CUSTOMIZABLE ROLLUPS with Neel Somani of Eclipse

Transcript

we're live my guest today is Neil simani Neil is the founder of eclipse it's a customizable roll-up provider for building decentralized applications in today's conversation we'll discuss what is a roll-up and how it makes sense uh in the context of modular blockchains we'll talk about how developers can choose between Roll-Ups app chains and smart contracts and one of the trade-offs there we'll talk about eclipse and da layers we'll also talk about eclipse and ethereum ZK rail ups and making sense of the modular architecture at a high level I'm also dying to find out why you think settlement makes no sense as a business and that includes ethereum so before we

get started make sure to hit the like button hit notification Bell And subscribe to get notified when I go live every week and remember that none of what is discussed here on the interoperative investment advice and if you enjoy this content please consider sticking with us we're live on Quicksilver Atmos osmosis and Juno just look for interrupt in the active set and I hope you're coming to Paris in a couple weeks because nebula's Summit is coming on July 24th and 25th it's two days of technical talks about Cosmos IBC the unit chain it's gonna be great and I'm really looking forward to it and I hope you are too so go to number dot Builders to get your tickets my guest Neil is coming up next right here on the interrupt [Music]

hey Neil hey Zab how are you good how are you going well thank you for having me yeah I'm really excited to talk about uh eclipse today you know we've touched on roll ups quite a bit here on the podcast we've talked with a lot of um execution layers actually so we did we did a couple episodes about that topic and we've also had Celestia uh but we haven't really discussed like that middle layer where I think Celeste uh Eclipse sits uh which you know of among other things is settlement so we're going to uh dissect this this layer of the stack today hopefully and talk about a whole bunch of other things but yeah before we get started like how did you get interested in this topic and wanting

to build eclipse so I was previously building an evm on Cosmos unlike uh like ether match or some of these other ABCI client supplementations this was intended to execute evm by code natively within the cosm was MVM so the wasmo VM would actually execute evm smart contracts alongside webassembly smart contracts the reason for doing this was Tara was doing really well at the time so I wanted to bring ebm compatibility to Tara wrote some blog posts about it so I was working on that for a couple of months and then when Tara D page then I stuck around the cosmos ecosystem but that project didn't make as much sense anymore given that there were evm Alternatives that already existed in Cosmos so I just went back to the

drawing board I was chatting with uh I think you mentioned the Celestia team so I was chatting with them briefly and Nick white was talking about just the value of Roll-Ups and what that can do as far as enabling new types of applications in crypto and I found that to be really compelling so I decided to build in the roll-up space and I was in Chicago so I was talking with a lot of Solana projects in particular and that's what I found interesting because even projects in Solana saw a lot of value in having their own chain and then having their own roll up in particular so that's what I wanted to build for it it's like apps that were really hitting the limits of throughput but also possibly needed other types of customization or economics that could only be enabled by their own chain

that's interesting because Solana builds itself as this chain with nearly unlimited transaction throughput but I guess what you're saying is as it turns out transaction throughput is not the only uh the only thing that people consider when building application the economics also have to make sense and you know aligning yourself with all the different layers of the stack like what were some of the things that um people building in Solana were finding Difficult about uh building on this chain that on the face of it was supposed to solve all their problems well at the time the two compelling use cases or two compelling examples were one was pith not so pith was an Oracle and Solana they were purportedly using

up like 10 of the throughput and they decided to spin off their own chain which is in basically their own svm Fork it's a full layer one that pith runs and they use that as their Oracle Network then they pass messages using Wormhole and then another example was powerlatcher which is also running their own Solana Fork so both in both of these cases uh the folks running the protocol either one felt like Solana's throughput wasn't sufficient or two they didn't want to pay that much in rent to a shared network like Solana instead they wanted to own the network themselves this way they can just incur that cost by running the validators themselves in power Ledger's cases of proof of authority uh and in Pitt's case I think that they they have a little bit less permissioning than that but the idea is

that these folks were going from uh like you mentioned a place that has basically unlimited throughput but you're a rent seeker and you have to pay Solana for every transaction to the most opposite extreme which is when you're the homeowner and they're paying a much higher fixed cost because now they have to run all the validators for this network and it's likely not economical to do so for an app that isn't running that many transactions but they have the the high fixed cost but very low marginal costs effectively zero marginal costs where every transaction is running on their own chain they don't have to pay anything so those are the two extremes that uh that monoliths and app chains were basically providing yeah I think the economics often get overlooked and we we look at blockchains I think a lot of the the narratives

around different chains and different uh ways to organize the modular stack Etc have revolved around fees and uh sorry yeah her revolved around throughput and not so much uh the the economics of running uh an application in a decentralized censorship resistant way what are some of the things that developers should consider when considering the economics of an underlying Tech stack the first thing I try to think about is what would the ideal fully decentralized version of an application look like so for example if you look at Helium Network they actually had their own layer 1 blockchain which they built

themselves prior to the existence of Cosmos SDK and it was a huge hassle for them to maintain it but in some ways it was closer to the itl in the sense that they were able to run everything on this chain they had this Oracle Network which now is actually centralized now that they've moved to Solana and that's going to take some work to decentralize it so start with that Assumption of what's the ideal version of it and then try to napkin math what are going to be the hot spots uh what are the places that you'll really access the same state because in general asset hot spots are irreducible costs and there's not really any way around that because if you have 10 people trying to access their right to the same piece of State then you need to have some economically Fair Way of

resolving that whereas if it's things that are actually just different parts of State then you can basically assign like a trivially low cost to that and then just kind of sum it up and think about is that cost something you need to collect on your own or is that something that you're okay with paying to some other intermediary uh do you want a hybrid solution where you can borrow the uh the security of something like ethereum in which case you might be able to leverage that Community but then of course you're going to pay a higher cost so you have to think about like where do you lie in terms of your preference for decentralization and what's the amount of costs that you're willing to pay whether fakes or marginal in order to achieve that yeah of course like the the um the Spectrum the on the like

decentralized the centralized Spectrum I think is an important one and is an important trade-off also when considering uh what kind of modular stack architecture to build on uh we can talk we can touch on that a little bit later I think but let's talk about the different layers of the modular stack and what is the role of each layer like this is something we've talked about here on the podcast but I think it's you know it's helpful in the context of this show to maybe remind people like what are those different layers and what is the uh the role of each layer yeah it's probably easiest to Think Through from the perspective of a roll-up so a roll-up receives a transaction and it executes it it computes some State transition and that's really what a blockchain is

ultimately right it's some State you apply transactions to it and you compute a new state so that's what the execution functions inherently doing it computes new States then the next part is you have to make those transactions available this is so that the rest of the network can also keep up and they can independently verify uh and compute their own State and figure out whether that state matches up with yours and then the last part which I meant this is why I wanted to keep start with the Roll-Ups uh which is that there's this last function that some folks dispute whether it's actually like a function of a chain but at some point like clients and other folks who are using that work need to figure out what's the truth and what's the version of the state route that we all agree on and that's the settlement process so in

the case of a ZK roll up uh once something is executed and a state route is produced then in order for that state route to be accepted by the quote-unquote settlement layer then you'd also have to provide some additional proof or evidence that that state group was computed correctly or in the case of an optimistic roll-up someone else like a verifier can obviously independently execute those transactions which were made available via the data availability layer and they can say I think if that scam group was incorrect and they can call for a fall proof and prove or demonstrate that that stay food is actually not correct and then it'll be rolled back so that's the settlement is kind of fraught because the question is like who's the one who determines settlement and who's the one who determines economical chain is it the

nodes who are running the chain and the users of the roll up or in the case of an ethereum roll-up people tend to think of it as there's some canonical bridge on ethereum itself and that's where the proofs are verified that's where fault proofs are verified as well and that uh bridge on this on the ethereum layer one is the one that determines the truth of the canonical chain but the reality is that if something really catastrophic happened on the roll-up or the bridge contract were compromised or something like that realistically the Roll-Ups Community would probably just fork and create to play a new contract and that would now be the source of Truth so it makes you wonder whether the source of truth is actually the smart contract living on the ethal one or whether it's the folks who are running the roll-up which more

realistically that's what my view aligns with so how does each layer interact with each other and what what is the like what triggers an interaction between the execution layer and the settlement layer and the settlement layer and the consensus da layer it's all pretty much driven by the execution layer and execution is a bucket term where for a rollout that typically comprises one sequencer or some set of some logical entity a sequencer which is ordering transactions then you might have some executor which takes that ordering which was committed to by the sequencer and actually computes the state transition if you're a ZK roll-up you need some prover or approver network which

generates the proof for that and then ultimately all these pieces are posted to the appropriate places so for a validium construction you might not even be posting those transactions to some publicly available da layer maybe it's some centralized da or a DAC or something but it's the execution layer that typically posts that block to the DA layer and then the state route and the additional evidence would be posted to the settlement layer or passed directly to peers in the case of a sovereign roll-up so it's all pretty much driven by this execution layer and so as we go from the execution layer to the DA layer or is there a spectrum of um decentralization versus centralization and risk of censorship

resistance versus you know yeah it depends on the part of the stack that you're looking at so execution really the key part there is the sequencer because anyone can execute a sequence of transactions once it's known so the question is which transactions were actually included in the sequence how are they ordered and that's what the sequencer determines and that's why the sequencer is often talked about as a point of centralization for Roll-Ups so if there if the sequencer were the only thing that could let transactions into the canonical chain then that would be a liveness risk because the sequencer could just disallow transactions to be included and therefore if you're an honest person you might be trying to get

your transaction in and it would never be included so that's why there's all kinds of other mechanisms like forced inclusion and other schemes that allow you to search your own transaction even if the sequencer doesn't want you to uh so that's that's the first part that could potentially be centralized and then the second important part is the da layer and that's the part where typically there's a lot of experimentation and there's a lot of different points on the spec on the decentralization Spectrum where the most extreme would be a completely centralized entity like Google cloud or an AWS or AWS S3 bucket or something you're allowed to do that right you can push your transactions to internet theoretically it could be censored there and the Other Extreme would be the ethereum L1 where it's quite expensive and it ends up costing something like

maybe 15 cents to post 200 bytes on a good day on the ethereum L1 and that's where the majority of it costs for an ethereum roll-up goes uh it's it's pretty much going to Da and that's why Celestia identified that that's the biggest bottleneck for roll-ups so yeah you mentioned uh or I mentioned in the intro your your hot take which is that settlement settlement makes no sense as a business and uh we were talking about this earlier and you said that optimism May net pay something like five bucks a day to settle on ethereum you can you can you expand on that and why why you think settlement makes no sense as a business including yeah well

just given that ethereum's like the most expensive layer one blockchain and optimism is a very actively this often is a main nut so this is the one that's you know probably handling millions of transactions it only posts a state route once an hour basically and so it's 24 times a day ends up being about five bucks a day and that's just not very much money it doesn't represent scalable revenue and it doesn't scale with respect to transaction volume and my view is basically that roll-up Frameworks at some point need to collect money from the execution layer that's the only reasonable plan they could try to compete on data availability but realistically the the quality of a data availability layer is just the security that it provides

which in the case of a proof of stake network is basically directly determined by the amount of stake that's uh securing the network so trying to build your own da layer compared to someone like Celestia who's a shared da layer for many Roll-Ups is likely not going to be competitive and then subtlemen like we were talking about just it's just a bridge really that you're posting State routes to and being the person who runs that bridge and collects that five dollars a day just isn't a profitable place to be even if you think if there's going to be thousands of Roll-Ups collecting five thousand dollars a day it's still not great and that's assuming you collect all of the you're charging the same amount of ethereum uh as ethereum and you're supporting all the Roll-Ups that exist which is unlikely so that's why it leaves execution meaning

that either you need to provide your own sequencer as a roll-up framework or you need to provide your own prover network if you're a ZK roll-up you need to maybe do the actual executor some way you need to capture value for that maybe it would be a shared sequencer or you could partner with Roll-Ups to the service and somehow capture value there but even I'd argue that if you're going to partner with a Rolex service more than likely the roll-up framework should just have their own in-house roll-up as a service which is kind of what we're doing at eclipse I see so what do is there like we we have there's an increasing number of data availability there so we have things like Celestia obviously you know

polygon uh Avalanche can can serve as a data availability layer uh ethereum so we have all these L1 chains on which applications can post uh their uh their state or proofs of the state is is there enough demand and do you think that that there's going to be enough demand to make it uh worthwhile for all of these uh data availability layers to exist like I sometimes think about like is all of this state going to be utilized to its fullest potentials such that like these things continue to exist be profitable you know generate revenue for for validators and remain secure

I think most people would be surprised at using an ALT R1 for data availability unless it's dedicated for that purpose such as Avail or Celestia which I wouldn't even describe as ones given that they're just a completely different product there's no execution they're not really competing with ethereum in that sense so I think I I tend to agree with you that Roll-Ups it's something I thought about too because like I want one Extreme as ethereum but a lot of apps that we support are high transaction volume but low value per transaction so it doesn't really make sense for those types of apps to be deployed directly to ethereum uh Celestia isn't quite live yet uh and we don't know what that it's a fee Market as well so we don't know

what the Celestia block space Market will end up looking like so the other alternative is some kind of DAC or something that's relatively centralized but then the question is like is it like more rational to just post those transactions something like Solana where it is also pretty much trivially cheap to post transactions to even if we're using it in kind of a uh unusual way which is just to store transaction data we don't want to use Solana to actually execute it we're using it as a data availability layer which you can use any L1 for and ultimately I just felt like people didn't really want that feature I don't know if people wanted to post to use uh Solana or one of these old l1's for ta because then they're viewing themselves

as a roll-off where one of the benefits is supposed to be that you're borrowing security from some highly decentralized blockchain like ethereum and now they're putting on something like polygon or I mean Finance was a highly controversial deployment of op stack oppnb so I think the technical folks tend not to like that solution too much okay so so do you think that you know the um data availability the data availability layers will in fact win out on um say ethereum as a a data availability there for Roll-Ups because there are

specialized it is uh you know or is there something else at play here I think you got a dedicated DNA layer just needs to provide much more data availability than ethereum because the The ethereum Da is just so limited even after 4844 and after all these different optimizations ultimately like there's just so much demand for D.A that that fee Market might end up being still prohibitively expensive for many applications so it's the last year's play has to be we can provide like five times more or ten times some like order of magnitude higher than what ethereum is able to provide and and then roll ups to some degree are forced to use it because it's the only other good da and what's also nice about Celestia is it's fully verifiable you can use data data availability sampling and you can make

sure that that encoding the read Solomon encoding is being done correctly and everything so um so you can make sure that it's not misbehaving basically foreign what other trade-offs should developers be considering when choosing a domain uh like a an L1 data availability layer on which to build their decentralized app I think it's really just cost and decentralization that's really what it comes to and it's the most important decision you can make for your app and what's nice is that if you're on a sovereign roll-up or some architecture like Eclipse then you can always switch that so that that part's kind of a little bit nicer technically you could actually do it for a smart contract roll

up too uh but the the disadvantage is you're still kind of anchored to some to a specific layer one if you deployed your smart contracts for supplement there cool so yeah let's talk about app specific Roll-Ups a little bit and you know we were talking about this before the show and uh you you uh you shared some some really cool articles that I'll link in the show notes and there was one article uh by um Electric Capital about understanding the the roll-up value accrual can you talk a little bit about the economics of of Roll-Ups and specifically like the applications themselves yeah so when you're on a roll up and you submit a transaction there's this base cost that you wouldn't pay if you were

on your own chain uh which is that you have to pay that rent and you have to post that transaction to some layer one and make it available so that's the biggest cost that anyone incurs when they want to roll up and that's the marginal cost then there's there's fixed costs fixed costs are things like in order to facilitate the roll up you need to have your sequencer running for example and that that's some hardware for an eclipse roll up maybe it's like two grand a month so that's 25 or 24 Grand a month a year which must be amortized or paid for through all of your transactions so if you have a million transactions it's basically nothing right if you have like 10 million that fixed cost gets amortized so in general it's less concerning another fixed cost is like the opportunity cost of the stake that's

being placed but for an optimistic roll-up maybe the executor has to put down some Stakes so that if you submit false State routes and they can be slashed um so that has some opportunity cost which I'd also throw on our fixed cost uh there's a bunch of factors like that and then uh and then they're back to marginal cost sorry I like made this kind of disorganized but then there's an execution fee so let's say there's a bunch of transactions getting the same piece of State like we were talking about earlier then the execution layer is going to charge a premium for that uh and then one more source of revenue for the roll-up or for the uh execution operator in general is Mev so that could be cross domain Mev across different Roll-Ups or it could all be within the same roll up or within the same chain

and that's one more place where the world can make money so there's a transaction fee part of that is cost part of it is the execution fee which is profit and then they have some additional money which wasn't even charged the user some sense it was kind of indirectly charged but uh that's the MEP that they can capture so that's the high level on how the cost breakdown for a transaction works yeah talk a little bit about Mev in robot space because uh this is this is something that I'm like not super familiar with like I I would at which layer of the stack does Mev occur it occurs uh during the ordering um so by the sequencer because they can remove transactions uh maybe they see liquidations happening the sequencer can stick their own transaction in maybe

they see uh something else that's for some kind of profitable Arbitrage between two different pools then they can take advantage of that Arbitrage opportunity and in the future maybe we'll see something similar to proposer Builder separation but for Roll-Ups where you have some big block builders that are basically bidding to include uh transactions by that from that sequencer uh and and you end up with a similar Dynamic to what we've seen in ethereum with flashpots so that's that's One Direction I can go in the reality though is as far as Mev for Roll-Ups goes or I mean for app chains is it's effectively zero there isn't that much Mev that happens for the vast majority of applications the couple times where there might be an exception something like the size of unit swap unit swap probably has quite a bit of Meb that

they could probably even recapture for their own protocol just given that there's so much trading volume that happens and Mev to some degree is just a function of the trading volume that occurs so because you know the more training volume there is there's probably more Arbitrage opportunities to make just between all those different trades and there's more constraints to satisfy so uh so that's that's kind of the way to think about it and then even further than that is cross domain Mev which is like even more far-fetched uh and doesn't really occur in practice yet so that's that's why I feel like a lot of these Arguments for shared sequencers around Mev really aren't the reason why uh they're going to make a lot of money ultimately the shared sequencer is going to make the vast majority of its revenue and the same with our regular sequencer

does which is just through execution fees right so it occurs to me that we're talking about like two different types of Roll-Ups here so there is a roll-up that can host multiple applications and in this case we might have Mev uh for instance like if you're doing a swap like if you have a like an mm built on that roll up and other applications that are uh interacting with that uh with with that with that at that amm and on the on the other hand we have like application specific Roll-Ups where here you're leveraging a roll-up sort of as as an app chain as a viewer your own app chain

um where is that that's that that uh sort of visualization is correct like you can have roll-up ecosystems as well as application specific role yeah I agree yeah there's like general purpose Roll-Ups and app specific Roll-Ups but I kind of view the med on those the same way because ultimately it's just like a set of smart contracts whether it's the same app or whether it's a few different apps that are even providing their own liquidity pool or something it can kind of be treated the same way Okay so does it make sense to build an app specific roll-up and like when when thinking about the economics and the complexity of doing that is it the same as building like it's not the same as like

building an app chain and going out and getting your own validator uh uh your your like doing a validated Roadshow Etc like you you do have different considerations when doing a specific roll-up versus an app chain yeah I mean the big difference is just that fixed cost difference so if an eclipse chain were a full app train for example instead of two grand a month for one validator and that would be two grand a month for 100 validators if it were a cosmos chain or for Solana itself it's like 3 000 validators so Solana Burns something like sixty thousand dollars a month and that's this is an underestimate because the reality is most validators don't just cost two grand a month uh then there's opportunity cost at the capital but so long is Burning uh like 60 Grand a month

wait is that is that even right yeah uh well okay yeah it's basically Burning uh like like an or like uh 3 000 times more uh than a single Eclipse chain is my point so uh so yeah that's that's the fundamental difference that you're paying this huge fixed cost when you're running your own L1 because you have to pay for all the validators that are running it uh and usually it's paid via inflation so you have to inflate your token or you've just somehow compensate those validators for their work uh whereas if you're a roll-up you just have to compensate this one sequencer or one executor which is which is typically much cheaper okay yeah I think the semantics are important because when people talk about Roll-Ups and blockchains and app chains it feels like these terms can be used

while meaning different things that might have different levels of sovereignty at different cost uh Etc so like for example on on eclipse on the website you say you're you're allowing people to build customized blockchains I was reading that I was saying like yeah kind of but it's you can also build decentralized applications that leverage um uh that leverage the security of like another system and that's not necessarily the same the same as like if you were building your own Cosmos SDK chain so I wonder like do you think that you know generally speaking as a space like what what kind of um confusion still exist around the semantics of all these words yeah that's

your blockchains of the catch-all term uh and yeah technically like you could even take a roll up and one day you could make it its own solver in L1 if you wanted to uh you can't really go in the opposite direction take it take a roll up and somehow make it a smart contract somewhere um that's that's a little bit harder but yeah I agree that basically it's it's about what does the user want and what type of app are they building if they're building a consumer app that's running millions of transactions and they're all low value there's pretty much no world where deploying to a shared chain would make sense because those transactions are just too low value to justify paying a transaction fee and being on their full layer one wouldn't really make sense either because being on a layer one blockchain means that they have to pay this huge

fixed cost which they likely can't afford so the only real solution is have some minimal number of nodes I have some settlement process something like a DAC for data availability and now this kind of application can finally be enabled on chain and a quasi-decentralizer arguably fully decentralized fashion so that's the kind of app that makes a lot of sense another type of app is just something on ethereum uh where they're used to paying high fees and they want to be a roll-up on ethereum where yeah you have that 15 marginal cost which is going to go down over time and for many apps that's too high but if you're a D5 app on ethereum your users are already used to paying like 10 bips or something or something higher than that so that that marginal amount of cost probably won't make a difference and you'd much

rather capture that yourself as a as an application maybe even capture some of the Mev that occurs within your app so that's another kind of like point in the Spectrum and another use case for having your own app roll up I mean do you think that like we're talking about Roll-Ups here and like this this modular stack and you know I I think like if you go up to the other end of to the far end of that Spectrum you know uh a roll up can graduate to becoming its own roll app and then a roll app can graduate to becoming an app chain and you know in some ways like dydx is an example of of that that that progress from like one construction to another um do you do you think there's still

room for you know vanilla smart contracts on a general purpose chain uh and is there a graduation path between smart contract and roll up yeah uh on the dydx example that's an interesting one because I actually our chief business officer Vijay was the head of BD dydx at that time then was a had a beauty at uniswap after so always a lot of discussion about those guys potentially moving to their own chain and I think a large reason why dydx uh moved to its own L1 is because the app roll-up infrastructure just wasn't there at that time so being on star core

wasn't really like its own app specific roll-up I was really like a shared chain in which case in some ways you're kind of getting the boat the worst of Both Worlds where you're not receiving the rent so you're still paying to someone else but you also don't get to um like you still like don't get to have the full security or just the full decentralization or the argument the ethos of being on ethal1 or the reliability of that so you almost want to hit something that's either like your own app specifically roll up or uh being on your own like chain which you own or being on a shared layer one chain in my opinion there's not really I don't see the volume being in one of these shared Roll-Ups and that was really just like an intermediate graduating graduating phase for a lot of roll-ups so uh so

that's the first point on that and then the second point is whether there's like a graduation path I I think they're that's like exactly what you're describing I think we're gonna see it time and time again where people move from some share chain to a roll-up to their own chain uh but what's interesting is examples that don't move in that direction like helium where they have their own chain to start and then they move to the reverse direction to Solana but I think if helium had started with the cosmos SDK it's possible that it never would have happened because I think a large reason why they moved to Solana is because the tooling and the infrastructure for their own uh their own proprietary L1 just wasn't there so it was a huge hassle for them to maintain it yeah I think that also I think we'll see

this reverse graduation happen again where you know the cost of securing a a a an app chain and you know the app chain thesis I think was was correct in in its uh in its ethos of having sovereignty over uh over the entire um application but now you know with with data availability and Roll-Ups I think a lot of those considerations are are are now less relevant but um but the cost will continue to go up so like one example that uh we've been talking about here on the podcast is Juno right so like Juno you used to have quite a bit of security when when the

token was worth 40 bucks and now Juno has something like 10 million dollars securing the chain and stake but it it hosts essentially kind of like diagnosis safe of of Cosmos uh which is meant to hold assets you know if if the um if the cost of of uh of attacking Juno is considerably low that puts those assets at risk it might be wise for Juno to you know people were talking about this on Twitter that Juno's are just like Rebrand to Dow Dow chain and then you know be secured somewhere else like move to Interchange security or like become a roll-up or something like that

right that would make more sense and it would actually be like quite useful for the community to have that um that Dao is a service kind of infrastructure but at the same time be secure so yeah I I think probably we'll see that reverse graduation as well as chains figure out that like hey the cost of securing this thing and maintaining that high level of security is going to be considerably high and you know some chains are probably not going to make it yeah I totally agree that's why I think if the role can't go to market it should be something like that which is Celestia's roll-up framework because it's basically taking an ABCI client implementation and turning into a roll-up which is perfect for taking a cosmos SDK chain and doing that or facilitating that reverse graduation whereas if you're something like an app

or uh that was on another chain or like Solana or the EDM or maybe you had your own bespoke L1 uh such as what power Ledger did then I think something like an eclipse chain makes much more sense as a graduation path rather than trying to force all your stuff into a framework that you're really using to begin with which is the cosmos SDK so so yeah I I think at another I mean use case for these show chains is just that uh I mean the fixed cost is zero so if you're a consumer app that's just getting started it really doesn't make sense for you to pay even like random ons or a couple Grand a month you just need to get started and start playing around with something and once you realize you've graduated past that point on a shared chain then you should probably go onto

your own roll up or your own your own app chain or something yeah this is super cool I'm I'm really learning a lot here about uh about you know like the mental model for for Roll-Ups and app chains so yeah let's talk about some misconceptions what what are the biggest misconceptions about Roll-Ups in your view I think the big one is uh what makes something ethereum aligned and what makes a roll-up ethereum align and people don't realize that it really is as simple as you post your transactions to ethereum and you have a bridge there and those are the two things that make something an ethereum roll up versus something else and there's there's nothing else it's not like yeah you could have an evm as well that that might help it might make you a little more ethereum aligned but

beyond that there's no special sauce uh and that's why this distinction between smart contract settlement versus Sovereign settlement is kind of a false dichotomy because ultimately even if you have a bridge that's anchored to ethereum like I was mentioning at the beginning of the podcast if some horrible thing happens or for some reason your community just disagrees with the result on that bridge they're just gonna fork and you're just going to deploy another smart contract that agrees with your new consents your new social consensus so uh that that's probably the biggest misconception to me and then the second misconception is people seem to think that blow-ups just make everything cheaper so if you deploy to roll up to Solana or something that would be cheaper but just from our argument and our the economic thinking

that we've been doing earlier on this podcast you can kind of deduce that a roll up on Solana can't get cheaper than Solana because now you're you're just you've just added the fixed cost for running the roll-up which must be amortized over the transactions and you're still paying for the same data availability which was ultimately the biggest cost that you were paying anywhere so you're not going to get cheaper than that yeah yeah I mean I mean that that totally makes sense right like you're building a roll up on top of on top of Solana you're going to have to add the cost of the data availability to the to the to the cost of settlement um and uh you know there's there's a post you sent me um by um by the guys over at uh a DBA

where they argue that like Roll-Ups don't exist can you unpack that a little bit Yeah so DBA is um one of our biggest investors uh and yeah his argument is kind of that ultimately the roll-up is just that state transition function and that's what defines any blockchain or any Roll-Up and then you have you all have to be aware of what those transactions that are going into the roll-up bar but beyond that it's not like roll-up has rollups has some special property where they're borrowing Security in any like miraculous way it just comes down to the assumptions you're making about how the community is communicating information and that's kind of what that blog post is getting at cool yeah I'll link to that in the show notes

so yeah let's let's um focus in on Eclipse a little bit here because we've been talking about this high level stuff since the beginning but I do want to talk about eclipse and how it works and how that network is is um is organized and uh who are the network participants so I I get the I get the feeling like maybe I'm gonna maybe simplify this a little bit I feel like Eclipse it can be summarized as a way to connect different roll-up VMS to different data availability layers it kind of sits there in the middle and it allows developers to choose the VM on which they're going to build their application and also the data availability layer on

which their transaction data will be posted yeah that's exactly what we're building yeah that's right like under the hood it just executes all that by code using the same virtual machine whether it's evm or fuel VM move whatever bytecode it is it's always using this Berkeley packet filter VM which is highly parallelized and there's all these nice properties to it um but and that's the only like implementation detail that's really relevant for users to know because it means that they're going to get a lot more throughput than what's available on other chains typically so yeah describe the architecture so what what does it look like and who who are the participants that are you know making allowing this this connection

between VMS and data availability there is to work yeah so we basically took the Solana code base and threw everything out but the virtual machine removed the vote transactions which is kind of a contentious topic but if you look at Solana's throughput or the number of transactions running through it there's like some base level of about 3 000 transactions that are just continuously run and that's because those are vote transactions so those are the validators that are voting and I I was talking about how there's about 3 000 validators earlier in the episode so that's basically where that's coming from so we removed that given that it's a roll-up and we don't need to do consensus uh and you have one instance of that that's running which is ordering transactions in the future that could be a shared sequencer such as espresso who we're

working with but someone orders transactions right now it's that virtual machine uh as transactions come in it's just doing first in first out it execute some produces the result and then it posts it to the Salomon layer so the settlement layer for eclipse is basically another roll-off but it's a stubborn roll up uh it's an optimistic roll up so you have some verifiers for that settlement layer and you have some operators uh and that and they operate in the same way they order transactions compute the result and they uh and they circulate the state route and then you have this these verifiers that are basically ensuring that they're being honest so you have two sets of executors two sets of verifiers and the blocks are alternately posted to whatever data availability layer you chose that could be Celestia it could be somewhere else

but what's nice about this architecture is you could remove the eclipse part of it so you could have some executor that's just Computing State transitions posting blocks to the L1 to the data availability layer and anyone can still pull those transactions from the L1 recompute the state of your chain and and they are also a participant of your network you never needed to touch eclipse and eclipse is the settlement layer is just sitting there as a convenience feature it's just if you want to rely on some other set of nodes in order to figure out what's the canonical chain then you can do that so so can we come back to the network participants so you have the the um the sequencers and the verifiers yeah so uh so right so this is assuming

an optimistic setup so you have a sequencer uh Slash executor which exists for every app specific chain and you have a verifier for each of those chains or some set of verifiers for each of those chains and the state routes that they're verifying are the state routes that were posted to another roll-up which is the eclipse settlement layer and that roll-up receives State routes and again has a set of executors a set of verifiers and that's what keeps the eclipse settlement layer operating correctly and then all these transactions are posted to l1s and these all ones act as data availability layers there's our honest majority it's just like a normal L1 operates so those those are the network participants uh in the future there will be more participants where the sequencer might be a separate entity

you might have Builders the bid for those sequencers for they might bid to have their transaction included uh you might have like RPC operators there's all kinds of other participants that aren't as like critical but this is what the system looks like right now okay so you said that there's a sequencer for every roll up on uh on Eclipse so yeah each and everyone gets a separate sequencer yeah right so so is is that sort of part of the service that uh eclipse is providing is it like if eclipses kind of sequence or like roll up sequencer as a service that's the right way you know yeah Roll-Ups to the surface is kind of a weird term and I kind of hate the term for for the reason

that you're getting at because the core of the business is actually in the case for most roles for the service is providing some kind of isolated sequencer as a service the alternative is they can also provide a shared sequencer and those would be competitive with each other but the thing is that a shared sequencer can there's limits to how many change shared sequencers can support because the shared sequencer is only as good as the atomic guarantees and the composition that can provide and ultimately in order to provide Atomic guarantees and composition you need to have some awareness of the state across all these different chains so you need to have some big block Builder that's executing all the transactions on all the chains that are being supported not unlike interchange security for other Cosmos Hub very similar like you need to cut somehow have someone who's able to

know the state of all these chains and that ends up being very expensive to do so each marginal chain the marginal cost should increase and the revenue for each change should be decrease because presumably the shared sequencer picks the most profitable chains to sequence for first and then each chain after uh thereafter is just less profitable so there's some break-even point maybe at like 100 or 200 chains or something where the shared sequencer is it's no longer profitable for it to support additional chains and those chains will have to be supported by isolated sequencers could that be part of some tiered service offering right where you get like share sequencer for like a lower cost the same way you might have like shared hosting you know in web 2 and then like a dedicated

sequence sequencer for applications that are willing to pay for it like does that make sense at all or I mean is it I think that's true like the you'd figure it should be something cheaper right because it's shared but but the shared sequencer gets a little bit of a bit of power there because they're going to be like why am I doing the work of sequencing for this chain for such a small amount of money so uh because they likely there'll be some governance process to be included in the shared sequencer such as the optimism super train or one of these other shared sequencers that is they're going to pop up uh and that's going to be the decision making process that they go through and then in the same way as for Cosmos Hub when they're thinking about whether they should provide interchange security for a chain in some way you're

kind of making a bat on that chain that you're uh that you're providing validation for okay yeah that makes sense um okay this is interesting uh when it comes so you meant you mentioned that this is a setup for for optimistic uh setups yeah so the only difference for a ZK roll up is that now you have some prover Network or you have some set of provers and those are those are the folks who are likely alongside the executor uh who are generating the proofs so there are these hybrid startups where maybe you submit a state route and it's optimistic until someone submits a state until someone submits a ZK proof and now it's finalized immediately there's all kinds of hybrids but in a simplified world you just you'd

have no verifiers in that case and you just have provers okay interesting well so how how are these participants incentivized like what are the or is there what are the tokenomics and um who who's paying these uh these Network participants so it's probably easiest to go from cost center to a revenue Center so the data availability layer is incentivized because um it's just a fee market so if there's more if there's more demand for uh storing or for making blocks available on Celestia then uh then the price will go up and that will incentivize more people to participate and if there isn't that much demand then likely fewer

people will validate until um until it basically becomes economical for them to do so so that that's the thinking behind that cost center the settlement layer uh and the verifiers the thinking is that people will want verifiers when they want the chain to run correctly so if I'm a gamer and I really have a ton of money locked up in this game then I would run a verifier myself as a player or maybe I would have my own RPC node which is keeping up to date and I submit for my transactions directly to that something like that so it's the participants of the network in the community who's running those verifiers so verifiers aren't compensated is basically what I'm what I'm saying with that sentence uh and then provers and people who execute

need to be compensated for that work so they would likely be paid some fee which would uh the top of the line is through the sequencers the sequencer charged the charges the initial fee to the user in order for the transaction to be included but that fee would be distributed among the executor and provers and whoever else is involved in the network foreign so the verifiers are are not um incentivized because they have an incentive to to verify so I think we we've kind of understood that they're somewhat excluded from the economics of uh of how uh fees trickle down from the application to the to the DA layer yeah right now there are but the tricky part is that it

introduces this like crypto economic problem which is it's only it only ends up being irrational for there to be like one verifier because if you have two verifiers now your probability of success is lower um but your your reward is not increasing for uh for being a verifier so it all kind of like in converges on these uh really simple numbers okay so let's let's use a a practical example I'm playing some game that's built uh maybe it's like a roll app uh and and it's leveraging it um it's it's uh leveraging eclipse and you know Celestia when when I I have a wallet that wallet might have

the the tokens from that game um like the games economic token how how is how are the different participants uh so the um the sequencer and then finally the da layer how are they getting paid are they getting paid with this in-game token or is the roll-up itself having to um do some sort of transaction to you know convert those tokens into you know Eclipse tokens so that the sequences are getting paid in the native token that they want to be paid with so that that that's unclear I think yeah it's more like the latter but uh it's that the sequencer receives fees and whatever the

L3 specified or whatever the app specific chain specified and it will do all these conversions in order to pay the TA layer and the Celestia token for example it'll do a swap for the eclipse token in order to pay the eclipse settlement layer I had to pay all these different participants okay so the sequencer generally agrees to be paid in the token of the layer three uh and it has to uh do this sort of Treasury Management in order to pay uh Celestia yeah another reason why the shared sequencer is basically taking it back because they're collecting fees in the native token too for each of these oh okay yeah okay this makes sense and where does eclipse make money in all

this so we're going to have to provide some kind of sequencing whether it's isolated sequencer as a shared sequencer whether we share Revenue with sequencers but that's the only place that makes sense in my view okay can you expand a little bit on your thinking there for foreign revenue from that as well I just don't anticipate that that will be the bulk event ultimately the big service that we're providing to people is this roll-up framework we're providing like all the tooling for deploying chains basically and likely some convenience features for sequencing because that's the hard part for running all the infrastructure related to a roll-up you

don't want to make it like go on outage if you have to upgrade it so that's that's what eclipse is doing and the eclipse Network would be compensated for that part I think that's the core part that a roll-up should be trying to capture value at overall framework so in these Market Maps or Market segmentations really like every roll-up framework needs to be in another category as well like I hate it when someone says like hey um opt op stock or eclipse is just on a roll-up Frameworks we should be there but we should also be under sequencing or under like some other area depending on where we intend on accruing value yeah can you talk about the roll-up framework a little bit Yeah Royal framework meaning just whoever developed those core components in order to run or

roll up so whoever wrote the sequencer logic whoever wrote the virtual machine the settlement logic that's software it's open source but these companies also aren't Charities so they need to somehow accrue value that's why I push back on the public good narrative a little bit because roll-up Frameworks can't be public goods only they also need to make money somehow otherwise these these projects didn't make sense to invest in they wouldn't it wouldn't have made sense for even people to buy the token right now yeah yeah um it's and so is is Eclipse compatible with you know any VM and or other team like I mean I know the dimension is building

this whole rolapp thing that'll allow you to do like a cosmos SDK as a roll up and evm as a roll-up what are some of the other uh roll up layer um VMS or packages that will will be compatible with uh Eclipse so right now we have this highly paralyzed evm and that's the vast majority of eclipse chance and then we also have an option for Solana VM so that's what we ran for injective the Cascade chain which is a Solana VM roll-up for their Community uh we have a signed agreement with polygon with BNB chain a handful of these other chains and those are all SPM chains so this is like a way for these different l1s to bring a different virtual machine to their Community they can better

capture Solana developers or developers who are more comfortable in that highly parallelized environment so so yeah that's that's kind of the reasoning behind why they would want something like that we have move and development too you have movement development too or who are you working with for that uh that's actually something that's Solana core themselves uh is actively developing okay so a lot of core is building a movie in yeah uh it's move by code for the salon event yeah oh okay interesting yeah what are your thoughts on like the move ecosystem I think that there's not much of an ecosystem at this point but move is a language I think is really cool and I think uh it's probably something that like it's a different

it's just a different way of walking and memory management and stuff so I think it's desirable to have a wider design space for different VMS and I think that's part of what makes roll up so great so I mean it's it's kind of like fuel in that sense like I think fuel is also a really interesting virtual machine yeah we hit fuel in epicenter and um yeah I also thought it was like a very interesting uh VM when when thinking about interoperability you know um obviously we've talked a lot about igc here on the podcast and like I think as a standard I mean it is really the only interoperability standard of those today and it is gaining Traction in the space help us understand maybe the

the um the the the the layers that are in that are interacting with IBC so you know if two Roll-Ups want to interoperate like two Roll-Ups that are built on Eclipse um would there would there be some uh some ability for those sequencers to to facilitate improbability between chains or would you still want to leverage something like IBC for the application so the applications can just send messages to each other I guess what I'm saying is the fact that two Roll-Ups are built on Eclipse is that facilitates some sort of interoperability that would allow those those applications to

forego using ivcu for like some message passing yeah I mean the composition is via the shared sequencer because that's Atomic composition but definitely next best is IBC and we're implementing that with polymer and actually there's a couple of efforts going on I think Skip for some period was thinking about having some sort of IBC validator set where they were somehow incentivized to actually relay the messages because I think that's an issue for a lot of these IPC chains sometimes the messages get stuck lingering there no one's actually doing the relaying and that can prove to be an issue in some cases I think they might have gone back on that plan though and then Jack sample and sent something kind of recently uh which is again like an ibca like Hub

type of thing where it takes State routes from Roll-Ups but to be really useful it actually we'd wanted to re-execute all the state transitions for the roll-up too so it actually achieves finality itself and once it just to like take a step back like what am I talking about if you look at like an optimistic roll-up optimistic Roll-Ups actually never get rolled back like the the chain itself for an Optimus role at the moment you submit a transaction it's included in that chain it will never be removed but what gets rolled back is the bridge so when you're bridging stuff that's when achieving finality is important that's why you have to wait this challenge period on ethereum because you're basically waiting some period of time until you have sufficient confidence that there was some honest

validator on ethereum that would have accepted a fall proof but if you have your own like committee like your own bridging committee you can just re-execute all those transactions for an optimistic roll-up all the way up to the tip and you can figure out right this is the correct State Route I know that this thing's never going to get rolled back and even if some Bridge gets rolled back they're just gonna get rolled back and recompute this same state route that we just computed so that that's basically what um what I would want for like an IBC bridge for uh for an eclipse roll-up and that's what we're doing with polymer and possibly with Jack sample is working on yeah we just had polymer on the podcast last week and uh still wrapping my head around that one to be honest

like I think it's um it's it's also I think sometimes a bit difficult to wrap your head around the mental model of like what what polymer will look like as a network yeah that's a really good question and like the economics of polymer I'm curious about that too yeah cool well you know I'd like to maybe take a step back here before we wrap up and um you know one thing I'm wondering here is as the Space is really you know embracing this modular chain modular design uh architecture and narrative do you think there's still room for like

monolithic chains and I'm thinking specifically for example like Neutron who just launched a few weeks ago and um you know sells itself as the most secure smart contract chain in Cosmos of course there's lots of other smart contract chains and we talked about Solana and ethereum obviously still continues to exist as a smart contract chain do you think these will continue to make sense over time is there still like a use case for monolithic chains so either in one case the monolithic chain especially a new monolithic needs to be prepared to take a loss and maybe there's some other reason why that chain exists maybe it's a public good or something or in other cases you need to napkin math what's the cost of running

this chain if you maybe have a validator that costs 20 grand a year or something you have 100 validators and you're burning 2 million a year so the question is are you going to bring in two million dollars of transaction volume what's your average transaction cost going to be your average transaction fee and how many transactions does that imply and do you seriously think you're going to get there and if you don't then it likely doesn't really make that much sense to be on your own full layer one blockchain as far as existing monoliths I think if there's just these Network effects that have already been built up I think it's something like the economics of Salon I could theoretically work out if they if they can really like 100x or transaction volume or something yeah I mean there's still like tons of

Monolithic chains and like lots more that are going to launch I mean like Aptos and sui and uh and like just in Cosmos there's like six or seven like evm and and cosmoasm and like soon some move ones and um yeah I think it's I think with the the economics is a big question but also the network effects and will they be able to attract sufficient amount of applications and um yeah um I think probably in five to ten years from now many of them will probably won't be around anymore or will have downgraded you know or say like this this reverse graduation to like some kind of roll-up uh yeah and if not then ultimately it's just the token

holders who are the losers because the validators will still get paid it'll just be the network will be inflated uh and anyone who just wants their token is gonna lose out yeah and so what are your thoughts on the atom economic zone idea and you know broadly like ICS interchange security yeah it's like another point on the fixed versus variable cost Spectrum uh for change to consider it I think the part that's a little bit like not so great is that you have right now we have to be Pat you have to go through this governance proposal and you have to actually like be accepted by the cosmos Hub I think another issue is just that I don't know how profitable like some of the chains have been so even from the cosmos Hub side I don't know if it makes

a lot of sense uh but I mean that's why I kind of like the concept of a roll-up where if you're if a roll-up is paying an L1 money or vice versa if the L1 is capturing value from a roll-up then they're both benefiting in some way because it implies that a transaction was actually posted to the Roll-Up and that some user did execute something and that transaction is not being written just as it's needed as opposed to an interchange security formulation where a validator is just continuously validating for this chain even if no transactions come in you're still burning that cost so it's I mean there's there's also advantages to doing it that way but um but yeah that's that's something that's I think that's a pretty key

economic difference between the two formulations yeah in terms of like ecosystem uh I mean I think the way the way I see the atom economic zone and increasingly like people are thinking about this is they see it as this um this this nation of applications that are somehow you know complementary to each other and will form an ecosystem of applications that leverage them right so they're like kind of creating their their little moat right around uh applications like Neutron and and and and um and Noble and like Skip and you know maybe maybe maybe Dao Dow chain right well as a as a

um as a utility in the atom economic zone and I think if that narrative can become stronger and um and really attract developers than it'll make sense for the academic economic zone to continue because you know there will be increasing Demand on on the token Etc on the economics will make sense but I think that that battle isn't quite won yet and I think it's you know Zaki was was I was talking to Zaki recently and he was saying basically like we have 12 months to figure out whether or not this is going to survive um the next cycle yeah I think Cosmos has a good shot I think the parts that confused me they were like how does it really increase demand for the atom token given the vast

majority of atoms staked so like the atom that's just sitting there lingering what's going on with that and like why is that useful and the second part is for each additional Cosmos Chan are they really adding value to each other or are they more like isolated communities because it's not like in the case of roll up you are all using the same layer one and that incentivizes more people to you know validate for that L1 and it becomes more secure overall there's no like Network effect like that so so yeah it makes me wonder uh what's what's the thing I think that's a more important thing like how do you build a flywheel of network effects around the cosmos ecosystem rather than like sovereignty is such a key feature of the ecosystems that might not even be

possible but rather than letting all these communities to exist in like entire Independence yeah I think that's totally right I think I think the the the the the flag wheel question is yet to be answered and uh yeah I hope I hope uh the community and the ecosystem and and the projects building there kind of figure that out soon so what's the road map for for Eclipse like what uh what can people expect in the next couple months and where can people go to find out more about building on Eclipse next couple months the biggest thing is our self-service offering so we're gonna have some because right now all the eclipse chains it might be surprising because we have

like 30 of these chains running but all these teams have talked to Eclipse they've like signed a service agreement they've gone through like a fairly manual process but now we want to make that whole process more scalable so that's what we're doing right now we're building out this offering where you just put in your credit card details or paying crypto or whatever you can get some free trial and you can get your own Eclipse chain and try it out yourself uh and that's a test net if you want to go mainnet you'll still have to talk to us because mainnet has some other considerations around it you probably want to get it audited if you want any customizations and there's other stuff that we'd want to warn you about but um but yeah that's basically the biggest thing that we're going to be doing the

other thing is historically we've done consumer which is gaming and social done a lot of physical infrastructure networks such as helium type things react wind uh we have a bunch of those but the last area that we've just started expanding into more is D5 so like constructions where block times are especially short where it might even be like fully ethereum aligned where the transactions are all posted there historically it hasn't made a lot of sense just because it's so expensive but there are ethereum apps and especially DPI apps that have graduated to the point where it does make sense because and they're they're just paying egregious amounts of fees for no reason so I think we can really solve a pain point there yeah something I forgot to mention

forgot to ask you about here is you know on your website it seems like you mentioned gaming quite a bit as as a use case for eclipse is that you know part of the go to market for eclipse and I guess my you know the follow-up question that would be how big do you think crypto and web3 gaming will become in the next cycle because it feels like you know we run an infrastructure fund we we do Focus quite a bit on like the infrastructure layer and I I've noticed a drastic increase in like the amount of infrastructure projects that are talking about how their infrastructure is going to power

web through gaming but I have like very little kind of visibility on the on the gaming space so yeah what's coming in that in that space yeah part of what sparked that was in the early days for collapse I was just chatting with projects and different verticals I found small D5 projects that wasn't really economical and a tea products in general it didn't really make sense though we are off the Record working with a pretty large nft uh project and we'll probably release a lot of paper on that soon um but the the category where there was a lot of interest was often gaming just because these guys are posting so many transactions on chain and one founder I've been at Worlds really convinced me because he's describing how he's using the blockchain as a public or open API for his community to build additional

composition on top of whether that's financialization failed their own user generated content which was a huge bottleneck for many games uh and it's kind of like a standardized API that people can build tooling on top of Evan was one of the developers of Unreal Engine 5. so he's like a traditional gaming guy and he just saw so much value in having this kind of product and so he was an early design partner for us and hammering on what is the economic structure need to look like for someone like him and also how can we make this generalizable for other games foreign great well Neil thanks so much for coming on the show today I've really enjoyed this conversation you've uh really helped me and understand the um the architecture and like how to

reason about the mental model for uh for Roll-Ups and um yeah it's been it's been a great conversation thanks for having me