Panel: State of zkRollups | 14th Oct 2023 | Sunway University | ETHKL 2023

all right thanks everyone for joining our panel today um can I get a quick show of hands how many people know what a ZK rollup is okay and how many people know ZK evm Okay cool so I think there will be some people who are not very familiar uh so I think we can jump uh straight into the panel with that uh in mind so I'm showing uh on the chart I'm not sure if the panelists can see I'm showing this very popular uh chart that vitalic created a while back describing uh different uh ZK EVMS but before that can I get maybe one of you and then the other can chip in just to give a quick explainer about what a ZK rollup is and then the other person can maybe extend that by explaining what a ZK evm is and what the different types are um maybe I can go first okay sure um maybe I'll first like introduce myself first so like uh my name is Kang uh I'm from Tao uh a ZK rollup um a layer to ZK rollup so yeah we are type one ZK evm so um as you can see on the chart there is like um type one type 2 2.5 three and four right so Tao is uh right at the top there with uh being type one so type one is uh ethereum equivalent that means we don't do any like uh code changes to the the the core protocol and yeah so what this means for developers is that whatever they um Deploy on etherum they can also Deploy on Tao without any like code changes to their smart contracts yeah maybe gu for those who've not looked at the type um there's somewhat important the a good mental model to build is that uh type one type two type 2.5 to some extent all those three don't uh require developers to change any line of code so if they've written I don't know a few thousand lines of solidity to write their protocol they can just copy paste it from L1 to a zero knowledge rollup onward so type three type four requires some amount of change depending on the Chain but this is looking more like Z Cas Ina um but overall um the outlay of the Outlook of the Z kvms today so scoll Tao polyon Z KVM kakarat linear they target a level of compatibility that enables any Dev to just write their code and not worry about the underlying roll up this is kind of how it should be then Tao being typ one has a bit even more compatibility is being compatible even with the nodes this is more like there there are many personas right there's the dev who writes solidity and then de deploys to a chain and then there's infra players so being type one you can kind of plug and play ether scan you're supposed to be at least or able to just use gu or W which are the software to run a node so this is nice and others at qvms don't allow for that but let's say for there it doesn't change anything okay um just to rewind a little bit uh because I think we skipped the question about ZK rollups for those who don't know what a ziki rollup is and why it matters to ethereum or just to blockchains in general uh can I get uh maybe can to kind of give your Eli five explain like five answer uh or at least maybe 15 for the people here to maybe understand Z they didn't know anything right so to know about why there y ZK rollups exist you need to also know about the the other type of rollup which is the optimistic rollups so the main difference between optimistic rollups and the ZK rollups is the proofs right uh a proof being like how do you verify if a block is uh valid or not an L2 block is valid or not so uh optimistic rollup um just optimistically believes that all blocks are valid until someone submits uh uh fraud proof or say uh like for example oh I I want to challenge that this block is actually um invalid and you will have to compute a fraud proof to um argue that this log is invalid but for uh ZK rollups we actually um validate every single block um such that we know that every block is is valid so instead of believing that every block is valid we we submit a zero knowledge proof to prove that um to prove the validity of each of these blocks yeah yeah I guess the the main attribute of a rollup is to save money on computation so that's kind of and then the ZK part is you save money but you don't compromise on security that's kind my so for any Deens who spent like upwards of 50 ether uh to make your transactions over the last couple years that's you probably understand why rollups matter uh so I think we kind of um so Kang mentioned Tao is a type one rollup uh so for yes for Kakarot like what type would it be so Kakarot can Target type two type 2.5 or type three this is really a trade-off on on what you want to achieve so for example if you're type one you can prove L1 blocks but this is maybe a different like activity so for kakarat I think we're going to Target type two which is essentially you have compatibility and almost equivalence um so you have the same gas accounting you have the same you don't compile um evm by code you run it so you have full compatibility on the language level you have all the pre compiles uh the only thing that changes is maybe the underlying data structure so um the the mopat that evm uses is MPT for with ketak as a har function you may want to swap ketak out with posidon pesan to save money so this is what scroll is doing this is what we're doing um it really depends what you want to achieve and in the long run in the short run in the long run I would say everyone wants to achieve type one why because in the the long game is to have this kind of standardization of all Z kvms in the short run we found out that proving kak and all others non non non ZK friendly things are very expensive and so as a trade-off you can you're free to roam around the types in some sense the only type that's a bit harder to roam around with is type four because what you're essentially doing so yeah for the mental model type four is really you take solidity and you compile it to your own language all the other types take solidity and other evm language Hof Viper ex compile it to evm machine code and then run it so as long as you have this you're kind of free in some sense to roam around and choose your trade-off do you want to be cheaper do you want to be faster do you want to be more equivalent all all of this comes as the price and I think uh one thing we didn't mention is that you're a founder of uh kakarat so I think a very interesting kind of question to ask is like um what was the back story for kakar like how did it come to existence and why did you choose to uh you know be this uh type two to type three kind of rollup and why build on Ciro VM which maybe you can explain a little bit more later but yeah so like what's the backstory like how did it end up existing okay so the back story is we're almost at the birthday because it started in October 2022 as a hackathon project in Lisbon in E Lisbon actually um the whole point and I was new to k as well I was not so new to zero knowledge but new to Koo and the promise was whatever you write in Koo so whatever targets the km as the the back end is provable so what sh one of the inventors of the km I think Dano hinted at it during his talk essentially if you write the instruction set so if you write the evm if you write an evm uh in goang and then add some RPC you have gu which is the most used eum node but if you write the evm in Koo what you get is a provable get or like a provable evm so whatever transaction you you run so if I do a erc20 transfer I can prove it so this was the dream it was weird and and it the dream was you can create a zit KVM fast without being a cryptographer this was kind of the the promise then everyone gathered in the hacker house uh and the ha house with 80 people I think we lived together for a week I think seven people were hacking on on kakarat Danilo Clon fresh pizza who to this day remain remains unded no one knows his name so yeah so in two months we were able to achieve 100% up code and that's that's kind of my second point is the fun part of Koo is it's quite a high level language it's as as hard as it was then what you're actually doing is and I don't know if people know zth here but doesn't matter but essentially what you have is um kind of a intermediary uh VM so you [Music] you different ways to write an evm like too Tao uses circuit so their team has cryptographers our team doesn't yet so I would say it's it's easier in a way to build with Koo and Kaka so so the back story is we we we gathered at a hacker house we hacked we saw that it worked two months we had a working evm then we kind of like solidified from the company and said okay let's see how far Koo and this approach of a zit KVM through a intermediary Zer knowledge v machine takes us and uh go to production with it so it's literally been like a year only yeah okay cool uh so uh next one's for Kang uh so you mentioned uh Tao uh is a type one rollup um so as elas mentioned earlier like a lot of of other projects are starting with like kind of the lower types you know the type 2 2.5 whereas Tao is like going all the way straight to uh type one uh what was the rationale there for like making that decision and what were some trade-offs that you had to make so actually the previous uh the the founders of Tao are actually X um from from Ling so Ling is actually a application specific rollup so maybe it could be classified as a type four or maybe yeah type four yeah yeah that's not a general not a generalized uh rollup but yeah but so from there we just decided to move on into a general um General ZK roll up I I would say and why why type one um we believe in like um minimalism to to the the developers we we want them to have like as like literally zero code changes like um when when deploying on a on a a rollup so that's why we chose to go for like a type one ckvm and were there any trade-offs that you had to make when you decided to do uh type one yeah so type one would mean that we have to uh cover every single um operation I guess um in um we have we have to write every single operation in uh as circuits to complete the the zkm that means the act more like the expensive functions like your KCK 256 we also have to um uh write those in in circuits and this would mean that our proving provs would have to spend uh significantly much more uh resources to prove a single block I would say and um but this benefits the developer because they don't have to um care so much much about uh H H cases yeah so it might seem small but for example if your um if your um K 256 hash function um or yeah if your K 256 hash functions replacement actually cost more gas on other ZK VMS um existing um protocols that use the existing hash function function um might actually exceed the block gas limit causing your transaction to like fail and so your depth might actually fail to work on the um yeah the other type of uh ZK VMS yeah okay cool uh so shifting a little bit uh so my understanding is both rollups are in test net uh at this point we um close to test net okay close test net okay well still test net then uh so I guess uh the question would be when minet or what would be the road map towards getting to minet um yeah maybe I'll take this one first So currently Tao is on uh its fifth iteration of test net U called jolia um up to minet which will be about U Q2 next year we we plan to launch um with the Proto sharting up up up upgrade in on ethereum um in our next test net we plan to um uh release like this contestable rollup um as part of like a multi-prover rollup so we want to not only um use ZK rollups but maybe we won't want to try um in in including yeah M multi prooof like some some blocks could be like uh proven optim optimistically and maybe every hundredth block we submit a validity proof so we don't have to spend so much resources on um like submitting ZK proofs and that might like make the blockchain faster I guess yeah yes um I guess it's always hard tricky to announce me inet uh our goal basically the steps we need to follow are um pass all the eim foundation evm tests so so there is kind of a the evm is an instruction set so you can you can implement it then is since this this the centralized what is the authority that can look at your execution or code and say okay this is an evm like what makes it an evm so my belief is that if you implement all the official tests that all the clients run but you adapt them to your Z ke VM kind of context uh when you pass it you have a working evm so you feel kind of safe to kind of go to an audit go to test net go to main in that so we're finishing that up now so I would expect Q2 next year that at least we'll have I mean at least December we'll have much more clarity that then I can look back and say okay I was wrong to say Q2 or I was right so I will say I think that I'm right to say Q2 now and I reserve the right to be wrong I think it's Q2 for for both next year okay cool Q2 minus one day hopefully hopefully um and then well the topic of the panel is state of ZK rollup so I don't want to like zoom in in just Kakarot and Tao um one obvious thing uh that we see is that right now when it comes to the actual Market uh pretty much I think uh L2 beat says it's like 85% of the market is using optimistic rollups that's arbitrum optimism based um from your points of view do you see um when do you see if ever uh ziki rollups would uh catch up to optimistic rollups and what would it take to get there but for example for on the statistics CU um yeah optimistic base and arbitrum have 85% of market share in terms of tvl but in terms of activity the top three is etherum that King stet for example in terms of TPS so then arguably you're asking okay who are making those transactions are they real apps or air drop Hunters I think part of everything but you could argue then that other rup should attract also airdrop Hunters um because optimism has uh retroactive funding so any any chain but so I think that's that's the first point so I think Optimist rollup have the Dominion because they've been here first and working first and partially because they launched an evm and said we will have grps later which is kind of really smart because they had the centralized sequencer they know that everything works and then people trust them so I think it was nice and now um after three or four years optimism has thought proof or is going to and then they will eventually incorporate zero know proofs that's kind of for the thing um the state of the market is z optimistic roll up dominate but ZK rollups are on the rise and I think the only factor is how fast how cheap and how reliable it is so we've known Z kvms and even Z knowledge robs to have pretty bad ux over the past year because the systems were kind of like coming to fruition mainly because you cannot launch a zero knowledge rollup and say I will have proofs later doesn't make sense so it was a it's it's how it is but I think very very soon so in the coming two or three years the landscape is going to completely change we're saying this because op stack are all already looking for proof and I think it's going to it's going to look like that anything to add K yeah so like whether ZK rollups will catch up with optimistic rollups I think it's just a a matter of time um I think first of all you got to have like good good depths on on the ecosystem so I I personally feel like why arbitrum blew up was because of GMX and um how it supports like high frequency Perpetual trading and I think that attracted a lot of Traders and does like a lot of tvl um I think and I think on ZK sync era there is a protocol called like Maverick protocol it's like a concentrated liquidity um protocol where you can set like bounds um and and also helps to shift your your liquidity according to the direction of a price so these sorts of like Innovation that help to solve problems and um help to make like Capital uh uh help to make Capital more efficient I think these are like like key factors to attract um more more users and uh liquidity to your to your platform especially in the long term like in the short term yes you can you can come up with like airdrop campaigns and and stuff like that and and the actually the success is the success of these campaigns are actually pretty big like even like a a Galaxy task which is like a so so Galaxy is like a a platform where you if you if you complete certain onchain actions or like follow Twitter blah blah blah you you get like certain points or when you complete these quests and actually for for Tao it it worked pretty well um in our third test net we launched a Galaxy campaign and it like caused our user base to grow by like two times which was crazy there was like two times more work for for us like in a short period of time but yeah so yeah back to my original point it's all about the the the apps that are are building on top of the in the ecosystem yeah okay and um so I'll just ask one specific question to each of you uh before we open it up to the audience if you have time uh so maybe first for Kang um Tao's uh been kind of like talking a lot about uh based rollup uh can you explain uh to the audience what a base rollup is and why Tao is doing that so like a base rollup this term was coined by um if I'm not wrong like Justin Drake uh so basically a just uh a base roll is a rollup where the L2 blocks are sequenced on uh on a L1 um and why we chose to do this is is also part of our per permissionless uh philosophy uh in order to have like a permissionless proposal and block uh proposal we need to uh be a base rollup and why a permissionless rollup is important is because we want to decentralize uh m we want to make sure that people can contribute and people can like earn fees from from the from the network and not just the the protocol yeah okay cool and for alas um so kakarot's building on top of KIRO VM um some developers have told me that one issue they have with Cairo is it keeps changing over time although apparently that the rate has slowed down um so maybe what are some uh challenges uh if any on building on Cairo and at the same time what are some benefits of doing that right um I think it's true that there are many Breaking changes very often I was talking to the one of the creators of Koo earlier and he was mentioning this trade-off between stabilization and and Innovation so it's hard so if you look at for example the forums on stet people are saying hey guys my application broke because one of the API changed in the latest patch are in the latest upgrade and then right next line someone is saying hey guys why why isn't this done already like a new feature so people are asking new features and people are asking for stabilization so I think it's hard the challenges for us is that we migrate often and also we have to think ahead talk to the teams that are making so we talk with starware talk with Stark net Foundation talk with the people that develop sdks um kind of voice are concerns what is nice is for example Stark Knight and then KO is able to kind of learn from the um um limitations of evm and then build on top so like good examples were Native account obstruction multical um and also this is also not known by many people but in St in Koo and more precisely stet there is this system of class so every contract is firstly a class and then you can instantiate your class so it means that for example in evm when you deploy erc20 that that's the standard every time you deploy the same so you blow to your chain with similar code so you duplicate duplicate in stocket actually what you do is once you declare the erc20 standard and then whenever you want to deploy it you you kind of tell the chain please new erc20 so it's kind of like more Lin so all these Innovations come at a cost which is uh instability the first three years so we're going through that process at first when I started Caro the language looked like assembly now it looks like rust um when we started the TPS was .

3 now it's 100 it's moving forward it's breaking things we are annoyed some days so yeah the K Dev is very annoyed one day on on on two and then the second day they're so happy because we have generic types or like so many things so and and I'm writing Koo so that people can write solidity so in a way I'm like um double brain and like betting on Koo but also I'm such evm moderately Maxi but I think evm is going to dominate the space mostly through Innovation so I say yeah it's a bit of Dr ja and Mr Hyde every day thanks I think that's very enlightening um I think we probably can have time for one um audience question um so if anyone has any please raise your hand um the first one gets to ask the question all right I think you have probably answered very well I oh okay we do have one question uh yeah can take this mic hello good afternoon thank you very much for your talk uh was enjoying it very much so listening to you is there any use case application or service that you are more excited about in the future using a ZK rollup at least from what I'm seeing I really enjoy um account abstraction on rollups because at least for that Cas Ina and kakarat and stet they were able to incorporate into the protocol account obstruction what it means in very very concrete terms is my wallet uh uses my phone secure enclave and not a 12 12 w seed so I never need to write somewhere 12 words and I can never self-re myself which happened quite a bit so this I enjoy for example so so this is one of the things I enjoy and I I think 4337 so it's the adaptation of account obstruction in main net etherum is good but it's not good enough so this is kind of what I'm excited about yeah I think to add on like account abstraction will open a lot of like doors especially in in terms of like gaming um if you combine account abstraction with low transaction fees you can actually have onchain games where users don't even have to realize that they are um transacting on chain and and like with with the benefits of uh decentralization and ownership of assets um yeah that's probably what I'm um looking to see not only on Tao but like in in web 3 in general yeah all right so I think that's about all the time we have uh please give a round of applause to the panelist thank you thank you all right thank you to our three two speakers and our one moderator we really appreciate an insight into the ZK ecosystem all right