Rollup Rewound: Reflections on the Past, Visions of the Future | KBW2023

thank you it's wonderful to be back in Korea and to see all the growth and energy in the Korea blockchain Community uh we're really happy to be here and to be working with you um I would like to talk about the growth and the future of rollup technology and I'll tell this story in three parts past present and future so uh let me start with the past my own engagement with this area started I think as many people did with a moment of recognizing the potential of this technology recognizing what it is that we could do and what it is we could build by taking advantage of the opportunities unlocked by not only by blockchain but by scalable blockchain and for me one of the crucial moments in that was this uh conference that we hosted at Princeton University in early 2014 where we brought together a lot of the people from what was then a very small community of blockchain Builders and um at in the in the time leading up to this and in the time during this conference I had a I had an aha moment that really changed what I was going to do over the now nine years since then and that was a realization that uh uh that smart contract technology in particular was going to do something that I had seen happen in a really big way twice before so the first one was something that I saw happen when I was younger and that was the arrival of personal computers personal computers put into the hands of so many more people the opportunity to build software to build new kinds of applications um and deliver them to people like themselves so no longer would uh computation be something that only a few lucky people could do or only a few lucky people could use but suddenly we're going to have this tremendous growth in opportunity and tremendous growth in accessibility of the technology that followed from the fact that highly Advanced uh Technologies like this existed and this of course turned out to be incredibly important the same thing happened again and this time I uh was fortunate to be able to participate in this next Revolution and this was the uh the moment um when the web became programmable what I have here is the hot Java browser this debuted in 1995 um and it was not a product that lasted a long time but it really opened the eyes of a lot of people about what was possible on the net because what this did was it took the web and moved it from something which had been a way of delivering static content to people and turned it into a platform that you could use to build applications and again a very large number of people were now able to build these applications and distribute them and a much larger number of people were able to use them and the result of this was an explosion of creativity and when we use the internet today what we are using is essentially the technology that was pioneered here so this opened up a tremendous uh field for Innovation for building for creativity and open it up to a very large number of people so with those in mind when the idea of smart contracts crossed my Consciousness back in uh 2014 I said aha this is going to be the same story again this is going to be a way to uh to take what was previously a way of doing a relatively limited set of activities namely moving money around and it was going to create a new kind of platform that would allow us to deliver new kinds of services and to open that platform to so many more people to build in short a globally shared transparent and trustless virtual computer something that had never been possible before we could uh interact with each other across the internet over distance um and that was wonderful um but uh this was different because this was a glob this was globally shared and because it was transparent and trustless we didn't need to worry about whose computer or whose server was running this uh this technology because we could see it and we knew that it was going to execute correctly so this was a huge change in the opportunity to build Technologies and it seemed to me at least then that this was going to be a giant opportunity to unlock creativity and to uh unlock uh new kinds of uh of applications but not long after this realization came another realization in the form of a question will it scale because the obvious ways of building that Global uh globally shared trustless transparent computer don't scale very well they involve taking all of the activities that are happening on a chain all of the smart contract activity and executing all of that smart contract activity on every single user's computer um and if we're going if we're going to have everyone's business happening on everyone's computers in addition to the privacy considerations there are some real issues with scale and so this became my quest uh starting in early 2014 to understand how could we scale this smart contract technology because we were going to need to be able to do that in order to make real the the the possibility of this globally shared transparent trustless computer so how are we going to do that and that led um before long to the creation of the first uh version of a layer two or rollup which was arbitrum um in the fall of 2014 a group of Princeton University students that I was supervising did a course project in which they built the very first uh sort of Proto version of arbitrum you can go on YouTube and search for something like uh Princeton arbitrum student presentation and you'll probably see this and here you have the students um uh here you have the students a screenshot from this video explaining what are some of the Core Concepts of of arbitrum in particular the idea of what's now called interactive fraud proofs um you see up in this board so um already we were starting to build we were starting to experiment we were a long distance from having a pro a product of course and I'll be honest this was a student project and so not everything worked perfectly uh the students did impressive work during a short time but this this again was another step in helping to show what's possible now as this work developed there were a couple of key ideas in it um which I want to talk about the first key idea is the idea of Layer Two that is in building a blockchain is very helpful if we can build on top of some larger more stable chain um you can then have a solid foundation on which to build the security of your chain and with that Foundation you can build things that would not be possible if you were building a layer one that Foundation that in the case of of arbitrum Technologies that's provided by ethereum is critical so we are not an alternative to layer ones we are a complement to it and layer one is very important for us but this key idea to build on the security of a larger chain and then the second key idea is uh although we want to rely on that um on that uh parent chain uh we don't want to rely on it for everything we want as much as possible to move the activity of this layer 2 Chain off of the main layer one chain move everything almost everything off chain and just use that parent chain that layer one to record what's happening um and to prove it so that people can trust based on the the uh secure activity of that layer one chain that the layer two chain is going to give the correct results so these are the core ideas uh behind Layer Two now um uh after this student project I went off and did some other things for a while when I came back to Princeton um I met up with a couple of then PhD students uh Harry kadner and Steven goldfeder um and we worked together to build a more serious academic prototype of arbitrum and that led to this paper uh which was published in the unik security Symposium in the summer of 2018 about almost exactly 5 years ago um and this gave a more complete design for arbitrum for a layer two uh based on U interactive fraud proofs and then described an academic prototype we had built which was a proof of concept that showed what was possible but was not really anything close to a product well from there it uh soon after that we founded the uh the company here we are all looking younger uh uh without 5 years of um of startup stress under our belts um but nonetheless the the three of us um then professor and then PhD students uh now co-founders um started uh started to build this seriously um and what we what we aimed to do at the time perhaps naively as firsttime startup Founders was simply to take our academic prototype and make it compatible with ethereum and then roll it out as a product well we had to do a bunch more to reach product Market fit we had to solve a lot of engineering problems I'll spare you that all of those stories um but while we were building all of this over the three years it took to go from founding this to actually having a main net product something else very important happened and that is credit to vitalic vitalic posted in 2020 this very influential post um about the rollup Centric road map for ethereum and the idea there is simply that um uh that one of the keys to scaling up the ethereum experience and the ethereum ecosystem was to uh make ethereum the best possible place to host rollups to to build ethereum into that solid security Foundation that would provide what would what was needed for rollups to grow and then to rely on rollups to a great extent to be the solution for scaling ethereum we were happy to see this and we were happy to work with uh with vitalic and with members of other layer 2 teams to try to figure out how to make this possible and that brings us up to the present now um the promise that layer twos would be able to provide better scale and better performance um than you can get on a layer one I think has been realized and to choose my favorite uh layer 2 arbitrum one um and compare it against using ethereum alone you see that arbitrum 1 has roughly 20x lower cost per transaction about 5x higher gas capacity at present and about 48x faster block time a quarter of a second versus 12 seconds now this is not to dunk on ethereum because none of what we did would be possible none of what we're doing today would be possible without ethereum ethereum plays a very important role so this is not so much L2 versus ethereum but L2 with ethereum versus ethereum alone and so we've demonstrated and other teams have demonstrated too that it's possible to get much better performance higher throughput much lower cost for people and that translates into new types of applications being possible all right so I'm not going to go through through this in detail L2 architecture has evolved quite a bit I just want to point out two things about this first um that the uh the functionality of a modern L2 basically breaks into two halves you could draw a line uh down the middle if you cut this down the middle everything to the left is what you might call sequencing and everything to the right is execution and indeed one way to think about the way a modern rollup works is to imagine it almost as two chains that you have a sequencing function which is intake of transactions that produces on the upper right here a sequence of transactions which is kind of like a chain so you can imagine this chain sequencing chain which only just notarizes that some transactions have arrived and then on the right side and down in the lower right you have the L2 blocks that are produced by execution by consuming the transactions from that that chain one at a time in executing them now importantly both of those parts are anchored in ethereum down here on the lower left and this is the security anchor and this is what makes a rollup a rollup because it's anchored in ethereum and that means two things that we rely on ethereum to post data so people have Avail people have guaranteed availability to know what is happening on the chain and we rely on ethereum for fraud or validity proofs and which one you use depends on which kind of rollup you use but any modern rollup provides at least data posting and fraud or validity proofs now uh the L2 ecosystem has grown quite a bit this is a graph of total value uh some total value on chain summed up over all of the uh major layer 2os and you see over the last year this roughly doubles from around $5 billion to around1 billion but perhaps even more important this is a graph of trans transaction volume the blue line is transactions per second on ethereum over the last year and the red line is the total transaction volume of all the l2s and what you see is the l2s starting about even with uh with ethereum and growing to be about 5x ethereum in terms of total transaction flow and by the way because l2s have cheaper gas typically those transactions each consume more gas than an average ethereum transaction so it's 5x in transaction count but it's a larger increase than that in the amount of computational activity happening on these chains okay let's pivot to talk about the future uh the future of course is the most interesting place um that and it's about where we're trying to get to all right so let's remember this goal of a globally shared transparent trustless virtual computer so how are we doing in delivering that well the first thing that we need to do in order to to deliver that is to not lose focus on the things that we've been successful at so far the successful L2 teams Focus every day on driving cost down in order to lower user fees and bring more people in activity in and focus every day on driving scale up and much of the important work that is done in this um uh in this area is the everyday grind of improving a little bit every day that continual Improvement and that hard engineering work is really important it's not all about major breakthroughs um a lot of it is about continuing to just work and work and work to drive these metrics um and that's important that we don't lose lose sight of that but of course we need to strive for more um moving forward incrementally is valuable but we also want to look for those breakthroughs or look for those Quantum improvements that uh where they're available okay so the question that everyone is asking and of course which also shows up in the titles of quite a few talks today is this question of how can we on board the next billion users what do we need to do and of course the answer can only be that we need to remove all the barriers to that happening if you believe as I do that this technology is naturally valuable and naturally attractive that it brings a lot of value for the reasons that I talked about earlier um the question is what are the barriers to the next billion users coming and what can we as a community do to eliminate those barriers the first thing of course is user experience people complain about the user experience of essentially all blockchain systems and their right to complain we are not where we should be on user experience we're not close to where we should be on user experience uh we need to do a lot better but we also need to recognize um that the reasons why our user experience is not as good as it could be are not only the shallow reasons such as the fact that we like to build systems for people like us and people like us are not normal uh in many ways um but also there are deep reasons there are deep reasons because the technology is built in a way that is convenient for building but is not always um but it's not always well-designed for users um we need to take an approach that involves a bit more user oriented design and this is the idea that the uh the key to the best user interface is not just about surface it's not just about what it looks like or how it's and how graphical elements are arranged the best user experience is driven by thinking very hard and very creatively and pragmatically about what mental models users should have about how a system is supposed to work and what users will expect to happen and how we can build those expectations and then fulfill them and this means that we need user experience considerations to permeate our entire design it's not enough to just put a surface over a uh over a system that is designed for for nerds it's really important that we think deeply about what users need us to deliver how can we make a model and approach that will work for users and then how can we build that and we need to do that also recognizing that users are not all alike uh and users uh in different roles in different stages of life in different cultures and parts of the world have different needs and we need to also understand this is one of the reasons why it's so important to have such broad participation in this community okay but one of the questions uh we also need to ask if we're going to get to a billion users is how can we onboard the next million developers because if we're going to have a billion users we need two things first of all we're going to need a lot of developers because a billion users require a lot of tools a lot of products a lot of support and it's going to take a lot of people to actually build and sustain that but also we need brilliant ideas and the more developers we have the more access we will have to smart people to people who have different points of view and so building those million developers is crucial to developing the breakthroughs of the future in addition to the routine work of the future so how do we onboard the next million developers well where are the developers today um uh and here I have a maybe not so happy story to tell at least if you're a member of the ethereum community um because uh here's a graph uh that shows the number of developers who use different programming languages um and Tool chains to build their uh to build their software on the left we have solidity which is the main language that's used for building ethereum applications the onchain part with about 20,000 developers uh in the middle is rust uh the upand cominging systems programming language in the broader world with about 3 million that is about about 150 times as many as solidity and on the right we have the uh the Elder of um of programming languages C++ with around uh 12 million developers and so how are we going to get over that gap between a a community of 20,000 maybe a few more developers and getting up to the million or Beyond right now we sort of have a house divided in the blockchain world you kind of have two options you can be in the ethereum world over on the left where you have the ethereum virtual machine or evm which is the main execution engine and if you're developing software you're developing in languages like solidity or Viper which are which are good but let's face it still a bit experimental or over on the right you can have certain other blockchains uh like salana or near uh where you're using different virtual machines as a core execution engines things like Berkeley packet filter or web assembly and then you can program in the more popular program programming languages rust C C++ and others so you can be on the most popular chain with the largest um uh and most robust community in in my opinion ethereum or you can be in an environment where the development tools are the best and most welcoming to the largest number of developers well so the obvious solution to this if if we could figure out how to do it is to give you the best of both and I'm going to tell you in a minute about how that works but essentially the idea is to build a single system that supports both the ethereum virtual machine and some other uh more uh more modern battle tested and scaled uh virtual machine like say web assembly and then on the bottom be able to program in any of the languages be able to program in the uh uh in the ethereum specific languages like solidity and Viper but also in Rust C C++ and others so build by building a single system that allows you to do both and allows you to build parts of your application in one Parts in the other and interoperate then we really open up a a new world and this is what arbitrum stylist does we're really excited about stylist we announced it last Thursday opened up a uh a test net which you can try um you can read the developer docs if you're a developer but essentially it uh is a real implementation of what we call evm plus this idea that the ethereum virtual machine and uh evm compatibility evm equivalence on a layer 2 Chain is not a ceiling it's a floor U that is we want to build on top of it not by taking away not by stopping anything that already works from working but by adding new capabilities so evm plus is implemented by arbitrum stylus um and um this allows you to run these two closely coupled virtual machines on the same chain so you can have Smart contracts that are written in solidity or Viper rust C C++ they can run together they can fully interoperate fully compose it's an amazing piece of engineering I'm incredibly proud of our team for having built it uh and I really would urge you to try it because it's one of those tools that if you're a developer you can't really appreciate until you try it anyway I think this is an this is a very important step forward um I hope that the ethereum community itself considers adopting stylus in ethereum and we'd certainly be happy to work with them uh if they're considering that okay so we have a bunch of ways of scaling up each chain oh and I forgot to mention uh while Shilling stylus it's also 10 to 15x faster than evm okay end of advertisement now um methods like this and methods like that everyday grinding to get become more efficient not to mention Moore's Law helping us every day allow us to scale up each chain so that we can improve the amount of activity that we can do on a single chain without breaking anything that's really important to um uh to scale each chain now um in addition though we need to move to a world of many chains um because uh we want to be able to scale to a larger scale and grow the capacity faster than we can grow the capacity of any one chain so uh many chains leads to the idea now quite popular of app chains the idea that someone who in maybe in a previous point in time maybe last year two years three years ago might have built an application to run on top of an existing chain now they have the opportunity to launch their own chain now this is not without cost and um we need to resist the temptation to try to turn everything into a chain because running a chain does involve more complexity and more cost a little bit more technical risk than if you're building on someone else's chain but many people will need what an app chain can provide them many people will want to decouple the economics decouple the governance of their system from um from uh an existing chain and app chains give them the ability to do that so uh you can imagine a world uh which we and it's not a coincidence that we we call our program and the arbitrum team orbit to do this um but the concept is simply that uh you will have not only chains revolving around the Sun of ethereum but you we will also have moons revolving around those but of course what we really want is something more like this um that is very large and where we can scale to a larger number of chains because scaling requires not only making each chain faster but also having more chains uh so uh we really want to increase this and we're excited and our team and other layer 2 teams are working actively on making this possible now uh one of the key ideas here is the idea of layer three that is if you can build a layer two on top of a layer one like ethereum building a layer three provides on top of a layer two such as arbitrum one provides additional advantages now you might be asking why would I want a layer three why don't I just build a layer two myself there actually is some really important advantages to layer 3es that I think will drive more and more teams to want to build layer 3es why build a layer three well layer three has lower cost than a layer two why is that well the largest the largest part cost of operating a an I'll call an upper layer chain is uh is the gas in your parent chain that you use so if you're building a layer two you're using ethereum gas if you're building a layer three you're using Layer Two gas which is maybe 300 times cheaper um and so by reducing what is the largest source of cost in operating a an upper layer chain uh you can do that by being a layer three instead of a layer two uh second a layer three chain can have faster finality than a layer two why is that because layer it's because layer 2 Chains are better uh are engineered to provide faster finality than layer one and also because by building a layer three you are in in a very real sense pooling your resource not only with the all of the organic traffic on the layer two you're building on but also with all the other layer 3es that are there so lower cost faster finality and then also the ability to evolve your protocol faster because layer twos are becoming more capable than ethereum not only in performance but also in programmability and flexibility ethereum moves more slowly it's a little bit uh more careful and that's rightly so ethereum has more of a history it has to support uh more a broader range of overall activities most of it through layer twos and eventually layer 3es so it's it's right that ethereum moves more slowly but the advantage then of course of being a layer three is you can use protocols that are built on top of this more in inovative faster and more flexible Layer Two technology and that pays off for you so layer threes I think are an important uh step forward okay so we see a future of many chains um uh with continual growth of chains from Layer Two to layer three maybe someday layer fours for those who want it uh that's really important some of those chains will want to be tightly coupled tightly coupled simply means that a set of chains decide that they are going to um a set of chains decide that they're going to get together they are going to share some resources The Operators of these chains are going to validate the other chains they're going to uh create a shared security environment across these chains and what they get in exchange is much faster and Tighter interoperation composability between those chains so some will choose to do this this is not without tradeoffs though because when you tightly couple to someone you are agreeing to expend resources to support them them just as they do the same for you and you're taking on a shared security responsibility so this is not uh this is not meant to be open to everyone willy-nilly but it will be very valuable for those who decide to Ally themselves in this way so what emerges from this is a vision of a set of self-governing voluntary Consortium chains where groups of chains will get together they're they'll Ally themselves and they will decide to work together share security and share resources in order to get faster composability and uh and and Tighter security coupling now I believe strongly that we're not going to see a world with a single Mega chain that uh with limited uh membership and centralized governance I think we're going to see a world instead of many voluntary self-governing Consortium chains uh so we're trying to build this a future of a globally shared transparent trustless virtual computer will it scale yes it will scale we have demonstrated that it will we're not done we're not close to done but uh we need to keep working on this goal there there are exciting times ahead we're going to be able to build amazing things together and um I look forward to working with uh the Korean blockchain community and many others all around the world as we build this Future Let's Build That future together thank you very much consam [Music] oh