ZK Stack and the Rise of the Rollup-State | KBW2023

[Music] Hello everybody welcome my name is Lassa Clausen I'm one of the co-founders of uh 1kx uh we're an early stage um crypto Network investor uh in fact we're actually the first Fund in the space that said we only want to invest in tokens uh it was very esoteric strategy in 2017 2018 when uh you had private Enterprise blockchains were a big thing back then uh you know blockchain not Bitcoin Equity not tokens Etc and uh from day one we thought really the big innovation is in open source token networks and U yeah I'm very excited here to be with Marco from ZK sync metalabs um disclosure we are very very biased investors in metalabs we uh invested in the first round in 2018 so this is when we started thinking about optimistic and uh zer knowledge based rollups and um yeah really saw or predicted that basically ZK based zero knowledge based rollups are going to be the future so I'm very excited to be uh talking with Marco today thank you lass and maybe quick introduction I'm Marco I'm the head of the business team and metal Labs before this I was the CEO of a fund management company of the Asian operations of a global fund management company and before that um I was a Quant you know the guys that do the um the modeling for complex derivatives so for for me crypto is both an atonement for all the sins of previous traditional Finance um you know work but also is something that I used to do all the time I started mining Bitcoin very early on and I've always been involved personally in crypto and I decided to merge these two these two things in my life and actually get some free time like this and uh yeah for those unfamiliar with uh with you and then ZK sync maybe in you know very high level terms what is what is ZK sync era what can I do with it so ZK sync is a layer two is a zero knowledge roll up what does it mean it means that we take big batches of transaction compute a proof that they are valid this proof is also very small and we send this proof back on ethereum which means that you get the same user experience that you have on ethereum but it's just much much cheaper you have the same security the same decentralization uh properties which are inherited by the layer one but the environment is very very uh cheap to operate everything else is pretty much the same if you're a developer you have a similar developer experience if you're a user you have a similar user experience we've always focused very much on improving and not just making sure that we were you know coping with ethereum so in reality the ux is slightly better you have subse confirmation and the developer experience is similar if you want it to be um but because we are we we have a slightly different design we can allow you to develop things that are more complex than those that can be done on ethereum and I think to to reiterate the you know how amazing zero knowledge proofs are um I would say also they you know the invention of blockchain Technology almost as big of an invention is actually zero knowledge proofs uh and maybe to give a simple example what you can do is you can have you know many many transactions if not millions of transactions and you can express all of those you know lots of data lots of things happening and you can express a mathematical proof that all of these transactions happened in one succinct little proof and that is what you store in ethereum so you basically get the security of ethereum but you know you have you know very very high scalability uh you know lots of speed and and I think just to to reiterate how much of a big um you know Innovation zero knowledge proofs are it's also the timing is quite fortunate you know there's been Decades of research into this and just some uh breakthroughs and some of them you know your company was actually part of that um you know it's 20 years ago 10 years ago 5 years ago it was too early for zero knowledge proofs and production and we just basically hit that inflection points a very very few years ago yeah it would have been impractical to do what we can do today just uh four or five years ago it took us four year to be where we are today today we are very very happy um the new proving system buum that we launched a few weeks ago is the most performant in the industry according to third party uh um benchmarks is you know one a little bit more than one scale of magnitude more performant than the other um than the next one um but also so we are making the computation of this proof very fast and very easy to do but we are also changing the hardware that is required to run this proof I think Sandy before mentioned Hardware acceleration is a very important thing in the in the space we have our own internal Hardware acceleration team and we' have been you know advancing this quite a lot there's a bunch of prizes around in the community our team has been performing in very very well in those um and we moved from running this proof on gigantic machines um that at the time were bigger than the one say open AI use to now the current system in production use the a100 machine similar to what open AI use our new prover um can use a GPU that is a consumer GPU with like 6 to 8 gabt of ram 16 gbt of ram machine are actually easier to find so that's what probably we are going to use but we are making the system very very easy to decentralize very easy to run normal cloud and all of these advances would have been impossible I mean we were speaking with very smart people in research that essentially told us that what we were saying was impossible and eventually you know they changed their mind a little bit yes I remember very well that I introduced uh your team and your company to a very prominent very large investment fund of the space um at that time you were raising money at $100 million valuation and uh one of the most prolific cryptography professors in the world said what they're doing is five years out um which is evm compatibility I'll touch on on that in in a second and uh and you know I could not convince them otherwise uh the professor said so and then basically one year later they invested at a billon valuation um so I think that's you know the other really really big sort of breakthrough that um that ZK sync um managed to do because basically everyone in the ethereum community or in the industry thought that zero knowledge proves for them to be evm compatible meaning that if you have an application on ethereum and you've built everything that it runs on ethereum for that to work on a zero knowledge based rollup that is going to be five years out and you sort of have to rebuild everything if you're an engineer if you're an application developer you have to almost rebuild the whole thing just to make it work on your ZK sync and um you know this is what's called evm compatibility meaning that you can just use the application that you already have on ethereum and you can put it on on the layer two and um yeah and I think that's where you also beat the estimates by basically four years because one year later you launched the um ZK uh you know an evm compatible uh um zero knowledge roll up yeah and uh I mean ZK Sync has been at the Forefront of this we launched ZK sync V1 which is now called light I think more than years ago now it has the most the oldest R rollup the oldest ZK roll up to the best of my knowledge we were also first to launching the first Z KVM and and you know we are now you know improving on the on the prover technology and on the GPU proving and hardw acceleration so we are try trying to stay at the Forefront of uh of the technology and um over the past couple of months we also announced that we are completely opening the entire Tech stock so we have a chain that is running today is called ZK sync era it it it is actually very well received by the community we are humbled and very thankful um today we are running at around 29 million transaction a month which is very close to ethereum at around 30 million but it's like almost twice as much the second uh of the layer two so we are by far the layer two with the biggest amount of usage we have more usage than all the optimistic cup all the all the school optimistic cup combined so the zero knowledge uh Revolution is coming and the user are adopting it in order to build era we use a bigger technology called Z stack so era is the first instantiation of zik Stack that we use to build our own chain but zik stack is bigger than this and the idea is that no single system that is monolitic including z Ina will be able to accept all transactions so it seems strange to say but we we are already thinking that we are not going to be able to cope with the demand that we are going to have relatively soon and so we built a general purpose technology that allow anybody to spin up their own chain these chain are modular so you don't have to think about a lot of the things that you otherwise have to deal with if you build your own chain so you don't have to think about decentralization you don't have to think about security because those are inherited from ethereum exactly like for every roll up and and and zky in particular and they're Sovereign so they're truly yours they have nothing to do with zyn era they simply share the same technology the code is already MIT AI licensed is free and open source is truly free and truly open source you can take it use it and we are here to help you to do so um and we cannot we committed by making it open source not to be able to stop you um you know if if you decide to do that so zik stack is a general set of technology that allow anybody to now benefit from Zol app you don't need to be um you don't need four years of your life like for us to get to this point you can do it in a very short period of time because now the technology is ready yeah maybe with that could you elaborate a little I think from day one you've been it's been a very very Mission driven and values driven project and I think maybe you can just elaborate a little bit on that and share the values of of the team yeah yeah so um metal laabs is is not really a I mean is a company in a formal sense but we don't think of it as a company we think of it as a mission driven project everything we do is there to accelerate freedom for all uh we want to make make users being able to be Sovereign in their interaction with other users and everything we do is an instantiation is a manifestation of this and it's informed by this by this by this principle so everything we do is open source and it's free very quickly after we launch it sometimes immediately at the same time we launch it everything is available we do our best to document it also as best as we can but of course that that can you know have some improvement but we we are really working toward making sure that our user and our developer can just take what's there and build upon it um and and the idea is to make the system very very open as well so today like many roll up we are running in a centralized fashion meaning that we are running the sequencers we are running the prover which are the two entities that one ordered the transaction and the other computed this zero knowledge proofs uh this is a very uncomfortable situation for us because it goes against the principle and we have committed already many times to try to be the first to fully centralized we will of course be super happy if someone else does it even further because even faster because we'll be able to learn I know how how they do it but we are totally committed to make the entire system to decentralized so it's already free and open source and that's the final step to really give it to the community to to own and Direct in whatever Direction it will see fit in the future and I can actually attest to this is not just sort of nice talk on a stage uh you know we're also Community friendly um you know I've been a situation where your co-founder was offered hundreds of millions of dollars by investors but they wanted very close to control of the network they wanted a very very large chunk they wanted 10% of the network and remember you know with with these Technologies what makes them special is that they're not owned by anyone they're not owned by three people they're not owned by three entities right the security comes from decentralization that just means that many many different people ideally the users the stakeholders of the technology collectively control this but it's not a small group of people one entity that controls this and so that matters a lot that when you're raising money you cannot sell too much of your token Network because this is when you lose your decentralization and then you basically lose the entire point of of why you're even doing this and you know I've been involved with conversations where large funds offered a lot of money um to the team to invest uh you know tens of millions of dollars and uh and they really stayed strong and they you know pushed back uh over several rounds of negotiations to really sell that fund only a very very small percentage of the network so you'll see lots of people up here claiming all these things and I can prove to you this is one of the few projects that didn't given the temptation of tens of millions of dollars and actually stay true to the principle yeah we are trying our best and and we think this is not just um um because it's the right thing to do it's also the selfish thing to do there was a very interesting presentation from um Denison from Tali at itcc where he essentially saying that you know the example is dinga and and Facebook no you don't want to create your own great product with great product Market Feit on Facebook anymore at least not in a gaming space because Facebook proved that they can just shut you down and and that the lack of credible neutrality which is in turn created by the fact that you have a small number of people that can control the system so by making it credibly neutral and credibly decentralize um it's both the right thing to do but it's also the smart thing to do it's it's a selfish and selfless act at the same time right because if you look at at most technology companies they stop growing they get to this point where nobody voluntarily helps this company succeed even more because they're so big they control so many aspects they can kill your company if they want to and then kind of just stall out uh and at that point all they're doing is monetizing sort of their existing reach their existing user base their existing technology lock in and I think with u with token networks you can actually become bigger than that because you never get to this adversarial position where oh here's the three guys in the boardroom against me who's dependent on the technology and um you know as an example you have uh banks that are still locked into IBM mainframes and are being sold the hundreds integration on top to get anything done um I don't know if anyone here is using Salesforce but you know after a while you're basically uh you know you're a Salesforce slave you can't get out of that thing anymore and anything you really want to do any additional feature you want is another uh integration that that you know you get upsold on and so this is really part of the the core principles why this technology or this industry can actually really be different because you're avoiding this sort of technology and vendor lock in through the C ization um and so you you know I'm intrigued about this upgrade of like using boam do you have more yeah do you have just like more details on that because that is I mean know potentially for us we in 2018 basically decided it's zero knowledge based rollups it's not optimistic rollups optimistic rollups were always considered a hack in the meantime because evm compatibility is 5 years out and and so we've you know have have great teams that build great optimistic Roll-Ups but it was always thought as something we need that in the meantime but really the end came is zero zero uh knowledge based rollups and and so we've got that now but then you've you've you're taken it one step further yes so we we we were the first to to launch it on on on Main net already and now we have upgraded to what is today the most performance proving system performance proving system in the world to the best to the best of our knowledge the the there is a lot of things so provs are a very complex technology are a huge tradeoff and and I don't think it's super interesting for the audience to go into the technical details here but maybe there is one thing that is very important for us one of the thing we really wanted to optimize for withum is the fact that he runs on Hardware that is consumer grade so one thing that made us very uncomfortable before is that we were depending on this very large specialized Hardware which in turn meant that whoever runs this that Hardware becomes a central point of failure for the entire system you have you have a liveliness issue essentially someone manages to shut down those five entity in the world that can deal with this very specialized of expensive um and difficult to manage uh pieces of equipment if you manage to shut them down you shut down the system which is very um against the broad philosophy buum is interesting because he runs on gpus that are consumer based So the plan is to run them on 16 GB GPU but you know in our test use Alp of that Al of that Ram as well it's just that 16 GB GPU are more available in the cloud and so it makes the system more resilient is extremely easy to spin them up you can spin them up in very very abundant uh amounts this also creates another very important feature is that today um if you have the usage that we have which is great in the rollup world I mean we have 43% of all the daily active addresses across all the layer two and we we do you know yeah 30 million transaction when the second does like 16 17 so our usage is is quite good compared to relap but is Tiny we are really not happy you know this usage should go up by three four five factor of magnitude to really support the world of value to move on chain and that just cannot happen if you have specialized Hardware we were already using too much at some point we had to negotiate with Google to get more of these pieces of Hardware that were unable unavailable no and and that's very very difficult you can't then go and sell this general purpose technology to another 100 companies because there's just no space for them now with buum there is space for everyone basically and then you have another exciting um upgrade you already running it in parallel now uh Alex spoke about this at at our Summit in Paris you have more details on that because I think the cost of transactions is it's going to be really dramatic the the impact right yes so we are running it in parallel that's another interesting thing is one of the technical feature is uh for um for many rollup because of the way they said set up the system and you know if if they have been not super careful or maybe they didn't have time because they developed in a rush you can set these things up in a way that is very difficult to upgrade them and we saw an instance of this with with one of the rollup that erases some accounts know they were very old and uh but this is this is something you don't want to do you don't want to enter into that conversation and it touches on the immutability and if you erase them and then you recreate them you just create all these precedents that are not that are not desirable um we spent a lot of time to develop the system also because uh we made it upgradable so buum is already live and is proving things in parallel to the old proofing system which allow us to really test and make sure that everything is ready and then we will go into a phase when we will continue to keep it in parallel but switch the real proving system onto buum and eventually will remove the old proving system because the new proving system is much faster it's much lighter on resources and it will allow us to scale to to to the next level and so if you're you have some price points where we are what's a typical token transfer transaction today what will be after the upgrade and then how will what is this concept of layer three cuz I mean you know we've got L1 we've got L2 now we're going to have L3 what is yeah what are what what is that so I think on the first question is a little bit difficult to say because in parallel there is 4844 and so a majority of the cost on any roll up including including ZK sync is the C data and ethereum is changing the cost of the C data by still an unclear amount we don't know we have I mean I heard the original estimation of 100x um as low as 2x it looks like today for anything that is not a simple transfer it's probably going to be say five six 7x so cold data is already 99% of the cost for for Zing so if that goes down by 6 or 7x transaction go down today a transaction is in the region of 10 to 20 cents which is cheap but not as cheap as we want we want to bring it to bring down this number to sub one cent we think that's a price point that starts to be attractive for large master master scale application and we think is feasible with bom and 48444 in in in a short period of time and then what what do what do l3s look like on on zkn right so we've in general um so rollup is just something that exists on top of ethereum a way to see it as an abstraction is a layer one smart contract that contains all the assets and is a mechanism to update the state of the entire system and then stores the the state of the system on chain two so it's just a layer one smart contract 98% of all the complexity lives in the server the things that produce the ZK proof uh and that's completely of chain so what you can do is you can take the layer one smart contract and deploy it on the layer two and now you have a layer three um we are calling these chains in general hyper chains we think is more interesting than the dichotomy layer two layer three layer four uh the idea is that you can launch your hyper chain in whatever mode you want they are totally soverign they can be a layer two or they can be a layer three they can be a layer three on top of your Layer Two or someone else's they can be a layer three on on era they can be launched in rollup mode where all the data goes in ethereum it's a little bit more expensive but super secure or you can launch them in validium mode where some of the data all the data goes somewhere else or in volution mode where some of the data goes somewhere else and some goes on ethereum so all of this customization are going to be available to everyone very very quickly um you may want to launch a layer three if you are very very interested in super scaling so the idea of layer 3es is especially if you launch them without the data availability going going on ethereum is that they become very very very cheap um but there is a problem with all this multiplication of layers so hyper chain for us are an interesting or ZK stack in general is an answer to the story of DB chains and the story of private chains that you mentioned before if you launch your D chain as an alternative layer one you have a bunch of things to deal with it's a super complicated Endeavor but there is two problem that are very difficult to solve one is that you need to create a bridge to ethereum and that bridge is going to be the thing that people will attack and will absorb 20 to 30% of your engineering time and is a damoc class sword is something that you know when it hits it destroys the project and probably destroys it forever um and the second is you need to bootstrap the decentralization property of your network otherwise is kind of just a database and Hyper chain is fundamentally better because these two problem instead compet disappear they come for free by being a rollup rollup don't really have a bridge although we call them Bridge they're not really Bridges it's just a l one smart contract that own all the assets so we don't really we have just n native layer one tokens which is fundamentally better fundamentally more secure um and you inherited the security of ethereum so again you don't need to care about about those those problem and you can focus on uh customizing your environment for execution exactly for the right execution that your user base needs and we think that's very interesting and the same is true for private chains which similar argument is that with a private chain you know before you were going to R3 or or to Corda to to hyperledger um the problem with this is that Tech technically they're highly scalable but practically they don't scale and they don't scale because they're not credibly neutral in the argument that we were trying to to say before um but then there's a problem so if everybody has their own hper chain then um you get this massive fragmentation of liquidity different islands with no you know very wonky unsafe Bridges to connect to them maybe or you have to go to that Island first to get to that Island that's next door yeah which is which is a terrible exercise you can wrap things and you created these chains of risk and and and liabilities that are very ineffective and remind us a little bit of traditional Finance know where lemon goes and then all of a sudden the Domino starts to to fall in a way that is very difficult to predict self- sovereignty and self- custody can isolate the system from that now make it very resilient but then you require a very strong communication Network so with ZK stack one thing that to the best of our knowledge the other team are not yet building or or not yet focusing on is we're building something that is called hyper Bridge which is a general messaging system system that allow you to send messages from any of these hyper chain to any other hyper chain in a way that is very fast few minutes not days not weeks um is very cheap few cents and is cryptographically secure so it doesn't need any human it doesn't need any trust in anybody uh it doesn't need any liquidity um so you can create very interesting tooling with this and essentially unify all of these chains uh together and you know I've I'm I'm being asked a lot this well when is when you know when is my when is my grandmother going to use crypto when is this mainstream adoption coming and I think we've been talking about mainstream adoption for too many years and quite frankly I think it only now starts to make sense for entrepreneurs to come in and start building actual end consumer mainstream applications um you know with because of infrastructure like this you can get to sub one cent transactions you know that cost less than a cent um you have account abstraction natively built in meaning also that you as a company you can actually pay gas for these accounts for these users so you don't need to course get some eth or get this native token to pay the transaction fee and and um we're talking with a um a bank a very large Bank who you know wants to run this experiment of creating a corporate banking product a corporate bank account and a consumer bank account for fully on the blockchain and um especially in the consumer account we are telling them okay now is actually a good time to really start building this anything before you know what was promised a lot in end consumer adoption really quite frankly wasn't there um the last big wave of mainstream consumers came with the 21 bull market which happened you know mainly for liquidity reasons uh but then you have an a normal person that buys an nft for $50 and then spends $250 uh transaction fees and you know they're gone forever they they never come back and so um technology like ZK saying hyper chains are actually what make us bullish that in the next three years I think we are going to see some you know end consumer applications that actually can support 10 20 million 30 million monthly active users yeah and and so if I if I can you know add a couple of points everything we do is to try to create an environment that will give you the developer and your user three broad things so one is this magical ux the ability to offer magic ux come from native account obstruction the fact that is native means well first of all it's great to see that ethereum is embracing it as well with 4337 most of their all up are implementing it too we may actually add the 4337 natively as well having it natively means that all the user automatically can benefit from it so a big issue with non native account like today on ethereum if you implement some features your user will need to deploy a new smart address migrate all the asset and then they can use this new feature and the migration are are a pain nobody wants to do them if you do the same on ZK sync instead as soon as you add a a a smart feature every user without doing anything at all will immediately have access to it so that's one thing that is very important for us the second is Limitless scale so zero knowledge proof uh enable um a change of Paradigm today blockchains until now have been Supply constraint systems where they look like an island and know ethereum looks like an island or like the city of Soul once you finish the space price sky rocket I think everybody can experience that normal roll up look like the invention of skyscraper you can build a 30 floor on on an island and you get more space but at some point people come and again you have 20 million people in so price the Skyrocket again um ZK roll up instead scale much more like Google like a server if we have more capacity to throw at gpus we can scale we can scale limitlessly this is a feature that I think is very important for you to consider when you decide what to want to use to build it makes the system future proof and the third is that uh any hyper chain that you make is not an island in as of itself is embedded into a network of either chain and so you benefit from this aggregated network effect if your neighbor builds a million user you have access very very quickly very very closely to those million users and and that's I think a great thing to be as compared to build your own a very nice shiny place on the moon that is isolated and takes a huge amount of effort to reach and to come back from so you know we know optimistic rollups launched a little earlier have some especially tvl I think they're doing really well then we've got zero knowledge rollup uh rollups we already have you said you roughly 40% of all transaction is 40% of all the daily active users and a little bit less in terms of transaction right and so already more than the two biggest optimistic rollups combined I think and so in 3 years where do you think where do you think optimistic versus zerob base uh zero knowledge rollups where do you think we're going to be so I think um both the big optimistic rup provider have said that they like the concept of moving onto zkp at different speed one you know one is developing with third party the other said that they can do it anytime they want so I think eventually the technologies will will converge and the optimistic rollup will all migrate into becoming ZK rollup um all right right you've heard it here in the future it's all going to be zero knowledge based rollups uh with that we're coming towards the end uh thank you very much thank you Marco thank