ZK Rollups and Beyond: Innovations, Opportunities, and Challenges

Transcript

hello okay so here we have two great players from the ZK VM space marim and Jordy maybe we can start by introducing ourselves hi I'm Martin I'm working for matter Labs we're working on ZK sync and other ZK Technologies hello yes uh uh well I'm Jord Vina I'm the technical lead at the polygon zkm and polygon cofounder uh yeah we're building the zkm we're building the AER in polygon and uh

we are doing a lot of ZK also in polygon yeah okay and I'm Danilo previously co-founded at kakar currently building zap so we'll start with a little bit now we have a history in ZK like two years ago ZK was very new now we have two years of History so we start some questions about the history then we go what's going on and then later like okay what's the next steps um so looking back two years from today ZK have been one of the fastest mov moving technology even comparing with AI um what what what do you think was the major big breakthroughs that allow us to be where we are today so one thing I would say is

that first we should think about how much we did over last two years we in this community turn to forget about the past we're always thinking about the next cool thing but if you think about like two years ago people were saying like hey ZK proofs will take years 10 years 15 years to do ethereum fees were in dozens of dollars the same thing for the ZK l2s and look now now we're like in sense subsense how do we do it by a combination of dozen maybe even more than a dozen small improvements coming from different places the fact that this was this research was done so much in the open

actually allowed us to get where we are and thanks to Jordi of course no and yeah and I agree uh and um I would here I would highlight uh a couple of things one's more visible and one that's probably less visible in the last two years one of course is the I mean the scalability in general I mean right now building proofs really complex proof in seconds is feasible I mean some and this this is the the reality that we have and I mean we'll talk maybe later but uh things like sp1 or vzo we are seeing even that the the the even the way of writing provs is is is is getting even

more simple uh uh on that but there is another probably more hidden thing that happened in the last two years that's more I would say um a maturity in security I mean we in the last two years we understand much better how to build these proofs what are the things that can go wrong what are the the I mean the security issues that can happen even how can affect these security issues so the the the the knowledge that we collect in the last in these last two years uh running the zbm having a permanently open uh back bounty on there having engineers

looking at the code rebuilding checking um this the deepness of understanding on this I think this has matured a lot can we say that um this technology is absolutely safe and I mean you never can say that of any technology but what I can tell you is that the confidence that we are getting on this technology from the security perspective is uh every time higher and I probably I would say that this is the the one of the things that I would say and and that doesn't guarantee that the Cod is free of backs and all that but that's that's the

feeling and that's the that this is a clear fact that happens in the last two years yes but one thing that did not happen and this actually the disappointment that I'm seeing is that a lot of people still treating ZK as magic they do not know what ZK means they are lacking the ZK intuition and I'm not talking about them coming up here and discussing with us the exact details of Stark versus snarks but more like understanding if someone comes to me and says I going to solve it with ZK doesn't even make sense yeah probably the magic when when when we say that the ZK the magic part of the ZK is actually the zero knowledge

but it's probably the part that we are not using in a normal roll up I mean that's more that comes more from the Privacy I mean being able to prove a computation but without having to reveal reveal the input of this computation I think this is more magic than just proving a computation because proving a computation at the end uh there is a very easy way to verify you just run the computation again and and this runs so the only thing that we are doing here is um that instead you to have to run everything again you just get this cryptographic proof to do the the equivalent to that so intuitively it's not that difference okay it's I think it's more magic when we are talking

about privacy when you are hiding these uh uh inputs here but yeah I I get your point I mean there is I would say that the problem is that and this is we have seen that is that uh people just trust in the tech and in the ZK is but I would say say that ZK is not an exception here I mean we have seen how many people just verify smart contracts and check the smart contracts uh we just trust uh EAS scan that they are verifying the right way but do we know if if the scan is doing well I'm not saying they not but we should verify more I mean we should check more we should check more

on the the technology we should check more on how the things are done and I agree I mean we see a lack of uh uh understanding of developers and people getting more dep in this Tech uh to to to understand this this Tech but again I mean is people there is something yeah because for me it's one one part is the laziness what you mentioned with ether scan this is like our laziness as developers we got fat and lazy originally ethereum blockchain was supposed to be the thing that you know know against decentralization against this assuming that everybody around you is malicious and we got lazy let me

explain you one one one story so so when we deploy the the the smart contracts uh well actually the appr the the verification of the smart contracts for the zbm we were we publish the I mean kind of a well instructions in how to uh verify the verifier I mean how we can check that the verifier I mean there is this source code that's open you have a a comit hash and from this commit hash you go to this verifier and you have all the source here okay and I mean this verification process was not comp it was relatively complex because I mean you need a big machine and require some time and some skills but I mean there was a full process there okay um in this

script there was a back there was a back I mean there was like something that was wrong so it was possible to verify with the script we verify nobody told us I think was like after 3 months or four month on that for me that was really frustrating when uh uh on that because it's do we really have a community that we just deploy these smart contracts probably and these verifiers probably one of the most complex things and nobody just even verify that just just me just just let it here yeah I I agree and I think looking back to ZK like it it shows the power of also

decentralization and and um the web tree space this progress wouldn't have possible if it not in the web tree context where um open sourness and Community is so important I think like if it was in the normal standard um Market it would take a lot more time to go from where we were to to to where we are um okay going to the next question um okay looking now back at the journey so polygon had U the polygon Hermes ZK VM and then had the cdk and today you have the AG layer um the zync team had the

zyn era the Z stack and today the elastic chains um I I see that your your your journey has been very complimentary and maybe could you Jordy uh maybe you can talk a little bit about EG and then U marim you can talk about uh the elastic chain and so we can have some comparison and some differences in this yeah I think that there is no many differences but uh the idea here of the AER is uh to have a set of sovereign chains I mean chains that can be very different in essence I mean they have a may they may have different tokens they may be validium or zkm they may have a calite sequencer or a desite sequencer

uh I mean they may even have a different state transition function uh I mean it can be one one is maybe evbm the other could be maybe a zwm or a maiden or something that's a a nap chain with some specific statation function on there so they can be very independent very Sovereign each one and the idea is that these chains they can communicate one each other that value can be trans transer in one each other with the less friction for the user perspective and here we are talking that the idea of the gler is to be we'll see what what's going to be the level but the idea is to be uh very few seconds for transferring value in a committed way I mean in a

safe way and trustless way between the different the different chains and that's very much that of course should be open to for the people to deploy these chains and this I mean this is the horizontal scaling I mean when we were building the zkm we were thinking about the vertical scaling mean just having like a fast processor high frequency and and trying to be as fast as as we can and here is more about having like a a network of processors many many uh about different processors I mean different networks different SS with different rules with different governance rules in each one and connecting all them

together and this is very much the layer and this is what we are building right now right because the way you can think about it is the the more progress there is in ZK it kind of uncovers more and more use cases like what Jordie was saying the fact that suddenly we can do a proof in couple seconds rather than a couple hours allows you to do these fast interrupts the fact that you can use things like sp1 to prove the risk the risk instruction set means that you can suddenly have far more flexible stms so both of the companies are approaching this from a little bit different angle the first approximation and some of the BD people kill me for this is the same thing the same way how you have like Diesel and gas engines for the first

approximation it's like it's an engine it goes of course the more deeper you go we differ a little bit in the approaches on like what we prioritize which path we selected towards but the ultimate thing is that you are able to have far more flexibility than what today ethereum gives you because ethereum for kind of allows you a lot but forces you into this evm transition function forces you into ethereum Behavior this opens up the the whole new possibilities but with this the whole new set of risks that's kind of coming back to our discussion a couple minutes ago about people not understanding what is within the ZK shield and what's outside because the thing that scares me actually and this

is one of the nightmares is that you know people will think that yes I am lounching on chain Fu because under ZK I'm I'm super safe and then they going to get rug pulled and then there will be a huge storm of but wait I thought this was under ZK why did it work yeah because ZK only covered this and you didn't cover that's that's the recipe so I repeat the thing for all of you folks here is what we what we're building with a layer with elastic chain will open up a world with new possibilities but please do your research do not be lazy yeah I wanted to um ask another

question like now looking at the future as let's say like Ager gets established and there's a lot of chains and then elastic chains also get established um I feel like is this going to be um just like today we have two chains is it going to be like still two different ecosystems or do you think it later on there's going to be some kind of interaction between these ecosystems because as of now like we could say let's say each chain is adap it would if we just just go the way we are it would seem it would still be like just two ecosystem do you think that the future is interoperability between an ecosystem or even a a a one level higher of an ecosystem um that will put aayers and

cdk apps together I'm just curious on what are your thoughts on what would come yeah my thoughts is maybe maybe not two maybe three four five I mean we are seeing that these kind of constellations or are happening we will see how many it's clear that the I mean I'm not sure because so I'm not sure if there is a space for thousands okay but a small I mean a small set probably yes and um the question would be is what's the friction between this inter interon installation or uh so if somebody wants to transfer fonts from one chain in the cdk to one uh chain of the elastic chain of Zing

what's going to be the friction for that and I would say the answer here is eum uh the common the common root is is is ethereum so uh here for reducing this friction we probably here will need the help of uh the L1 uh ethereum we will have to go through L1 it can be an aggregated it can be an aggregated uh transaction is something that can can be Consolidated um quite fast but the from the user perspective I mean the problem in ethereum is that the finality in ethereum is uh couple of theas that can be 12 12 minutes or

something like that and 12 minutes for a user is a lot uh I know know that in the in the Y there is a lot of research for trying to lower these numbers and I think this is fundamental or is going to be fundamental in the future especially for these uh uh let's say interc constellation uh transfers uh uh on that so I would say that first remember when we talking about the ZK it's not only two of us or two of our companies right that's this is the ecosystem this is the community and the community is not just an empty word the thing that I predict is that

when there is a need people will find a way and usually with all the interrupts with all the bridges it's a trade-off between what's your latency what's your security what are the risks now our role is to just bring the overall like this parito Frontier like down meaning find it even cheaper even faster but it will always be this trade up because even if it's one someone will be like Jord but I want 100 millisecond one that might be insecure but I want it faster right so there's always will be these Corner cases and and and it's important that there is many projects trying that I mean the this s layer is not an easy uh project and it's very tricky and and uh when you think deep about that I mean

there are some implications so and there is different path to build that uh we've seen that for example when we're talking I mean there is different words here I mean we talking about uh um um shared sequencers and we're talking but at the end we are talking about this interchain communication uh somehow and there is a lot of options synchronous versus asynchronous uh using ZK not using ZK I mean there is a lot of variables and ways to do that so having different teams trying different approaches the ones that are doing better I think this is great for the space because the same that with the zkm I mean we learn one

each other we learn a lot from zky we learn a lot from scroll we learn a lot from a starnet we I mean each each of these networks they are adding some value in something you will take it and sometimes the value is the mistakes that they are doing I mean is like this you cannot do that way but that's okay I mean that's that's a learnings there are good things but things you see what's going on you improve your system and you're just trying to improve yours sometimes it's new ideas that comes from yourselves sometime are good ideas that are come from the others and this is the way to grow the the the ecosystem so that's why it's important that there is

a set of projects trying this this is a new thing I mean it hasn't we haven't built that before it's something that the humanity haven't done that I mean and we have the K ZK Tech is quite new now again I mean I think is fundamental but for this for this the role of the ZK uh proof for this uh interchain communication somehow is uh fundamental and key the fact that in the ZK you can prove something immediate I mean you just generate the proof and this is a proof you don't need to White for uh some gaming timeout or something else this is in the at the moment that you

are transferring fund sending something to the other network being able to prove that I'm here I'm committed it's correct go ahead use it this this ZK is fundamental uh piece but I mean it's available since uh few months yeah I I agree like um I think this um diversity makes the ecosystem more resilient just like we need a lot of uh different implementations of the ethereum node I think all this it's important to have different teams great teams implementing different um so and then talking about resilience um what do you think like today we haven't found any big bug or

anyone wasn't able to like have like hack it or or like do a big hack but I feel like this is if time was infinite this is inevitable like at some point it might happen there's a probability to happen um is there like how would that look like from your thoughts and today is there a measure so if that happen um we have like a counter measure for it or is it something that because it never happen it's it's hard to predict how to say it the way I would compare it is like you know in 1905 when they pushed out the first car going 5 km per hour there were no car accidents so

they could assume that there will never be car accidents right how they solve it with cars they started putting security measures air buug seat belts whatnot and the similar thing will be happening to ZK I'm very happy that you're so optimistic about the amount of bugs that are exist in ZK I'm less uh but I mean uh last two years we've found many bugs mean it's not something oh no it's no no we found many bus I mean the thing is that we are responsible and because I mean and I think I think I'm talking also uh in the name of ZK and all the ZK projects I mean I think that we have been very responsible in the security we have running a lot of audits

we have running all these trailing Wheels with uh some setups for example one of the things that give us a lot of warrantees is that we are uh we have a centralized provs I mean we are running the pro this is not a big thing because at the end of pro you cannot do much I mean you either can prove or not okay and actually in the smart contracts if we don't prove anybody should be able to prove and so the pro being centralized Pro here is not a big uh it's much worse than centralized sequencer I mean that's big big as the right but this C from the security perspective makes a lot of difference because these for example all the soundness uh errors um uh I mean it it in the case of

a sess error uh I mean you you you cannot exploit that doesn't mean that we are not assuming uh security risk yeah with that because uh here is important to mention that for example when there is a soundness uh we have an internal risk so uh what happen with the developer that has the private Keys that's sending the proof uh if they can exploit to that so I mean it's really important I mean having a sness free uh ZK is important but the I mean here the the training wheels this some sort of centralization that we hav't been applying in these

networks has been fundamental for not uh having to explain uh uh a disaster or something bad but there is a lot of evolution as I said at the beginning I mean one of the big uh improvements in the last two years is the confidence that we from the security perspective we have in in in in the system so one thing I would add is that the more companies we have working on ZK this also gives additional security things because we can as well easily say like okay in order to proceed you have to provide two ZK proofs from two different provs right so this slowly and steadily adds these security or you know from the panel before I don't know who won te or fhe

but like all these systems are also providing these additional security guaranties the same way how you would put two or maybe three locks on your doors even if you think that one one lock should be enough you going to add the additional two just in case the same thing here really depending on the amount of security you want to have in your doors or in your systems thank you I I think now we can go to talk a little bit more about the future of where do we think ZK can can can be in a few years and from from your perspectives what is missing for the from here the X 10x 20x breakthrough like how can we get even FAS so we got I

think maybe 100x Improvement or like a lot of X improvements since two years ago like and I think a lot of that came from research do you think research is still there's a lot to be done or or there's more yeah here um of course this is a future um here we're just taking our crystal wall and just talking okay but here what I would say is that uh uh uh at least in the Starks Starks is the most used I mean it's the fastest technology uh I think that most of the projects are migrating to that somehow and but the Starks we are touching a little bit the limit of the Stark and this is very much because the end Starks are based in Hash Hash functions and uh

I mean hash functions are quite well optimized and and and the minimum thing that you can do in a in a winess is um running a hash function which actually more or less what you are doing in a stark I mean you are doing an nt2 but but that's that's uh so we are touching I would say we are touching the the we are touching the limit uh uh very much here I think that from the research perspective I mean from new things that are happening I mean some for example some gkr in some parts of the Stark or the Venus or I mean there are some new techniques for some papers that allows us to maybe and here maybe we can get one order of magnitude but I don't think that we can get much

more unless there is some a huge breakthrough in in the tech that you never you know that never happen but I don't think that using this current Tech we can get much more that one order of magnitude where we can get two orders two orders of magnitude of on top of that is in the hardware acceleration and this is I think this is the phase that we are entering uh right now we know that there is different projects that are uh currently developing AIG and fpj and different Hardware solutions for building ZK and here um talking about at least one order of magnitude and I would

say uh reasonably talking two orders of magnitudes on top of what we have uh this is uh perfectly perfectly and again and this piece is fundamental for this aggregation layers uh uh thing because this is can accelerate a lot and that would can give this latency until now the Z case we haven't been optimizing the the proof generation for latency I mean we have been optimizing optimizing for for through put or if you want for cost I mean it's how much the proof cost I don't care if I take it more but but the cost at the end the total cost the total dollars per per proof is is slower but when you are optimizing the latency

okay I'm willing to pay a little bit more but I want the proof to be fast I mean I want to to do the proof in one second or in two seconds here is where Hardware acceleration makes a lot of uh difference this is going to be fundamental in the upcoming months uh years and I think this is going to be the big thing that will happen in the in the next months years so yeah the one thing I would add to this is look folks what you're hearing basically is that this will cost almost the same as CPU so you'll be able to ZK proof whatever you want so now this is up to you again I'm kind of repeating what I said before to

get the ZK intuition to understand what are the new cases that you can apply the zero knowledge for as Jordy mentioned before this also part of my disappointment is that we're not using the zero knowledge part of the zero knowledge we called it zero knowledge for a reason and here we are hoping that some of you maybe after this panel will be like huh maybe I should do something about that I should do it to introduce maybe a little bit of privacy maybe a little bit of this yeah I remember that at the beginning mean just maybe three four years ago there was a debate if we should keep uh using the term ZK or changing or just finding a new term for that we we we already decided to go for

ZK because actually is is the tech it's the naming that we have been using on that but also worthy to mention this uh uh this because uh ZK and this is very related to to privacy and I think this is really important uh to mention and I agree that uh using ZK uh for privacy is also is also important yes I think we are out of time so um so get off your phones learn about ZK and start using it in your next project thank you yes thank you Jord thank you much