One Rollup or A Million Rollups?

Transcript

what's going on guys welcome back to another episode of the rollup today's episode is with Terry the head of strategy at eclipse eclipse went main net today for developers with the Solana virtual machine this Solana VM settles to evm rollups and so a lot of people have said this is the best of both worlds Solana performance with ethereum security we grill Terry on the ins and outs of the eclipse design we talk all about why Eclipse could be the one rollup to rule them all we also talk about the alternative pieces in the app chain thesis and why a lot of applications are going through the app chain life cycle and graduating to be

their own app chains we talk about all this as well as shared sequencers and what the incentives are to decentralize or migrate to Shared sequencers sit back enjoy and first a word from our sponsors k is solving the longterm data availability problem where a lot of the data availability providers today in the modu space when you post data to them they actually only hold that data for about 12 to 16 days and then it's basically they can go into archive note or often times it's just kind of put away into the abyss and what coen's trying to do is build this long-term data availability solution which can

plug into some of these modular data availability providers and provide this longer term Perpetual data solution so what we try to think of this as is when you post transaction data onto Celestial or Vil or any of these modu da Solutions or even ethereum coala is able to provide a long-term data storage and data availability platform for that data after the time period blockchains are Billboards not databases and you know we need these providers in the space to be able to kind of hold on to the longterm data that is providing from these providers and these blockchain State this is where coent comes in they just had a nice Rebrand check them out on Twitter as well as the website thank you

for being here hope you enjoy the rest of the show what's going on guys welcome back to the rollup today is Eclipse main net for devs so devs are certainly going to be doing things on the eclipse Network today h fresh off of a Rebrand today I think we're going to talk a lot about some of the technical content that Terry presented at majo Summit it's particularly interesting I think me and Nick from Celestia were sitting in on the on the talk and uh there was a lot of math equations on the presentation so a lot to kind of digest here Rob GM to you and uh good to be back yeah GM indeed um big fan of math equations um didn't get a chance to see Terry's

presentation at modular Summit but I'm pretty stoked to learn more about it now um also congrats on mainnet it's a long time coming it's uh quite an accomplishment so we're stoked um and yeah Terry welcome to the rollup good to have you thank you for hosting yeah man you got a pudgey penguin how long have you had that penguin for oh uh it's been like two years I think um actually got it as a gift uh so shout out to the GPS of 1kx oh nice peter is a is a legend yeah those guys are great and they've been investing in applications long before that was you know consensus yeah I mean that's their thing now um and it's been and they you know

they did pudgy Penguins the seed round and whatnot and they're extremely bullish just talking to Peter about that so that's cool Rob and I are penguins uh Rob changed his profile picture recently but he's still Penguin at heart pH Penguin at heart so Tera how did you join Eclipse what's your kind of defi uh story um and then we'll just jump right into it yeah for sure um I'm one of the few the class like 2020 and 2021 that actually came into crypto through gaming right there was a huge wave of people that positioned um or you know had the thesis that all gaming on board the next ban users like I was actually one of those users um so I used to play axi Infinity um before it became kind of a a play to earn farm and I was pretty into classic

uh and in the process you know I ended up buying AXS and transacting on Ronin uh and then going and down D SL you know more so the gaming and nft rabbit hole um I just ended up writing about it what I was interested in and suddenly you know some people reached out asking if I was interested in a research and investing role ended up joining 1kx um there ended up doing both gaming nft Finance research but also infrastructure research specifically in ZK scalability and then kind of throughout time you know we were looking at the progress of zvm the progress of Da layers realized though that you know execution was a bot like there were serious problems that applications were demanding the Bas layer solve and saw eclipse kind of the uh the place where all this was coming

together right the greatest uh where I felt like the thesis was playing out and so ended up joining a clips about two months ago three months ago at this point nice awesome and uh pushing devet today for yeah main net for developers and then uh main net for users soon as well so congrats on that and yeah let's dive in so uh my first spicy question to start the Pod is with regards to this idea between one big rollup or a million rollups and how Eclipse went from being kind of a big uh proponent of this idea of modularity and kind of scaling block space and scaling new chains to kind of coming back to oh wait if we just have a really performant VM settling to

ethereum we don't need all these app chains and all these rollups how did we get here yeah the first realization comes from the fact that you know if you're an L2 fundamentally I'm with the opinion you're selling execution space uh you're selling block space but specifically execution right because you're Outsourcing cons sensus and data availability you're definitely not selling that then what are you selling right you have execution that's left you're not even selling settlement so then if you're selling that you should be the best you should be the best place where people can get execution and that means it should be highly performant and it should be at a low cost and it should not be congested right and if that's the case then there's no reason for you to horizontally scale by creating duplicates of yourself uh the other realization and the second portion is

that we went deep and studied dsvm um and kind of some of the benefits of the svm and one of the biggest benefits that at least we took away with is that it could support multiple concurrent hotspots and if especially if you had very large blocks then you could even increase the number of concurrent hotspots that were possible and if you take these two elements to combin together uh then the out outcome is that way why would you scale horizontally by like copying yourself and then incurring additional overhead and trying to piece those together and or you know um you know create friction for Bri in friction for users friction for different applications when you could just be you know one rollup so it it is one rollup however the the thesis in mind is

horizontally scale with parallelization in concurrent uh block space that has localized fee markets because of the svm that doesn't kind of impact one another so you can give each application within Eclipse its own kind of fee Market or block space uh via parall parallel parallelization but rather than having them all kind of sharded in different execution environments and threads which require bridging we can just do it all in one uh but it almost has the same outcome as doing as having a ton of app change but just better interoperability yes exactly is that kind of like the right mindset yeah that's absolutely the right mindset right like because I've talked to a lot

of applications and when they move the software block space when they move to an app chain they really want two things right they don't want congestion and specifically they don't want their fee markets and their users and their revenue impacted by someone else um having a lot of usage right like if the igen lay if team is having an airdrop I don't want my D5 protocol to suffer and my Revenue to suffer um and so uh we talked to a lot of applications where like oh they were on polygon but polygon gas V were rising in order of magnitude and over time that made their users um unsatisfied or it made them churn some of their users and so they had to leave for their sofware block space on the other hand the second reason a lot of people leave is because they want value

cruel uh and whether that means specifically like we want to be calm to an L2 token L3 token uh or um you know we theorize that in the future people are going to care a lot more about Revenue therefore we want execution fees Plus meev either one a lot of people have you know thought okay if we move to our own app chain we're going to get additional value and that has largely played out as well with things like dgen chain or you know DMT the meme coin those are interesting cases um so our thought is if we can provide those two in one L2 there's no reason for apps to leave so so yes I agree in theory but in practice when on salana when the goto airdrop happened and the Chino airdrop it was very very congested transactions failing you had to go on Jupiter and

increase your gas size to get a transaction through so the the the svm on eclipse on ethereum is going to use that same architecture as salana so why how is this going to be any different if you have an application or two on Eclipse that grows in popularity how are you not going to suffer the same kind of Fate as you know these these these slight uh congestion periods in salana or are you even worried about that yeah so two ways and two answers there um number one a lot of congestion and congestion problems historically in Solana have been in the networking stack and specifically the transaction cost of salana is so cheap such that it is incentivized that if you ever see an opportunity for inclusion of a transaction you're incentivized to spam as much copies or different copies in

fact of that transaction to many leaders as possible or rpcs as possible then everyone forward their transaction to the leader uh and the leader is overwhelmed with so many transactions that they cannot actually process them and they end up dropping the packets right that is the predominant flavor of congestion and um issues on the salot Network that we have experienced in say like the past six months that flavor has largely been um patched mainly with stake weighted qos uh and Upstream qos methods that RPC providers are currently implementing uh and of course there's also next iterations of stake we to qos that are less centralizing uh people are working with inclusion weighted qos D weighted qos and different qos mechanisms and that will over time be solved uh in the

salonist networking stack the second answer is and real quick Terry what is Q qos quality of service right it's it's like a way to filter and throttle um bandwidth from people that are spamming your connections right and we'll have a piece out that kind of goes through the endtoend life cycle of salana transactions that explain exactly what happened at every single step uh but if you want to read about it hel is as great content uh but let me get to the second point the second point is if you're an L2 um and a lot of people don't really take advantage of this fact I'm not entirely sure why your safety guarantee is one event or either it's zero event if you're a ZK roller right someone can just verify the ZK proof

know that the sequencer you know has not equivocated or has not lied about the resulted state if that's the case we think there is room where you can just go extremely hard on sequencer throughput right sequencer Hardware um capabilities specifically like using even Hardware acceleration at the sequencer level uh and specifically fire dancer is built in a way that there are portions of the networking stack that scale sublinearly with the number of cores that you throw at throw at it so for example their quick implementation uh or their signature verification stack uses an fpga right so we're of the opinion that we see a possibility where if if you have a rollup and you are able

to provide it infinite Hardware uh specifically in current implementations of you know certain salon of clients specifically fired answerer there's room where you can actually you know scale uh to multiple orders of magnitude more throughput and at that point we'll see right okay you implemented the networking layer fixes uh that cor salana already has and very very smart people are working on you implemented a different client that scales also uh through with um kind of the amount of Hardware that you provided and the sophistication of it uh and then you see like can you support the the transaction throughput required of uh today and tomorrow and maybe next year and the next year year right we don't know yet um but we position we think that you can there's a there's another project out

there that kind of I I think is taking a similar approach correct me if I'm wrong but this idea of like very centralized block production but then decentralized uh block validation which is Mega eth is also a big proponent of this idea and so like taking a ZK rollup for instance is is this idea of just like really maxing out the sequencer Hardware that first part we're just we don't really care if block production is centralized we're just going to get as much throughput as possible and then in in the ZK case because anyone can verify that we can all all we can have a very decentralized block validation process because well anyone and everyone can

validate a ZK proof is that kind of a similar idea here yeah absolutely absolutely um and specifically like both in Mega E case and R case we have a verifier architecture by verifier I mean more like L2 full node so for example what you might run if you're a part of the optimism Network that uh here is you know State divs and State roots from the sequencer prior to it being on the settlement layer uh both of our verifier architectures are generally lower Hardware requirements than the actual sequencer itself right the verifier only needs to receive the purported state route from the sequencer uh and then in order to verify the state rout that is on the settlement layer download the data availability transactions and then

apply that uh to the previous Epoch uh finalized State rout right so it's much less Hardware requirements than having to you know connect to different rpcs connect to different um people that are sending your transactions and then having to actually process those right so we're thinking and mega eth is along the same path uh that not only do you have verifiability on the smart contract layer uh so in their case through a ZK proof in our case through an optimistic settlement window but you also have the ability for people to run quote unquote like full nodes that actually verify the correct State um of the rollup in a way that's much uh less Hardware intensive than the sequencer so your thesis is basically like juice that up to to the

Max and and rely on better interoperability than the let's scale out with a ton of app chains and ethereum l2s and then go through the coordination process of shared sequencing and kind of like going down that that uh like the economic and uh like I guess social consensus uh consequences of scaling in that model so maybe we can like talk a bit about that and uh I would love to kind of further understand your your thesis on kind of the the cost of uh of synchronous composability or just the cost of using shared sequencing from a developer's point of view and a user's point of view and and kind of get your broader uh take

on this uh idea for app chains as well because I think that we might like Clash heads a bit on this thesis but um you know I I I really enjoyed the presentation so would like to kind of bring it to light for why you believe that this kind of cost is is a bit too much at the current state of scaling yeah let's talk about it more from an economic perspective first right because from a developer perspective I think evm to evm cross atomicity is going to be rather simple um I think it's more so going to be like there's going to be a bit of effort on kind of like unifying the the times or like the block times and kind of like the mechanism in which like different block time transactions happen and call each other and but again anyway scrap all that and let's talk about more

so the economics right like as I mentioned one of the two reasons that applications are moving through their Sovereign sequencing or app chain is because they want to ACR value and specifically execution value and MV if you currently have a say an app chain with a PBS mechanism say for example then you as the sequencer operator retain the me that you generate or some portion of it that is given to you via tips uh From the Block Builders uh plus you retain the execution fees right compared that to uh but in the share sequencer design um say let's take the case of node kit and espresso I'm largely referencing these two uh and

specifically shout out to those teams and Noah from not kit who helped me a lot through the math and the design as well when you opt into a shared sequencer design uh that is these two like new generation share sequencers you receive me from your own domain through your own PBS auction but you also receive a portion of cross domain Surplus right in noit you auction off the top of block of the aggregate domain and then you distribute any cross domain access or so like for example there's like a cross domain transaction in the top of block that is auctioned off in the PBS auction then whatever that fee is paid or whatever tip is paid from that transaction you auction or you give prata to the two entities or three entities that are involved in the cross

transaction so there is some sort of surplus that's generated by everyone being interoperable for example some sort of Arbitrage between the domains uh then you prata give that uh to whitelisted people um that's the no design the express is a look more complicated it's more so that you know everyone runs their own auction and then there's a aggregate auction and the aggregate auction excess of the addition of the you know sub auctions becomes part of uh becomes like pretta distributed by the sub auction prata um values right it's a little more complicated uh there's a little more economic design I think they're still in the in the middle of hashing it so to summarize um if you're your own app chain you you get to internalize your

meev but and your execution fees if you opted to shareed sequencer you get to internalize your own me your execution fees and some abstract portion of the Cross domain Surplus by being interoperable with other people but you also pay and fundamentally you have to pay because they're maintaining their own you know um they're maintaining their own consensus Set uh they're running their own infrastructure so there must be some fee for them to operate you're paying them some small fee in charge for that yeah in exchange for that but isn't the fee that you would pay them coming out of that Surplus that you get from being interoperable with the rest of the networks on that sequencer so it's at the moment unclear I would expect so

that it comes from the cross domain Surplus and a margin on that take rate but at the moment it's unclear and the reason why it's the reason why there's a bit of um uncertainty is also because like largely in the espresso design I'm not entirely sure uh and they have to Hash this out I'm not entirely sure that you can actually realize a cross domain Surplus um in a prated fashion right specifically in their design let's go over kind of how it works there's two there's two uh auctions there's the Jus in time auction and then there's ahead of time auction in the ahead of time auction uh it's kind of like execution tickets right people have a reserve price and then you can bid for the right to propose uh in the future for

different rollups as well as different combination of pups and likely what's going to happen is there's going to be a highly sophisticated entity uh you know that has the ability to you know connect to a sex uh and run sexx Arbitrage and Arbitrage across multiple domains and constantly bids their expected value um at the uh execution at the at the ahead of time auction right and that's largely going to be constant uh and maybe per perhaps like a small portion of the reserve price and then in the Justin Time auction it's a PBS auction uh and likely the same actor is going to win the same auction uh you know um as they propose and build blocks for multiple domains that they first purchase the execution tickets for there is a bit of a bit of um issue when you do

like composability uh in an open and permissionless way uh and that's why nkit has Whit listed designs and that you are exposed to this idea of Searcher operated rollups right so for example if I'm rece if there's a 20 total cross doain Surplus and you get it prata depending on each bid in each domain right say if I bid 555 the cross domain service is 20 then we get you know 13 one3 1/3 then there is an incentive for a Searcher to spin up their own infrastructure and their own rollup and constantly call it in a cross domain transaction right so say there's rollup D now and rollup D the Searcher bids six for it because they all know that other rollups are going to get a bid for 555 then suddenly uh well that's

not okay say it's like 333 and then like five so the aggregate okay this is getting a little long but the aggregate is still less than the cross domain then because they have a larger portion of a bid but even though it doesn't necessarily provide any value but because there was a bid for it then more of the Cross domain Surplus will flow into the Searcher operated infrastructure search operator Ro and therefore so the general idea is therefore if you want to have some sort of distribution in a permissionless composable cross-domain environment that depends directly on the uh amount of value that is purported and generated from that domain that is very difficult is unclear that you could actually distribute the uh the cross-domain

Surplus and it is unclear that there will be enough of the Cross domain Surplus to actually distribute to real operators the same problem still exists in a general purpose L2 but it is a very difficult problem to solve yeah and I mean it's also like uh without permissionless uh without the permissionless nature of shared sequencers it kind of removes the purpose of the entire idea in a sense but also it sounds more like an incentive uh mechanism problem than it is like a a conceptual problem in the sense that uh if there's a a search or operator rollup that's just going to maliciously bid they will there there will either be in the early stages of espresso and nkit

estri Etc a mechanism to kind of Slash or punish that or it will be iterated on over time such that there will be like this constant war of like it's almost like you're in Discord like you're trying to like hack the Discord and then they they like patch the hack and then you try to hack it a different way and you patch it again so it's almost just going to kind of be like this NeverEnding iteration of people trying to bomb the shared sequencers to up the total amount of me to then be distributed back to themselves um but I know espresso has like an interesting model to kind of solve this as as as you were saying with with execution tickets but yeah it's it's it's largely unsolved but kind of coming back to like the original the original uh point of view here basically what you're saying is

that there's no way that that app chains running on a shared sequencer will be able to uh earn as much H Surplus than if a rollup was running its own shared sequencer or its own sequencer uh because it just doesn't make sense because you were subtracting the amount of of or you're subtracting some sort of feed to the share sequencer Weare as if you were running it on your own you would be earning that fee so uh we should just build all in one execute one L2 that's highly performant but then if I'm an app builder how am I going to get any me from the sequencer on this L2 that is just internalizing it for the from the operator basically and I'm just

an app builder like why do why should I want to participate in that system when I can at least participate in the system where I at least get something does that make sense yeah yeah yeah that makes a perfect sense so first of all maybe I want say like three or four things first of all I want to go back to the share sequencer and that's why like noit works because they just Whit list of domains that you can interoperate with right and so it's more so like Progressive social consensus accumulation where different entities are like okay we trust each other we want to White list and that's I think completely fine because a lot of chains and app chains are longtail versus the places I want to interoperate between are largely like large domains like I want to interoperate between say optimism Bas and arbitrum and it is

perfectly fine and probably that they will whist each it is unlikely that you know I want to interoperate between D gen chain and Zora and those are longtail chains and you know if they don't Whit list each other even though you know they're highly respectable entities if they don't Whit list each other that's fine so in a large sense I think the no kit design is going towards something that is um just like just okay enough such that it works so I'm a fan of more so the white listing but yes uh the second point is there is an incentive design issue in terms of do I launch my own share sequencer or do I interoperate with other people by opting into a third party uh this is largely true for large companies um and specifically it's because shared

sequencing is not too difficult of a technical uh technical lift at least that's my understanding of it and that's how you know Justin Drake talks about it that's how shared sequence or design companies talk about it uh it's the economic incentive compatibility issue that's very difficult but the building of the tech of the Shar sequence itself apparently it is not like an enormous enormous like gargantuan Herculean effort and so if that's the case then optimism arbit term you know these large companies they have the resources to do so but then they don't necessarily have the incentive to you know like all interoperate between one Shar sequence or shelling Point uh it's that they would launch their own finally let's get to the point where why would I stay on a general purpose al2 this is why I found

it imperative right it is imperative for a general purpose execution environment to give back the value that they created otherwise they simply cannot compete it's not a factor of we don't want to go this way or we do want to go this it's a matter of if you don't go this way you will not survive right so you have to go this one you have to give back value you have to make applications first party decisions because applications are now thinking oh why don't we go this chair why don't we they have infinite choices and in fact people are willing to give them you know massive eight figure grants if you're a well-known Marquee project you're going to get a eight figureure Grant from ZK sync to deploy on ZK sync right so then how do you match that you have to give back a

portion of Revenue you have to promise that you know they will retain the value to [Music] create cartei is your home for app specific rollups at the Forefront of developing new tools and platforms that enhance the capabilities of blockchain Technology cartei enables applications that are flexible and built to scale build on a full Linux environment and use traditional software libraries but in blockchain development go to cy. that's cy. to learn more about building your own app chain today I tend to agree I struggle to see

the regulatory uh burden but I we don't need to really dive into that right now um but I okay so you're going to give back the value to to the app developers in some form of Grants tokens Etc okay but that's still with a centralized sequencer right like like is there a shift like Eclipse launches with centralized sequencer there's no incentive to ever decentralize like it is that not problematic in it of itself like like like and even just broadly outside of eclipse like even if you're still giving back the if you're giving back the the me to your application Builders and your and your users and your community like then do you ever need to deiz the sequencer or like how

does how do you think about those two things like kind of separately or are they separate yeah so this is interesting let's start with an L1 right so good thing about an L2 is you have completely centralized sequencing and therefore oh maybe not a good thing but like in the in this case of value distribution right all the value occurs to a single entity a single wallet and therefore it's easier to distribute in opinionated way take for example an L1 right the value distributes to individual validators that individual validators then have to distribute the value is that possible it is unclear the bar chain model uh what's interesting about bar chain is they're taking an opinion approach where each validator signs or opts into the social contract of the be chain L1 such that they are

subject to distribution of their coinbase rewards and their validator rewards that come from you know building a Blog and proposing a block in the network uh dependent on rules that each validator sets for itself right in the case of bar chain you have something called The Cutting Board uh which is a contract that dictates programmatically where your bgt inflation rewards go to and then within that contract uh it is distributed to daps and specifically pool and liquidity pools of daps uh that each validator defines and opts into for themselves right so that's the L1 method the only not issue I would say but like there is a Marketplace for bribing on on

be chain so for example if I'm an application if I want to get a larger portion of the bgt rewards I simply bribe the validator with my native tokens and then now I get most of their coinbase rewards and smaller daps uh then you know cannot compete so even that they might generate like a portion of rewards uh that might not flow back to them or if they generate a large portion of rewards they might be mandated to uh then bribe the validator who then gives back their rewards to S themselves so it's kind of like a circular economy um so you have this kind of fundamental tradeoff between okay do you have open validator set or do you have a fair um value distribution mechanism and this is a very hard problem uh at the moment we're of the

opinion that okay it is strictly better to have a centralized sequencer and have value distribution uh because from the perspective of an application if you're an L1 where you have a complete decentralized uh sequencer set so for example um like be chain uh but you have value distribution in a way that isn't a creative to me versus if you have a centralized set with you know safety guarantees sensorship resistance liess but you have additional value coru we think this is strictly better but we are in the process of ideating ideas around um you know how do we make sure that we can have a decentralized sequencer set with maybe like a a distribution attestation committee um that has external uh uh kind of external trust um given it from you know say for example

restak uh that enforces the distribution in a fair way right it is a difficult problem T quickly bring this to sort of the the app chain thesis again and whether or not they would run their own app Chain versus like run it on the general purpose rollup of eclips I think we we are talking about this in in like a new application sense where someone comes to the chain and they have to choose like am i g to build an app chain or am I going to build an application on a on a rollup but I think in practice what we see more often is people start as applications on the rollup but then they choose to leave and build their own application once have distribution and they have the uh incentive to do so and

so when when you're talking about like you know keeping people onto onto the rollup um if they do you know and and a lot of times like when when applications aren't de developing their own incentives you know that's when they would choose to build on the application early on they get a grant from ZK sync whatever or eclipse and they get you know a lot of funding to start off but then how do you ensure that uh the validator set is contributing this incentive back in a fair way is it is it a matter of like like you're saying the open validator set versus the decentralized validator set bar train has done this by kind of like leaving it up to the validators to choose how they redistribute the incentives back to

Applications you know with Eclipse kind of like going on mainnet today how is this working in practice is it kind of like the bar chain model or is it kind of more like uh a centralized model where everyone has to agree on the same place to reallocate the incentives that are generated by the applications so a couple ways I'm thinking about it number one something is strictly better than nothing and so if you receive something as part of being part of an application part being part of a general purpose L2 it is strictly better than you know an opinionated a non opinionated no redistribution I I think that's true of like the the early applications that don't have anything where they would come to the general purpose to get something first but like once an appli

has generated some distribution it has some uh some incentives that it's generating on its own what's the incentive for them to just stay when they when they do have something right they they have to be incentivized more to stay than they would be to leave yeah and let me get to that let me get to that and okay U the second thought I'm having is very similar to why people who have the ability to be mobile between different countries live in different countries the reason why we're in different jurisdictions and States is because we optt in necessarily to kind of the value capture and the value redistribution of that specific jurisdiction so for example you know you might move from New York to Florida

because you don't want additional state tax and you don't agree with the way that tax capital is spent in the state of New York and kind of in a similar sense um you know you can compare arbitrum versus optimism say versus Eclipse where arbitrum optimism they're right now having 100 % tax on your execution fees on your M and in exchange optimism spends it on public goods funding they fund you know things related to the super chain things related to optimism arbitr in exchange funds say for example gaming initiatives on the Chain so if you're a gaming you know project on arbitrum you're happy to stay there because your tax dollars go to good use over time the jurisdiction you opt into uh in the rollup will be at opinionated

opinionated um vote for the value capture and the value distribution strategy of that rollup in exchange also for the distribution and the liquidity and the you know all the network effects that the base layer execution layer can provide okay now finally how do you make applications stay on your roll uh number one it is really expensive to become a rollup and people vastly vastly vastly underestimate this right I'm talking rpcs indexers block scanners liquidity providers PS the on the orders of tens of millions a year well not a year maybe like tens of Millions for that sticker cost to actually spin up a roll up and operate

it for its fa three to four years right it is very very very expensive uh and so number two like if you can provide a layer where there's sufficient user sufficient distribution and sufficient infrastructure and you constantly work on providing those to your applications you constantly work to provide you know oh let's bring in additional users let's lower user acquisition costs for daps let's bring in additional liquidity uh let's make sure that we have a robust you know public goods infrastructure and robust system then we think and this is hypothesis and this is kind of the bet that eclipse is making we think there is room where you know apps are both incentivized to come and apps are

incentivized to State yeah I'm curious if this is like also part of your infrastructure take rate thesis decreasing it are do these two kind of go hand inand where like okay this idea of um infrastructure kind of occuring all the value is sort of potentially going away into more of like this application occurring value thesis or like how does this idea that you're that you're saying about okay we're going to try to keep apps on a general purpose L2 and incentivize them to do so how does that thesis relate to this idea that you have about the infrastructure take rate decreasing and perhaps we can just start with what that really means and then kind of pull them together yeah and specifically I'm talking about chains um

I think there are the the other portion of infrastructure like modular infrastructure you know block scanners rpcs indexers we can talk about later I think they are in a very different position from chains but ultimately right now chains do not own the end user relationship applications own the Ender relationship and if you own the Ender relationship right you're sticky uh and therefore you can ACR value um the only difference is most of the time applications have been paying that as taxes to l2s or l1s uh but applications want something different in fact there's two large trends that make this very obvious um and overall uh both of those could be called Sovereign sequencing but one is app chains uh lot of apps are moving to app chains because

they want their own value C but the second one is some apps are moving to um controlling the sequencing of their transactions but it's not an app chain so for example Sorella if you guys are familiar is kind of the titular case here or vertex where instead of submitting um so ethereum transactions or L2 transactions you submit just like signed orders uh into the application the application then takes in all those orders they sell the rights to reorder the orders uh or they sell the rights for example cow swop right you sell the rights to the batch auction and order the batch auction uh and Searchers can then inject transactions in there of Their Own subject to certain rules enforced by the

application and the auctioneer so for example you can't front run in s then the tips that are paid to the auctioneer or the runner of the auction is then distributed in opinionated way so for example in sella uh it's given to the passive LPS of certain pools that people are trading in so it's a way to like internalize lvr and then return that to the LPS so apps want to control their own sequencing and this is what's going to uh be the direction of applications on the L1 so application L1 uh because you know I think the bar chain model still needs a little bit more explaining in terms of how that's beneficial to Applications so applications on say for example ethereum L1 are going to go this direction okay they're going to sovereignly sequenced and then applications on app chains and model

going to go this direction which means ultimately what's left for a general purpose L2 right there will be no applications on general purpose L2 right why would you be on one um and that's kind of the the impetus behind infrastructure take rates are going to zero right you can only compete uh if you're actually giving away all of your value uh and therefore you compete on scale you compete such the such that U kind of things I mentioned earlier which is what brings in the uh Network effects of being on the same layer which is users liquidity and composability you make that valuable such that actually your Tech rate times the value of this is equal to what they would have made by leaving your chain it took us 40 minutes to realize that we have the same thesis

but we're just approaching it in two different ways with afchain Sovereign sequencing or just go on a super performant general purpose L2 and receive your fair share of me generated or basically sequencer revenue generated so uh uh that's a that's a that's a good thing to note and I also see that this is gonna just keep happening like anisha's ecosystem with like their VIP incentives they're just going to be giving uh incentives to applications to stay within the ecosystem because of the the fees generated by the L1 token emissions Etc um you know and it's just just going to keep scaling out so I think that's a really really good point which is that applications are just going to want to have Sovereign

sequencing but there's different ways to do it and it's all that's like kind of the beauty of this whole modular thesis is we can kind of just experiment with whichever ways that we want so I I think that we're on the same page and I'm excited to kind of see both angles play out in terms of like so I I don't think the question is necessarily one big rollup or a million rollups it's one big rollup that distributes its me to its applications and its users in a fair way and is effectively a nonprofit organization to some capacity right or just to some some terminology perhaps that's not the right like the right term and then it's app chains which do their own Sovereign sequencing and interoperate with some sort of shared sequencing layer or interoperability

layer uh you know chain exraction and the sort so uh yeah I think I think the question of one big rollup or a million rollups is is almost like it's it's one big L2 with this model or a ton of app chains with this model and I'm not sure which one's the right answer but obviously eclipse is going towards the one big rollup model and we'll see other kind of app chains launch and go into the shared sequencing networks and so uh needless to say excited to kind of see these play out and keep kind of developing the thesis yeah it's two sides of the same coin another thing that you talked about Terry in your presentation that was super interesting was this idea that interoperability is a social consensus game more than technicals or economic game which I I find to be uh maybe a little

bit misleading in the sense where I think the the core interoperability problem right now is the technicals it's we got to figure out the intent-based interop that's going to be the fastest we need sub2 200 millisecond interoperability between evm to evm we're getting there with the cross and some of these other designs but then we extrapolate that out to other execution environments and then also the user experience of all that so that's all a technical and ux problem but but you're saying that that's not even the the core of it the core of it is it's it's largely a social problem for a social consensus situation maybe you can kind of uh uh unpack that and what that means the way I see it is there is no incentive for optimism to want to interoperate with arbitrum by opting into a thirdparty shared sequencer and

or a shared sequencer that is operated by themselves so fundamentally if I'm optimism my incentive is to create my own Shar sequencer mechanism because I already have a huge ecosystem of different applications and app chains right and therefore I can charge right as we went over kind of the percentage of the Cross domain Surplus as fees they're doing this they're doing this with the super chain amongst the rollups that they have on the super chain exactly and if I'm arbitr then my incentive is to launch my own chck sequence because I have all these orbit chains I have Nova I have Stylus all these different you know chains and applications that in my ecosystem already if I'm ZK sync same ZK stack and so none of these ecosystem systems have an incentive to say hey I'm optimism I have a huge ecosystem I'm going to give

all a portion of the fees generated from all my ecosystem activities back to arbitr and the arbitrary share sequencer in which case there are these large domains I want to interoperate with they're just not interoperating with each other and that kind of frustrates me as well right because I use Optimus I use arbit I use base I use zync um and I would love if I was able to submit you know some Atomic cross domain transactions between these uh but that's the incentive G um and that's why I think shared sequencing is very difficult mantle is a layer to blockchain with one of the biggest onchain treasuries in all of crypto and they have a 100 day campaign called metamorphosis live now what this is is it is a campaign for their liquid staked eth token M eth as well as their upcoming liquid reaking token CM e where

you can utilize this token on the mantle Network in a variety of different apps some of your favorite reaking protocols and defi protocols and earn yields in the form of powder which will eventually be turning into a governance token Mantle's been around for a while we like them because they use their own form of Igan da uh they have a good ecosystem and they're growing so check out this 100 day metamorphosis campaign and hope you guys enjoy the rest of this video Everclear is pioneering the clearing layer the clearing layer can almost be thought of as splitwise when you go on a trip with friends but for the blockchain what I mean by that is when users are sending cross chain messages there's often times fillers and solvers who have

balances left on different chains that they need to rebalance and reset their positions this is where evercar comes in ever is pioneering a collaborative approach to the chain abstraction future and they are willing to work with interoperability Partners solvers and other providers in the space to build a interoperable and chain abstraction future check out Everclear ever clear.org and welcome back to the L up yeah I mean it's definitely a Winner Takes most or all market and uh also I we're having a podcast with Jill from espresso it's going to be all about this like why and i' I've been tweeting about this like we're building these walled Gardens these clusters of rollup ecosystems themselves okay we're gonna

have really good interrup and economic composability and Vibes between them but like like there's like these B like these borders between them that were just going to like interoperability providers like hyperlane or like Union or some of these other teams are going to try to build out Bridges and such but it's just yeah the the way that these walled Gardens these clusters are being built is I'm trying to figure out how they're going to interoperate with each other more so than like within each other because it seems like okay the super chain orbit ET is great like the interoperability there is is fantastic but you know kind of bridging that outside and then even taking that into like other types of uh virtual machine and execution environments it's kind of a question mark still it definitely is

um and I think a lot lot of proponents are saying that it's okay so it'll be atomically synchronous between or atomically synchronous for inclusion between these um Shar sequencer environments and then it'll be some sort of asynchronous uh kind of just like Bridge infrastructure between these large but that that looks exactly the same as it does today right that's exactly what we have today just faster with and and you're calling this an engineering question mark Terry you're calling this a a a social question mark but I think fundamentally it's an incentive question mark because these rollups are only going to do so if they earn more than they would from from running their own centralized infrastructure so until we have some proven model or projection

that says like you know arbitr makes x amount from its shared from its centralized sequencer optimism makes y amount from its centralized sequencer but if we run on the shared sequencer because of the Cross rollup transactions we can all make X plus y plus Z and then we can all split Z and make at least as much as X and Y until we get that and we know what Z is we're not no one's going to be incentivized to make that switch I but even if we know Z is what Z is everyone will be incentivized to take the Z themselves like every until they're on a shared sequencer right they have to be because cross roll up transactions are what

generate Z but if you're the creator of the the C the shared sequencer then you get you know some fee on Z so then everyone wants to create their own shared sequencer so it's like okay why would I op yours because because you would still get some arbitrum would get some uh it would get y plus some subset of Z optimism would get X plus some subset of Z it's still more than x arbitrum still gets more than y yeah node kit or espresso or someone whoever runs the shared sequencer is still getting some portion of Z but they're still redistributing another portion of Z back on top of X and on top of Y yeah absolutely but if I'm arbitrum and I have my chain and my share sequencing layer then I can get a portion of Z and

the additional distribution from Cross domain but Z is much bigger if you're also interoperating with optimism and the rest of the the networks into that more that more General share to sir that's where I feel like it's a social problem because arbitrum could say optimism come to me optimism would say arbitr no come to me and then they would just be at a standoff it's a Vibes problem man yeah and could all decide to be makes most sense to do that right but then why would they all go together and be like no we're going to go to expressive there has to be some shelling for Z to become the biggest like the cross roll of transactions to become available and profit and and generate like the most incentive for the rollup to split some unbiased shared sequencer

layer that is credibly neutral perhaps big going to be like a winner takes most or Winner Takes all Market I think shared sequencing I tweeted this today I was talking to a couple VC's guys VC guys about some unnamed uh shared sequencers that aren't like you know I guess Kem of the KREM and they're like just doesn't make sense to invest in this because it's going to be a Winner Takes most or all Market because of what exactly where we're what we're talking about I mean I'm I'm starting to believe this TS toward based sequencing as whole another podcast you're not wrong like base sequencing specifically was created in my opinion like specifically like uh base shared sequencing with

pre-confirmation was created because oh all the execution values leaving the L1 how do we bring it back bring it to base sequencing the sequencer actually Harvest you know the L2 M and the L2 execution fees well the problem there is um if I'm an L2 why would I go to Bas sequencing and give up all of my fees and my to the L1 proposer right there should mechanism around this and um I forgot the te Aspire Labs is doing this like execution ticket mechanism where um you know a subset of the L1 validators buy execution tickets for the based rollup sequencing rights and therefore that internalizes the um the me and the execution fees that you

expect the expected value uh but you still have you know a subset of the L1 sequencers sequencing you so it's kind of like um base sequencing but with a subset of the proposers versus like the other way to go through it is kind of start with one centralized provider and then increase that through governance right but largely you know you end up meeting in the Middle where one is uh here's a large set and then you take a subset and then you you know um have a mechanism to retain value by you know uh ahead of time auctions and then here you have um a subset you know of some large set which is you know the possible proposal in the world and you kind of

choose a couple through governance and then you have them rotate yeah spiral Labs is is is some Alpha Terry's dropped a ton of alpha on this pod today so we're we're just gonna relieve him of his duties and uh appreciate appreciate your in-depth uh understanding of these kind of kind of tough to grasp Concepts but once you kind of grasp the larger idea it it it becomes uh clear as far as like where we're going and as well as like the incentive alignment uh situations that are present but um generally Terry I really appreciate your time thanks for coming on man been a pleasure been a pleasure here as well thanks for usting congrats on mainnet Deet M DEET M for

Developers devs