"Liquidity Fragmentation Isn't a Technical Problem, It's an Economic One" - Josh Bowen of Astria

Transcript

One of the things I think that's like interesting or maybe like a wrong starting point when people think about like fragmenting liquidity is there's just like a blunt reality of like liquidity costs money. Liquidity fragmentation is an economic problem, right? Like people are going to pay for liquidity and they're not going to like share the liquidity they paid for, right? They're going to be like, I paid for this liquidity. I'm going to monetize the liquidity. Why would I give it to you? Right? And people are like, oh my god, they're like fragmenting the liquidity. They're like, you were free riding on the liquidity. like maybe you are in the wrong and go pay for your own liquidity kind of thing. We'll solve the like ability to move nominal value between chains. Probably not in the way people necessarily wanted it to go, but if you want to move dollars between

chains, you'll probably have like a pretty deep liquidity market to do that in the next 6 months. [Music] I'm here with Josh Bowen, co-founder of Astria. Astria is a decentralized sequencer layer for the modular ecosystem. And we're going to be diving deep today into how Astria provides sequencing for a lot of apps across DeFi, including uh their new uh Flame uh decks, which is built on Celestia. But first, before we get started, please make sure to subscribe, hit the like

button, and the notification bell to get notified when new episodes drop every week. Hey Josh, how's it going? Hey, Sebastian. Going well. Thanks for having me on. Yeah, it's been uh a long time since I wanted to do this. I first saw you, I guess, like in end of 2023, some Celestia event in Istanbul, and I was like, I got to get that guy on the podcast at some point. And you've been somehow on my list for like this long. So, I'm glad I'm getting you on now because Astria went mainnet in October, a couple months ago, and so it feels now there's actually things that people can use and actually products that are live that people can interact with. So, yeah, what's that trajectory been like? I mean, you guys have been building Astria for quite some time now. How's it feel to be like, you know, a couple months

into mainet? You know, it feels it feels good. You know, we've had, you know, a few hiccups kind of here and there, but overall we're relatively happy with kind of things when, you know, we launched mainet, you know, kind of stealing a term from like my friend Henry over at Penumbra was, you know, we launched the mainet when the protocol was finished, less so when like the product was finished. And so, you know, we've had this kind of like slow roll to have kind of like, you know, usage and adoption. And that was kind of expected when we said, look, from an engineering perspective, the thing with V1, we were comfortable tagging it that way. We had changes, but like less breaking changes. So, we got it out. It's been running. We haven't kind of dropped blocks. We haven't had huge complaints from kind of

validators that are running the network. And so, overall, we're quite happy with the launch. And then the focus kind of going into, you know, Q1 as kind of, you know, okay, we launched like end of November. We went into kind of or end of October. We went into November kind of holiday season a little bit. Everyone is obviously a little bit tired kind of getting the main net out and so we kind of tried to ramp back into Q1 and say okay what are we doing here and a lot of that focus has been around kind of you know Flame and then Forma one of our kind of partner projects which is currently not running on kind of Astria's mainet network they've been running on their own you know sequencing network same software you know it still uses like the Astria stack but off of

its own kind of independent you know um smaller validator set and kind of going through that lift and shift migration so that ideally at the end of this quarter we're kind of where we wanted to be from say like my presentation I gave at you know modular summit in Brussels last year about okay that kind of gives us like what I've been calling like the foundation of what we want we have the sequencing layer we have our bridging to celestia and we have flame as our defi rollup and we have form as our NFT chain and it kind of gives us the foundation that we can then iterate upon and kind of provide you know somewhat status quo kind of defy you know and and NFT projects within the Celestia um or or larger modular ecosystem and so we're quite happy with where we've gotten now

nice I I actually just kind of remembered that Forma is also uh an Astria product. So I guess we can also talk about that. Um I really like the the art on uh on Formma. It's super cool. what what's what's Astria trying to solve as a problem really like you know if you look if you take a step back and look at the blockchain ecosystem and and the way things have been modularizing in the last you know three or four years what place does Astria want to occupy within that broader modular infrastructure stack so I think the the key focus for Astria you know going all the way back to like when I joined Celestia like end of 2021. So

like a very long time ago, you know, when I was working on like what was called Sevmos at the time and what has evolved in kind of the roll kit project, you know, but the general kind of goal was to have what I would call like a Celestia native rollup. And what that meant was, you know, I viewed Celestia as this kind of desirable scaling solution where I structurally differentiate it from something like Solano where the goal was sharding, but what I mean by that is more like fractional block storage. was the idea of like you know okay what a blockchain is is getting many many nodes in a network to have consensus over data and like a specifically like a time stamp of that data right defined by a block height and there were kind of two ways to make that scalable you say all the

nodes in the network are going to be big and we're going to optimize the hell out of like you know latency bandwidth etc right this is like the increased block like increased bandwidth reduce latency kind of like salon meme um and they kind of bit that bullet and said we're going to have big blocks and then the celestial side was saying okay user using enra coding as kind of like the the key thing in this data availability sampling we can have a larger or or like a like a similar like duplication factor, right? Like total replicas of like the data in the network for like you know redundancy reliability but with smaller individual node requirements via these light nodes. And so instead of having every node storing the whole thing and having to be a beefier node, you have a larger number of smaller nodes which each have like a fractional

store of the overall data. And so I viewed those as two kind of categories. And then to like kind of Astria point evolving out of Semos the goal was saying I don't want Celestia to be constrained by Ethereum and say oh we're just going to use it as like what is now termed like an altda solution to Ethereum because fundamentally from both like a technical and kind of a political view. I saw it as that would limit the potential of Celestia as this kind of alternative ecosystem if it said well can kind of only run as fast as Ethereum's willing to keep up with it but there's potential competitive pressures with kind of Ethereum and so that has evolved into you know Semos and Asteria you know as kind of like the the kind of you know logical successors there we started as like a settlement

layer and we still serve that role and the reason for that was that you know there were these debates in kind of early 2022 around does Celestia want to have a bridge in it essentially or does it want to act as a settlement layer for roll-ups or does it just want to do the DA and at the time kind of you know Mustafa and the rest of the Celestia kind of core team was pretty strongly opinionated that like no we don't want to like the term used then was like enshrine a settlement layer they wanted to have kind of like a market of settlement layers that kind of come up and say okay one of these will kind of provide you know settlement or this kind of bridging hub for them right and so that was what Aster set out to do initially to a large extent we still do

that we are like a settlement layer we bridge TIA and other IBC assets sets via IBC onto Astria and then from Astria we have our own rollup bridging protocol upon which like Astria roll-ups can bridge from and then going to the point of kind of both decentralized sequencing and shared sequencing how we got there was you know one again thinking in this kind of mid 2022 time frame we were relatively early into this kind of modular role of exploration we had centralized sequencers but they were not like they were uh they're relatively recent right they were in the like order of like a year or two kind of out and so the idea of all these centralized sequencers were decentralized was very much a still kind of consensus take. It just hadn't been finished implemented yet. And so we saw decentralized

sequencing is important and then the going to shared sequencing as like the kind of final piece here was from that kind of 2022 time market going down. We talked to a lot of validators in the kind of Cosmos and like interchain ecosystem and they said look these longtail networks aren't really like cost-ffective for us to run validators for and so we were seeing a contraction in number of networks that validators were supporting and so at that point we said okay if we want the roll-ups to be decentralized and we take the there will be more rollups than like sovereign L1's then we assume we have an even greater kind of longtail scaling problem here so we need to have a system a design that amortizes the cost of the decentralization across each of these rollups. So all of them can have decentralized sequencing,

but you're not trimming this long tail very aggressively and having like a winner take all where only the large players can be decentralized. And so that's how we got to shared sequencing. And the term that would be most accurate from like a technical perspective of how Astria roll for because they would be based rollups of the Astria sequencing layer, which is also a settlement layer. I know that's kind of longwinded, but that is generally how we see it and kind of the the products or the kind of value we're trying to provide in the modular ecosystem. In the last I guess three years that we've been talking about modularization and making blockchains more modular. Celestia has been live now for I guess probably about two years. I forget the exact mainet. It was end of it was October 2023 I think was the main I think it was October. So it hasn't

even been two years. Yeah. So, what's the what are the main lessons we've learned? You know, going back to those initial ideas about making blockchains more modular, the uh you know the the the the lazy ledger white paper, some of the ideas to come out of Celestia and the rollup ecosystem. What would you say are the main lessons we've learned in terms of what worked or what we expected to be the case or where have things not gone the way we were thinking or you know are perhaps suboptimal now that we have this very complex infrastructure stack. Yeah. So I think the key thing like if I had to pick one thing it would be like proofs took longer than we thought they would or like proofs are

harder than we thought they would be. And so this both applies to like fraud proofs and so you know the initial example of like trying to do fraud proofs is like a like a singleshot fraud proof, right? And the terminology is going to clash a little bit, right? Um you know with like optimism's first implementation like that didn't work and so they're like all right we have to do this interactive verification game that you know arbitrum had, right? but then getting that into like a fully permissionless setting with something like bold or like I'm going to yeah I don't know what optimism is currently calling it like canon is like the term that I have in my head but that's relatively outdated so apologies to the optimism folks but getting this proper permissionless kind of interactive verification game it was just like a hard problem and when I was working on

it a little bit at Celestia it was starting to like become clear was a hard problem and that's one of these things where because it's this like economic game like any kind of edge case can like compromise the security of like the system as a whole and satisfying that like long tale of like kind of like security vulnerabilities. It's just like a fundamentally hard problem. I think both teams kind of like working on it are like they're obviously like extremely competent teams. They're great teams. They're where I kind of learned all of the fundamentals about how rollups work come from, but that took them longer to do and it was a harder problem to solve than maybe was thought in 2021. Same thing on, you know, ZK or validity proofs, right? both they were a little bit harder to do and they took longer to do but also the the cost kind

of reduction didn't come down as fast as I think people wanted to right you know today it's not super trivial to do you know just take a take an EVM chain just pass it through you know a ZKVM whether you pick you know succinct or risk zero or one of the other ones right and get a proof out every block for call it a 30 million gas EVM block on a two- second block time or something it's just non-trivial from a cost perspective to do that um but Also from a latency perspective fundamentally my view of kind of the key distinction between a rollup and again this is very very astria biased take of a rollup with a decentralized sequencer and we say it's a decentralized lazy sequencer borrowing some celestial terminology. The key distinction between that and an L1 is you still have a a network that is

coming to consensus over the data underpinning the rollup. But what that network's not doing is it's not coming to consensus over the specific state route of any given rollup. So like you know we have a shared sequencer you could have you know a hundred rollups running on the same essentially like message bus here each one is deriving a subset of that data and running that through like an execution path whether that's an EVM a move VM an SVM whatever right but then from like a bridging perspective or a user security perspective they're interacting with one of these discrete state machines and generally in like an L1 you have the the consensus network actually giving you kind of consensus over a state route of that actual execution in a rollup design in like a based rollup or a

decentralized sequence lazy rollup whatever what you're assuming is you're not doing that for each rollup and instead you're going to rely on some kind of proof system where maybe you have an optimistic mechanism where some semi-rusted actor is making an addestation of here is what I claim to be the valid state route and this is to some degree a preconf and then you have some challenge mechanism if that person kind of like lied about that you have some withdrawal delay to kind of secure the bridge that that is kind um you know interacting with or you would have validity proofs produced on top of that state route and that validity proof would essentially be a replacement for a committee kind of thing and that's structurally in my view the difference between like what is a rollup and what

is like an L1 and fundamentally just the proofs were harder to build than I think people expected in 2021. Okay, interesting. So like the idea here I I liked how you describe Astria as a message bus. I think that for me is a is a much clearer terminology than some of this very esoteric terminology that we have in in crypto. I wish everybody would just start naming components in blockchains as components in PCs. It would make I think life a lot easier. But yeah, so a message bus a sequencer is essentially a message bus. It sort of decides which messages get passed first or in this case which transactions get settled first. In the case of a

decentralized sequencer, early on when people were talking about sequencers, I've I can't remember who coined this, but someone said to me on a podcast once like a decentralized sequencer is just another blockchain or like there's no such thing as a decentralized sequencer. At some point, this the scalability of that message bus starts to hit limits. If there are perhaps hundreds or maybe thousands of rollups leveraging that decentralized sequencer, it also starts to hit limits and we start getting the same scaling problems as we have with regular L1s, right? That uh that were smart contracts and computing for for bandwidth. H how do you think about that sort of moving the the the scalability to a different layer of the stack and how can we address that issue?

Presumably you you guys are using I think you're using tenderint that scaling problem is probably a long ways away but how do you think about scaling that message bus so that we don't start hitting those limits in the future. Yeah. So I guess there's like like fundamentally right one of the things about like a message bus is you know and again like the salon of people are somewhat correct in like their terminology too of like increased bandwidth reduce like latency right those are really the two variables we're kind of looking for and if you look at like you know go all the way back to like you know Bitcoin right it's like I don't know how large a bitcoin block is like four megabytes or whatever now right with the segwit I'm going to get my terminology wrong on like six minute block times right and so like okay six

minutes is very long okay Ethereum has like 12 seconds now we're like ah we'd like you know whatever like 2 seconds or 1 second or 500 milliseconds or whatever Right? So that's reducing the kind of latency, but fundamentally your constraint is just like bandwidth, right? You know, and if you have a decentralized network, there's some level of um you know, it's kind of like a hard wall. And what I mean by that is, you know, okay, if you're using like a like a Byzantine fault tolerant network like tenderint or, you know, hot shot or something like that, right? Byzantine consensus, right? You have however many rounds of consensus you need across however many nodes. You have whatever your scaling properties of commun between those nodes. And then you have like the speed of light just says like ah you can move this much data you know

this quickly right across the globe and you have this many nodes in the network and it just gives you this bound where I think from some level on like the latency side like we're pretty close to like the low bound and what I mean by that is like okay like you know you're looking at like monad with monad BFT wants to have like 500 millisecond blocks but then they have like this like two slot finality delay is they're they're kind of re-riggering exactly how like the the BFT kind of voting scheme goes but there's a little bit of like you know we're not going to get to like 50 millisecond like block times on like a decentralized like actually getting commitment just like no one's going to go tell you ah we're have 50 millisecond like block times right because you're like yeah this is like speed of light

just says no right and so you know in that level right there's this kind of constraint I think in specifically like a like a tender kind of comment BFT we're not quite there it's kind of well known is and I like harshish terminology but there's not like a criticism of anyone on like a technical sense like comp is like phenomenal public good like extremely important extremely valuable right it lets people get off the ground with like you some like the first practical Byzantine fault But everyone's kind of aware that like ah the like flood gossip is kind of like bad and like the block propagation the transaction propagation is like very inefficient from like a usage of bandwidth in the network kind of thing.

And so that's kind of next of like how we're thinking about kind of scaling but it doesn't get us what I say like it doesn't get us like a magical land of like infinite kind of scaling thing. It just kind of brings us much closer to this kind of endgame status quo of like this is how good we can do. we can push this much data with this node requirements across this many nodes at this latency and we're looking to do that and like you know optimistically we've written it in kind of Q2 to start looking at that we probably only poke into it a bit in Q2 and maybe Q3 we start implementing you know I've talked to like you know dev over at you know osmosis a bit about this the celestia guys have their vacuum spec on this the monad BFT guys also have you know raptor

cast as like an option for this so there's a lot of stuff kind of in that that's just like what is the raw throughput foot of data. I think again going to the proof thing, one of the things that's been interesting is you know to some degree kind of thinking about like how much gain do you get by being like lazy or in again to reference like monad thing they would call it like async execution where they're essentially separating the consensus bit from the execution bit and structurally to to me that looks very similar to like a lazy sequencer where the lazy sequencer is just responsible for consensus. you have a roll up full node that's going to do execution of that right and actually like you know you can separate that but it is this question of like actually how much are you gaining

and I think a little bit of it's like from like a latency perspective or from a bandwidth perspective like not that much so a little bit of this this kind of question in my head and it's been something I've been thinking about since East Denver and talking to people is I think we're relatively close to you know what I would say is like a relatively hard wall in that like we're not going to get like three like orders of magnitude improvements on like overall bandwidth we get in a decentralized system and right now our effort is kind of saying okay but like go slam into that wall right like maybe we have questions about like maybe that wall is pretty hard but like at least get to the wall and then we can reevaluate from there and say okay what are the really tricks here but it's a little bit of

like you're not going to get that much gain of like oh I've like you know trimmed latency out of my like execution bit because you're just going to run into this like the latency is defined by like the speed of light and the number of nodes in the network and you're not going to get 10x better and maybe you trim 20% % but at some level it's like like who cares if it's like 280 milliseconds or like 260 milliseconds. There's just a degree where like you can't even make a marketing pitch about those numbers anymore. Yeah. I I wonder you you talked earlier about Cosmos chains um and the the cost and overhead of running a validator network. How easy is it for existing Cosmos chains to move over to becoming an Astria

rollup? And have have some chains expressed an interest in that or done that already? H how how easy is it for for you know a Cosmos SDK chain to become an Astria chain? So they'll say it's like not like a batteries included thing. What I mean by that is like how a roll up on Astria works is we have like this comet BFT kind of network and then it's like okay how does like a like a state machine and so in in a comet BFT or like Cosmos chain right you have their own comet BFT network and they communicate with the Cosmos SDK over like ABCI right is like their interface boundary between the state machine and the consensus engine right we essentially separate this but we have our own API we just call it like the execution API which is like a slightly overloaded with like east stuff but we

didn't want to call it an engine API we didn't want to call it like ABCI And it's our own kind of implementation there. And essentially what this does is it's just a hey we produce a block on the sequencing layer and then we're just going to like send the data for its namespace and we just say okay you are a node we're just going to send the data right you can either you know pull it right you can pull the data out of the network you know we have like a websocket thing we provide a sidecar component that is like a this is a thing that knows how to read from Astria as well as Celestia. It has a concept of a soft commitment from Astria and a firm commitment from Celestia and then it pushes that data assumed to a like trusted entity on like a local host kind

of thing. It uses like a cloudnative sidecar pattern. But what we haven't done is we actually haven't implemented what I would say like the server side of that on the state machine for Cosmos SDK right now. we have implemented for GU and that is how Flame and Forma use it and so it's not like I say this like it's not that hard to get in Cosmos SDK and that like you know one of our engineers did like an implementation in like I don't know like 3 days or whatever and it was like a thing that worked and produced blocks. It was obviously not something that you're like I am going to go put you know a million dollars or a billion dollars on this and like ship it to prod but it's not something we've priorit because a little bit of our focus has been thinking about

kind of how do we I guess I'd say like it's not just like battle test but a little bit of like okay the criticisms we've got from like investors and a broader ecosystem around Astri is like are you guys doing too many things and like there's a little bit of truth to that right of the scope creep but a little bit of our focus is saying okay we're going to make sure that like we have a pretty battle tested platform where instead of going and implementing all the possible frameworks so anyone can kind of migrate on. We're going to kind of acknowledge that like you know there's 20 of us that's both a large team and a small team. We don't fundamentally have the resources and we haven't prioritized having an extremely large number of rollups that we can kind of provide customized help to. We're

going to focus on making sure that our system works very very well for a smaller number of roll-ups with our current focus being flame and forma. And then we're going to say once we have high confidence in that, we can kind of, you know, go out to people and say, look, instead of having to have a somewhat kind of like vaporwareesque like long- winded sales pitch of, you know, again, similar to this kind of 2021 era like rollup thing of like, oh, we'll do this and then the proofs will be good and then 3 years later we're like, ah, it was a little bit trickier problem. We're just going to say, look, we don't have to like tell you, we can show you. We can be like, this is how well Flame runs. This is how well former runs. would you like to get structurally

the same functionality out of this instead of kind of trying to be you know more speculative in our kind of like presentation or sales pitch to people. So again, long-winded answer, but right now it's like if someone is like a highquality like technical team in like the Cosmos space and they wanted to kind of migrate to Astria, we think we could probably get them something that they would comfortably work on as a DevNet they could cleanly iterate from in easily under two weeks and then it's maybe like I don't know call it a quarter worth of work to get something that is relatively confident and then you know audit time in there depending on who you want to go with and how you want to schedule but it's not something where we're providing that kind completely out of the box right now. And a little bit of that's because we don't

want to essentially like bait someone in and say, "Hey, I'm gonna go build on yours." And then they're going to realize we don't have enough resources to kind of support them to a degree at which they would kind of feel like they're getting a high quality kind of user experience or customer experience. Right? Since Celestia launched, you we haven't seen a massive number of rollups deploy on Celestia. I'm not sure what the exact number is, but it's, you know, it's it's in the I guess like low double digits. And there's been some criticisms or some observations that Celestia is not generating that much revenue. You what do you think of this slow adoption as as I see it of Celestia as a DA layer for rollups? What do you think needs to happen for Celestia adoption amongst

developers to really increase and be able to compete with the likes of you know Salana and and other uh really high fast high throughput monolithic chains. Yeah. Yeah. So, I think there's like there's like a couple components here, right? One is like the marketing thing. And you know, again, this is going to come across as like slightly a criticism, but I don't want it to be like overly like that. Like, you know, I love ECum and Ecram is like a master of his field and I think P is like appropriately respected, but there's a degree of like a lot of crypto is somewhat of like a similar game to like tech in that like at some level you do want to pre-ell your stuff, right? And you want to sell it as like you want to be ahead of the curve on the marketing before the product. And I think Ecum did

like a masterful job of that. But what you got out of that is right like at the start of this podcast you're like oh actually Celestia hasn't even been out for like two years but your like gut feel was like maybe it's like two and a half trending on three right and part of that is because that's also just cuz like crypto moves at a speed that is beyond comprehension but I also don't think you're necessarily like an outlier right I think a lot of people's kind of like default like kind of subconscious mental model as Celestia has been like live longer than it actually has right um and so there's like one side of that where there's a little bit of kind of time dilation right here of it's actually hasn't been that long. Same thing when people are like, "Oh, you've been building Astria forever." And I'm

like, "Ah, like I was like ideulating Astria in like mid 2022. First check was in in like October. Like I had four guys in like February 2023. Like, you know, okay, it's been like, you know, we've had like four of us have had like a 2-year anniversary." And I was like, "Ah, it's like that's a long time, but like it's a distributed system. It's an engineering problem. It's like scaling a team, whatever." Like that's not that long. So, so there's a little bit of just like you know people think things move a little bit faster than they can and there's a degree of like look if you want to build a novel thing it takes like some time and I don't want to like excuse like ah people are slow to some degree like Astra was late right like one of the reasons we haven't seen a lot

of things on Celestia like Astra was late right we were trying to provide some core thing and we took longer to build the thing than we were expecting to for like just a variety of reasons to the broader kind of like revenue question I think you know Mustafa had a good post on the forum last week week before last week again time is like you know an enigma here you know on kind of like okay do we just increase the kind of like revenue numbers, right? Do we just like 10x the fees or whatever, right? Because right now, you know, the fees are like trivial like to we we don't really like like yeah, we have like a spreadsheet, but we don't really think about the cost of like writing like data to Celestial. Like we write data, we have various monitors, we make sure we're writing all the data to

Celestia, but we're not really worried about like what that's costing us from like a balance sheet kind of perspective. On the one hand, that's good, right? on the other hand, you know, there's this inversion with like Celestia, of like that means they're not generating revenue if we're not thinking about as a customer the kind of cost we're paying there. And so they can kind of invert that, but it's also a question of, you know, I I think looking at Salana right now, right, they're in the kind of depths of this like SIMD 228, right? Like debate as like an ecosystem and there's this question about like issuance and like do you want a market rate issuance and etc et all the internals of it, right? But I know it's this kind of big debate on changing issuance just you know yesterday I gave a presentation to the team um you know

internally and one of the things I was talking about is like what are the things that generate revenue in crypto and and really it rounds out to you have token issuance you have you know what would be called like rev by like the block works guys or like you know the DBA guys like John Charbano right um and and that really means like you have like base fees and you have like priority fees and they split out into like you have base fees priority fees and MEV tips my general view is like you know and this is how Celestia thinks about it you wouldn't have priority priority fees like you knew about me when you were making the network. It's just like you know there's bidding for priority is like and that's what MEV does, right? But you have base fees and priority fees and if you look at that your priority

fees are going to be 98% of your like fee structure. So you have token issuance, you're going to have your rev which is really priority fees/mev and then you're going to have like front-end app fees and then you're going to have like swap fees and then to some degree you're going to have like lending markets but a lot of lending markets don't have like a fee switch on where ETH is the dominant lending market environment. you know, a had a proposal for that, but there's this degree of like not that many things as categories generate kind of a lot of revenue in crypto. And so I think that's fundamentally something that has to be kind of evaluated as like an industry as a whole is, you know, what are the things which generate revenue and what are the things that are kind of sustainable? How much are it cost to run

a thing? Who needs to generate what revenue? So there's that that marketing thing, there's that. going to maybe like the adoption piece on Celestia. You know, I think it's a little bit of a question of like again I come from like a like a cloud world and more of like an enterprise kind of software thing where it's a lot of kind of B2B sales thing. You know, I think Celestia's done quite a good job of kind of courting um you know folks like like conduit and like Caldera and these kind of like RAS ecosystems trying to launch a large number of rollups. But the thing that always sticks in my head and so we launched a RAS very briefly and we decided not to like lean into it. We like all right we made it work but we're questionable about the economics is in cloud the thing that they said and this

is very out ofd like seven years old or whatever but it was said that like you know 95% of all cloud spend is the top thousand companies on earth and like the key thing is just in a lot of these markets you're going to see a very power law thing and then the difficulty in enterprise software markets with this power law thing is the longtail might not be profitable for you ever and it's kind of this venture style betting on well maybe I'll lock in this guy and then they'll thousand next their revenue and now they're one of my big customers. But when you're sourcing kind of enterprise customers, they have a lot of pricing power because they know, hey, I'm half your revenue, man. Like, and therefore you're they're going to bid down your margins. The classic example of this would be like Netflix and

Amazon. Like Netflix pays whatever, like $2 billion a year or something to AWS, but like they're not paying the sticker price on any of that compute. They're paying like, you know, onetenth if that of like sticker price on things. And so I think there's a little bit of like we're seeing things be more revenue focused as an industry across the broad thing. I think that's something that we didn't really as just like a broad industry think too hard about for the last like three years or so, right? And now there's kind of this evaluation of like all right, what is the actual strategy we're trying to take towards revenue? And I think it's a little bit of an open question. So I know again long-winded answer. I didn't really answer your root question. I'm not existentially concerned about it right now because to some degree, you know,

what we're doing at Astria, what we're doing with Flame, what we're doing with like Forma is, you know, okay, other people could do other things. We have a kind of like confidence in what we're doing. If Celestia rates go up, then like we'll be okay with like Celestia higher rates. Again, we can cover that cost. If Celestia rates stay the same, then like that's also fine for us. From a broader ecosystem, I do think there's an open question going to one of your earlier things of like, you know, what didn't play out about module, what didn't play out about like rollups. I think right now there's a question of like how power law is this right to some degree right if you think about you know when you see people say oh I compare salon versus eth and dex volume or revenue or whatever people often say oh

did you include all the roll-ups but you don't have to include all the roll-ups you have to include like three or four rollups and then all the others are kind of like a rounding error here right and there's a question of like how power law is this going to be who's going to compete with bass to some degree right like from like a funding perspective right you can look at like monad as like this largest funding and they raised like 225 million And like I think they're doing great work. They're building a lot of innovative thing. But like Coinbase doesn't give a about $225 million that they're like we'll burn that like every quarter if they think base is like a big long-term play kind of thing and no one else has that level of firepower from just an ability to fund a thing. And so I think that's

the existential question as industry is how power law are these networks going to be and then who's going to have the actual war chest to if people say this is going to be big but it's going to be classic business market where there's going to be a big winner a number two that might be half the size or 70% of the size number one a middling third and everyone outside a top three might as well not exist. If we see that we're going to see a very very different market environment than you know oh we'll have 10,000 rollups because maybe it's great to have 10,000 roll-ups and maybe the more decentralized ethos of you know sovereignty and whatever but if people are thinking about revenue then we might actually see a lot more business style dynamics where it's going to be power law. Yeah, I um I tend to

think that there's going to be a lot of consolidation in the next couple of years and that there will be, you know, a handful of of chains, if not maybe ecosystems that capture the majority of the users that capture the majority of the liquidity. It just makes sense, right? If you look at things like aggregation theory, um that um No, I'm a huge Ben Thompson guy. Ben Thompson. Yeah. Yeah. stretcher right his name was escaping me you know aggregation theory I think also applies to crypto I think right now the ecosystem is still very nent and we haven't seen a lot of that but it's going to happen I mean it is it is

happening a little bit with the liquidity right a lot of liquidity is being kind of captured in bitcoin captured in the top assets and there are dynamics at play that make it such that it's you know because there's a lot of because there's VC funding for a lot of chains it's possible to get your chain off the ground, but once that VC capital kind of runs out, we see a lot of those chains become zombies and and not really amount to anything. You know, over over the next couple years, I think that's going to play out in a more substantial way. Which brings me to my next question, which is about, you know, liquidity fragmentation. Where where do you see how do you see that that playing out? And what what indications do we have that fragmentation is getting better or that we have like less

fragmentation, if any? And now that we have this pretty, you know, mature stack of modular components for for blockchains that allow us to scale, do we look back and say, okay, like the scalability of solve, but okay, we have like all this fragmented liquidity, or are there actually solutions to this problem? So, I think one of the things that been I've been thinking about this a lot, right, and I'll try to I can go down many many tangents, but I'll try to keep this somewhat kind of turf, you know, as much as I can. One of the things I think that's like interesting or maybe like a wrong starting point when people think about like fragmenting liquidity is there's just like a blunt reality of like liquidity costs money.

There's like an opportunity cost of like liquidity and so a lot of people are like oh my god you're fragmenting the liquidity and you're just like well like no one has like a free right to your liquidity. Like I think you know like like like you know like disclosure investor whatever but like I think the Cella guys are doing this thing where they're like you know there's this weird thing in ETH where like in like the MEV stack there's like people that are LPing and then like people that have like order flow like unis swap and like MetaMask and like whatever and like they sell order flow to like the builders and then the builders you know arb the like lps and then you know they send most of their profits you know over to um the block proposers which get it and there's this weird like why are these all

different parties kind of thing a little bit it's like path dependent And it's just like there just is a bunch of liquidity in like east USDC pools on like east mainet. And like when we think about it at like you know Aster and we're like I'm going to go like bootstrap liquidity on Flame. There's a degree of like all right I'm the guy who made the chain. I'm the guy who like you know has the like me infrastructure. I'm the guy who like deployed the uni contracts like set up the order flows on like our front end but I'm also the guy like going in like incentivizing liquidity beyond the chain. And so like I see all the parts of this and so there's a reality that like liquidity costs money and like the person who's going to have like the most liquidity is like at some level just like the person

who can monetize the liquidity best. And so like again we think like power law thing and so there's a little bit of and I think this is like common across like a lot of different areas in crypto where like people look at like a snapshot in time and they're like oh I'm going to like you know kind of like over um like extend based on like you know what I saw at a snapshot in time and you're like there was a very strange like non-re repeatable past dependence like the greatest one would be like like you can't do like a new proof of work chain right like you can't reproduce Bitcoin because like Bitcoin started mining when like no one knew what the hell this was. There was no thesis that this was going to be like a big thing worth like a trillion dollars or whatever. So like it

was a pretty slow trickle of kind of like you know people coming in discovering Bitcoin doing mining. If you launch like a proofof work chain today that anyone has any thesis is like going to work or be desirable like immediately right people will be stacking it'll be like complete power law and people will just like mine the hell out of it and like you know any kind of layman will have like no chance at doing anything right and so similar things there's like just because the state of it was like hey I'm a guy who has like order flow and I don't have to pay for liquidity because it just kind of sits there on ETH cuz it just has forever when you're starting again you're like no you don't get that deal you have to go buy your liquidity and liquidity costs what money

And so you have to go monetize that liquidity. And so then we look at fragmentation, there's kind of this view of like, oh my god, those people are fragmenting liquidity and like from like a liquidity provider perspective or like an efficient market thing, you're like, well maybe that guy just has like a better thesis on like how to spend the liquidity kind of thing. And there's this trade-off of, you know, users might say, "Oh, but I'd prefer all the liquidity to be like in this one spot and that would be better for me." And you're like, "Yeah, sure." But like when we think about standards, right? You know, there's like the XKCD meme people linked a lot at like East Denver. The way you like get a standard is through dominance, right? It's not through like, you know, yeah, you can have like your cabal coordination, EU, whatever, but

generally how you get a standard is like one person just wins so hard, everyone else is like, I'm going to stop bothering to even play that game, right? And I think that's if we're going to see liquidity like consolidate, it's going to be because someone wins so hard that everyone else just says like, I'm not even going to play, right? And I think there's a lot of kind of not dread, but that there's like a existential concern about like, well, no, no, we don't want to solve the problem that way. We want the liquidity to not be fragmented, but we don't want it to be because like one guy has 90% of the liquidity and we all have to pay rent to them if we want to use the liquidity kind of thing. So I think there's a little bit of this maybe disconnect between like the blunt

reality of how these systems work in like an efficient market versus what is maybe some optimal desired user kind of how the world works thing. Okay. So, so your your your take is that if I can just summarize this is that there there is more to the there's more to the idea of liquidity fragmentation than what we most people think because the cost uh the cost of liquidity doesn't get taken into account. Um people throw around this term of like liquidity fragmentation as I guess as a given given the infrastructure and how it has been modularized and there's just like these different kind of incentive pools but in fact it's more of a question of what it costs for liquidity to be sort of

sitting at on one platform at any time at any given time. Yeah. Yeah, I guess if I had to summarize it in like a little bit less like people thought liquidity fragmentation was a technical problem and were trying to propose like technical solutions to it. Liquidity fragmentation is an economic problem, right? Like people are going to pay for liquidity and they're not going to like share the liquidity they paid for, right? They're going to be like, I paid for this liquidity. I'm going to monetize the liquidity. Why would I give it to you? Right? And people are like, oh my god, they're like fragmenting the liquidity. They're like, you were freewriting on the liquidity. like maybe you are in the wrong and go pay for your own liquidity kind of thing. And then I do think there are technical improvements, right? I think bridging has gotten a lot better. I think the

interoperability has gotten a lot better, but there's also this again uncomfortable thing and the the you know the the decentralization maxi is going to be mad about this. But like in my view, the moment we had both ETH as a dominant ecosystem and Salana as a dominant ecosystem and it was very obvious that like neither of their coins was going to be like the dominant coin on the other one, right? Like you're not going to have like a do like ETH is not going to be like the dominant coin on Soul or whatever, right? So now you have two assets, right, of size, which each have their own ecosystem. Immediately that meant the standard asset for all of the things between those ecosystems was the dollar. And like that's just how the world works because again, money has a network effect. And look, man, like

China's not beating the dollar in like international markets. Like I'm sorry, but your blockchain is probably not going to beat the dollar in international markets. Like people are going to transact in the dollar as their like intermediary ecosystem asset. And so I think we'll see liquidity fragmentation in like a ability to move like nominal value between ecosystems will be resolved but it will not be resolved in a way that is satisfying to the decentralization maxis right it will be this hey I want to move a million dollars well there's a lot of like dollars in this system there is an intent RFQ orderbook whatever crossbridge that is willing to take whatever collateral on this side and give it to you on other venue and they are going rebalance either using USDC CTP CCTP or you know whatever like a

layer zero tether zero whatever thing or if you see you know all the stable coins that we're going to see come out like this year that are yield bearing they're going to have some crosschain chain rebalancing redemption mechanism and you're just going to have a bunch of solvers who are just going to be like I just calculated my risk adjusted spread I have to take on dollars across chain and that's how you're going to move a lot of money across chains and People are like, "Oh my god, central." Yeah, but like that's going to give you a lot of efficiency in moving nominal amounts of capital between ecosystems, but then within a given, you know, application or chain or ecosystem, you know, you're going to be like, you look, we can move all the dollars around, but then if you want liquidity in your decks or

whatever, right, you're just going to have to pay for that and you're just going to have to pay a relatively competitive market rate. And again, a little bit of you like that's a good thing. Like as you know, it's like there's this like very Peter Teal thing, right? He's like, you know, competition is for losers or whatever. He's like, you don't want to be in a competitive market. As a user, you want to competitive market because you have products. But as a builder, competition sucks because you're just margins are getting destroyed. And that's a little bit of the phase I see crypto in where like, you know, for a long time, USDC circle, whatever, has gotten like functionally all the yield on their like USDC, right? And like we saw like Noble Dollar come out like yesterday, right?

And I like love the Noble team, right? Um, and we're going to see like in the next six months like a bunch of stables and they're all going to give you like I don't know like 4% or whatever like T- bill minus you know whatever 10 20 30 bips or whatever and so circle's just going to be like ah I guess our margins are just going to go down by 4% annually in like a six-month window and that's going to suck for Circle but it's going to be great for everyone else cuz now your dollars are more yield bearing right and so that's my broad thesis again to the interrupt like we'll solve the like ability to move nominal value between chains probably not in the way people necessarily wanted it to go, but if you want to move dollars between chains, you'll probably have like a

pretty deep liquidity market to do that in the next 6 months. Yeah. So, let's talk about um let's talk about uh Flame a little bit and some of the stuff you guys are doing on the DeFi side. Yeah. So, um what we kind of done with Flame is a little bit of like kind of similar to a lot of your questions here, right? I think there's a lot of questions around like what generates revenue in crypto. You know, the question that's been stuck in my head forever. Like I remember being asked this on a panel at like like 2022 when I was still at like Celestia or something and it was like you know where is value going to acrew in the modular stack and like it was literally those words were asked on like panels like constantly right and I think there's still like an open question

right is DA a commodity is DA not a commodity is settlement layers where the value occurs is it at execution is it at the app layer is it the protocol of the app or is it at like the front end of the app again aggregation theory a little bit of flame is us cheating and what I mean by that is I'm just saying look I don't know how to answer that a priority. So, I'm going to vertically integrate a bit and I'm going to make sure I have exposure to the whole stack, right? And so, we said fine, we'll build a settlement layer. It'll also be a sequencing layer. We'll also have our own rollup on top of it. We'll deploy our own UNV kind of like clone on top of it, right? I don't call it like a fork because like we very very intentionally

did not touch it. Like, thank you unis swap. They made a great protocol. I did not want to touch that code. It's great as is, right? And then we like run a front end on top of it. And then we're also, you know, super happy to be working with like the Morpho team and we're going to have like a Morpho deployment. And if I go way way way back to like mid 2022, like the first like pitch talk I had for like Astria. The vision we had when we were still like a settlement layer and we weren't thinking about shared sequencing yet was I had these like these like four boxes and I had like Celestia and I had Astra on top of that and then I had two rollups on top of that and one was the DeFi rollup and it literally in the box it said unis

swap plus a and then the other box it just said NFTs and so to some degree like that is what we've done like I have forma that's my NFT box and sure I have like unis swap and then I have morpho in like this other box and like the reason we've done that going back to kind of a revenue is like what are the things kind of generating revenue in this case really it's swaps on spot and it's lending and lending revenue is a little bit dependent on the fee switch and we'll see like what a does with like their governance fee switch proposal whatever right and we wanted to kind of say okay let's go make sure we can provide again going back to this celestian thing not dependent on Ethereum as like a protocol or network at all let's make sure we can provide

the status quo base case of what people do in DeFi what is the majority of volume of TVL of like user activity let's make sure we provide that for the Celestia ecosystem and that's what we've really focused on kind of getting out with Flame and you know that's kind of what we have now. We do have a bit of optimization to do quite frankly like the front end that we have is like not very good right now and so like that was again in the scope creep that was like the last thing where we had like you know one of our employees been with us forever you know he's like great you know Jesse but like he was like half his time was building like a front end kind of thing and then we're like all right we need a front end the new you know

administration's in you know they're throwing out the lawsuit for you know MetaMask and like unis swap we're not worried about running a front end for our protocol anymore and so we staffed you know a front-end team starting in kind of January with two more people on and so we're rebuilding a front And really what we want is to say, okay, let's do this Celestian native DeFi. Let's see how much kind of latent demand there is for interacting with TIA as an asset. You know, one of the terms we thought about using as like a marketing term like, you know, do more with your TIA, right? Right now people say, oh, there's not that much to do with TIA right now, right? And so we want to make sure they can have like this status quo of what can you do with a given asset in

another ecosystem. What can you do with ETH? What can you do with Salana? Make sure people have a venue that is Celestia native where they can do that with TIA, right? They can take their TIA, you know, they can liquid stake it through, you know, drop or they can go through Milky Way or they can go through Stride. They can get that on Flame. They can get USDC from Noble, right? You know, we're really excited to support like USDN from Noble. So, we have a yield bearing asset and then we provide, you know, liquid. We have a bunch of tea liquidity right now, right? You know, one of the big things is like, you know, we're not like a marketing focused company that we're working on kind of in Q2 a bit, but we have, you know, something like $9 million of like TVL and an AMM. If you go swap 10,000

through um you know TI USDC through Flame right now, you're going to get very competitive rates with um Coinbase on the same thing, right? You know, it's depending it's marginally worth. The pools need to be arbed a little bit better. We're working on having kind of, you know, top of block like arbitrage lane and then we're going to start working out, you know, um attracting searchers in the next couple of weeks here so that we have these more efficiently arbed and thus accurately priced pools. But really fundamentally, we're trying to provide a competitive swap environment for that. And then you know we build on top of skip connects Oracle. We really you know love the skip team and we think they did great work with that. That's going to be a price feed we feed into like morpho and all of that will be represented in this kind of

like you know super app kind of like front end for flame. And really we would just want to make sure we are providing at minimum the status quo of what is kind of the standard DeFi tools of spot and lending in the celestial native way and that's what we're kind of doing with slain. Very cool. And uh do you want to talk a little bit about the NFTs? Yeah. So on oh no like forma is not like us like we work very very closely with like forma but it is like technically like you know a different team. I think the biggest thing right you know that form's done really well is they're not how to word this right so like I did a lot of like NFT stuff on Salada in 2021 and that was this very 10,000 PSP very memecoin esque right the memecoin had a

lot of like similarities but it was gambling right form's done the opposite right they're really you know it's I really respect they're like very artist forward like they have like noted like name brand known artists like I've talked to people like oh I know that person in not a crypto context like I like that artist and I'm like yeah they issued an NFT collection so I think they've like legitimately a lot of really really good art and I think not kind of sacrificing that principle right not leaning into the you know overly speculative excessive kind of gambling thing has done great right we have the mammoth collection right you know any kind of given like um you know ecosystem wants to have this kind of flagship kind of NFT collection we see the mammoth we love the mammoth we think that's been kind of a great thing for the ecosystem

to kind of have again in this like very Celestian native fashion right okay we have this kind of stack that goes to Celestia it's not half over in Ethereum or like half over in another Cosmos chain, right? It's just this Celestian native stack. And so again, I am the former guys. I don't want to like overly speak for them, but it's also been really really exciting to work with them as the first kind of customer of Astria and being able to kind of build something out and kind of getting this kind of, you know, arms length, you know, working with them to learn how can Astria kind of provide them a chain that is good for them to use that allows them to do what they want to do with this kind of art focused NFT chain. Cool.

Well, Josh, uh, thanks for coming on. I mean, this has been a really fascinating discussion. I've learned a ton about Astria, but also just sequencing in general and how to be sort of thinking about the role of sequencing in the space. Um, anything else you want to share before we wrap up? No, I mean, I I think, you know, my my kind of like chill here would be like, you know, I encourage people go try out Flamian, flame.astria.org. You know, we think we have, you know, what is a relatively compelling product. The front end will get better in the coming weeks, but we're more, you know, we're very excited to have people kind of like start kicking the tires and being able to do a little bit more with their TIA than they can today. Yeah, I'm excited to do more with my TIA. I'm currently liquid

staking, but I'd love to be able to do more as I I'll definitely give it a shot as well. Thanks for Thanks so much for having me on. [Music] [Music]