From Concept to Reality: Modular Blockchain Architecture | KBW2023

[Music] testing hi um thank you everyone so much for coming here I'm really excited to chat today with Nick um at Celestia so for those of you who don't know Nick white um has had his electrical engineering degree from Stanford um and then he was a co-founder of Harmony and he spent the last two and a half years building Celestia where he is COO and so we're going to spend some time breaking down his views on modular blockchain how he's seeing that ecosystem um so maybe just to get started um how did you you know how did you first get into crypto and what Drew you to uh Celestia specifically well actually one of the first crypto conferences I ever attended was here in Korea uh it was deconomy in the year 2018 and I just quit my job I was living in Hong Kong and I wanted to go full-time in crypto so I flew up here I didn't know anyone I didn't have a ticket to the conference I just kind of like mingled around and talked to people but I knew I had so much conviction in uh blockchain and uh that was kind of the beginning of my journey and I ended up joining uh Harmony uh as a co-founder and moving back to the Bay Area and basically from the very beginning my uh desire or like goal was to solve scalability of blockchains because I knew that the underlying technology had so much potential but until we could actually build these trust minimized systems that scale uh we couldn't actually build things that would be useful for people and so that was kind of my goal at Harmony we did staking and sharding kind of implemented the eth2 road map at the time um but then sharding kind of clearly became a dead end and uh at the same time I discovered this new white paper called lazy Ledger which uh defined a new modular uh approach to building blockchains that to me solved uh once and for all a lot of the core problems with blockchain uh infrastructure including scaling among other things and so that's kind of like the how I got started and how I got to where I am um so although Celestia has been around for a while and I think this idea of modular blockchains has been around for a while I think it's definitely entered um it's entered the meta more recently and I think a lot of people are still trying to wrap their heads around what does it mean to you know why do we need a modular architecture like what do it mean to have one um and I think one of the um the way that Celestia has kind of explained itself is pretty interesting so the uh the quote is if Bitcoin is a calculator and ethereum is a computer then Celestia is a cloud computer um could you unpack that a little bit for the audience like what does what does that mean and and you know how does cestia function in the ecosystem so that analogy is is trying to help like break down sort of the evolution of blockchain architectures from Bitcoin to ethereum to Celestia and the new modular Paradigm and so we call Bitcoin like a a calculator because basically Bitcoin was a blockchain that only had one function it had a very limited number of things that it could do basically you could transfer Bitcoin and that's kind of it it's kind of like email and if you wanted to uh build something that did did something different like a decentralized exchange or nfts right you would have to build an entirely new blockchain and program in that logic to that blockchain so a calculator is also very limited in the scope of what it can do right it can just add and subtract and multiply numbers that's it uh and so then when ethereum came along it generalized this concept uh to make a blockchain that you could actually uh run all different kinds of programs on top of it so if you wanted to build an application that didn't yet exist you could write a smart contract and could run on ethereum so ethereum sort of generalize it to a computer like a computer is something that uh you know it doesn't just add or multiply you can load new programs you can download things and run them and they do they run new applications now the problem with ethereum is that as as just a one single Standalone computer it kind of comes pre-loaded with the evm and so it's kind of like imagine if you had a laptop that only ran Windows right but you wanted to run a Mac app all of a sudden you can't actually run you're limited into the scope of what you can run and also ethereum is sort of like a uh a single computer that everyone in the world has to share and now imagine if everyone was trying to use one single computer it would get it's like running every app on your computer eventually it kind of like maxes out uh the resources right of the computer and it slows down or crashes and that's kind of what's happening to ethereum uh in various bull markets and even even now with the high gas fees and so where Celestia comes in is it's it's a cloud computer where in the modern web today most web2 applications run in the cloud meaning that someone like Amazon or Microsoft has these very scalable data centers that have vast amounts of very cost uh effective compute and you can actually build like rent parts of that compute to run virtual machines or like servers to run your application and that server can have whatever kind of operating system it wants it can scale up and down on demand and um that this is essentially what Celestia and the modular architecture is instead of a data center you have uh celesa which is like a Data Network and it provides raw rather than raw compute it's raw block space and then developers come to Celestia they consume that block space and run virtual blockchains called rollups that can be customized and scaled up and down depending on demand so that's that's like the that's how the analogy makes sense yeah um so I mean obviously scalability I I think since the beginning has been one of the biggest issues that people have tried to tackle and there's been a lot of different approaches there's been the approach of hey maybe we should Shard um like sharding has been approach um using you know more hardcore Hardware has been another approach and Celestia is obviously a different approach can you kind of compare and contrast these different approaches and and why you think Celestia's approach is is the correct one here yeah so um I guess I'll start with charting and touch on it briefly because charting now is kind of like faded into irrelevance I'd say um but the the what charting tries to do is actually in some ways similar to the modular stack where you take the like one single Committee of validators and then you split them up into smaller groups and each one of them runs their own chain and then those chains like actually kind of checkpoint and together and try to share this common network of security the problem with sharding is that the way that you rotate validators like everything about it is this very cumbersome and complex and at the end of the day it's like it makes sense as a white paper but then when you try to implement it it doesn't really play out as well as you think and also more importantly it doesn't actually solve the fundamental problem of scalability which um I need to maybe spend a little bit of time defining so um what's challenging about scaling blockchains is that what makes blockchains powerful is that they are verifiable computers meaning that people anyone around the world can run a node of the chain and verify everything that's happening is following the rules that everyone has agreed upon and when you the reason that that is hard to scale is that when you increase the number of transactions in the network netw work every person running a node has to then download and verify more and more transactions and so if you increase the number of transactions too high no one is able to run a node anymore and now you you just have Google running a massive blockchain that no one can verify so all of a sudden you lost the whole core value of what blockchains are about and the problem with charting is it doesn't actually make it easier to verify the chain as it scales now another approach to scalability is um to build to just basically not even solve scalability and just make your blockchain really really big so things like salana or like these chains that uh are are really fast and have a lot of throughput they tend to just what they've done is they've raised the node requirements so they they make it so that like we're as ethereum you can verify it on your laptop for salana you need a very powerful computer with a very high like bandwidth connection to verify it and so it's not really solving the problem what they do do that that actually does help solve scalability is that they make optimizations that allow you to use the machine of your node more efficiently so you can parallelize verifying execution you can uh sort of like gossip data faster so there are elements that do solve scalability but at the end of the end of the day a lot of the scale up of these kind of systems is just increasing the node requirements so now how is how are modu the blockchains different to this so modul blockchains are different because they employ two new technologies that fundamentally SCA scale the verifiability of blockchains so the first one is one that we were all familiar with in this room I would think called rollups and so what rollups do is they make it so that rather than having to verif ver ify every single transaction one by one you can verify a single proof that all those transactions are valid and so you compress you reduce all that computation down to verifying one thing so there could be a million transactions instead of verifying every single one you just verify one single proof so that's a huge you know uh advantage over like normal monolithic chains and then the other component to scaling uh modular blockchains is a technology called Data availability sampling so in order for any blockchain to function you need all of the uh transaction data to be public information it needs to have been published to everyone who is running a node and um unfortunately the only way to verify that data has been published is you have to download all of it yourself and of course that doesn't if you increase the number of transactions then every node has to then download that much more data so you're now still not able to scale so what data availability sampling does is it allows you to verify that data has been published by only downloading you know 0.1% of the overall block data and so rather than having to download a million transactions which could be like a gigabyte of data or something you just download a few kilobytes but you have a statistical guarantee that all that data was published so those two things together actually make it possible for end users to verify uh a modular blockchain on their phone and so like I I don't I actually bought a phone that's running a Celestia node with me uh and it's really cool like everyone this this is the to me the endgame of of um so this is a node that's running a Celestia uh like verifying light node and so in the future this is this is sort of the the what we imagine uh blockchains to look like is that and end users on their phone are verifying the chain directly um that was a really awesome breakdown thank you for going through all of that um so I I asked um I asked Twitter what questions that they had for you and this was so I'm going to sprinkle a couple as they as they come up in the conversation so this is one that someone had which is um could you explain what are the ux so the user experience and latency trade-offs um you know that that basically exists by passing on devs and users uh that are passing on devs and users due to modularity or like what are what are the trade-offs that people have to consider due to modularity yeah this is a good question I think it came from Ilia at near uh so there's so the thing about modular blockchains is that you're stacking you you're splitting what used to be one single protocol and one set of nodes into SE several different protocols that are layered on top of each other and uh so it does make the system kind of have more dependencies I guess um now to the end user there isn't like the the the when when it comes to like latency the main latency that matters is the uh speed of finality of the underlying data availability layer like Celestia or ethereum because you you won't ever have complete finality of your transaction until it's included in the base da layer and you've sampled and verified it so um that doesn't really so just depends on the underline D layer in celesa it's only like a 15-second block time so it's not like it's kind of onar with existing block chains um and then beyond that actually let's say you wanted something faster than 15 seconds you can still do things like having uh soft confirmations from uh the rollup sequencers so like if you ever used optimism or arbitr you know that like your transactions get confirmed almost instantaneously and that's because that finality is coming from the sequencer guaranteeing you that that TR transaction has been included and executed so you can still have but then like the only the final final confirmation right happens on on ethereum L1 in in for those for those rollups so like the user experience doesn't necessarily have to suffer uh at all really in fact it can be as good or better than a lot of Monolithic chains um in terms of the uh developer experience there can be more complexity because you have to choose you know which components to put together and you have to fit them together um however the there's a lot of work that's going on to basically first of all establish standard interfaces across the different layers so it becomes very plug Inplay and then there are teams like uh Caldera or vestara or conduit there's a bunch of different teams doing what's called rollups as a service which is sort of like a uh a white glove or like a a sort of hosted uh rollup platform in which developers don't need to know anything they just put in a bunch of parameters and click a button and they get a rollup deployed uh on demand so I think the developer experience is going to be like our our end goal and I think where we're already almost at is to make it as easy easy to deploy a rollup as it is to deploy a smart contract so that like you know there isn't really a difference between building a rollup and building like deploying a normal dap on ethereum in the long term how do you think about the first use cases or the first types of rollups or or the first um participants in that that's going to be leveraging modular architecture do you think there's kind of a specific type where it's like hey that's you know these are these are the types that make sense to start with and then obviously you can scale out and there are other users as well but how do you kind of how do you see who do you think are the people who need it most or like most likely to adopt it first that's a very good question so to understand who are likely to adopt it we have to talk about um what the what the advantages are really quick so like one of the first advantages we talked about is scalability so with a modular blockchain you can have a lot of throughput and low fees which is uh useful for some use cases you also get greater customizability so because of the fact that each of these layers can be swapped in and out there's a and there's a bunch of different sort of uh versions of each layer to choose from developers have this superpower of flexibility and customization that they don't get in a monolithic chain so they can you know just like choose the exact stack that is perfectly optimized for the system they can even for example customize themselves like write custom logic at the execution layer to make things like just specific to them uh so there's that level of of control so customizability and scale are two really uh big advantages of the uh modular Paradigm now also though are like the disadvantages like so starting out what I would say is when you if you deploy your like a new rollup the thing is that it won't be as connected right if you deploy a smart contract on eel1 right it's it's natively interoperable it has access to all the liquidity on eel1 Etc but if you deploy a rollup um it's a new chain and so it needs to have bridg in order to effectively have connectivity with the rest of the ecosystem and so I think that's that's one downside so kind of I guess where where I'm like now to like go back to your question of like so who would be most who would benefit most I think it's uh projects that benefit from scale customizability and don't at least for now have much as much of a need for like uh direct connectivity with the existing uh stack so some of the things that I've seen along these lines are there's a team called Argus who I think is speaking tomorrow um they're building a stack specifically for onchain gaming so unchain gaming is a good use case because games need a lot of scalability and low fees um also gaming is a use case that requires a lot of customization so some of the things that they've done is added uh a very specific way in which to to Shard the game world and they've also uh made it possible to have a very high tick rate so that if feels when you're playing the game it feels like instantaneous um and then another example is like Manta which is a project um that will be Deploy on Celestia that is a uh an evm based rollup but they've customized the evm to add specific op codes to make ZK verification more gas efficient and so um and and they're they're more targeting like really cool like privacy use cases or uh you know having verifiable onchain Randomness for for different games like they have a a poker game that you can play on chain um but like defi for example well def also benefits because part of customization is that you can capture more value or more me um but you want to be able to be at the stage of a defi app where you're like dydx where you have like an existing user base an existing liquidity base so that launching your own like rollup you will that liquidity will like migrate with you it's like it might be hard to like have a cold start of liquidity if you're a new defi app yeah yeah um how how do you think about how bridges will work and you know it's obviously something that will eventually have to emerge um assuming that interoperability is going to be an important thing that we you know we we uh we keep and again like at to your point not everything needs to be interop maybe some games can can exist very well on its own um but how do you see Bridges emerging and and you know what are what are some of the challenges there so bridging is going to be really important in the modular stack because you have all these different rollups uh that are kind of their own chain and in order to connect the applications and the users and liquidity across those chains you need you need to Bridge and um right now what that and the funny thing is that the superpower of rollups is that they have the ability to do something called trust minimized bridging so in like typically between two separate L1 chains you have to make a trust assumption on the other chain when you Bridge and uh what's really cool about rollups is as long as they are deployed on the same data availability layer they can Bridge with each other without making any trust assumptions on the other rollup now uh unfortunately there's still a lot of work to be done to make rollup bridging uh sort of practical uh in a lot of cases I think like once like if we can get ZK rollups to production and at scale I think that will solve a lot of the problems because like bridging ZK RS is is simpler and faster than uh bridging um uh optimistic rollups in many cases but there's teams like for example hyperlane that uh I'm really excited about because what they're doing is they're making it possible to deploy a bridge on your rollup by just deploying a set of smart contracts and so um you know the worst thing what you don't want to do is have a world where I'm a developer I'm building my rollup and all of a sudden as I build my rollup now I have to do BD or convince some Bridge team to integrate with me and that they they become like The Gatekeepers of my ecosystem and so we want to have bridging that's permissionless and very fast and easy to deploy and um Pro projects like hyperlane make that possible where you can just spin up your rollup deploy a set of smart contracts and then Connect into this whole like uh interconnected network of of different chains and it's not just only specific for rollups but also other l1s so maybe along that vein in terms of you know what needs to be built um to have this really robust modular architecture take off what are bridges we just talked about what are some other things that you think um need to be built in the ecosystem and um in in you know alongside what are what are the things that you're really most excited about that's coming up good question well um this kind of goes back to the uh ux problem that you talked about uh we talked about earlier one aspect of user experience in sort of the future modular world that we envision is there's going to be tons and tons of different blockchains and um now the the thing about that is even if let's say bridging is secure and and relatively seamless um there's going to be so many different you're going to have assets on all these different chains you're going to have applications on all these different chains uh and as a user it's going to be just really really complex it's like a this whole web of different dependencies and and things going on and so we need to find a way that abstracts a lot of that actual complexity uh so that you don't even maybe know that there are separate chains under the hood right and I think to me one of the things I'm really excited about on that front is this uh emerging uh sort of idea called intense which is that you know a transaction the way that we interact with blockchains now is through transactions where you uh basically specify a set of inputs right to the chain and then the chain goes and tells you what the output is and an intent is is kind of the reverse of that where you specify what the output or the outcome should be and then you authorize someone to act on your behalf and and determine like the inputs to make that uh possible and uh what's nice about that is that as a user you just say this is what I want to happen and then it kind of like magically takes place and so you don't have to like on the other hand like if it's a transaction it had to be like okay I want to trade I want to take these tokens from this chain and Route them through this chain to that chain do a swap move back buy this nft you know then move the nft over to here or whatever uh and and then you have to specify that exact whole trace and that's really hard whereas instead in an intense world you just say hey I want to buy this nft I'm willing to sell these various tokens for that you can choose whichever ones you want and then at the end I want this nft to be moved over to this like nft Marketplace chain and then boom it just happens yeah um another question from Twitter um if you had to take the Devil's Advocate view then what are some reasons that you think modular blockchain would not take off this is a good question so I would say the Devil's Advocate side against modular blockchains is either that Atomic composability is so important that everything just wants to live on one single chain that's one bare case although in theory you could still have uh like one massive rollup on top of a like a one massive da layer like Celestia and you kind of would end up on in the same place but then at that point there's no real reason why those shouldn't just be the same protocol um another uh be case is that um people the like we believe that the the customizability that people want mostly is that sort of like the execution layer and maybe the consensus layer I guess but like that that you don't you don't need to tightly couple these different layers of the stack because you can kind of get the benefits of each one in an isolated way um but like if like the bare case would be that okay well in order to make this one execution tradeoff for optimization we're going to need to actually control specifically the consensus Network part of this or the men pool specifically and so uh and th those kind of things are are harder to do in a modular stack because those those different like layers are not really uh they're kind of independent so those are those are the two like bare cases I could see okay um and and maybe now the flip side of that since we you know we're we're kind of reaching the end of our session can you paint for the audience what you think let's project forward you know 10 years into the future what do you think assuming modular architecture takes off what do you think the world kind of looks like in in in that in that um if we go along this path well I I love this question because uh I'm kind of a dreamer and I've like since like you know five years ago when I first got into the space have like wanted to see this end manifestation of of blockchains come to come to life and uh so it in Celestia specifically our vision is uh kind of encapsulated by three words so the first is uh gigabyte blocks the second is a million rollups and the last one is 1 billion light nodes and so uh I'll explain what each one of those mean and then kind of what it means for this like broader vision of of where crypto where we want crypto to go so gigabyte blocks is referring to the fact that we want to eventually scale Celestia to have gigabyte siiz blocks um which would just be a massive amount of of data throughput and hopefully could scale to you know billions of actual users uh and that's just I'm sure it can go beyond that but that's sort of like the the Milestone that we've set for ourselves then a million rollups is referring to the fact that we want to see an explosion of of applications and and rollups and chains um similar to what we saw on ethereum uh with with smart contract and we think there's just so much more Innovation and exploration to be done at that layer and also within that if people were actually able to deploy a million rollups that means we've made it very easy right it's like right now it's it's probably not that easy it would be very hard to imagine uh a million different rolls being deployed but like we if we make it as easy to deploy as a smart contract like every developer could just go online and write a few lines and and on a command line and and have a roll up and then the last one is a billion light nodes and this is one that's very near and dear to my heart which is that we want to build a culture where all the users of crypto are actually verifying the chain on their phones or on their browser or whatever device they're using they're directly verifying the blockchain because that's what we came for that's what we've been building from the beginning is a like a a user verified user owned internet and so um to make that happen we have to integrate the celestial light note into different wallets um we need to build the behavior for people and just in general get adoption so that's kind of like the the end game and the end game is like like I said a user owned user verified internet where we uh have more like sovereignty and power and autonomy and um the whole world is connected in a network of cryptographic trust um that will scale like the economy scale our ability to cooperate uh I think lead to a more peaceful and prosperous world so that's that's our vision for the future of blockchain awesome well thank you so much for the time such a pleasure chatting with you and and learning more about Celestia thank you for breaking down everything to its granular parts and sharing your vision um yeah this is it this is the end of the session but thank you so much Nick thank you great questions see you guys