zkSessions: Mapping the L2 Landscape - Park Bench Panel

Transcript

um this is the park bench panel it's a format that i've done a number of times in person back when we could do that and what it basically consists of is you have sort of a starting set of panelists and then you start a conversation with the audience and as people ask questions and share opinions you kick off some of the panelists and you bring back on you bring on new panelists and so this is a really exciting way to almost like get a feel for the larger community's thinking and that's what i'm hoping to do this time but as i've warned everyone here and a lot of people out there this is an experiment and it's the first time i try to do it online so it might be real clunky or it might be awesome and i hope it goes great

but as a starting point i'm actually going to run at least the first few minutes of this as a kind of more traditional panel i have a few questions and comments and ideas and i think this starting set of participants is really interesting because we have will matt and john each kind of working in like slightly different parts of the roll-up space and with potentially differing ideas about what should or should not be you know the the path forward in that respect so um since you haven't you none of you have already given talks if you can do a really quick intro to who you are and what project you're from um that might help the audience

understand who you are i think you're all muted right now we let the who goes first okay um so i'm will i'm the founder of diversify we are i'd say an app specific um scaling solution on ethereum uh we're using the starkx validium we're the first operator of that and we enable trading and transfers very low cost on layer two and we use stark proofs but actually the data for diversified controversially is like held off chain um in a validium style setup which has a lot of benefits in terms of much lower cost and privacy but

i know many people think also has some like unacceptable skating security trade-offs but actually we think that for a lot of applications including um lots of types of d5 that's actually a sort of major improvement over um using a pure roll-up that's my uh freaking train that's your two cents all right john do you want to introduce yourself sure can you guys hear me uh first of all i'd like to point out that it's very funny that you said controversially about off-chain data but anyways i'm john i do protocol research at fuel labs we're building an optimistic roll-up on top of ethereum which is primarily focused on bringing a

new execution environment to ethereum instead of replicating the evm inside a roll-up we're essentially building a much more optimized environment for both simple transfers and in the future smart contracts matt thanks anna so i'm matthew from loopring um i do business stuff over there uh we as uh similar to will and and diversify we are an application specific zk roll up but in our case uh yes zk roll up not validium uh for the trade-offs that we'll probably explore here um another difference being that we use like ourselves as the roll-up provider

loop ring is the zk roll up i'm not using a starkware or another provider so we're kind of at the lower end of the stack and operating the products as well sorry for the barking going on here um and yes we we allow kind of low cost um apparently instant trading and transfers on uh with without ever sacrificing ethereum level security that's uh that's what we're going for okay a question to all of you and i think this won't be too controversial but it is what is an l2 solution and what is not an lt solution i don't know if john you want to weigh in on that

what would you define an l2 solution as i guess since you targeted me i have to go first uh but i'm gonna i'm going to use the definition that giorgio's kind of popularized namely how do we separate an l2 from a separate l1 with like a bridge where side chains would be a separate l1 and specifically what we want out of an l2 is we want it to be trust minimized so users can always remove their funds from the l2 and in the case of like a smart contract style l2 they can always send transactions because you can imagine like if you have a loan like like a maker loan or something on this l2 because it's a smart contract support

you can't just like withdraw your loan you have to be able to process transactions in the l2 right so it's not just withdrawing but essentially what you want to be able to do you want to be able to always move your funds and you want your funds to not be able to be stolen and if you have those two guarantees then you can say their funds are as secure on l2 as they're in l1 sidechains don't give you this uh rollups do give you this uh and the general term that i've been using to encompass kind of what is the minimum set of trust assumptions you need is this expression trust minimized where i would describe the minimum set of trust assumptions as we trust the l1 is censorship resistant and that's basically the only thing you

should you you have you have to trust uh if you have to trust more than that then i wouldn't call that trust minimize and then i wouldn't call that an l2 so an example of something that isn't trust minimized would be just a plain side chain because whoever controls the side gen consensus can just steal your money so l2 isn't necessarily in my opinion of these just relegated to side chain style stuff or rather side chains with additional properties like grow ups it could also be channels i mean channels are also l2 as long as l1 central ship is listen you can always get in your channel disputes and and then you can be guaranteed your funds will never be stolen does validium and volition those type of paradigms also fit in the l2 definition will you're welcome to answer that well that's not

the world yeah i think my definition of l2 would winning would specifically focus on the exit conditions around the l2 and that ultimately funds held in l2 are relying on the only the only the security parameters of ethereum in order to or in order to sort of guarantee that they can exit funds but i think all all air tools do give up some um some sort of censorship resistance or trust at the moment at least for trust assumptions in that um whether even if you're using a full roll-up it's if you're using optimistic roll-ups ek roll-up um at the moment operators have the ability to censor we just heard about sequences and i think that's a very important sort of piece of work but

there are sort of various um at least during operation uh sort of security and assumptions that you have to make but what they all have in common including valedium and volition is that um if operator goes offline or if uh there's some other sort of set of conditions that's broken um you can exit your funds using only the security of ethereum and that's that's the ultimate thing you can fall back onto um and that's where my different definition may be slightly different and so in that case i'd say that um palladium certainly is an we're already starting to get a few questions from the audience and so

sort of define for the audience members what what i'm looking for basically if you ask a straight up question i will just ask it to the panelists that are here but if you ask a question that is loaded with an opinion or a straight up opinion i will then invite you to come up onto the stage just so you know that's sort of what my criteria is right now um for example this i think is actually a question mark just asked it sounds like you consider an l2 with governance not a real l2 could you elaborate or is that an opinion no i think that's a question i think that's fair matt did you have something to add or i i had

so i mean i agree with john john's been but i see where will's coming from too with the valydium and volition i don't want to um uh put off mark's question but a point that john said which in my day it's the as like i say you could always rely on layer one security to eject to withdraw your funds the most fundamental security is like to be able to get your money out of somewhere but john makes a good point like if you have a loan on maker you need to be able to send transactions is that little additional criteria only for pvm capable z cure all it's like like loopering which has a single relayer you could always eject and your funds are safe but a relay our relayer need not process

your transaction it's not resistant if if the roll-up is not evm compatible does the criteria drop or must you always be able to force all types of transactions that the roll-up supports enough i'm sure john guess i don't know if that's a bit convoluted but um i'm curious if your criteria is stricter than uh yeah i think you see what i'm saying must any type of transaction be able to be forced through are you asking me yeah yeah i think so sure we can get back to mark's question because i feel bad for mark but we'll get we'll get back to it shortly uh because i mean this is related this is related but

okay uh so this this goes back to some i think we had some minor conversations on this on twitter between me and matt uh but essentially the kind of adversarial scenario i'm thinking is if you for instance have very high throughput roll-up what you can call the zk roll up right it's doing a thousand transactions per second it has millions of users the problem is that if it's permissioned then one person the operator whoever you want to call it that can just stop the roll-up and they'll force everyone to withdraw now sure it's not like a mass exit scenario where everyone has to rush to withdraw by some time but still just the very fact that you now have this huge volume of users that have to each make an l1 transaction to

withdraw is like that's that's super expensive right it's that's something that one person can make one small action and force a bunch of other people to now make an expensive action i would still call that l2 though uh i would say the minimum requirement for l2 is trust minimization and permissionlessness is like something that's really nice to have that you want your l2 to be a replacement for l1 but i would still recall something that's just just trust minimize even it's permissioned on l2 like a plasma for instance plasma's permission but i'll still call that l2 let's continue then with mark's question now which is if it's if like does it need to have governance all right so do you consider an l2 with governance

not a real l2 is actually the question there yeah so uh i don't know i see alex also asking a question so to mark's question i think it's also brought up on stuff we discussed briefly on twitter but it's basically uh this relates to on-chain governance so so you have some dao that can control upgrades to a contract uh the reason i wouldn't call that a real l2 is you know it's no longer trust minimized by what i call trust minimize namely the only assumption you're making is that l1 is uh l1 is trust is censorship resistant that's the only assumption you make and potentially i would also say a trusted setup for the zk circuit i would also put that under trust minimized uh but if you have like an entry in governance now user funds can be stolen

by a majority vote by potentially maybe like one tenth of the token holders even because lots of people just don't vote because they're lazy uh and i wouldn't i would call that a much stronger trust assumption and in fact i would say that's the same trust assumption as if you just have a sidechain and if you approve of a state sidechain that's the same threat model is that like a majority of the consensus of the sidechain can just you know upgrade and take users money uh mark says is it a weird mountain fantasy and practice maybe it is uh but this is more like us being pedantic here as opposed to what is what do practical users want okay i'm gonna i'm gonna ask one of ryan's questions here what do you guys what do you think of alex's predictions on fee

growth that's a question [Music] what was the predictions did i miss it in an earlier oh it was alex yeah it was alex from matter lapse talk if uh to uh two thousand gigaway i think oh no yeah anita goes to ten thousand um and you know honestly i think you know this this is a this is a genuine um real suggestion in that ethereum fees will keep growing on layer one there's no reason why they wouldn't demand if if this system is successful which i think is already like continuing to happen it's only gonna go up for layer one security um and at the moment we're wasting their one security on

all sorts of other transactions contracts which don't make sense to pay so much money for um but that's going to keep happening because this is a global computer that many people want to put transactions into and so i think it's completely realistic that that will eventually happen and so most most end users will want to interact via l2s i i agree um very feasible like we saw a thousand way a couple weeks ago and like look at the space we it's like tiny we all know each other we're using it for certain applications sure it's it's whales getting transactions through but imagine when this is uh just completely normalized on on layer twos and

you have these layer twos these like institutions competing for that block space uh i i see it going very high and uh i mean eat price i guess is a different sort of prediction i'm also uh as bullish as alex apparently but um yeah i think that's that's very feasible so users will necessarily live on these layer twos um do you think this is kind of a question oh mark said holding a global lock to the internet financial systems could be expensive so that i'm going to define potentially as an opinion

straight up it made opinion so i'm gonna do the the thing that i kind of promised to do which is switch off someone from this panel and bring on mark and matt i hope it's okay for you since i had sort of poached you last minute can i switch you out absolutely absolutely thanks so much for being on the panel you're very well mark just add uh jump on if you can so we'll give him a minute just like a hunger games thing or once you're out then you're done no you can always come back on but you have to here's the whole thing questions will be read to the panelists opinions like really clear opinions and

mark you your first question actually had a bit of an opinion in it so i i was like i should i was like oh should we already switch it out but now you made a very clear opinion so now you are on the panel but yeah there are some other questions that have just been put out here um so the first one that ryan had said was about this about alex's fee predictions but this i think goes mostly in the direction of lydium but if fees do grow this much will valydium get much stronger or grow faster in practice due to the fee savings kind of like is volidium and that model actually the most efficient i'm kind of looking at you well but up

to anyone you're looking at a mark on my screen so um yeah i definitely i mean i think the the cost is like at the moment the the main i mean a huge advantage for that idiom and um that's becoming more and more so even you know since diversified launch which was around eight months ago um gas fees have just gone through the roof and the cost difference is huge and for end users it's huge and in the sense that it's completely feasible that you know in another six months um even interacting with like pure zk roll up ends up being too expensive for a lot of people to interact with d fight um and i think more more people will take that design choice but also the

solutions for off-chain data availability will improve so it won't be um as simple as just having like what diversified does now would have a committee of seven who will hold that data and um therefore you have like a certain amount of guarantees which are better than just having a centralized data availability but there will be other ways of publishing that data ensuring that there are economic incentives around releasing it and around um you know handling handling all of the edge cases um but i do i don't think that's just in my view the ultimate end ideal case would be that everyone has pure cryptographic security via um only appears uk roll up but i think that's becoming

already sort of infeasible with current cost and therefore there'll be sort of better and better solutions around um data availability um sky asked is late this is for john this is actually just a question as well is lazy ledger used as a supplementary layer for any sort of rollups today sure i can briefly answer that uh so this also ties into a question marked uh mark asked which is uh well the trade-off without governance is no upgrades so can we actually do better than having no upgrades or having upgrades with governance this is where lazy ledger ties in and roll ups so laser ledger is a blockchain designed

for i'm optimized from the ground up for being a scalable data availability layer so you can think of it kind of like e2 with all its started data availability trends and then you have some separate execution system for example in the case of ethereum you have the eth1 evm chain that processes stuff processes transactions and then you have a separate chain for data availability so these ledger can be used instead of e2 for such purposes and we have been in talks with various teams that are building roll-ups for this laser ledger is not live yet so nothing live yet but how does this relate to upgradeability uh so the ethereum world computer model is basically you have these immutable smart contracts that live in a

single virtual machine and you can't really upgrade these without on-chain governance or without like admin keys the cosmos model is like this internet of blockchain system where each chain is sovereign so each chain can be upgraded via hard forks and this is one of the amazing properties of blockchains in general is that you can do forks soft and hard so if there's a dispute between two parties for whatever reason or two groups of people you can just fork the chain uh native which is something you can't do in the traditional in the traditional world right if you have a problem with your country you can't duplicate your country and all its government and all its money and all its assets and let the best one win right you have to come to an agreement everyone has to

come to agreement to a compromise uh with blockchains you can just work if you don't like bitcoin you just fork it into bitcoin and bitcoin cash right maybe you don't like the direction ethereum goes with refunding people from the dow hack you just fork it into ethereum and ethereum classic and let the best one win uh so the uh the cosmos system is this uh the system of sovereign chains and the only difference in the sovereign chain and this world computer model is these applications can upgrade themselves through forks and let the best one win but the downside is they don't have shared security when you have all your applications in a single virtual machine they have shared security uh because you can be guaranteed that you know one application like union swap and one application like maker they can be guaranteed that both of them

uh to corrupt any of them you need to corrupt the thing the single ethereum chain uh but it turns out you don't actually need them on the same virtual machine you really only need them on the same data availability there that's what we've learned in recent years and it turns out the lazy ledger is a blockchain optimized for this purpose you can have a bunch of applications on their own server and chains like in the cosmos system have them all post their data availability to a single shared security data availability layer of laser ledger and then you get the best of both worlds you get upgradeability through sovereignness so you don't end up with you know this no upgradable world computer model and you get shared security which is what you get out of ethereum so this model i think is the future of blockchains

but is does that kind of answer but i think i think the question here was is it being used currently for any sort of rollups today that was the the question right like i guess i kind of deviated into explaining why that question the background behind why the questions is is irrelevant to mark's earlier question but to to specifically answer that specific question the answer is it's not used today because lazy ledger is not live it's undeveloped okay got it that was kind of the next one is lazy ledger live um yeah so mexcool said that's one of the key selling points of cosmos know that all apps need to be need the same level of security

so that i don't know max cool do you are you up for coming on the screen and joining this panel that's not quite a question if so hit the blue button up there um and you can join oh actually i think i have to knock somebody off oh yes okay so i can go since i've talked quite a bit i had an opinion there he says uh lazy oh yes that's true okay okay so i'm gonna i'll kick you off john stay in the chat i hope to see you chatting and uh i will i'll uh i'll let max join so this is how this is what we're learning about the platform about how the

the timing goes with the switch okay and actually you're right will that i think alex's question here or yeah i think this is a little bit leading as well so will i maybe i'll oh yeah so mex yeah you've added great and will i might boot you as well is that okay you're muted play i'm gonna assume so okay cool so alex you join as well alex asks how does it enable forks this is actually a question about lazy ledger but maybe you can help answer this mark uh lazy ledger models could only be used as l1's yeah i think john would probably be the

expert to answer this one yeah if you're right this is a very interesting we'll see if this is working with the with the questions let me try to find another question up here because i mean i or i have some brendan's question do you view the end state for roll-ups especially optimistic roll-ups as a single roll-up instance to rule them all or as many roll-ups kept to being small enough that users can easily verify every transaction that occur on the roll-up oh okay that's pretty good yeah i mean if there's just one roll-up it's kind of like it's like a trade-off space right if there's one roll up then you get composability with

everything within that roll up but you don't get a ton of like scalability and you don't get like the ability to like opt in to like a particular roll-up if there's many roll-ups then um you can choose to opt into like validating like the particular ones so um it's a little bit more scalable with like many roll-ups but um then you lose out on the like composability um until we have like good kind of cross roll up primitives um which may or may not be a thing alex what do you think well i think if you apply the philosophy of

inequality uh to roll-ups then you might also want to balance them out and give everyone equal chance to serve users so like we should all try to to see like what roll-ups are under-served and like to move things around a lot of uh a lot of gas will be used like arbing between roll-ups um one question that i've been kind of like thinking about a lot lately is like what is like an unacceptable amount of gas to be used for arbing or is like any amount of gas like like let's say that like 80 of the gas budget is used for arbing like is that too much

or is that just like the market is using 80 of the gas budget for arms between roll-ups like um that's just something that i'm trying to figure out if there's like an upper bound for that we should target like as a community you know what i mean um you mean like arbitrage between the roll-ups where you you like people are gonna trade and balance out the prices yeah well like to honest answer i think that we will uh actually like see the the world more gravitating towards a few big roll-ups because uh unfortunately inequality is inherent in this world and the predator principle and the power law are like ubiquitous uh and uh

this will eventually the situation where you will have a lot more uh transactions happening inside those few ecosystems which will be on top uh and i'm not sure how things will work between the roll-ups i'm actually fairly skeptical that we will have uh good mechanisms of unidirectional so like if the if the trade zelda is uh biased towards one direction then the state channels won't work or payment channels won't work and you will eventually have to go through layer one and this will obviously follow the same route as uh layer two to layer one so it will depend on on what kind of roll up you

use uh it i think it's gonna be complicated so i think this will be something that happens rarely and more frequently people will just stay in in their ecosystems there's an opinion there from jordy who said multiple rollups are important for evolution like multiple blockchains and jordy's actually just about to join us as well so i just added jordy there hi charity hello everyone yes that's i was to say that that uh this this comes to my mind when when people say that okay bitcoin will bullet all and you don't have any option to uh

other blockchains to exist or we would not have ethereum if we are thinking that way so i think that uh for the good of evolution and specifically at the beginning we are still bringing a lot of research uh and a lot of work that's coming and we don't we don't really know what's the best solution we are competing here some way with optimistic roll up and seeking all up we don't even know what's right uh what will happen and how they will work we don't know even if it's possible to interpret between them we don't know the space so it's important i think at least in the beginning to have a an ecosystem i think the interesting thing is this interoperability between roll-ups if we are able to in some way to work and this is the cool thing here

is at the end is ethereum because at the end ethereum on the bottom is what in some way can join all these roll-ups and can facilitate a lot the uh interoperability but having many roll-ups will allow space for a lot of innovation a lot of testing a lot of learning and i think it's important uh at the current stage maybe 10 years from now um who knows uh but i think it's cool it's good now to have uh many roll-ups i i fully agree and actually i really like your talk about mass migration so mess up my questions as you call them uh and i i really think that they are crucial to any roll up

like um jumping back to the definition uh of john uh i think that like being able to fork away from a uh a roll-up to a different uh roll-up or two to different player to solution is actually what makes it a uh permissionless like true decentralized layer too and it should be like important property of every solution there's a quite there's a question slash comment here from brendan do you think that l2s will be antagonistic to l1 or to value accruing to eth in the long run mev and l2 is captured by roll-up operators not miners or validators and

network effects around specific roll-ups might allow roll-up operators to capture value even or even leave an l1 entirely taking users with them and brendan i'm going to invite you to come up if you're up for it no no no we need all the you know all the that's working in l2 even if you are an operator on l2 if you are on username too if you are a developer that's doing something on l2 you need to understand that you are running on top of l1 so uh it's crucial for the space to have a very good uh l1 and to support l1 and the 2l1 to evolve and it's important that l1 continue

evolution and give and think so it's interesting to see for example the last one of the last uh papers of vitalik just saying that uh you know that ethereum is going to be like the volatile focus or roll up uh centric and this is important to to understand because uh layer two needs layer one so having a layer one that thinks in layer two and there is this symbiosi symbiosis between uh the two layers it's important for the for the growth of the of the ecosystem i just just wanted to to mention that i believe that

mev long term will not be this the problem with this as we see it now so like mv is not forever and will not always be here because we have actually primitives that uh can fight it specifically with their knowledge probes and with uh vdfs and time locked encryption i believe that very soon we will see the systems materialize and completely change the the entire discussion there around this so users will be pricing will become a lot more transparent and fair yeah i i was just wondering in the context of like eip1559 it just it seems like there might be

sort of this tension because mev is seen as like compensating miners for lost revenue due to decreased fees or to burned fees and it seems like in a world in which um meb is instead captured by by operators that um like that scenario might not play out and uh like uh some of the value that might be going to providing security on the main chain might instead be going to providing security on on l2 and and um yeah i'm just curious like what the the moats won't be uh long term whether uh if you have a roll up that has uh sort of like a lot of user

activity that stays and sort of exists in the roll-up whether it might make sense for the roll-up operator to provide that availability in some way that doesn't require it during long-term well this is what my talk was about like that sorry you will have to share some like you will have to split data availability and maybe implement your own sharding in some ways and you will always be probably less secure than the layer one in this regard but you can allow your users to have flexibility and choose how you uh you balance that i think that

the security of layer one will always be determined by the market so i honestly i don't see how eap 1559 makes like any fundamental difference like the the miners will just like see look at the the the amount of money they get from the tips and from uh the block rewards and they will adjust the the hash rate accordingly or the hash rate will balance out so like you probably want to avoid any rapid changes like any sudden step functions there like it would be better maybe to to make some smooth transition but eventually it's going to be the same thing like just some some part of the

of the fuse is going to be burned which is good for ethereum supply um yeah yeah yeah i mean that that oh sorry oh the uh that that might be the case i i was just uh sort of noting that the argument for eip 1559 uh is that meb will sort of uh like fill lost revenue for miners and and like clearly in in if more activity is moving on to l2 that won't be the case but yeah i want to ask there's a question here for everyone on the panel the internet started back started 1869 what year do panelists think

think we are at right now as related to l1 and l2 technology development [Music] the beginning of the beginning of the 90s maybe end of the 80s beginning of the 90s people don't know people if we see from the number of users that are actually uh using two solutions well we have some we have family adopters it's not zero but it's not massive options for sure even in the ethereum space so most of the people in the community of ethereum is not using a layer 2 solution right now

that's a good way to think about it that's interesting question here and this is uh someone who's just curious i think can you use l1 wallets on all l2s i think the answer is this will depend on the this will depend on the l1 uh on the support of l1 wallets i'm sure uh all the wallets will support the main roll-ups in l2 very soon there was a question here for alex which was how does how is transaction ordering handled with zk sync

well right now we have uh instant confirmation so the the operator the sequencer provides uh confirmation of the transaction immediately to the user and with the uh introduction of multi-validator consensus it will be the the role of the validators but we want to keep the block sizes very small um and so the the validators will just be able to reorder within the blocks uh and eventually we want to get rid of it entirely with introduction of timelock encryption in the case of hermes in the case of firms we are we are launching with multivalidator uh from day zero so uh valid it works

like a normal blockchain like a normal blockchain so validators are uh free to choose what uh transactions occur and we expect they include the most profitable or the most profitable uh transactions one that get better rewards for from the operator so do you use some consensus mechanism yeah we use a proof of it's a kind of proof of donation it's a proof of work the days that you are bidding so you can buy um you can buy a slot and anybody can buy a slot and uh it's a bidding so it's a it's an auction so anybody can beat and the winner it can forge as many batches in that slot as they can but where is the bidding

taking place in lay one what do you mean layer one oh every smart contract yeah you buy it so you can buy it like you have i see see so but wait wait a second i don't understand so you what happens if uh one of the valedictories is always gonna buy the all the slots can that yeah well it's uh well actually uh it could it would well it it could happen when it's fully decentralized so it's a winner or all available but would because very costly it's like buying uh it's like buying an ethereum uh it's like buying an ethereum uh so that's a minor just buying all the blocks and mining empty empty so you're you're playing the

lottery there yeah yeah but here is what we are doing is uh for the especially for the bootstrap what we are going to do is just half of the slots will be decentralized and the other half of the slots will be from the run from the boot coordinator so it's like anybody can beat anybody can forge so we can keep the censorship resistant uh and the worst thing that can happen is that somebody just buy one slot and they don't force anything but then the boot coordinator will do it but we expect that to remove we want to remove all the governance in there and we expect that when there is enough volume in the system uh this should can be fully this can be fully can be fully decentralized because it's like it's not working how do you know that they skip their turn

what's the timeout because on layer one you can like you need to give them some time you and do you have to wait for the previous blocks to complete before you can produce your one you can so it's just you are the owner of 40 you are you are the owner of 40 slots and 40 blocks for that during that 40 blocks you can afford as many batches as uh as many batches as you want and after that 40 blocks it's it's an extra slot and if whoever wins that slot is the one that has the right we have an extra thing is that if nobody forces uh if nobody forgets a slot for um number i think it's like 20 slots or

something like that then anybody can forget so as a user how do i know to whom to send my transactions to all of them to you you send to all the you you know who is the winner of the bidding of the next slots so you stand well [Music] okay a question here is is ethereum l2 bigger than bitcoin channels already it is l2 so sorry is ethereum's l2 bigger than bitcoin channels like lightning i think just purely by by locked value i think l2 already surpassed uh lightning and we see a lot

a lot more adoption a lot more interest so like it's it's clearly on uh the speed of growth is is a lot stronger okay um a question for you from john uh and this actually this has been sort of mentioned in timothy's question i believe too it has to do with like compatibility this concept that you had said before so can you give us some more details on how you enable solidity compatibility is it an evm circuit or a transpiler or something else yeah this is this is a great question uh so what we do is uh we have a generic universal circuit which can prove an execution trace of a program so

this is similar to a tiny ram model which is pretty pretty wide known and we we put all the heavy operations out of the circuit and we delegate them to specialized circuits and this is the trick which allows us to have uh close to full evm compatibility because in the vm for for a tiny ram you need to have all the operations to be of equal uh equal cost right because like in in the circuit you have like uh for those people who are not not closely familiar with your knowledge circuits they represent a mathematical function so you you can have they you really look

like a circuit you have some inputs then you do some computations and then you have some outputs and they have fixed so you cannot do anything dynamic you cannot do loops of variable size you cannot do recursion so how do you represent the program which can take various paths through the execution so the way you can do this is just to have steps on each step you will prove that you checked all the instructions against the current instruction from the memory from the program counter and you actually executed all of them and then you just pick the one which corresponds to the instruction you pick the result like if it's in the addition you pick the result of the addition and put it to the to the register if it's subtraction you

pick the result of the second part of the circuit and put it there and you have some overhead there right so like if you have 30 instructions then you have to do 30 times more work than ever then than every single one of them but in evm you have some instructions that are like hundreds thousands times more expensive than some others like if you compare storage for example to a store to to edition you can't really put them in this rigid circuit so what you have to do instead is to find some way how you can balance and we do it with recursion so we have specialized circuits which are optimized for this uh for these operations such as sketchup hash or

uh ecdsa signature verification or a storage access uh or some some other compiles and then we use as many of them as necessary for this particular block and we connect them on top of uh external circuits and we we kind of we can aggregate variable number of blocks which are necessary for each of the application so i'm not sure how understandable that was but that's roughly how it works with zika vm and it looks really well so we are making a lot of progress and uh the estimates are really good so the costs will remain negligible because the we also made quite some progress on

optimizing this um specialized circuits for hashes so for example i think uh well sha-256 went down from 25 000 gates in r1cs to 5 000 gates in plonk with specialized planck uh uh gates and so on so yeah this is kind of the answer there's a question here for the l2s which are using proving systems which require a trusted set if you deploy and there's a bug in a few and it's found i guess in a few years time what problem could this cause in the worst case scenario i think both of you are the perfect

people to answer this actually what kind of what do you mean what do you fear the most kind of thing the problem is is it's not so if you have a problem for example in the circuit don't matter if you have a trusted setup or not you you have a you have a problem anyway uh and you need to well it's like a smart contract if you can fix it you do it if you cannot it's like a smart contract uh you are you are you are you are you are gone um in hermes we are taking an approach of uh you start with some compromise with something that's a little bit

centralized uh and the idea is to turn it off uh as as to turn it off as we go it's like a balance between security and decentralization and there are some balance there it's like we are still like non-custodial okay but we can we can have some kind of escape hatches and things like that that the idea is to uh turn it down to them turn them off uh as soon upgradability of the smart contracts is the other thing is like uh uh we want to create a system where that's not upgradeable that's uh decentralized that's governance less but uh at the beginning this is this is hard so here we need to take some compromises to to to launch thing alex you are more or

less following this approach because your your circuits are upgradeable as far as i know yes sir so i i would add that indeed if you have a problem in the circuit that's equivalent to the problem in a smart contract so and then we have a lot of examples in ethereum where contracts were faulty and people lost money or frozen money so this is this this is exactly the same thing because contracts are just extension of sorry circuits extension of the contracts we have been running three different audits in small smart contracts and circuits and uh we have a bounty there we are trying to keep the same

parameters uh the same security parameters uh for circuits that us as for smart contracts is harder because it's like a lot more more newer technology in there so it's harder to find people to fully understand to review but we have been working a lot in this security and even that you'd never have a zero uh you never have the full warranty that everything is going to go right maybe there is some back and uh things can happen but you know this is uh this is a space uh uh we need to pass and and yeah it's like it would be easy not just not to do anything but if you want to launch if you want to push if you want to move forward in this

technology you want you want to take some risk we are trying to minimize service we're trying to do good risk analysis we are trying to do as best as we can but the risk is going to be there always and yeah this is the risk and this is something that any user i would say in any blocks chain any in in the space but especially in this new uh technologies is something this that we need to we need to be aware that you know we are playing with you know last very fancy cryptography very fancy uh very new code very new new ideas and this is something we need to take in account that we need to uh think about and we need to be very

aware of that now i would just add that relying on audience is is probably not not not sufficient uh like or at least we need to to step up the game with audits and make sure that people understand what is actually being audited what what assumptions are there and what has at least this code being deployed to uh to the contracts that the auditors check so yeah yeah it's like and even even for example it's i would give i would tell you something we have been we've been hiring up we have to spend a lot of a lot of money without it and they check it oh i can't hear jordy yeah i can't hear

jordy jordy i don't know why we just lost her oh no can he hear us i think maybe his oh his headphones might have died and already maybe we should yeah we can hear you we can't hear you um i think you'll charge i'm going to actually ask will will had a question here and also i do want to invite there's there's a bit of uh room on the uh on here so if somebody wants to bounce up onto the chat actually if you have something to say just uh like if you have something to say and you want to answer some questions just add add yourself here so while they're joining i'm just going

to say that okay i am skeptical of audits as the final source of truth they're helpful for the teams to to actually find bugs and make fixes but they are not a certificate of quality you actually have to wait quite some time uh with a lot of money locked to make sure that these systems are are working as as necessary so i was about to ask will's question but i actually want to ask lassa's before because it's directed at you and since you're on here and we can hear you um could be it actually relates to jordy's project too could the evm be transpiled to another roll-up example could you have hermes inside zkvm um well

you can uh but let me think quickly you can so like the very expensive operations like whatever is expensive in evm and let me think about this this is actually actually yes so it's it's actually not going to be that expensive because we are only talking about other medical operations so yeah yeah yeah sure why not absolutely jordy are you back oh you're muted right now you muted yeah hopefully we can hear you when you unmute oh maybe your microphone isn't picking it up it should be in the screen the the mute button but

i can't unmute you and marcel just said geordi you went out just as you were talking about cost of audience odd is the interesting part your headset died so there's basically two things one is finishing your thought on audits if we can and then maybe following up on what alex is saying about this now we still can't hear you do you want to reload yeah no we still can't oh yeah there yeah john's giving you some tips in the chat there um kev i mean i we need some new people to join though now it's just me and alex fireside but which is awesome but uh okay jordy's joining again but i

do if there's anyone else who has a question or wants to jump in please do so hello i can hear you now yes we're sorry okay okay do you want to finish your thought on um the audits and then think about this idea of combining your two projects yes i i heard what alex says before and it's like yes audits are essential are important people have to do audits uh but they are not the they are not the panacea they don't they don't give you the full warranty on that uh we for example it's like hiding out it has been very helpful just for even for questions that the auditors just ask and you know just one thing comes to the

other and it's just a matter of understanding your code if uh learning your code and going through all that i say they just and you know even with that is is is hard that's the only way right now is the only way to to to move forward uh uh at the end i don't know this is our policies just try to do it the best we can just try to do the the the best so the best is safe that we can knowing that there is a risk then let's light a risk and it happens something it's whatever it is but we need to put our best effort uh to for this to be safe okay now could you merge hermes and zk sync

i'm sure i'm sure i'm sure it will mainly because uh hermes and zk at least right now they are upgradeable at the beginning so we can we can always uh there is a lot of opportunities and things that we have not even think about the important is to at the beginning it's just the people to start using these collapse people understand understanding the advantages and see the different approaches that this this new technology have and the users will ask for these coordination transfers and we will see it's a

lot of new things that we need to discover i do not have a crystal wall but yeah i'm sure that's not it's going to be a space you know it's not going to be just one roll up to the laps of your laps it's like it's going to be different roll ups and people this technology will will in some way will be uh it will be accessible to many other projects and i'm sure it's going to be a lot of very specific roll-ups that will appear in space in the upcoming years cool i'm now going to ask will's question which is where is this is for zkvm where is the state stored for evm compatible zkvm does it all end up needing to go on l1 and who pays for the storage

i guess this is towards alex but maybe both of you yeah okay so with for zkvm you mean like storage slots of ckvm so like uh yeah let me just find exactly the where is the state stored the status start off chain just as for uh with any other zq roll up so the validator store store the the state and the users can follow the state if they want and also store it and um so it's going to all end up needing to go on l1 and who pays for the storage that's the second part so the by by saying that something goes on l1 in zika roll-ups we mean that it's been broadcast through l1

and being uh distributed to to the other validators and the users now it's interesting we're actually considering introducing state rent from from the beginning in the case uh sticky sync because uh like the without state rent the problem of endless state growth will will continue and we at least want to think through the interfaces and provide the developers with guidance on how to structure their contracts in such a way that the state the transition to stage rent will be seamless for them because we can we can do it from like we start from scratch and we we don't have any legacy um and but once you have state rent uh it's

actually the the cost of this storage is very predictable very long what is the word you're saying there is it state ran state rent state rent yeah you actually you pay for a certain period of store in your state uh and not just for storing it forever when you make a transaction didn't hear the question can i repeat i think jordy just said he has a lot of faith yeah in the ethereum 2.0 for that

availability well uh ethereum 2.0 will start with proof of stake and it's going to be great if we will see proof of stake uh launching in anytime soon and i think that sharding will follow with much longer delay than like it's like it's we will probably still have to wait for a proof of stake even to arrive but we will still have to wait much longer for sharding to arrive and it's not yet 100 clear how sharing will work so i wouldn't be that optimistic in this regard i think we need to plan for me like let's hope for the best but we need to plan for the worse uh for

the worst and uh be ready to like work with a limited uh throughput for the next couple of years yeah um here that the thing in that availability and it's something that's interesting uh that's a little bit different about consensus is that availability you just need a single uh a single um honest uh actor in the system it's not like you need two thirds you just need a single one so this i think it's a big difference that's quite okay i'm a little bit against this um against this um uh sets you know this um grow up future these groups yeah uh um yeah the validium kindness for

example you have seven uh seven people seven people that are just soaring i'm a little against but uh personally i think that this can give you so these intermediate solutions like validium can give you uh enough air uh to white for uh ethereum 2.0 uh uh ethereum 2.0 uh um that availability if required so that's well i i i'm not advocating for validium i i don't like validium i think palladium is a threat uh because like it will be centralized much higher degree

i'm advocating for ck porter yeah but i agree but i think that i agree but i think that the satanist story as a transitory solution uh until as a backup transitory solution until ethereum 2.0 deploys that availability it could uh it can help in the development process not in the decentralization the development process of this space that's just you know this is we'll see maybe it's required maybe not we see how this uh how much that that availability is required uh there are different needs but that's something that uh i would consider if uh there is a delay a delay in the data

availability in the serial 2.0 but i i really think that ethereum 2.0 uh in less than a year we can have it but you know i'm not i'm not a core there and i'm not on the theory for the developer but i want to i want to come in here because there's this question that actually i had listed something like it myself in the prep questions which was could one l1 roll up for example ethereum or i guess something built on ethereum come to consensus over state transitions on the roll-up that requires state from another l1 roll-up example on cello i don't know exactly what that means but i'm guessing it could i don't know if you can see it i i have a strong opinion on that i would

say that it's not possible and this is exactly the same reason why it's not possible to have uh uh so like i'm skeptical about lazy ledger it would be great if john is still here to join us because we we have a space and and talk about this i'm really interested to hear what john will say uh uh but the the the fundamental thing about uh lair once is that they are forkable it's like john was absolutely right you they are forkable and because they are forkable you can have two different uh chains with different rules where different people disagreed on what the canonical chain is the majority went on on one and the minority one on the other one

now if you have a system where that uh that offers a bridge between let's say ethereum and bitcoin uh how is this bridge supposed to work in case bitcoin forks or ethereum forks like you you will have ambiguity built right into the system you can do this like if if you have a canonical state on one and the canonical state on the other one uh like they are unforceable and um any spending on the one will be irreversible like if you have a message which is passed from one consensus to to the other chain you cannot really change anything after the

this message becomes finalized and so with with lazy ledger uh you specifically have a problem that if you're if the majority of or like the minority of your users think that the data is unavailable what are they going to do like if you implemented it at a layer one level they could just do a fork and say okay we don't trust this majority of the data stores they don't do their job properly they don't share data with us which is going to go on a fork and continue there from the last known state but how are you going to do it to player two

so like the the only thing you can do there too is is something like zk porter where you have to trust that you have an honest minority but if you if you exceed that they can always corrupt the data there is nothing like so really curious what john had to say but apparently he's not here anymore i don't think he's here i mean i i think he he might have he might have it on but i don't think he's actually hearing us unfortunately but i would say you know in a lot of ways i feel like we could continue this conversation forever there's lots of questions still in the chat but i would say this i just sent an email out to everyone

to head over to gather.town i think now would probably be a great time there's a lot of very specific questions based like you know specifically for you geordi or specifically to you alex so please head over there let's continue the conversation and to everybody who came to this zk session thank you so much for coming to uh map the l2 landscape this is not everything obviously there's lots of other l2s coming online who aren't here so i think we'll have to continue that conversation in subsequent events as well but thanks for every thanks everybody thank you very much it was a great event yeah cool all right see you over at gather then