[English][Pop-x HK] How Does zkVM Empower L2s

Transcript

This is like a smaller group. Um, I hate talking for 20 minutes straight. I'm not that smart. So, if you guys have any questions throughout the conversation, um, I much prefer you to like learn about the topic and like for people to understand it while we leave today. Um, so feel free to ask your hand and we can, you know, stop and talk. I much prefer that sort of style. Um, quick introduction. My name is Jacob Everly. Uh I'm the product lead for the protocol that rece uh building called boundless. Boundless is really focused and we'll talk about it more but we're really focused on bringing ZK to every single chain. Um you know in the past it's been very hard to build CK protocols. We had this really nice thing called a ZKVM

where you know you can write Rust and sort of generate ZKBS u but still very hard to deploy. So we sort of want to bring the infrastructure necessary to deploy these applications um to every single chain and make it as easy as just coding high level language like you're ready to go. Sort of like a real quick show of hands. Does everyone know what an L2 is? Good. So L2 is you know sort of this thing to like scale up Ethereum. Does everyone know what ZK is? Good.

very ahead of mature. Um, so you know, CKL2s have been around for, you know, three or four years at this point. I would say they've they've matured quite a bit. Uh, and when you saw this like first um iteration of CKLUS, they were built these things like circuits. and circuits are something you know we've helped sort of like nullify the existence of the the analogy I always use is if I ask anyone in this room to code the game of chess you can probably ask AI to do it it's not it's not very difficult um the problem with circuits is we have to represent programs as polomials um and if I ask you to represent if I

ask you to code the game of chess but um you need to represent presented as polomials. It's very difficult to do. Uh you need a PhD in cryptography. I was actually doing this at one point. I quit ZK because I thought it was impossible. Um I was like this is going to get no adoption. You can barely do anything. It's very slow. It's very expensive. Um super the circuits um then cause rollups to sort of have this EVM equivalent style. Um so to make the circuits faster, you can sort of make them small.

Um, and what people like ZKYNC was doing was making it much smaller to make it faster, but then they weren't fully EVM compatible. Um, so you can imagine that a smart contract running on Ethereum. It actually doesn't work on CC. They actually congrats to them. Two weeks ago, they're finally full EDM equivalent. Um, for the past four years, it hasn't been the case. there's been really great limitations on just how much you know execution what the gas and and all of that created this sort of like developer fragmentation um uh you know like foundry which I think is like sort of the superpower um doesn't work with a lot of these L2s so now you're asking people to you know you

sort of have this really nice network effects with the EPM um and you're losing it and you really saw sort of like the stalled adoption of of building on L2 over the past year. That's where we come in. So um what we sort of did is we wrote this circuit. This is very handy data that I can get into but we wrote this circuit and the circuit looks like a CPU and what does a CPU uh do? Just executes arbitrary bite code. So now for like any user you don't have to write a circuit anymore. you just compile down to some you know common bite code like we use 32 plus 5 which is what uh Rust is and you

can just generate CKPS out of the box so not only do you not have to write circuits but you also inherit a bunch of really nice libraries um you know Rust has been around for like 10 years 12 years something like that um but now it's sort of like the foundation of cryptography of the world um you know it was sort of started in this niche thing but even like all the biggest tech companies are all changing. So, Facebook, Amazon, Google, um they're all switching to using Rustolegraphy. So, it's a really strong um foundation of libraries, which is just really nice for you because imagine you want to do a catch hash or you want to do um some sort of like curve. You don't have to write the

library yourself. You just, you know, grab off the box and do the AI language and work. Um, in the ZKVM, we sort of I think this is the the key piece I would pay attention to, which I don't really illustrate here. The executor, so we sort of execute your program once. Uh, and then you're going to pass this process off to like some third party prover. Um, and a really nice thing here is we can sort of prove that that prover has actually run your program. So you'll just get um that tiny receipt at the end and sort of your output. And and that's how roll-ups work. So in a rollup we have sort of a state transition function um and the pro can just sort of run this function can you know go through all the

transactions in a block and then just post the proof on Ethereum update the smart contract bridge and you're done. If we get back to the point of you know Ethereum equivalents and being able to use normal code there's a very popular Ethereum client called breath uh which is actually written in Rust. So when we build a CDK level which say um uh we just run the client straight up. So we get full EVM equivalents it's very easy to maintain. We know you maintain it does um and you just get a fully EVM compatible L2 out of the box.

You don't have to build anything, right? Like ZKC, Starkware, um, Scroll, like these companies raised hundreds of millions of dollars uh to build this thing and now you could deploy one in and out. Um, so the abstractions have just gotten so much better past few years. Cool. Hi. Now for a long time uh there was a drawback. Um we you know I I talked about sort of circuits earlier and how they were sort of being more precise to lower cost. Well now with us we're inheriting a whole instruction set

which can be quite expensive. Um when we proved our first Ethereum transaction two years ago it cost$1 US. That's very expensive. and we want to do thousands of TPS, thousands of dollars per second would never work. I can happily say the number is much larger. We can now do over 8,000 8,000 transactions for $1, which is cheaper than any blockchain um under any sort of load. So now we can just do right um much cheaper execution. Um and you know how how have we sort of done to do this with circuits? You have to change them all the time. So you can't really optimize them. Um with the

ZML we have a very stable platform and we sort of get compounded gains. Um so we bring in GPU acceleration. We optimize the circuit. We already have we start to bring in like you know GP or A6 and FPGAAS. We can just continuously drop the price. We've been around for like three and a half years. We haven't gone a single quarter um without a 4x gain. Um, and we have some stuff that we're very excited about this year. We just launched V2 of our circuit. Uh, we're working on D3. We're working on a new system. Uh, you know, we're working with some really popular ship manufacturers in China to build A6. So, the cost is just going to

continue. Yeah. And the next really cool thing is is that middle point um with with current and CKL2s to generate the block is is quite slow. It can take anywhere between 10 minutes and an hour. But with us since we sort of have this general purpose stable platform we can scale very very very wide um just by adding more machines and the machines they don't have to be these ridiculous things. I think to run a to run a prover on Polygon, it takes over one terabyte of RAM, um 200 something cores and like a 4 terbte stripe, right? I don't know where I would buy that machine. Um that's

that's quite it's quite pricey to get a 256 core CPU right now is just like ridiculous. Um so yeah, I'll stop there. Are there any questions? There's no way. Is that clear? Yes, that's clear. Um, and we we see the results of that. So, we actually uh did a demo uh we ran base as a CK roll up and as you can see uh the spreadsheet. So, I I said uh you know roughly two years ago we were at $1 per transaction and we're currently at uh almost

one,000 of a cent which is a um we can do this like triple you don't need that many GPUs. You need like 30 to 40 u to like really keep up. So it's it's just it's just coming so far it really makes you know operating these rollups so easy. Um and it's something nice with payload is you know we can sort of deploy this in a few hours for um you know I sort of mentioned this you know earlier but like the the costs continue to to plummet um one thing that I'll talk about a bit is you know we're sort of building an open market all the proving currently is done by us and we use AWS on demands on demand is very expensive

we've you know done some testing with like onrem hardware current data center and we can reduce the cost by another 10x. Um and then additionally we are going to get accelerated hardware. We're talking with some big ASIC people right now. Um once we move to our new proof system in a few months we'll finally be able to build one and the new system will be a 30x reduction. The the ASIC will be a 10x reduction. So you're just going to continue to see the cost. Sort of what we jokingly say right now is cost has been a really big tension point in in ZK systems. And this is the last year where we're going to talk about um it's going to become like sort of small what?

Yeah. To like uh do the the the ZK proof to move the chain forward. Okay. So like this is not including the onchain cost of like actually doing the proof on the chain like verifying. Yeah. So that's a good point. This is execution cost. Um, cool thing about boundless is we do like native aggregation. We can actually get a single proof to request the proof, someone accept the proof and then verify the proof is only 35k EVM gas which is it's pretty much almost a transfer. It's it's it's very tiny. Um, we we we've developed a library called steel where

we can use events really easily. now. Um, so yeah, we can get it get it very cheap. That is with like sufficient volume to be fair. The volume won't be there day one, but I think in six months that's your onchain footprint. Um, which is bonkers. There is an additional 20k EVM gas there to update like your smart contract, but that's the world of design. Can't fix that. But but still like quite cheap, right? 35k EVM gas today is probably 35 double check that very good

question. Anything else? Yeah, that's a that's a good point. um team's a little interesting right now because um um you know you have like a veil and Celestia and Iden DA and none of them have hit really hit price discovery um so they're sort of like you know how like in EVM you have like a gas target um well they have like a DA target and they haven't hit it so maybe DA 120 is more expensive than CPU um but I I know right now like D is pretty much free as well like I celestia and

avails. Any more questions? Um so we talked about cost. Um then the next big thing we're sort of after is is latency as well. So has anyone heard of real time proving here? Sweet. So it's going to be a really hot topic. You want to be like cool and in the know. This is going to be like a big conversation. Um it kind of goes along with uh native rollups and whatnot. Who's heard of like someone mention asynchronous composability between rollups? That's not me. Okay. Um uh so yeah, so this idea that um weirdly

enough L2s today sort of view Ethereum as like a black box. It it seems weird, but like like smart contracts on L2s don't really have have a sense of like what the current L1 is. Um and this is bad for obvious reasons. Getting to and from the L2 is quite annoying. The additional thing is you know Ethereum is about scaling with multiple L2s and L2 to L2 communication is even more complicated um and not not very possible today and you know some immediate impacts today is like that to bridge from who who's bridged from an L2 to an L2

recently what what you used to bridge okay so you probably paid some sort of fee and it was probably pretty snappy here like you know them across budgey and it's quite good but the thing is is that fee what what are you actually paying for right so you sort of have this third party who's willing to put up with your capital and they sort of deal with the 7day delay it's actually very expensive to these so each of the big rollups today are bleeding hundreds of thousands of dollars a day to these bridges and right this is just getting sucked out of their ecosystems because their their users were losing money right and the users want to go to different roles to do different Um and with real time proving that actually changes. So um real time

proving is about proving an in each block let's say 39 gas under 12 seconds. And what does that allow us to do? Uh let's say I want to transfer from OP to base. Okay. If I can prove the OP block and the base block under 12 seconds and then submit it in the same block on Ethereum, I can essentially have atomic swaps between these worlds. And this is really nice. We have to stop paying these bridge people. Um, we have to stop paying, you know, large market makers for capital opportunity costs within these systems. Um, so it'll be a really nice solution once our ZK system is there to just finally be able to really

easily send messages um between these rollups. And I guess what we're not that far. Um, today internally we've done a block under 30 seconds. Um, you know, I say we constantly work on these like 4x improvements. I'm not we're not guaranteeing by the end of the quarter, but one of the internal teams right now is really focused on this problem and we hope to be to real time proving in a practical sense um by Q3. I will say be careful uh someone's going to claim real time proofing before it actually works.

Um so until someone does it in production, it doesn't count. Uh is what I would say. Um but yeah and and again we get this nice property because we can just scale very very wide in our system. Do you have a question? Yeah. Any questions about real time proving this a hot topic? There's a research paper that Justin Drake wrote about native rollups. Fantastic point. So I I honestly I don't like the name real time grouping because real time grooming to me is like oh you're like uh the real time of like the execution program which we sort of can

do right now. Um I'd be much better if they changed the name to like ETH real time proving but real time proving is specific to each blockchain. Um so Ethereum it's actually not 12 seconds it needs to be like 9 seconds because you have to do like propagation of the block and get everyone sign it essentially. Um, but yeah, so Ethereum is 12 seconds. If you think of it like on a chain like Salana, it's got to be much harder, right? They have a block every 200 milliseconds. Um, Salana is less relevant with this exact goal because their rollups look so much different.

Why do you have a roll up on Ethereum? Storage is super cheap on Ethereum. Execution is very expensive. On Salana, they have the exact opposite problem. Execution is very cheap, right? Right. When I go to use radium, I click a button. But storage is very very expensive. Uh actually to mint an NFT on the Salana today cost $15. Yeah. It's very very expensive because uh Salana is really fast because they sort of keep all storage and RAM at all time. They have to make people pay for it. Um or they'll run out. It's also true.

Bitcoin. Bitcoin is Yeah. I mean, we're for Bitcoin. Um, Bitcoin is interesting in its own perspective. You can actually verify a proof on Bitcoin now using BitVM. You sort of do it across three like three. Um, you know, we're working with like five or six Bitcoin blowups. Uh, at this point um they can all like sort of aggravate the proofs and change it and have synchronous composability in a sense. There is a problem of like once you desync the the roll up block time and the settlement block time you get some sort of get in right you're still constrained by the the settlement layer but yeah

yeah but it's very slow right because the Bitcoin block is or the Bitcoin finalization is like 60 minutes so you have you have composibility within 60 minutes that's probably feels pretty slow versus Ethereum is trying to get to they already got a 12cond block. they want to do single slot finality and then once you have like single slot finality in real time proving um you sort of have this like asynchronous group of problem that I mean people have been working on this for like three years now it's been a it's been a while a lot of propos awesome question um now I say all of that you know our system works today but I would say it

works okay and we want to make much much better. For example, if you if you want to deploy a ZK system today, you you're relying on a single party. Essentially, you got to go to an individual, shake their hand, and say, "Will you generate my proofs for me? I don't like this." You know, it's what we offer people today, but but we we we can do better. Um, in some cases, it's it's still too expensive. Um, you know, we're always in this hot pursuit of of drop prices. Um, but some people want to do some crazy things. a partner the other day. Um, you know, an Ethereum block for us like 1 billion cycles. Someone asked me for a one trillion cycle the other day. Well, that's kind of expensive, but in six

months it could be totally gone. Um, deployment times are too long. You know, we sort of looked 12 months ago and every single one of our partners was building a prover network, was building an aggregation solution, was building verify our contracts. Uh so we want to take that away right rust deploy production ready ZK applications that's what we're exactly what my project is um so balances is is the universal zk protocol um and it's sort of made up of these three components so we have two parties sorry you know we have two parties here it's like the the protocol the app developer okay and then we have proofers

and then we connect them through a market which we use smart contracts um for protocol development, you know, the promise we want to give you is you're going to have the best verifiable execution environment. Um not just today, but for tomorrow. Um so, how do we do that, right? First of all, you know, we're starting with the risk zero CKVM, but there's other ZKVMs out there. Um there's some great teams working on this technology. Our CKVM is not the best for everything, so we want to support multiple CKMs. Um, additionally, we want to support, you know, uh, multiple proof systems. Um, we're working on binius right now. Binius is going to be really good for very big

proofs. Binius is not good for proving on an iPhone. Um, we think there's value in generating big proofs. We think there's value in doing client side proving and sort of privacy use cases there. So we want to be able to support it all and we want you as a pro, you know, an app developer to not have to worry about choosing a protocol based on the best or the like a zk. We just have the best offers for you and you know again write rust deploy zations. Um add additionally we want to go past rust. Um we're going to support some other high level languages in 2025 but we already support a bunch of popular you know other blockchain EVM. So we can support EVM. You can actually run solidity inside the CKVM today. We have a library

called steel where you can do this. Um you can run salon virtual machine, you can run move VM, you can run bit VM or you can run you know some other alternative VM. So you know there's different reasons. So you use different virtual machines. You have those reasons. I want to support you. Next is is is this idea of a proof of market, right? conduit between you and the best infrastructure providers out there. So you know you're getting test. Um so you might have heard who's who's heard of a prover network, right? So provers um maybe you've heard them in the past 12 months. They they come up with mechanisms to guarantee proof

generation. But I ask for a proof um and you're you're going to generate. I want to go a step farther. um I want you to ask for a proof and I guarantee you that I actually deliver it to your protocol and I don't get paid till then. Um that's exactly what we do about this. We we colllocate a smart contract to your application. We sort of have these market contracts on on each of these chains and the prover doesn't get paid till they give you the proof. So this like you know this is why they're incentivized to do native aggregation. Um um and if you as a developer if you're sort of using proven network you got to grab it you got to then verify it have to pay for verification it's additional risk for you to do that it's also just

harder um I want to get rid of stop thinking about infrastructure just focus on the application next is that there's no single points of failure at day one when we launch it's fully decentralized if my company risk zero dies day two It will carry on. I'm sorry. Yeah. Yeah. Very uh hello. Um yeah, if we die day two, you'll still work day three. You don't

have to worry about us existing. Of course, we're going to continue to improve the protocol um you know through governance proposals and whatnot. Um but we cannot afford to launch with training wheels. You'll see a lot of protocols um launched today and you know there's a multi-IG or there's um a small group of people that can sort of revert changes. Um currently we're going to power who's curve layer. Yeah. So layer has this function in their token essentially where they can slash that operator um and we're powering they're using our tech to do the slashing essentially calculations this very large calculation you could never do on chain and the

problem with them so we look at rollups um if the proof shows up if the proof doesn't show up the chain just dies okay everyone's mad why is my chain down but it'll boot it back up you're not at risk with ID layer it's Um, with Ian layer, if the proof doesn't show up, the funds are at risk. And when you have single points of failure, right, you're just inviting things to get hacked and targeted. So, sort of throughout the whole protocol, there's only a one of uh trust assumption, meaning if one person is online, your protocol and your proofs will be generated. Um, next is, you know, we offer the best latency. we can

essentially have your proof delivered in or not delivered but accepted and then generated um in the next block. Um we use the best price discovery mechanism uh that can be done in a decentralized uh environment. Of course the best price discovery mechanism is uh the second price auction. You can't do that in blockchains because everyone sees each other's bids and then games are played. Uh so we use a reverse touch to give you the best price, right? that's all you need to do. Um and there's no redundant work. Uh so on the market uh when I submit a request for a proof um someone's going to lock it in and then only then only they are allowed to um generate that proof. So you know ZK

proving can be expensive. It would make it much more expensive. We were all racing. So has anyone heard of Alio? Alio was like a CK blockchain. They did a ZK proof of work, but they can't really generate proofs that well because um it's just way too expensive because they have everyone racing to generate the proof. So the throughput is that and then as an app developer, you pay more money which I don't like. I want cheapest proofs possible. Um and this last thing is sort of something we call proof sovereignty. Um if you have to use other proof generation solutions, um you know, you're probably introducing trust assumptions. There might be a bridge to transport the proof to you.

There might be another blockchain in a validator set um that now you have to pay for actually um and we don't like that right smart contracts work. I colllocate them to you and we don't introduce uh any trust assumptions also if you're will you can run market contracts uh on your chain. Um and then you really get this sort of sovereignty aspect. Lastly is, you know, I want to work with the best provers, the best, you know, cost to again, not just get you the best price, but um get you the best quality of service. So, we're on we're on test net right now, uh and uh we have 10 to 15 rivers in test. These people are securing mining billions each

year. Uh so uh have you heard of like Bitmain? Bit main they make the A6 for Bitcoin miner. They're on our test net. We're working with people that are the biggest validators on Ethereum. So you sort of know that you're going to get this best quality of service. People have tried doing decentralized I would sort of refer to this as like decentralized infrastructure in a way. Uh, and one problem I I think the graph is a great um purpose here is like great design, reasons to be decentralized, but it didn't work. And like if your thing doesn't work, no one's going to use it.

So want to make sure you get the best experience. Um, people also have experience in bringing accelerated hardware to market. Um, and then it's an open prover set. Um, not just for livveness, but also for innovation. Um, I'm a believer in like open markets. If I make that proof set open, then people have an incentive to innovate and then it all ends up trickling down to the uh to the user which again I want is the best price for you. Sweet. So you know summary we get this like very nice industrial strength in gen one. Um you as a user get the most efficient market. So cost plus latency. Uh and then it should be the best development environment um based on your preferences and use

case. Sweet. Thank you. Any questions? OP man production. Yeah. So, Kyua is like our like op framework essentially. We're like well already have that. So, we just took OP's code. We just run in CKDM. It actually it's very easy. You deploy one new contract and you do two onchain

transactions. I could do it in 10 minutes. I wouldn't do it in production in 10 as a demo. I can do it for 10 minutes, right? Like let's not have systems that get upgraded in 10 minutes. That's asking for trouble. But it's very easy. It's fully super chain compatible. Uh we're talking to some like very large builds right now about changing. Yeah. So what you'll see uh is a lot of actually every single L2 at this point.

So even legacy ZK systems are sort of changing to our approach of using a CKVM. So ZK sync just announced a ZKVM based on this hero. Um you know scroll is building a CKVM now. Yeah. So they sort of built those like custom circuits at first, but then like our approach more so there's they're switching to like a risk 5 CKBM and then they'll get the same benefits essentially. So you'll see all big rollups switch to a CKVM based rollup this I would say as B this year

because we also make optimistic groups uh OP groups much cheaper. So if you want to run the OP fraud proof game or arbitron. So OP to run the fraud proof game. Any guesses how much it costs to run the game? $100. Okay. Any other guesses? Sweet. So you have to put down $1 million and then it costs hundreds of thousands of dollars over multiple days to play the game. it's very very expensive. So what we did instead of that is we play the whole game in Zc. Um and now in our version it cost uh you

have to put $100 down and then um the proof costs $5, something ridiculously cheap. And with this um since our game is so much shorter, you can actually get much better finality. Um, so if you're willing to accept that Ethereum is not going to social for theum has only ever social forge one time, I personally am okay accepting that Ethereum is not going to social. Um, and you can have an OP roll up with like three hour finality. You just have to make sure you can get one transaction on it. What about other things that expensive? Yeah, so Arbitron's actually more expensive. So, anyone want to guess how much ETH it costs to run the game uh

in Bolt, which is Arbitum's like, you know, I love Arbitum. They're fantastic. Very good tech. They're the first ones to deploy like a permissionless project. Guess how much ETH um it costs to run the Arbit. It's one million E. So the only one people are like Vitalic and Lazarus um is very expensive because the bond because there's this thing called a whale attack um where someone with a lot of money could delay very often. So they put a really high stakes to prevent someone uh anyone using them. Um with us there's no whale attacks. Anyone can

start a group for very or anyone can start a dispute for very cheap and anyone can generate them very easily. uh much better. It'll be a very nice sprout if you're familiar with like L2 beats. Maybe you've seen people starting to talk about like stage two um rollups. Uh Kyua will be stage two very soon. Uh it's actually not really up to the tech anymore. It's about other people getting rid of security councils and I can't force someone to get rid of their security. Did I answer your original question? I think so. Yeah. So I think I think I answered your original question of people are just sort of moving to the CKD approach. It helps optimistic rollups and it will help CK rollups. I

think a lot of people will change to CK produce very soon. It saves them money today and then as that cost continues to plummet. uh we'll see more and more people come to the CK vote especially as we unlock asynchronous visibility they'll have a really large financial region uh reason to change 800 say 8,000 uh connections in $1 and you say a real time um real time um proof proof So are they the same um are they the same um BK proof size? Yeah. So

all the proofs are like constant or um the proof sizes the actual proving changes. The end of proof is always the same. So are you asking like how much proving needs to be done or are you asking like what the final proof is? I mean and you don't care about the TA. So the the final proof always the the the minimum size. So I I say uh when you say um 8,000 transactions in $1. Yeah. It must have a batch size. Yeah. Right.

So 8,000 transactions in $1. Can these le proofs be generated in real time? Yeah. So very good question. Um, two things to note here is proving cost is sort of linear except when it comes to proving roll-ups for some reason. The more you batch um, the cheaper it actually gets in our system. It's a little different in other systems, but in our system you can make it cheaper. So we are doing quite a large batch here 500 blocks. The block time on base is 200 milliseconds.

So that is 100 seconds. Um so that would work for like real I would say. Um so if we were to do like a smaller block that would be like let's say 50 blocks instead of I'm getting there. Sorry it's 50 seconds. Um so let's say we were going to do like a 10-second block. Um that would be like 50 blocks. It would be a little more expensive I would I would expect. Um but still relatively cheap. I would say maybe I'm really shooting from the hip here. Um yeah, I mean do you have the data? Um I I don't. These tests are really hard to benchmark with ZK systems. It's like quite difficult

actually. Um but I would suspect maybe it was 20% more expensive than that. So still not bad. So maybe we go down to like 6,000 transactions at all. The next question is about the TA. Yeah, I think you should should care about the because you're not building the the the general VBMS. So the data cannot be compressed, right? Yeah, of course. If you want want to provide the DA, you have to uh I mean the data will be huge. You have to find DPS.

You know that that's a great point. I think Celestia and like IDA are doing great work around this. Kyua is integrated with both. So if you use Kyua, you can also use Celestia or but if you don't the DA on Ethern I I I see what you're saying. I see what you're saying. Yeah. Um I would love for the Ethereum Foundation to prioritize making Ethereum have more bluffs. Okay. I can't do that work unfortunately.

I think in Petra and then yeah so like much more but completely agree we need more DA on Ethereum and I don't know if they're going to achieve that. Yeah. It also be great if we could like split blobs. Um so currently like most blobs they're not really like that full essentially like uh like arbitrum will take up like 10% of a blob. Yeah, currently not because they have no such I mean general VKVM that the transaction the TPS there is quite low but if you we have we do have such a high TPS ZBM L2s the blobs will be

full 100%. Also I mean we saw like was it three months ago the blobs were full for like one day and they got really expensive. Yeah, it was like really bad. So, completely agree. We need more blob space on Ethereum. Great question. Great. Any other questions? Do you have for example public want to use as a so using you to deploy different parts of a single automat. Yeah. So um you know we

give people the choice. I think there's a reason to use Celestia. There's a reason to use IMDb. I don't make that choice for people but we can still like you know do all the proving and then aggregate their proofs uh together on eth one to sort of share their so we take a break for like 10 minutes and we Thank you.

Thank you. Very good questions. Much prefer this than going to talk in front of a big crowd. Are you leaving? Uh I'll probably leave in like 10 minutes. I have a my talk. Yeah. So guys, so it's kind of like a small group for people to like communicate uh deeply. So feel free to check through that. Yeah. Oh, and if you guys want to check out Boundless, we're on testnet right now. It's uh boundless.xyz or if you just search