Fireside Chat With Marek Olszewski and Leland Lee

Next up, I want to bring on two special guests to have a amazing discussion on their brand new partnership. So, this is our next and second last fireside chat where the two panelists are going to be Mark from Stella and Land from Sus. Please give them a big round of applause. [applause] Thank you. Hey guys, it's good to be here.

Yeah, we have a lot of exciting news to share with you today.

Yeah. How you guys doing today?

Everyone bring the rain. Amazing. So, as you can see, we are two fanboys today. Um, and uh, yeah, we have some pretty exciting news to share. Um, you know, we sell has been working with Sync now for for a year.

We wanted to talk a little bit about what we've been cooking up. Um, and also sink has some pretty big um updates as well that they've been announcing recently. So, I want to spend some time talking about that. But maybe before we do that, let me give a quick intro for those who don't know us. Um, I'm Eric.

I'm one of the co-founders of Cello and Self. Um, and uh, yeah, fun fact, I love Argentina. This is my fourth time here. I actually proposed to my wife down in Patagonia. So, it's really good to be back here.

Uh, and yeah, let me hand it over to Le.

And I'm Leland. I'm the team lead. You can be succinct as one of the fastest CKVMs. It makes it really easy for any developer to build CK applications.

Cool. Uh, and speaking of building CK applications, um, you guys have just launched a pretty big, um, upgrade to Sustain. Um do you want to start off by sharing a little bit about that?

Yeah, of course. So we recently released the next version of SP1 which is our CKVM called SP1 micro cube. And we've been able to prove Ethereum in real time. Meaning we can prove 99.7% of Ethereum blocks in under 12 seconds with only 16 GPUs.

And why this is so important? It means that under, you know, $100,000 of capex, you can basically prove Ethereum in real time at home instead of using a massive data center.

That's really cool. And we should probably spend a little bit of time talking about what real time proving can accomplish and why it's interesting. But before that, maybe just as a little bit of background for folks, is it worth just talking a little bit about what is being proven? uh maybe talking about risk five a little bit uh just to um yeah get everyone on the same page.

Yeah, of course. Um so I think it'll be really helpful to discuss how does Ethereum work today. So you have a builder or a validator, they create a block, they combine all the transactions together, do some computation and say, "Hey, this is a new Merkel route." And you send the Merkel route to a bunch of other validators. Now other validators now really quickly have to reconstruct the same block by replaying all the transactions.

And as you notice you have, you know, hundreds of thousands of validators all doing repetitive work. And the question is how does it impact scaling if everyone's doing all this computation? It's a lot of work. There's a lot of data being flowing around throughout the chain. So how can we scale effectively?

So with real time proving it's the model is quite different. Instead you have one individual they create a zero knowledge proof of an entire execution of set block and they just propagate this very small piece of data to everyone and now everyone else they just need to verify this small piece of data which is very very quick instead of taking you know you know seconds to reexecute the whole block. So the benefit here is now you can really scale the gas limit in Ethereum without centralizing the chain at all.

That's really cool. And then I guess the other thing that you guys do is you're you're not proving EVM opose. You're actually proving the risk 5 instruction set which means that any implementation of um of the Ethereum execution client can be compiled to risk 5 and can be proven with the new system.

Correct. So currently most teams are proving it um which is Russ implementation of Ethereum client. However, we've also done some work with Aragon, which is a C C++ implementation of UDM, which is actually like 2x more. So, we'll release some new numbers very soon about that.

Wow, that's really cool. And I think I don't know like you just said it so casually, you can improve in real time. Uh, you know, the only 12 GPUs. Um, but I mean this is a pretty monumental achievement I would say. I mean, if you had asked me two years ago if this would be possible at this point in time, I think I would have said no.

Um, not only are you doing it, you're proving again risk five, which is actually significantly harder to prove than if you were proving at a higher level of abstraction. Uh, it's easier for developers and so that's why it's so interesting, but it's actually very difficult to prove that quickly. So, yeah, I think maybe it's worth giving these guys a round of applause. I mean, this is a pretty incredible accomplishment. And you know, it's exciting because I think you can use this for scaling Ethereum, but it's also interesting because I think you can use this for interop.

Do you want to spend a little bit of time talking about inter?

I was going to say like you can actually use it to scale L2, which is really relevant to Sal.

Yeah. And and we'll definitely talk about that. Um but maybe just to finish on the interop side because I think it's it's interesting uh certainly from our perspective too. Uh once you can do real-time proven then you can um start doing atomic execution of transactions across two chains either an L2 and the L1 or across two L2s uh because you can prove the correctness of these uh two transactions in the time that it takes to construct an L1 block which means that a um block builder can actually guarantee that these two transactions will execute concurrently and they can include them in the block in a way where they know for sure that you have atomic execution between these two transactions. And as soon as you have that, then suddenly you have synchronized composibility, you can actually start having transactions or smart contracts in one chain call functions in other smart contracts synchronously.

So not asynchronously but actually synchronously and even perhaps get the response back from the synchronous models back to the other chain. And so this is the holy grail for scaling Ethereum, right? It's being able to write code that executes across multiple chains um in a way where the developer experience is similar to writing on one chain, right? You don't have to worry about it as asynchrony. everything can happen at the same time.

And so that's pretty exciting and I think it's going to enable a lot of really cool things for scaling Ethereum. Um, which is why I'm especially so excited about it. So yeah, congrats again.

Thank you so much. And talking about scaling, Celler recently migrated from zone L1 to L2 and is now adopting ZK to its core technology. Can you talk about sort of the journey of Cello and how he decided to transition to L2 and then we can discuss the ZK component?

Yeah, for sure. So, I guess how many of you have heard about Cell transition from L1 to an L2? Uh, okay, most of you. So, as you know in in March, we we transitioned um from the dark side uh so to speak to the light side. Uh and it's been really exciting ever since.

uh and you know we did it because of mostly ideological reasons. We wanted to be we started building Ethereum and we wanted to find a way to fit into Ethereum uh once the scaling road map um was created and once we saw that there's a way for us to do so uh and it's been pretty amazing ever since uh and we started as a DA L2 an optimistic L2 uh which means uh you know we started using OBS stack with IDA for the data availability uh and that was important for us because we have a lot of real world use cases on cell. We needed to keep transaction fees low and so we we felt the need for uh making sure that we were using a GPA but that complicated our ability to secure the network. It certainly made it a little harder for us to reach stage one uh security. Um, and so we've been working together with Sync uh over this last year to to get to that milestone, which we're really happy to share is is just around the corner where we're planning to upgrade our mainet to use OKNC light uh next month around December 10th.

So that's going to be a huge huge win for us on the security front. It's going to basically bring us to something that is like a stage one rollup. These stages don't typically apply to LTA O2s, but if you ignore the fact that we're using LTAD, we can think of this as being equivalent to being a stage one, it's a pretty big milestone for us. Um, it's going to be our fourth fork this year. Uh, and that's just a testament to how hard the team's been working, but also a testament to how amazing it is to be working alongside uh, other great projects in the Ethereum ecosystem.

By working together, we've been able to, I think, accelerate our velocity, which is pretty pretty great. So, we're seeing a lot of benefits from that transition from L1 to M2.

And then what was the rush develop for deciding to use this sake? what sort of before we dive into like how it works, what benefits does it give to your users and developers building within the cell ecosystem?

Yeah, great question. So, when we first announced that we wanted to transition to an L2, we got bombarded by every L2 stack out there, all the optimistic ones, but also all of the uh ZK ones as well. Uh, and we we looked at all of them. We actually went through a a very involved evaluation process. uh a lot of these folks went to our governance.

So they forced a governance vote on which stack we wanted to pick and that meant that we really needed to do a lot of due diligence and and we looked at all of the stacks and we were very excited about the prospect of ZK. We we have a lot of conviction that ZK is going to be the best way to scale Ethereum long term. It's going to be great for inter has all these benefits but we felt that a lot of the existing ZK stacks around two years ago um would have been difficult for us to migrate on to and this is primarily because has some custom features. develop compatible, but it has a few extensions that allow you to pay for gas and tokens, for example, and and other things. And we knew that we would have had to port these features to one of these CK stacks if we were to pick one of these CK EVMs.

And so instead, we we thought, hm, we're seeing a lot of improvements to proving systems. Perhaps one day in the future people will be able to prove you know some machine code ISA. We didn't know it would be risk five or mints or something but we felt that you know the improvements that we were seeing improving cost uh were dropping fast enough that we felt wouldn't be too long maybe two to three years uh before this would be feasible for something like cell to adopt where we wouldn't have to handw write circuits or write very low-level assembly that gets converted into these uh circuits. uh we could just write our code in Rust or maybe in Go and simply compile down to one of these machine code ISAs and then have the prover work on that. Now lo and behold that day came a lot sooner than we thought.

I think uh we were chatting maybe a year ago when came to our office and said hey we have a system that actually can work on uh what you need today. Uh and we were just blown away. We didn't think that feature would would come that quickly. Um but we're excited that it did. And also we're excited that we I think picked the right direction two years ago.

Had we picked one of these um custom ZKVMs, I think it would have been a lot harder for us to one migrate to ML tool, but also two to maintain our code base. Now we can simply you know maintain our rest client our rest fork and and compile that to risk five and then have susterver operate on that and so we're very very um happy with kind of how things worked out and we're just super excited that you guys were able to deliver on on that promise uh reduction improving cost so quickly. It's kind of crazy if you think yourself two years ago, the technology you see here were also dramatically different. Um, you had to write as we actually alluded to these like really lowlevel like circuits or DSLs. A good example of this is let's say you're using Excel and using Excel is very easy, but if you want to write anything more complicated, you need to use something called VBA, which is quite frustrating to use.

And so two years ago, all the ZQ rules use something similar to VBA. you know, it's very difficult to make any changes or modifications or customizations. So the path that Mary took was let's use something that's like much easier like developer experience and let's use it stack and lo and behold one year ago as a sync we didn't work and we realized like hm what if we could prove the op snack inside of a zkm and turns out this is actually possible and we built something we like to call succinct. Uh so movies to say now has been live in production for over one year. It powers currently three and soon to be four rollups in production such as polygon katana mantle type dismantle and followup.

So these first three are fully executing and also very easy to upgrade because they're all written in rust. Um, for Zello, Zello is taking a similar path, but they're using something called XDK fault proofs. And so the benefit of this system here is that instead of when there's a dispute, instead of having this like interactive game, it can be settled in a single transaction, which means your finality window to be much shorter. Maybe we can talk about that for a bit. Like what are your thoughts on reducing, you know, the 7-day window potentially something?

Yeah, definitely. And maybe just to add a little color on the system like one of the reasons why we're so excited about open sync light and I think we're going to be the first chain to adopt it in production. I'm pretty stoked about that as well. The beauty of OBS light is that you don't have to in the in the normal case you don't have to prove every block. So this is an optimistic um uh roll up for L2 technology.

It works just like any other optimistic uh rollup. The proposer proposes state roots every now and then and if there is a fraud detected then someone can dispute that uh state route and uh a game ensues to to decide who is right. Uh in traditional optimistic rollup you have these uh interactive very complicated bisection games those take a longer time. uh with ECC collite you can just construct AC proof that proves that the state transition from the last proposed state route to the current one is in fact valid you can construct this proof in around 20 minutes it's not that expensive to to create uh and critically you don't have to scale your deposit for creating those uh uh this for challenging state group proportionally with the amount of TVL and change it secures um because of the way that because of how cheap and easy it is to create these EC groups. So, it's pretty exciting.

And what's really nice about this is you don't have to pay the cost of creating the Z fault groups for every block, right? As much work as they've done to make it uh significantly cheaper and significantly faster, it still costs on the order of tenth of a cent to a cent depending on the uploads of the transaction. And for our users, we were a little worried that that would increase the cost, potentially double to 10x the cost of a of a typical SA transaction. And our users who are primarily in emerging markets might find that a little a little expensive. And so with this, we don't actually have to pay that proving cost unless someone is challenging the proposer, which means that we can keep our transaction fees the same as they are today while inheriting this extra securement.

So, it's pretty exciting. Um, and then you had a question around exit times. Um, optimistic chains are famously slow for withdrawing funds out of their native bridges, right? On the order of seven days. And part of this is because these interactive group games take a while.

Now, they don't take quite 7 days. The other part of it is because a um 61% attack on Ethereum could choose to censor the participants in this game and because of that um people have chosen to wait a lot longer and Vitalic has personally promised to use social consensus to hard fork out the 51% that's attacking the network and he says that that should be for every seven days. Now, that thinking has evolved recently. I don't know if you guys saw the Teleg post about this, saying that actually it probably could be a lot shorter. And so, where we're thinking of lowering our exit window by a fair amount for this upgrade, we haven't yet decided on the exact number, but it's certainly going to be less than seven days.

And so, that's going to be good for for our end users. It' be good for for people building on top of as well. That's really exciting for users. Like being able to deposit on the drawer like really quickly really improves the user experience. Um cell is definitely the leader when it comes to L2 adopting new technology being one the first made that chain to adopt sync light.

How do you see others evolving as time goes on as the KD becomes cheaper and cheaper? Yeah, I mean so I can definitely speak about our plans which is to eventually move to regular openness things so that we can actually improve every transaction and uh do that in real time. Um I think you guys are working on lowering your cost even more and I think that will certainly make it possible for us to uh to do full proving of every block without without the optimistic piece. We're excited about that primarily because this enables real proving and interop between auto2s and all of that full composibility that we talked about or synchronous composability that we talked about before. And we're very excited about plugging into the rest of Ethereum and just making it feel like a single a single chain.

And we think that other alts will also be able to do that. And we think that, you know, it won't be too long before Ethereum will finally feel like one unified experience.

So, everyone uses the CK and all these chains are just like somehow merged together in terms of UX. That seems like a wonderful feature.

Yeah. No, I think that will uh be the holy grail, right? Ethereum. What do you guys think? Seeing a lot of nods.

Cool. Well, I think that's all the time we have. Just wanted to thank you all for being here and yeah, come find us if you have any questions. [applause]