DoubleZero Founder: Crypto’s Scaling Problem Isn't What You Think

Transcript

People forget how much of what we call the internet today is not actually the internet, right? It's all private networks. The kind of trade-off argument becomes one of like, okay, like are you trading off your life or your freedom for uh your network? I think it's very telling that none of the major liquid staking players in the Ethereum space were able to make their products compelling on Salana. Welcome back to the Rowup. We are your hosts. I'm Andy. I'm Robbie and this is your home to learn from the biggest guests, the brightest minds at the tip of innovation. Our goal is to help spread the collective knowledge of everyone in our community. We are breaking crypto out of its echo chamber and taking it to the real world. Sit

back, relax, and enjoy a quick word from some of our awesome partners who make this show possible. [Music] The intersection of AI and crypto is moving fast, and soon an agent will be managing your onchain transactions for you. But without guardrails, it's a disaster waiting to happen. Newton locks agents to your intent with cryptographic proofs and real penalties if they slip.

Learn more at Magic Newton on X. All right, Austin, welcome to the rollup. Um, live in New York City. Hey, thanks for having me. We made it through the rain even though the Uber took us to the wrong spot. Both of us. It's kind of funny. I don't know like what happened, but both of our Ubers dropped us off on like three blocks that way. Well, we we made it here now. We have got the new uh art from our guy Brendan. Shout out to Brendan. Got the Roly Pole in here. Um and thanks for joining us, man. I first time on the show, so I think it's it it'd be mistake to not hear about how you got into crypto. who's who's Austin and um how you got into distributed systems, man. Yeah, I mean getting into crypto uh technically 2013, but not

really. That was mostly buying some Bitcoin and then promptly losing it on a exchange that got hacked. Um, and then kind of being out of the space again until 2017 when I was working for a fintech company that was kind of in the process of failing. But one of the last things they tried to do to not fail was build what we would now call a DeFi application, but back then was just to borrow lend token um on Ethereum. And project never went anywhere. The company kind of pivoted and wound down a bunch of other stuff. Um, but that got me hooked on the concept of it because like bunker coin was never really my thing.

Um, like the whole world's going to so we should own Bitcoin type of thing. Yeah. It also is like oh great like we're going to have internet in our bunkers. Like it just it didn't you know uh it wasn't interesting to me and I didn't wasn't really that focused on whether it was going to be valuable or not. I just wanted to spend time on things I thought were worth spending time on and Bunkercoin wasn't that for me at the time. Um but coming back a few years later to Ethereum and having smart contracts which you know everything's named dumb in crypto be like smart contracts what are those oh programs okay yeah I know what a program is um and so going through that process for me was like oh there's something here and uh I want to go work in this industry very cool and I I remember about maybe

like six months ago or so I watched a podcast of yours with a from Celestial Block Works one of their shows talking about strategy, marketing, brand building. And you're the former strategy or head of strategy at Salona, which you know, notably has done very very well in terms of community, brand, that type of stuff. Um, so I'm just curious about your experience there. How did you go from this kind of technical realm into the strategy realm? Um, and what does a good head of strategy do? Yeah, I mean, it's a title that means kind of close to nothing, I would say. like like head of strategy is either like at some companies it's like financial planning at some companies it's what you give

like the co-founder who you kind of wish didn't work there anymore the title of it and some places like Salana it actually like ah it's a kind of meaningful title for the thing but okay uh that's the problem with titles is they mean different things in different places um you know look Salana the superpower of Salana was minimally viable foundation basically right like minimally viol viable intervention. And so you see a lot of places out there where it's like something's happened. What should the foundation do? And it's like well maybe the answer is nothing.

And most of the time maybe the answer is nothing. And so a lot of the work that was done at Salon Foundation, which like it's hard to say in a bare market, it's a lot easier to say in a you know bull market or a good market is it's the job of the foundation is to do as little as necessary. And so a lot of the work I spent time working on was like what are the things that either like no one else has an incentive to build that need to be invested in by the foundation with the treasury that it has. Um or why should we not do something someone is telling us what to do which is kind of an interesting thing because we're in this like very future forward industry that's always thinking about the next big thing and where do we go from there?

But like a great example of this is like post FTX, right? There's a ton of projects that are suddenly kind of underwater. A bunch of NFT projects that are like, ah, our NFT the NFD hype cycle is over. Our projects are kind of dying. And a lot of them kind of come to the foundation and are like, we we need a bailout. It's like, wait a minute, like what are you building? Yeah. And it's like, oh, it doesn't matter. We just need a bailout. Right. And and it's really and there's a lot of DeFi projects that this ended up happening to as well. And it's really kind of brutal because you can bleed out, you know, it's not like the Solana Foundation had a huge US dollar treasury at that point either. And so, you know, a lot of the work ends up being like we actually

don't need to play kingmaker. We don't need to save this thing or like actually in these situations we really do need to step in and do something or participate in this market or do something along those lines because it wasn't there was no other option on the table there. And so, um, you know, I would often describe my role as like the least popular person of the foundation. Like it I it was a real skunk role. Yeah. Saying no. Yeah. Everyone wants to give away money. Like if you're if I tell you like, "Look, your job's going to be to give away money." You're like, "This is the best job in the world. Everyone's going to Everyone loves me. Everything feels fulfilling. Like it's not my money I'm giving away, right?" Uh it's kind of like um you know but that's not how you build a good ecosystem over the long

term because then you start putting the foundation in the middle of that right and when the foundation becomes a kingmaker like the network has become I wouldn't say poisoned but there's a certain level of like um well I have to be friends with this person or I have to give this person a grant or if I don't give this executive director like some allocation in my round like I'm not going to go anywhere and stuff like that and and that's the type of stuff that I I think really can kill the ecosystem of a network. Um, and so it's it's about making sure that the work that's being done is for the furthering of the network and not for the furthering of the people who run the foundation. Yeah.

Yeah. And I think that had to have applied more so than ever post FTX soul trading like $10 for a while. All of crypto Twitter dunking on Salana. Yeah. the DeFi apps, the NFTts are struggling. And then, you know, I I like to think of it as one of the greatest comebacks from that, you know, in in the rise of 24 25 up till now. Um, what were some of the levers that you guys pulled in that deep bare market that ended up having an outsized impact that you may have not thought would have, right? like things that that you just did that end up having that that kind of big breakthrough or breakthroughs um which may have not been obvious at the time.

It's a good question. You know, I think it's more I would say it's it's almost we had pretty good amount of confidence in most of the things that we were doing. I think it's actually more surprising the ones that weren't successful. Right. So there's things like um you know GTO started in the bare market. Yeah. Right. I don't think anyone thought Jetto wasn't going to be big. Yeah. Right. But I think one of the things that that we learned in the bare market was a lot of these EVM teams that wanted to expand to Salana, specifically on the infrastructure side, weren't up to the task at all. And they came in, they talked to big game, they asked for big grants, we put them behind deliverables and milestones, and they didn't hit any

of them. Yeah. Um and then a bunch of them left Salana. Um, and it was the homegrown teams that like came up through the grind, right? Jetto, Helas, like you know, all all of those types of Jupiter, all those types of teams that were able to like really just capture the market, right? And and and build not like the Salana like new DeFi on Salana, but build like a fundamentally different type of DeFi. Like what Jupiter does and the way it does it is fundamentally different than the way aggregators work.

on other networks and like yes you can squint and say like oh well it's the same thing it's like yeah okay Erade and Robin Hood are the same thing if you squint at them right but they're clearly not right like the the UI experience the way you interface with the products is as much of the product as the underlying technology stack and so um I'd say that one that was one of the big learnings was really like the the homegrown native teams if you know nothing about two companies except one is moving from one ecosystem to another and the other one started here, the one that started here is usually going to be a better performing project. Yeah. The native the the ones that are there for the right

reasons. Yeah. Rather than just kind of hopping for grants or just for other Oh, yeah. You know, other opportunities and I think we've seen that time time and time again. Yeah. But there are exceptions, right? Like pumpf fun started on base, right? Um and they did not find any product market fit and they found product market fit with Salana. Yeah. Um and so you know it's not a hard and fast rule but some of the teams we were most excited from about coming to build on Salana were like veterans of EVM DeFi and then the things on Salana like never went anywhere. And then you know you have like things like Sanctum and um like none I think it's very telling that none of the major liquid staking players in the Ethereum space were able to make their products

compelling on Salana and the Salana space is owned by an entirely different type of much more vertically integrated much more profitable um liquid staking provider system. I think some of those people would say that the developer experience on Salana was a headache and that they wouldn't want to come back, but who knows? Well, no. So, this is actually a good like this is like a really good test because this is like and I don't mean to be too mean to these devs, but like this is the difference between a real job and a crypto job, right? Like no one complains who works for like a Tradfi firm that like, man, the the dev experience of the New York Stock Exchange sucks. Well, too bad.

It's your job to write quant trading systems. Like, you think they you think they like writing everything in C, right? Right. Like, and so there's there's a little bit of this where it's like, um, I love this industry so much, but it's just like, man, we got to grow up sometimes because like if you fumbled the bag on a hundred billion dollar market because you thought it was a little more hard to develop on Salana. Yeah. Yeah. Like, come on. Yeah. Yeah. Come on. There's a lot of great reasons to have been doubtful about Salana during that cycle. Definitely. Right. Um I I think what we what we what I I saw at least was like the the the biggest blessing Salana got post FTX was a swift, speedy and decisive trial of Sam Bankman been

freed, right? Like if that had drawn on for years and there were data leaks coming out and sort of this always like well what if Sam had the secret stash of soul somewhere or something like that would have been so detrimental to the network. But getting that like cleared off almost one year to the day was just like an incredible feat of prosecution. It let everyone put that behind them and that was basically what you saw was like the VCs came back to the Salana the second the trial was done. Yep. Yep. So kind of fast forwarding to now um Salana's grown its ecosystem you know tfold it feels like since the since the bare market in terms of users products applications but you've decided to kind

of take on a different feat and a different problem which something that that like us us in this kind of like modular horizontal scaling camp push hard and you know we we may have differing opinions as to how we solve these problems but sure the problem exists in that we got to basically increased bandwidth and reduced latency across the board for Salana, for all of crypto. So, you've came to double zero. Now, we'll we'll get into double zero, how that works, and maybe have a couple arguments about scaling, but before we do that, I'd like you just to break down your perspective of the fundamental blockchain scaling bottleneck. like what is the core of the blockchain scaling you know dilemma and and why does this exist in this form of distributed

systems so it depends a little bit on what your objectives are right but if your objective is to build uh a network however you're defining that could be a c you know a uh an L2 could be Salana could be a rollup whatever that is intending to serve the global market with some amount of fairness baked into how you serve the global market. Um, communication is your limiting factor on scale. Full stop. That was not the case in 2019 and 2020, right? In 2018 and 2020, if you if you ran GE locally, good luck getting it over 400 transactions per second, right?

Like the core software packages just were not and the core validator clients just weren't performant enough to get any any amount of high throughput. now but we've seen huge amounts of rewrites of pretty much every major validator system out there today to the point where you have very fast mostly paralyzable EVM you have extremely fast Salana you have sooie which is quite fast you have a which is quite fast you have lots of different systems nowadays that are quite performant the limiting factor across the board is how quickly you can propagate vote and synchronize state between all of these systems and the public internet is not good at The public internet is an incredible tool for reaching any point on the globe. But if you think about what

happens in let's take a decentralized network first and we can move to a centralized single sequencer network. But in a decentralized network, what happens is a block is built in one location and then that block has to be transmitted out to every other validator in the system so they can vote on it. Let alone there's inbound transactions flying all over the place. Um so your basic your center of data usage is moving constantly around the world. Yeah. And so if you can't communicate out the state the network can't advance, right? And so you can think about this as like the light cone propagation from a block that's produced on a distributed decentralized system. That is a very big problem, right? because each time a new

block is produced at a new validator, there's a there's a propagation delay from each of those systems. And if you're building a block in uh let's say New York City, there's a heck of a lot more internet connectivity infrastructure for transmitting that block out than if you're building a block in Buenosares. Also, Buenosares is very far away, right? Like we think about like, oh, it's it's light. It's, you know, it's like no, like the internet's slow. Like it can take you 700 milliseconds round trip to talk to Singapore. M that's nearly a full second. Yeah. Right. And you can run into far more delays in internet traffic that make it take even longer. So in the Ethereum L1 or the Bitcoin, there's no need for double zero because the blocks are so long, right? 98% of an Ethereum

block of that 30 seconds is just for networking and coordination and voting. And so if you're comfortable with producing one block every 30 seconds, public internet's okay for that application. But that's part of why Ethereum does, you know, 21 things per second on a good day. It's it's not designed for performance. And even if it were, you run into simple like state propagation problems. And so this is new though, right? For for years, the problem in blockchain was the software wasn't performant enough. And now we've got to the problem, which every successful software company gets to, of our software is actually pretty good.

It's the underlying infrastructure that's the core problem, right? And it's the it's in your in your perception, it's it's this data propagation problem of the the passing of information across all the validators or nodes in this redundant fashion. It effectively at the speed of light. Yeah. or faster uh in in some like collocation scenario or even in like a centralized sequencer scenario. And your your take here is that essentially we need to build something that provides the underlying infrastructure for these softwares to be faster. That's the double zero like problem, you know. And this exists in

trifi today, right? Like I think the thing that people forget is this has existed for 30 years in the traditional financial and traditional internet space. Yeah. Um, no major financial firm uses the public internet to trade on because it's too slow. Right. Right. They they started off running their own fiber. Right. This is Flash Boys. Yeah. Um, they switched to microwave links. They switched to shortwave now. Uh, and they're using dedicated private networking to do this because it's much much much faster than the public internet. Much more robust. You can use technologies like multiccast which allow you to move. It's basically hardware acceleration of data replication. So right now, I mean, even in the central sequencer model, you're limited by the

public internet. Right. Right. And people like, well, how are you limited by the public internet? Well, think about all the data flows that go in from when you open Phantom or MetaMask to when it's actually included in a block built by the L2 single sequencer. You open your wallet, you have a call to the RPC server. There's a whole bunch of back and forth there which is pretty slow over the public internet. Then the RPC has to send like we take like arbitrum for example it has to send something to an arbitum node. That node then sends it into the sequencer. The sequencer then has to include it and then the wallet has to act then sorry the node has to get the data from the sequencer. The node then has to communicate the data to the RPC that

says yep this was included. Then the RPC needs to go to the user and say, "Hey, good. Your transaction actually landed, was included." And so every single piece of that is network latency. Like if you actually break down how much of the time is a compute problem, it's less than uh less than a tenth of a percent. Of the actual amount of time it takes to get a a transaction, doesn't matter the complexity of the transaction onto the chain is actually the computer doing math. the rest of it is just waiting for the network side of things to do stuff.

And so if you look at kind of like what what is the biggest time savings we can get in blockchain, it's not making the compute faster, it's making the networking faster. So it's not a matter of reduction of block times to as low as possible necessarily. It's a matter of this latency problem more than anything. And they they're all interrelated, right? Like so for example if you if it if it takes right now 800 milliseconds to propagate a block and get blocks voted on and get that sort of sent around the world for a cluster. If you go from you know if you go down to 100 millisecond blocks well suddenly your finality lags by at least eight blocks.

Technically your confirmations might even lag by eight blocks. You go to 10 millisecond blocks now you're lagging by 80 blocks. Yeah that's a huge amount of kind of variability that you have to go into with these systems. And so if you can make the networking layer faster, you can decrease block times or you can decrease finality, right, by the number of blocks. And so there's a ton of advantages and it's really up to protocol designers what are they optimizing for, right? If you're optimizing for systems that produce single block finality and you can keep your current latency, well, you're going to you're going to get that finality in a block and that's great for certain types of DeFi, right? If what you're optimizing for is like responsiveness of system, that's a that's kind which is

kind of more like what what apttodos is optimizing for that's a different and then what base is optimizing for with pre-confirmations. That's a different type of optimization that you're doing. But both fundamentally are limited by the network. M so then if we're thinking about that in this lens I is there that much difference between a shared state machine with a highly optimized infrastructure layer as you guys would say an N1 layer for extremely low latency versus a a a base layer for maximum throughput with tons of like individual applications that exist in their own state but sh but share that same optimized base layer. like it's it's one and the same in a in a sense it it can be. The problem with a single

sequencer is it's regional. Okay. Right. Like you don't have anyone in Singapore trading against the New York Stock Exchange, right? Why? Because there's a 250 millisecond delay every time you want to send something in, right? Why is Coinbase mostly only used in the US and Europe? Why is Binance mostly used in Asia? like these all of these exchanges are regional the same reason financial markets are re regional latency right uh if you are colllocating at the New York Stock Exchange and I'm sitting you know even here like seven miles from where the trading servers are at a massive disadvantage to you let alone if I'm on the other side

of the world and so the problem that single sequencer systems are going to have to tackle is are Are they happy being regional? Right? No one in the Western world uses BNB. No one outside of the United States and maybe Europe uses base. And there's reasons for this, right? You're at a you're at a programmatic systemic disadvantage um when you're dealing with these types of systems. Now, there's ways to potentially address this. One is like, well, we can have multiple sequencers and or the sequencer can rotate. And so, you know, 50% of the time it's built in North America, 50% of the time it's built in Asia. Sure, but like you still have the problem of there's always incentives to colllocate with the

sequencer, right? And therefore, whoever's colllocating with the sequencer will have faster access not just to writing into the sequencer, but getting state updates. So, if you think about it like if I am a if I'm trying to trade DeFi on a system, there's two things I need to know. Well, there's two things I need to do. Get the current price and respond with an action. So that first one, I need to get data from wherever it's produced. Whether it's a block produced by a validator or a block produced by a sequencer, I need to ingest that and index it. Well, if I'm sitting next to that computer, I'm going to get that data very quickly. If I'm sitting in Singapore, it's going to take me at least a quarter of a second to get that data. And then I when I want to

take an action based on that, I need to send that into wherever that sequencer block builder is. M so this is the problem with single sequencer systems is it's not that I mean like you can you can make arguments about decentralization or control or whatever but even if you get a end stage single sequencer rollup with perfect security perfect exit windows perfect fraud proofs zero program upgradability the sort of the holy grail of a stage two rollup you you are not going to there's no way to make that a fair system there's always a geographic collocation preference And that can happen on a system with multiple nodes as well of course right which is why permissionless validation

and permissionless block building is really important if you like hyperlquid's a great example right hyperlid is not a single sequencer system it has 11 or 12 validators I forget but they're basically all run in Tokyo at this point yes and so while there is more ability to run a node um trading against hyperlid from New York is not a great experience because you're always getting front run by people who have fresher state and closer access to submitting transactions. Yeah. So even in the example of hyperlquid or you know hyperlquid to in a further decentralized manner which you can kind of draw a through line towards Salana you're still going to have a a a you have a different set of problems than a single sequencer

has right the the single sequencer location geographic issue is is very clear and obvious with with regards to proximity to it but the the data propagation and latency issue of the of a kind of decentralized validator system for a shared state is going to be across whoever is in control of the system um propagating that data. So in in this case double zero you know you have telecom companies in in the world right now there's you know several large telecom companies that manage all of or a lot of the of the information passing from different uh different hardware computers. Same thing will happen in

validators across a shared state uh you know high high performance chain. So how are we thinking about you know that trade-off with regards to okay there so for single sequencers we've got this kind of like regional colloccation proximity problem y in this decentralized validator but a high performance network we've we've got this kind of monopolization of information propagation problem. Uh I mean you still have that in the single sequencer model, right? It's just like now it's based on what data center you're running in. So there's about 20 tier one ISPs in the world which are sort of like the the largest global network providers. Okay.

And everything runs through them today, right? So and they have very good deep packet inspection and if they want to they can just drop all the Ethereum transactions. We don't think about this, right? This this is this is what this is like the whole thing about like the left curve answer to this is Bitcoin, right? You have blocks you can validate on an FM radio broadcast from a satellite in space where you can do offline validation and you can just walk the state yourself. Yeah. But like at the end of the day like we should not kid ourselves that if the OECD countries decide Ethereum is illegal, Salana is illegal, shut it all down, Bitcoin's basically going to be the only thing that survives that. Yeah. Right. and and and that is like not one state taking an

action but that is like 80% of the world's economy taking an action to get rid of the thing and it's because it all runs over the internet and you know if there is enough like at the end of the day if there's a guy with a gun telling a company censor this thing it's going to get censored correct and there is no such thing as an internet that is not run by a company right and so at the end of the day like yes we maybe have people set up pirate radios and but that's just that's Bitcoin at that point, right? Um, and so when we're talking about this on the zero side of things, our solution to this is multiple independent network contributors. And so like we very much accept that if all the OECD countries

say no blockchain, zero is not going to be censorship resistant to that, but also neither is any other ISP in the world, right? You're going to be running validators in like North Korea and South Sudan and like that's it, right? Uh that's not a great solution for anyone. I don't think that's the answer. No. Um and so our answer here is is multiple independent contributors. So for example, Tokyo, we have five different lines coming into Tokyo. They're all on independent cables or most of them are on independent cables. Um and they're from multiple independent providers. And so if one provider says, "Hey, like uh you know, we've got this big thing coming up with this regulator. They don't like crypto as much. we're gonna turn off crypto service, you know, or

something like that, and they decide to back out of the network, we have multiple other contributors that are going to be able to continue providing connectivity. And this is sort of an extension of the principle of like L2s only have a censorship problem because you can't run a sequencer yourself. There's no censorship problem in Ethereum, the L1, right? 60% of blocks are OFAC compliant and you can still land a transaction that's non OFAC compliant just fine because there's actually a lot of nodes out there. And so yes, maybe 60% of the time you can't get your transaction through because these validators installed a block list, which is true, right? That's how it works today. Um but 40% of the time you can. And so that's actually a pretty

good ratio, right? That that that's a coin flip whether my transaction is going to get through or not. Um, that's not bad in terms of censorship resistance. That's pretty good. The minute you move to a centralized sequencer, you have no censorship resistance, right? If if if uh you know um or what you're what you were relying on is like extreme goodwill. Yeah. If if tornado cashache was deployed on base and got sanctioned, we would have had a very different story than tornado cashache being deployed in Ethereum and getting sanctioned. Right. And I I think it actually applies to a lot of other chains than just Ethereum. Um, sure. But yeah, I totally but Ethereum is like a very good example of of how you can still have a good portion of the network

act engaging in censorship, right? An OFAC block list is a form of censorship. I think it's a very acceptable form of censorship personally. Yeah. But I'm not mad at any validators that are doing that. But it is a it's a great example of how you can still have a network survive. Yeah. Innovating finance, global, institutional and onchain. Deep liquidity to drive capital efficiency. Now maximize yields on institutionalgrade liquid staking with M. Unlock trillion dollars of capital via Fitbitcoin. All of this secured by Mantle's treasury of billions in stables and ETH capital pools. This entire

ecosystem is coordinated by a vast network of seasoned VCs that are aiming to create the next generation of banking and financial technology. This is Mantle. Okay. So the there's to kind of close that out. There is a list of trusted ISPs currently. They obviously are under the same principle as any you know company like D0ero who if there would be some sort of nation state or power uh enforcement on these networks then you know the the kind of trade-off argument becomes one of like okay like are we trading off your life or your freedom for right uh your network right and so that's obviously kind of much uh

more grandio scale on the on the more practical scale for you guys multiple independent contributors at the validator level and at the kind of network level to be able to provide um with some sort of service level agreement and yeah likely like some some form of token incentive of uh with like slashing for uh get preventing malicious behavior and then that is a sufficient enough uh network model to kind of build out slowly to enforce and solve the problem that I was originally getting at which is like okay well sure we have centralized sequencers with like a really performant base layer like a Celesti or something, but you guys are still going to have to worry about, you know, who's in control of passing that

that information. So now what you're saying, but it's a little bit like saying, well, the super performant rollup still has to worry about who's running Celestia. Like yes, Celestia can rug you too, right? Celestia is at the end of the day if that is your if that is your DA layer. Yes. And if they if the Celestia validators collude to say you know no more roll up X Y or Z like okay that's it. Yes. Right. And so at the end of the day censorship is an important analysis to view this on. I think the difference though is like0ero is not a settlement layer. It has no state. It has no smart contracts right all it is is a communications layer. And so from that perspective, like if you're mega eth and deploying on top of it, um

you're not like deploying in the crypto sense. What it means is you're locating your sequencers there. You're locating your servers and they have, you know, we can get real basic, they have two Ethernet ports coming out the back. Yeah. One goes to the public internet, one goes to zero. Yeah. Right. And so if you have um you know, maybe they're going to run multiple sequencers around the world and rotate them. Well, the communication framework between those multiple sequencers, maybe they're all on zero and suddenly you have the handoff between them going from something that's like kind of sketchy to do over the public internet and very slow into something where they know, hey, look, when we hand off from the New York sequencer to uh the London sequencer, we know it's going to take

like 64 and a half milliseconds for that handoff to occur. and it just becomes much easier for a network engineer to build that handoff system. Um, and so, you know, what we're not trying to do is replace any L1 or L2. We're fundamentally not building an L1 or L2. No, we're we're building an alternative communications pathway to the public internet. And this can be applied within blockchain as well as theoretically without, right? Oh, yeah. Yeah. I mean, people forget how much of what we call the internet today is not actually the internet, right? It's all private networks. Yeah. Um Meta is one of the largest investors and builders of private networks, right? And they do this so they can wire up the Philippines

so they can sell people ads. Um and and this happens time and time again. Software companies get successful enough they have to start thinking about hardware. And I don't mean like Apple build the iPhone. I mean like Facebook got to a size back when it's still called Facebook where they were like man renting data centers is really inefficient and we can't actually get what we want in these data centers. We got to build our own data centers. Few years later they were like you know what we got to build our own fiber. We have to build our own fiber network. Um and you see this time and time again with these big companies. Netflix did the exact same thing. Netflix operates a, you know, private network now for delivering data for movies out to

different places in the world. Um, financial companies do this all the time, right? Google has a giant fiber network. Amazon has a giant fiber network. Azure is a giant fiber network. Yeah. It's just crypto that's left out because historically we've had all these very reasonable questions like, well, why would I trust one company to run the network? Yeah. And the answer is you're not trusting one company to run the network. you're trusting, you know, at this point six and eventually more independent network contributors that are all part of the zero network. So even if these validators are plugged into zero, there's still like this last mile problem for them where they have to connect to somewhere they have to be plugged into somewhere how like which

leaves them vulnerable for some latency delays or connectivity issues potentially. What's your thoughts on kind of solving that? So you know in a ideal world in the end state we have fiber connectivity into every data center where people are running nodes. Okay. Now realistically it's kind of hard. There's you know numerous data centers in New York and so do we need to be in every one of them? Well no we need to be within like the metropolitan area and then we use basically software tunnels for people to connect in nearby but that metro connectivity is still pretty good. Okay. Right. So like if uh if we're talking about like a data center, you know, in New Jersey talking to a data center in New York City, that's a very fast connection because

there's so much kind of like fiber capillary connection within a metropolitan area. It's when you're going city to city or especially when you're crossing an ocean that latency gets really really bad or if you're crossing like a mountain range, any of these kind of choke points you might think. Yeah. Um and so from that perspective like uh we I think we can still offer very strong service to people who are not running validators in the same data centers that zero has fiber uh running into as well. Very very interesting. So, so do we expect these these L1s then like Salana others to have a ex like an extremely

tangible improvement in the UX as a result of using D0ero? If so, what is the how many how much of the Salana stake per se would need to be plugged in for like for me to go on Jupiter and like trade Fartcoin and be like, "Wow, that was definitely faster than the last time that I did it." Yeah. So, it's kind of funny because I think we're honestly at a point where I'm not sure you from a UX perspective are going to be able to tell. Okay. Because it's like I would say that Jupiter feels as fast as Coinbase today. Okay. Hyperlid feels as fast as Coinbase today, right?

or Binance uh you know today. So it's what it's more about is capacity under the hood. So there are certain pieces of physics that we can't get around with software. One being the size of the earth and the speed of light. Yeah. Right. Like we're never going to make it faster than, you know, 200 milliseconds to get to Singapore until we invent, you know, nutrino pointto-oint systems and start shooting particles through the middle of the earth. That's like the real endgame here, right? uh is uh you know circumference is a lot bigger than radius or yeah maybe then uh L1's will be able to compete with with horizontal scaling but you know we we'll see but like the thing we're the thing we're kind of

looking at here is like you know uh the validator clients for these networks can be much faster than the mainet performance today but if you go to like breakpoint last year uh the fire dancer team demoed a six node cluster running on four continents that could do a million transactions per second that was running on the double zero network. Right now you will not see a million transactions per second. You know there's a whole bunch of things that cluster was not also doing right but just as a raw throughput demo it shows you kind of a bit more of what's possible. I think we're pretty, you know, we can say with a high degree of confidence that like you get, you know, uh, more of the Salana cluster on zero, devs are going to be able to start

pushing performance limits up. So, it's not like so what happens today is Salana is limited by the protocol spec, right? The protocol says you can only have 50 million compute units in a block. Well, those are all used. It's like the Ethereum gas limit, right? Same thing, right? It's like bass has to like doesn't matter how fast bass's sequencer is. It was limited by a very low gas limit. Well, they start pushing that number up higher. Suddenly you can do more things per second on bass than you used to be able to do. And all they've done there is basically say, okay, um the validator is allowed to build bigger blocks, right? And so in some ways, I don't know exactly what point enough stake is on double zero that developers are going to be able to start raising

limits because I'm not the one raising those limits. Yeah. Right. I don't have that authority. That's up to the Salana developers. Yeah. But they've already said they want to double the number of CUS by the end of the year. Go from 50 million to 100 million. That'll be the first time that CUS have been increased in Salana ever. Ever. Yeah. The network has I was curious about that. Yeah. The network has had 48 million CUS since it launched. They tested a 4% increase. It went great. Um and so now the goal is get that up to uh a 2x the end of the year. M um these types of things though like they're much harder without double zero and so because if you think about it like let's take like a very you familiar with like turbine or raptor cast on monad so these are basically state propagation

tree systems and so if you if you said like monad's like a really kind of easy example because it's EVM world and you know there's less negative feelings maybe from etherean people about monad than salana Um dude, we are we're open here, man. You know, there's we we could talk about whatever you want to talk about also, right? So, um what happens in all these systems is a block is built and it needs to be propagated out. And so, if you had, um that validator and said, "Hey, your validator has built a block. Send it to 2,000 people." Yeah.

The block is five megabytes. Like, that's actually a lot of data, right? And so you would be going in here's your block, here's your block, here's your block. You're doing out blocks, right? It's actually a very slow process. And so what happens is there's these tree systems on Salana. Um, one validator builds a block. Yep. They send it to a root node. That root node sends it to 2,000 or 200 friends, they send it to 200 friends, they send it to 200 friends, and suddenly you you get state synchronized around the world in five hops. Raptorcast on Monad works very similar the same way. um the the uh system on base they just I'm forgetting the name of it that new system they put out that's based on the Salana shred

architecture right um and actually uses a very similar type of system all of this what it's doing is mimicking something called multiccast so multiccast is a technology that pretty much everyone listening has actually encountered in their day-to-day lives how AirPlay works or Sonos speakers work it is a it is packet replication technology that says I want to send this data from this one source to multiple destinations. The thing about multiccast is you can only do it on a private network. It doesn't work across the public internet.

And so this is why if your phone is sending you know a sound to 12 different speakers in your parents house probably if it's 12 speakers but you know a house party or something. Yeah. Right. Uh you know like or your you know that sort of thing. Your phone is not sending out 12 copies of the data. Your phone would overheat, right? It's sending out one copy of the data. The router is saying, "Where does this need to go?" And it's replicating all the data out. So imagine this applied to blockchain for a second, right? If you and your validator build a block, instead of having to send it out to 200 people or send it out to 2,000 nodes, you just send it out once. Yeah.

You say, "Here's the block." And every single validator can just say great I would like to receive this data and then the network automatically replicates the data and gets it to all the places it needs to go. Now the cool part about this is this replication is not done on the front end. So if my validator let's take a a dumb example my validator is in London and there's 10,000 validators in New York that need to receive that block. I'm not sending 10,000 copies across the Atlantic. I'm sending two copies across the Atlantic and then that data is being replicated as close to the endpoint as possible. And so multiccast is like a massive step forward for performance and it's used today in TRDFI

networks. High frequency trading firms use multiccast for price feeds, right? Think about this from like a like there's a new price for Tesla stock that comes from the ticker at the New York Stock Exchange. Who do I send it to first? Right? that like if I'm if I'm sitting there doling out price feeds, there's a real advantage in who gets it first. So instead, they use multiccast to basically make copies of that data and then they time align it. So it all basically goes out at at the same time. Um, and one of the real powers of that is it is that it creates a principle of fairness, but for when applied to blockchains, it means you don't have to wait for all of this communications over between the validators, right? It also

means that each validator needs less bandwidth because in the current model, you need to be able to send data to all of those nodes and you're actually holding the network back, right? If you don't propagate your data fast enough, the network can't move forward and they're likely to just drop your block and be like, "Look, it looks like this validator just didn't build a block for this slot. We're going to call this a dead slot and just keep going." Yeah, right. That happens all the time where it's like, "Oh, it's your turn to build a block." and your validator just misses its its shot. And so multiccast is just one of many technologies that we can do on0ero that you can't do on the public internet that will just make it much more faster to move that data around the world. And if we can make that faster,

suddenly we've offloaded a ton of work from validators themselves. We've made data propagation faster and validators and RPCs and indexers can all get data at the same time because remember there's no additional overhead for multiccast right this is this whole thing that goes on right now where like if you are a market maker you need to run a highly staked validator in order to get fresh data and if you don't have access to it your data is stale and so you're going to get picked off on marketmaking um you need to run an RPC that's bolted on to the sequencer to make sure that you get the freshest data coming off the sequencer and then you have to submit

transactions back in. Um you know this type of system if we're able to move it over to multiccast becomes something where suddenly like any old trader who wants to join the network and get a multiccast feed can do so and they're no longer at this structural disadvantage relative to like a big trading firm like Galaxy that's trading against them. Now, there's other reasons they might be at a disadvantage, but at least it's not structural from the blockchain's perspective. And so, multiccast actually can do a huge amount to increase the fairness of data access on these systems. Okay, so this makes a ton of sense. I I have two questions about it.

First one is what is the level because what what I'm thinking about here is that if we can if there's a system something like eraser coding where like if you if each each validator doesn't necessarily need to see every block but as long as there's some sort of like random verifiability then we can all kind of agree that this is done properly. But in in this multiccast kind of is sounding more like we we need to trust the copy that is being sent out to everybody but that copy has already been agreed upon by the validators or has reached consensus. No no so so what the val what would basically be broadcast out and and this is the beauty of public private key cryptography is you don't

have to trust the network layer for anything. Okay. Except deliverability. Right. Right. It just needs to get there. Yes. And so what happens is I have signed a block that says, "Hi, I've built this block. Would you like to vote on it?" Right? And and that goes out via multiccast and you say, "Here's my validator private key. I'm going to sign this as my vote and send the vote back." You know what? The votes can now also be multiccast. Okay? So I can So even if let's just say like the validator that built the block is 250 milliseconds away in Singapore, right? But I see well I see all the multiccast data flying around. Looks like actually a quorum has already voted on this block. Now you may not know it yet because your node is still waiting for all the votes to come

back to you. But I can say great this block is actually good. I'm going to start taking actions based on that data. And so we're not adding any more uh trust assumptions in the security model than already exist which are okay not which are all basically that like someone hasn't stolen the private key right multiccast doesn't change what data is coming to you or when it's just a different delivery method method right okay and that is at more advantageous than something like eraser coding or these kind of like random verifiable ways of of making sure that the data has been delivered for for what for. So so eraser coding is used in areas where you

may have a lot of packet loss. Okay. And so um eraser coding is really useful for like you know if I if if I'm missing one little bit of data, right? Maybe 1% of the packets get dropped. I can now still reconstruct what that block looks like. Eraser coding can work on multiccast too. They're not exclusive uh to one another. All multiccast systems have some sort of a repair function where you can basically request another copy of it. So the way this is done in trad networks is you have an A feed and a B feed. And so those are two multiccast basically sender groups. And so the data is sent twice and you can know one of two things. Well, if if some of the packets were dropped from one, you know the other is good. If the if the prices differ for any reason, right, it's

usually used in pricing feeds. You know, huh, something's off with some of this bit of data. Okay. But again, all of this stuff can just be signature verified, right? Because this is the beauty of this industry already works in the assumption you can't trust anything. Yes. And so we're not actually like introducing any more trust assumptions into that system. There's also repairability too where you could say, "Hey, I missed this thing." And then, you know, that that just falls back over normal TCP IP repair at that point. Um, you know, Salana's repair function in the extreme actually kind of looks like Bit Torrent. Um, you know, and you can request these things over gossip.

Repairing missed data is is actually like a huge problem on the internet today for any protocol, which is why there's this big eraser coding that's being used both in Raptorcast and in Turbine. The problem with eraser coding is it's not free. Yeah, you now have to add a bunch more data. It's not twice the data, but it is more data. And so the nice thing about multiccast is it's just a massive reduction in how much data an individual validator needs to have. Awesome. I hope you guys absolutely crush this because if if you guys do it sounds like and it seems like we are going to have a massive breakthrough for some of these fast getting like super fast super cheap L1s across the board not just Salana and then potentially up apply it to other other you know verticals. It also

applies just fine to L2s. Yeah. Right. There's no there's nothing about what we're talking about that is L1 versus L2 specific. No. Right. I I think that if you're running a centralized sequencer, state propagation is still a really big issue. You want to make sure that fairness to all the RPCs. Yeah. Uh you know, there's lots of applications for across this over the board, especially when we start to get networks that are built with multiple sequencers. It's just time for crypto to have its private network moment. Like it's it's it's infrastructure. It's it's it's pre-built customdesigned hardware moment. Yeah.

It's a little bit like imagining if like when Ethereum was built, they were like, "Well, so we're going to build Ethereum, but we think it's really important that it settles on Fedwire at the end of the day." Exactly. Right. It's like, "No, no, no. We got to like we got to restart this base layer." Right. Yeah. Well, Austin, um, timeline, roadmap, what's on your mind for the coming months into summer and into Q3 for0ero and then we can close it out. Yeah. So, we have a test net up and running now and so there's about 30 Salana validators on it. We just launched uh we we actually didn't even talk about all the FPGA filtration work and the bunch of the defense layers duplication and these things. Yes. Yeah. All that stuff that's built into it, but there's this new mode which is IBRL mode that we launched on

testNet which basically allows you to use your public IP address on the network. Um that may feel like a small thing, but it means there's zero modification of a validator client. Um you could run Celestia on this thing today. You could run Monet on this thing today. um you know and so there's no modification of client required to sort of get started. Um you don't get the filtration systems unless you actually do a little bit of additional work but again it's all pretty simple for folks to do. We are targeting a September mainet launch. Awesome. So that's exciting. It's uh it's a lot of engineering from here to there. We've got some incredible folks working on and that really is the superpower we have is like the engineers working on this

project are like people who've spent the last you know decade as network engineers for high frequency trading firms and other sorts of high performance private networks. Well, I hope you guys absolutely crush it. Thanks for your time today. It was absolute pleasure. Thank you.