Rollups are decentralized, right? Right?? - Tyler Tarsi

foreign [Music] thank you [Music] foreign [Music] foreign foreign [Music] good evening everyone it's it's Saturday evening uh so if you're still here you're a trooper um I'll get right into it so I'm Tyler and this is about Roll-Ups decentralization and what I call the opportunity Frontier this talk is going to be semi-technical I'll briefly touch on some technical components but the stock is really for anybody interested in contributing to scaling ethereum and maintaining its most important properties so just a little bit of background on me I'm the co-founder and CTO of recursive at recursive we're building defy infrastructure and work heavily with rollups hence this talk um if you want to learn more about me or recursive you can find me afterwards or check out our website the QR code on the left is for me on the right is for recursive and I'm going to talk about something that we've done deep Dives on so uh it's one of the most important things for the crypto Community to get right over the next couple years and it's about the current state of Roll-Ups so given the title of this talk you probably came here because you care about scaling ethereum in safe and secure ways so let's ask the question are Roll-Ups actually decentralized and the answer is no uh Roll-Ups are just not decentralized today but let's take a step back break this down and talk about what really matters when we talk about the decentralization of roll-ups um so I'll get into why I decentralization decentralization matters for Roll-Ups but first we need to understand what the goals of Roll-Ups actually are um it differs by project of course but roughly speaking um ethereum is kind of expensive and kind of slow around a year ago I remember spending over a hundred dollars to send a single transaction on mainnet um so goal of roll-up is to make ethereum cheap and not so slow um of course that's easier said than done but broadly speaking there are two ways that you can do this the first way is to make ethereum itself more efficient and this is already happening this is something that the ethereum core devs are working on all the time um and the second way is to do as much as we can off of ethereum but to try to keep everything else the same so the first way it gets us improvements but the second way gets us orders of magnitude better efficiency both are really important but I'm going to be focusing on the second here and I want to dig into the second bullet so keeping everything else the same what does that really mean um it's easier it's easy to get faster and cheaper transactions but it's hard to do that while maintaining ethereum's other properties in particular it's trust assumptions so can we scale throughput while maintaining security and how do we measure what security actually is um user sovereignty do you have complete control over your funds uh in your accounts and then protection against malicious mbv um which we'll get into some of the details of later so given these goals what does decentralization actually mean for Roll-Ups well there's no need to bootstrap a set of validators um we already have ethereums typically decentralization means that we split up a task across thousands or tens of thousands of validators and trust that an honest majority of them are doing it correctly but the whole point of Roll-Ups is that we don't actually need to do this we already have ethereums so given that we have ethereum as validators let's set up a system that achieves those goals that we previously mentioned with as few as possible trust assumptions Beyond ethereum as validators so what does a system like this look like well Roll-Ups have evolved very quickly over the past couple years but generally speaking the architecture that we're looking at today is the following components first we have transaction data this is something that lives on ethereum so when you have a transaction that you want to include you want its data to live on ethereum otherwise you're making trust assumptions Beyond ethereum's validators um second we have execution once we have transaction data and we can see it on ethereum mainnet we need to run the computation that tells us the result of your transaction when this happens on ethereum this is what makes it slow and expensive so we want to take it off of ethereum and that raises the question how do we confirm that this was actually run correctly and the outputs are correct and so that's actually the job of the third component here which is the validating Bridge so after your transaction is executed um the validating Bridge confirms that the outputs are actually correct and there are multiple ways to do this the two being actively developed right now are optimistic fault proofs and validity proofs or ZK proofs and then finally the fourth component in this system is the sequencer and so we talked about transaction data if we're all sending our transaction data to ethereum mainnet it needs to be ordered somehow so the execution component knows how to sequentially execute our transactions um and we'll discuss this more in a little bit but this is actually a really powerful role in the roll-up system so this General architecture works for optimistic and ZK or ZK Roll-Ups and it gives us a scalable environment with minimal trust assumptions um so it's kind of funny uh decentralization typically means using thousands or tens of thousands of validators to run something but Roll-Ups don't need that they just need this system to work properly typically the trust assumption Beyond ethereum is validators is one event so um we only need to trust one honest party to do this execution or validation correctly um so decentralization is actually not the goal of Roll-Ups even though uh well decentralization is a means to obtain computational Integrity the nuances of the differences between computational integrity and decentralization aren't super important and beyond that well they're important but beyond the scope of this talk so if you're interested in chatting about that feel free to grab me afterwards so for now we'll just call it decentralization um so how close are we to the end state of Roll-Ups that architecture we just discussed well let's look at the data um this is pulled from l2beat which is a fantastic resource for tracking the security and risks of using l2s um here are the top five Roll-Ups by tvl all of them have components that limit the decentralization of their systems today so we see State validation uh which is those fault proofs or ZK proofs that prove to the validating bridge that everything is operating correctly um we also have upgradability so could the teams that are operating these Roll-Ups upgrade the smart contracts that are that are running them um third what happens if the sequencer stops working uh and what happens if the validator software fails so all of the Roll-Ups today fail on one of these components um obligatory comment that fuel V1 actually is operating with all of these systems in production um but they've actually deprecated their their V1 system so it's tvl is practically zero at this point but it's still a very impressive system and they're working on their V2 right now um so I'm talking about why computational Integrity why decentralization matters but I want to ground this in reality do users actually care so looking at this data we have ethereum tvl which is around 108 billion um Arboretum and optimism are the two largest Roll-Ups by tvl and they're around one to three billion in tvl uh so you know correlation doesn't always imply causation of course but uh users do Care Capital will only migrate to l2s when they are more secure than they are today um no one's going to trust their life savings on l2s in in the current state that they're in uh nor should they so uh as a quick aside um building state-of-the-art secure technology is really hard uh this is not a day to get any of these teams in fact the point of this talk is actually the opposite uh we should be really optimistic about the opportunities that this creates for Builders across the ecosystem um so roles today fall short on several properties that are critical to scaling ethereum sovereignty upgradability faults and ZK proofs and Mev safe design and users care about this so we should all see this as an exciting opportunity to contribute to the ecosystem there are plenty of areas where new teams and Founders can help build solutions that improves state-of-the-art trustless compute um so these are some problem areas that I'm really excited about please reach out if you're interested in discussing them or building possible solutions these are just some examples and there's a long list of roll-up infrastructure that's both exciting and important for people to be working on um first how do we ensure that Mev is extracted in a way that's healthy for users Mev is maximal extractable value it's the possible value that can be siphoned from the network from a privileged position in the network this can be done in healthy ways or unhealthy ways but it's highly tied in with the second Point here which is given the sequencer's position of power in the network um can we uh do a couple things so as we discussed they get to order transactions and today sequencers are operated by a single team so how do we prevent them from extracting value from users or at least extracting value in a healthy way giving it back to users or otherwise using it for something like public goods which optimism is currently exploring um there are auction mechanisms first come first serve and others that were very early in designing recently there was a lot of debate about the arbitrary sequencer so check out their Forum or chat with me afterwards if you're interested in discussing it but they had um 100 like I think it was over a hundred thousand connections to their sequencer software um and they were were designing a proof of work mechanism that would uh privilege certain parties to get access to Mev unfortunately that's not happening right now but um there's a huge mechanism design space that we can uh you know that we're still very early in designing um and third uh how do we prevent code bugs from causing users to lose their funds so those validating bridges are subject to rules defined by code that can remove or not remove funds um they use optimistic fault proofs or ZK proofs and these are huge code bases some of the most state-of-the-art software we have um vitalik is prose proposed a multi-sig-like validation or validating Bridge system so in order to prove a state and withdraw funds there can be a validity proof a ZK or a validity proof a fault proof and a human multi-sig where if there's a bug in the fault proof a ZK proof and a multi-assay can override it there's a bug in the ZK proof a fall proof and the multi so you can override it but implementations of this area to exist so again very early in the design space here um so those were all challenges that are in our face today uh here are some others on the horizon that will become important in the next couple years so how do we build uh resilient proof generation systems specifically with ZK Roll-Ups as DK Roblox proliferate uh proofs are currently generated by a single super node um if that super node fails um the Rope cannot progress effectively and users funds can become stuck we can build open networks for participation in zk's proof systems but these designs are very nascent um uh second can we utilize those ZK proof systems as they become more mature to bootstrap new ZK applications So currently if a project wants to use UK they need to bootstrap all of their own infrastructure proof Network or nodes and as certain systems become widely used how can we establish DK infra standards to minimize necessary development costs for news 8K teams um third how do we scale these same security properties to L3 ecosystems that evolve on top of the base L2 Roll-Ups so fractal scaling is a whole other topic but we can actually build l3s on top of the l2s in a similar way that we've built l2s on l1s with some constraints but so how can we use all of the infrastructure we built for l2s to make the development and Adoption of L3 systems much quicker um so in summary Roll-Ups present the most promise that we have to scale ethereum uh and there are many challenges yet to be solved in their development so we should be excited by the opportunities to help decentralize them um please reach out to me or anybody from the recursive team if you're interested in discussing these problems um or interested in relevant resources that's all thank you and I hope you've enjoyed the conference [Music]