited trust where you can do this thing and then this address can do this thing and this address can do this thing you know within boundaries and like none of the addresses can do this thing without a vote and it lets you write custom rules that make sense for your Dow in a way that all of the members of your Dow can understand what the rules are you're not writing a set of smart contracts that hopefully add technical checks to enforce the rules of the Dow you just read what you wrote and anybody can interpret them and then this is a little bit outdated we are like hiring uh we hired a head of sales and marketing manager though but one thing that's missing on this we are looking for an Uma product manager and there's a lot of cool stuff that's out of the scope of this talk around llms that we're starting to do um which I know that like Ai and crypto is a meme but the way that Uma Works being a natural language based Oracle it actually fits really well so there's a lot of interesting things that you could do around using llms to interpret and answer questions I've been like back testing a lot of poly Market proposals to see if the llm gets it right with like web search enabled there's a lot of interesting stuff and I think that it's relevant to osnap2 like if you're writing these custom rules you could get the llm to check like is this rule going to be enforceable is there anything ambiguous about this it's going to be a really fun year on the product side so if you are a product person uh kind of experienced in this space or know somebody that would be interested in this let me know and we do have kpi options for referrals I forget if this slide gets into it or not you'll get like an on-chain kpi option where if this person sticks around for a while or makes it to like this stage in this stage you'll get a greater and greater payout so find us somebody who is uh good at this and we'll give you some tokens that'll do hopefully the max payout when they're there for a year yeah that's all my slides um I know that was a lot uh I guess we could go to questions what's up thanks for the mic first of all the most important question out there I mean like he's really Uma named after Uma Thurman or no okay great for our company off-site in Arizona last year that like our designer made these shirts okay cool actually um I've been advising some games you may know them I don't know if you worked with them like the Opium team you've had the project when you've been actually providing there's an oracle for a SpaceX launch and they've been oh yeah yeah so I mean so I looked at Tuma more in a ways like providing feeds for like real life options or something like that obviously now you're stressing more this Dao [Music] products so to say is this just a result of some kind of more Focus going in this way you see not too many applications say for the price feeds or like derivatives or something like this what's your view on the general so to say Market request for the Oracles in general how you describe that yeah I think that we're responding to the product Market fit that we're seeing around natural language stuff like prediction markets and insurance and then governance these are not things you can do with the price feed Oracle that easily it is funny internally we built we spent like a few months and build a Perpetual product and then we were like uh it's too we just weren't happy with it for various reasons like the ux was bad like we didn't like the way that like uh liquidations were happening with funding rates and we decided not to move forward with a Perpetual product and just like work on something that uh seemed like it could you know pick up more um I do think that the governance stuff is a really big problem and we kind of calculate internally how well we're doing based on total value secured and it's kind of a hack to increase that because Dow treasuries are so huge so if you're like optimizing for that number because that kind of indicates how much trust people are putting in your Oracle and it's really even greater than the amount stored in the safe because if your Dow treasury gets hacked and like all those native tokens get drained that's bad for you in a lot of other ways too like that could potentially kill your project and it is just like a huge weak point in the overall ecosystem like multi-cigs are not the way that these things should be done but you can still build options and there are people that are building options on top of Uma right now and we work with them so it's not something that the risk Labs team internally is doing but you can definitely still build like options and you know success tokens and kpi options have gotten more traction just because they're like weird exotic ones that are also hard to do with the price feed we've just kind of like avoided the price feed area because it's like a Battleground in there and just focus on stuff people can't do yeah cool thanks um just one question before uh I asked the bigger question so if I understood correctly when the there's a proposal and um that's when the like the Uma um dispute Community uh jumps in and verifies that the proposal text matches what uh the transaction that's going to be executed on chain is going to do right uh yes it can be verifying both um I mean it can also just check that you didn't just like sneak in a proposal with some transactions that don't actually match the transactions were approved because on Snapshot you're building the transactions like you're literally voting on that transaction bundle and really we would like for like that to get caught at the snapshot stage and that's kind of where llms could jump in because llms could interpret uh like combine the text and then give an opinion of whether the outcome of the transactions match what was in the text and give a warning to the snapshot voters and if the snapshot voters are still like oh but you know we like it anyway like this is fine then that's fine and that's none of uma's business um but yeah we could dispute based on those reasons if there's a clear mismatch the reason I'm asking is like the uh you know the tornado cash incident that happened a few weeks back yeah um that a malicious proposal you know was passed and the the the attacker was able to you know take all the governance tokens so um it's in in this case like it's up to the Uma dispute Network to find the issue with uh like the discrepancy between the the language and the like actual transactions being executed right yeah exactly and we do want to have like more like uh Bots that are able to do that like assisting human validators um but yeah human validators are kind of pouring through the transactions right now and comparing it to what the description is and making sure that things look okay um and there's a lot of ux improvements that we want to make within the snapshot interface and it's been kind of cool working with the snapshot team because we just have our own little pocket of snapshot right now where we're just like pushing our oh snap code and there's a lot we want to do to the transaction Builder to make it more transparent what's happening and then also uh you know using llms and AI to check that the transactions match the text or at least give an indication of there might be some ambiguity here and like you should take another look at this thank you thanks man um hi uh I'm I'm Tim I'm I'm actually in the stablecoin business so uh probably a quick technical question about um you know price fits of uh General Uma products um in terms of latency actually so uh the uh one of the concerns when speaking with oracles is about how uh how quick they can deliver data about this and from from centralized the centralized exchange is mostly centralized obviously so yeah I think that's that's the one whether optimistic kind of architecture allows to to to to to to get decent latency here yeah I think the architecture lets you be about as fast as it's theoretically possible to be the main thing though is around having a robust enough dispute or network that you feel confident that any bad proposals get caught because you could be there like every block like here's the new price and you have like uh walls in the mempool like we've got really good disputers they're going to catch this before it goes through something or maybe like one block or something like dispute by the next block or it's over that's theoretically as fast as you can go because you can set custom challenge Windows the big thing is uh are there enough incentives do you feel confident enough that disputers are online and watching and accurate and also are you going to catch those weird edge cases of like prices appeared to move in some way that looks fake and you need time to like analyze what that means and maybe trigger some alerts of like this Market looks like it's being manipulated or it looks like there's an API error here or you know some kind of dislocation so it's all about setting the challenge windows and bonds yeah and actually speaking of that I'm just thinking if there is any kind of incentive system for providing this kind of verification as fast as possible or it's like no matter if you do it say you know one hour to like say you have four hours to approve this and whatever time you take basically the output is the same I don't know maybe there is something to provide more incentivization for acting faster or it's not it's it's not viable no I think that it's possible um poly Market actually gives rewards for the person who proposes so there's an incentive to propose fast um but then there's a large Bond so there's a strong disincentive to propose incorrectly and most the polymarket questions are like who won this basketball game um which is usually very easy to answer baseball questions trip up proposers sometimes because there's like a double header and you'll have like a proposer that's like not familiar with baseball or like you had a double header and like one of them was rescheduled from some other thing and like somebody is like I'm gonna get that 10 bucks as a reward and then they like lose a 500 Bond so uh like that does work like people built Bots that they pre-load with answers to propose like right at the end of a game like once it's resolvable so you see a lot of competition around that um and I think that you could do a similar thing with disputers but it's not just about like verifying at some point you want to prove that you're always online you know like if somebody's trying to attack your protocol on Christmas Eve because that's like that makes the most sense to attack then people are like offline distracted our bot's gonna show up and dispute stuff that's bad are they going to like call people and wake them up if they need some more help like prove that they're online all the time and uh you know I it's really weird that like AI got really good in the past few months because I'm like saying stuff that sounds really weird to say like without it was just like oh we use AI for this but you could have like llms that are monitoring stuff that are proving that they're deliberating continuously post that on ipfs and be like look I like analyzed this and there's like five other AIS that are trained to act in a certain way and we're debating each other and here's the transcript of our debate for every single proposal that has ever happened you know um you know that's like a good way to prove that disputers are online uh so that is something that I think would be cool to build out I think that we are probably going to look at something like that internally because we think it makes the Oracle 100 times better to have like all of these proven online monitors for everything yeah cool thanks it could also be like um like a longer term fee model around Uma like if you want to pay for that monitoring somebody's got to pay for the Bots or something and it's like okay pay like ten dollars a month for monitoring or like more if you're doing like really complex like weird stuff that you need like custom monitors built or something [Music] thanks everybody