On The Brink with Castle Island - Austin Federa (DoubleZero) on Building a Better Internet (EP.664)

Episode Date: September 8, 2025

Wyatt sits down with Austin Federa, the founder of DoubleZero to discuss the hardware systems that underlie blockchains and the internet. In this episode: Building a newer, better fiber network How and why much of internet hardware infrastructure is antiquated Why software innovation has taken precedent over hardware innovation Vertical integration and who has access to the best hardware systems Why hardware innovation requires startups and new entrants Why blockchains demand faster, reliable bandwidth Why a token framework offers a better incentive mechanism

Today, I sat down with Austin Federa, co-founder of Double Zero. Double Zero is a new network of physical infrastructure to make blockchains and ultimately other internet technologies run faster. Austin started Double Zero last year after spending four years in senior roles at the Salana Foundation, solving infrastructure problems of how to make Solana an exceptionally fast and cheap money movement layer. Double Zero was born out of that firsthand experience. Austin brings an incredibly informed view on the intersection of blockchains, internet infrastructure,

and market coordination. It was a pleasure to have this conversation. I hope you enjoy listening. Matt Walsh and Nick Carter are partners at Castle Island Ventures. All of these expressed by them or the guests on this podcast are solely their opinions and do not reflect the opinions of Castle Island Ventures. Guests and host may maintain positions in the assets discussed in this podcast. You should not treat any opinion expressed by anyone on this podcast as a specific inducement to make a particular investment or follow a particular strategy but only as an expression of their personal opinion. This podcast is for informational purposes only. Brought down by bad mortgage investments, Lehman, which has 25,000 employees, will be liquidated.

The federal government loans American International Group, AIG, $85 billion. This is a different kind of market, and the Fed is asleep. The federal government is stepping it to stabilize Fannie Mae and Freddie Mac, the two mortgage giants that have been threatened by the housing crisis. The Bank of England has pumped 75 billion pounds more into Britain's ailing economy with a new round of quantitative easing. You print a couple trillion dollars and all of a sudden, people start to worry. So out of this worry, we have something called the Bitcoin.

Austin, thanks for sitting down with us today. I have heard a lot about Double Zero and excited to dive in as well as about some of your experience. Would you mind just kicking off with what is Double Zero for people who might be unfamiliar, vaguely familiar, or trying to familiarize themselves? So Double Zero is a crypto infrastructure project, but it's actual infrastructure, not just software systems. Because usually when crypto people talk about infrastructure, they're really just talking about software.

What we are building and have built out is a high-performance global fiber network, optimized for the types of problems that high-performance distributed systems have, as opposed to the types of problems that you see on the public internet today, which are more about reach than they are about performance. So we have 10 independent contributors

that are all contributing fiber to the double zero network, and there's a blockchain-based protocol that sits on top of that, that helps orchestrate link coordination, reward payouts, and the distributed management of that system. There's an irony because I think in AI and in other facets of technology data centers have really started to take popular mind share, and people are focused on the hardware infrastructure,

but it feels very much like something people miss on the crypto side. If you think about it, pretty much every large software company is in hardware now. And they're not in hardware because they want to be. Facebook did not start building its own servers and then building its own data centers and then building its own high-speed sub-C fiber because they wanted to. They got into it because the existing market

wouldn't service their needs for scale. And they're not running sub-C cables through the Philippines out of the goodness of their harder because it's a plus EV business line. They're doing it so they can sell ads. They want Instagram reels to load more quickly so you keep scrolling so you see more ads And at a certain point of scale, the way to solve a problem of their being slow internet in the Philippines

is just to go build it yourself. And time and time again, we see software companies making this transition from a pure software play into hardware that supports their software, which eventually becomes a competitive advantage. And for years, crypto just wasn't there because a lot of the major use cases were financial and those use cases we know and love and are great, but at the end of the day, we've got gotten to a place where if crypto is really going to start competing with traditional finance systems and other types of high performance distributed systems, it needs to be using the same tools and technologies. We don't think of Google as a distributed systems company, but it is.

Google is run in dozens of data centers all over the world. Now, it's all run by Google and it's all linked by Google's private fiber. But you wouldn't get the experience of YouTube or any other Google product if they didn't have this high-performance fiber network under the hood. But in crypto, we can't just say, oh, great, we're going to use Amazon's network or Google's network or someone like that because you wouldn't have to trust that network. If Double Zero were built as a centralized company with a global fiber network that we ran and managed, you shouldn't trust it. The reason you can trust Double Zero's fiber network is it's got 10 independent contributors, it's got censorship resistance.

It's got all of the things that you would expect from a blockchain-based project. And also, the Internet's a lot more centralized than people realize, too. I completely agree. And that alternative would completely defeat the purpose of blockchains in the first place. You might as well run databases at that point. Yeah, it depends what you're using blockchain for. There's this knock that I feel like some people have sometimes about, oh, blockchain systems are just databases.

And you're like, well, first off, yeah, everything is just databases. The New York Stock Exchange is just a database at the end of the day. The real unlock with blockchain is it's the most perfect system of accounting we've ever created. It doesn't require an external auditor. It doesn't require trusted execution environments. It doesn't require external validation of system integrity. It is internally consistent in how it validates data. And it has all these other characteristics of being able to be run anywhere in the world and being

permissionless. But part of the reason we are seeing more corporations, adopting blockchain technology, even if they're doing it in a fully centralized way, is because you can cut out a lot of the Web 2 detritus that we've had to stick on to these internet protocols to make them secure. We all know SOC2 is a joke and doesn't actually mean anything, but it's like this arbitrary stamp of approval that the business people stuck on a bunch of software because we fundamentally can't trust software because the way that Web 1 and Web 2

software was built, was that trust was an external property brought into it. In blockchain, trust is an intrinsic property to all these system designs. And that means that Visa's now got a program exploring building an L1. Stripe has a whole program building an L1. Circle obviously is a whole program building in L1. And it's not because they want to launch a token and raise more money. They're all very valuable companies in their own right. It's because they can actually reduce operational overheads by adopting blockchain technology. Completely agree. And we have those on-chain systems, which will continue to merge under the guise of those companies, they're going to be looking for that level performance optimization that we were talking about

where you had the cloud computing rush and you just had this major improvement to underlying performance in the traditional tech world. So in that main is the right way to think about double zero that you have this distributed hardware network sitting underneath and then an on-chain coordination layer on top. Yeah, that's the right way to think about it. And depending on level of abstraction you go, there's two or three different layers. But one of the things that we think is really important here is this is not a system just built for blockchain. So it's always been possible to build a multi-party high-performance fiber network. These things actually exist. Go West is an example of this. It's a consortium run by a number of HFT firms. The thing there, though, is access

to that fiber and that microwave tower system is quite controlled and quite expensive to get access too. So you can think about the cloud revolution was not a revolution in technology. It was a revolution in business model. At the end of the day, a hypervisor with a 20% performance overhead is a pretty uncompelling design for system architecture. People forget that the cloud is actually quite slow. It's very compute inefficient to run stuff in the cloud compared to like a bare metal service system. The advantage of the cloud is the management layer on top of it and the ability to go to AWS and say, I need 500 servers and have them deployed within an hour, that's something you can't get through another type of system. So in double zero, we don't have a performance overhead like the

cloud has. We are giving folks direct access to this fiber. But if you wanted to go and build a global fiber network yourself, it would take you two years minimum. And you need a staff of at least five people who've done this before to actually be able to go out and do the site sourcing and all the contracting and agreements. It's a very complicated process. But with something like Double Zero, you can just connect in and say, hey, I've got a server running in CC2 in Tokyo, and I need to connect that into NY4 because I need to send a bunch of data that's either low latency or I need to send a large volume of data that the public internet can't really support. And you can do that basically on demand. And our first application of this technology stack is

blockchain, because blockchain is the place where you have two important components. You have demand, and you have capital. So blockchains will really benefit from high performance connectivity, and blockchain operators receive pretty good amounts of revenue for operating their systems. So if they can squeak 10 basis points of performance out of the thing, they're going to do it. So it has that same characteristic of a high frequency trading firm, but it has a distributed systems problem that a high frequency trading firm does not. The incremental benefit to blockchain performance is very valuable. I want to double click on a point you made because I think it's really important,

and under-discussed. Can you touch on the market dynamics for why high-performance fiber networks have not permeated to being used by everyone, why everyone doesn't have access to them? How has that existed to date? And how does that inform where internet systems are high-performance, less performance? How is that shaped that? So we're very used to talking about network-effect systems when we think about anything in blockchain or web two. High-performance private networking is. an anti-network effect business. If I am jump trading and I have a high performance link from Chicago to New York, I don't want Citadel or DRW on that link. I actually want to keep it

just for myself. It's a competitive advantage. Exactly. And you don't see that in data center construction for the most part. You don't see that in server construction. It's not like Navidia is getting proprietary chips and just using them for themselves. They're selling it to other people. Their business model is built on distribution, whereas a lot of the traditional financial space, the business model is built on proprietary access or proprietary data or proprietary performance. So even if you're a company like eBay, you are running a private fiber network, but you have to be the size of a company like eBay to actually have the capital, to have the overhead to hire all these people. It is exactly how when a company makes a transition from the cloud back to a bare metal data center, they need to

hire a ton of people because that is not a turnkey software managed service. The real world is just a lot messier than software. Cables get cut by ship anchors or they go down for maintenance or you have a switch that burns out in a data center and you have to do a failover. Someone has to at the end of the day go to that location and move a thing and touch a thing and replace a product. And all of these reasons basically create massive overhead. The other piece too is you're dealing with 100 different countries and a number of different languages and the different legal systems. And the problems of international payment transfers are nothing compared to international data transfers. The amount of different regulations you have to go through, there's stuff that's totally not allowed in the EU, that if you do it in the US,

is totally fine. There's other things that the US wants you to do, that the EU will be mad if you do. It's a very messy, complicated system, in part because of government overhead, and in part because it is just something that is in the physical world and things in the physical world are much harder to automate or scale than stuff in the software world. So typically what happens now is you have these carriers and most of the carriers are owned by private equity firms. So what they really want to do is throw off 8 to 9% returns per year. So they're not investing ahead of time in a lot of infrastructure. Internet cables are basically run at 90 to 95% capacity 24-7. You can think of it as just in time bandwidth. Really truly, the way that these companies think of the internet is as a

pipeline and supply chain problem. It's similar to how energy grids are operated. There's no incentive to have a ton of burst capacity because all that burst capacity is is a cost when you're in a business that's on cost optimization. So one of the things that we're able to do with double zero is just change the entire revenue model. And tokens are really required for the alignment necessary in this type of system, but also we're just massively expanding the tam of what access can be to these types of systems. So our hope is that we have, and we've already seen this on a small level with our contributors, but we expect over the long term, people are going to be installing new fiber cables under the ocean just to participate in double zero, because right now we have this divorce between

the physical layer and the value add monetary layer on top of it. If I'm ordering a semi-trailer to move stuff from New York to San Francisco, it doesn't matter if it's full of wine or if it's full of gravel. The cost is the same per pound, even though what's in it is a very different value. So because in blockchain, we have to reach hundreds of points around the world,

no company can actually build out enough fiber capacity to do that. There's not many financial centers in the world. A trading firm can wire them all up. They can't wire up every data center where blockchain stuff is run, let alone every data center where CDNs or AI technology is run. So because of that, they need to be involved in a network like this.

And the minute you have that type of system, you're able to actually have a revenue model that supports portions of value transferred over the systems, not just dumb pipes. Yeah, that all makes a lot of sense. I have several points I want to ask you about in response. The first one is probably if you're in and around the business of more performant fiber optic networks, But even in performant hardware in general that sits underneath these systems, you're tussling with the incentives behind either owning this vertical stack where in the jump example, you profit through some business model that you built on top of your max performant layer because you can outperform everyone or you're productizing your performance.

You're choosing one of those two, which is like keep for yourself versus sell as a product because others will want it. Yep. And in that view, this seems conducive in. certain market structures to vertical monopolies where people would want to own the most performant part at every layer of the stack. You guys have a great layering on your website. I encourage people to look at where he talks about the seven layers of the blockchain stack, for example. I could see a market structure to that point where someone would want to own every

part, especially if they feel like they have a competitive advantage, say on the hardware layer. Do you see that as likely to happen, let's say in the blockchain industry, for example? So the natural end state of many of these systems you can think through as vertically integration. That is the natural end state of basically any economic system is complete vertical integration with the absence of human fealty. There is no reason a company in a dominant position should ever lose market share. We see Coinbase doing it to be realistic. Totally. The Coinbase thing is an interesting case study. So what we see here is that the scale someone needs to actually vertically integrate all of those systems is quite high. And the more vertically integrated a system you get, the lower

performance each layer in that vertical stack tends to be. This is not a law, but this is a generally observed concept, is that the more vertically integrated a company is, the less attention they pay to each individual component. And this is really true in anything that is. is in the financial space. And because at the end of the day, the relative performance gains that at each piece in the stack that you can get don't impact your overall bottom line

as much they do for an independent actor. I think this is one of the things that's not well understood with folks who are like, break up the big monopolies. There is a component of something like a Google where you can argue if it's monopolistic or not. What's very hard to argue is Google has the best offerings.

Almost across the board, everything Google offers, there's a better alternative to it. This is true of JPMorgan. This is true of Coinbase. This is true of AWS. This is true of pretty much any fully vertically integrated system. This is true of Tesla.

All of these components, they may have a one-stop shop for everything, but what they're able to do at each layer of that stack is minimized. Now, the reason that this can be a problem in the traditional space is because you have arbitrary barriers on interoperability. And those arbitrary barriers and interoperability are things like, iMessage just doesn't work on an Android phone or the classic thing there. In crypto, we don't have that. And Coinbase is trying very hard to build a fully vertically integrated product

where now there's wallets on base that you cannot move back to Ethereum. But fundamentally, the assets are still fungible and movable. So we see this all the time now where you'll have data centers that have a large amount of Solana stake in them. And then a competitor will spin up and they'll be a little bit cheaper or they'll be able to have about 10 basis points of additional MEV revenue. and you can just watch the validators migrate to the new data center. These are very portable industry.

This is part of why the validator as a service business or the RPC business is a pretty bad business statistically. As someone who used to work for Bison Trails, these businesses are very hard because the portability of your customers is extremely high. So I think we are a ways away from having to worry about that type of vertical integration in crypto because it's mostly open source software and it's mostly highly portable systems that you don't need to rely on some external trust validation for. I may not be comfortable moving my money out of JP Morgan into a startup bank,

not because the JP Morgan has a better product, but because the name of JP Morgan carries a certain amount of gravitas with it. Realistically, that's only a problem when you're dealing with systems that require trust. And if the system doesn't require trust, like in the case of blockchain, you don't necessarily have that portability limiting factor. I like your assertion that vertically integrated businesses won't be, frankly, at the cutting edge at each layer. I'm curious for your view on how industry positioning and regulatory capture plays into that, though, because I think of something like telecommunications in the 2000s, which somewhat famously was very difficult to break into for companies which were trying to innovate.

And there was a sharp reduction in the amount of venture funding that went into it, I think just because people started to give up a little bit. So granted, the vertically integrated larger players won't be the most innovative. Do you worry that it can become too difficult to disintermediate them even with a better offering? So I think the telecom regulatory capture wasn't actually on the regulatory side. If you look at what happened there, the access barriers were basically all at the state and local level. They weren't at the federal level, which is not what you typically expect. Banking regulation is captured at the federal level, it's not captured at the state level. The problem that we ran into starting in the 2000s is we just got allergic to building stuff. The same problems that mean you can't build a pipeline,

you can't build transmission lines, is the reason you couldn't build fiber lines, is that these systems tend to all co-locate with one another. You're putting in transmission lines, great, we can put some fiber beneath it. Or you're putting in a pipeline, great, we can run a bunch of fiber alongside it. And what we saw is just nimbism taken over and not letting you build anything. And, I think that's a different type of problem than the typical thing we think of as regulatory capture, like in the financial space where it's just the banking lobby has made it such that it's almost impossible to get a new banking charter. And that's a different type of regulatory capture than I think we see in the telecom space. What about the issue you have where to be crude, disintermediating

AWS would be inherently difficult because just the level of capture that they have with maybe strong army or compelling partners to continue to work with them? I actually think the mode of AWS is the tragedy of computer science over the last 20 years, which is that people don't know how anything works anymore. And this is not AWS. I actually don't think there's a lot of strong arming from the cloud providers. There's some sure. But the cloud provider is actually a fairly competitive market nowadays.

It is the fact that people just don't learn how anything works anymore. If you talk to a modern Web 2 engineer versus a high frequency trading engineer, they're almost entirely different species at this point. The work and concerns of someone who works for a trading firm, they would never be caught dead virtualizing anything because the performance overhead you get is they don't even use kernel networking. The overhead that you have to do because of all these layers of software abstractions on top of them, you listen to Kevin Bowers, who's one of the chief research scientists over a jump,

talk about how a modern processor works and the way modern compilers work. And a solid 20% of their job is just, stripping away all of the crap that people have built to make programming easy so they can actually talk to the hardware. And you think about this as they have this whole thing about tile architecture. And tile architecture is not a new concept. It's actually an old concept. But if you look at a modern processor, it has 24, 32 cores, 64 cores.

And if you just try and run stuff on it, the management software at a very low level just randomly assigns different tasks to different CPUs. And when you're trying to optimize for performance, you have to think about the physical signal pathways within the chip and how data flows through a chip and how the infiniband on the inside of a chip actually manages data flow. And so if you have one process on this corner of the chip and one process on this corner of the chip, it actually takes a meaningful amount of time for the data to get from one side of the chip to the other

because it has to get signal boosted as it moves through the silicon layers. The speed of light is actually a factor at this point too. So you get into a place where we're bleeding out so much performance in modern systems because people don't know how they work and they're not thinking at that low level anymore. Well, they haven't had to build it. It's been there for them. Yeah. The moat that the cloud providers have is that most modern software engineers are lazy. And it's the classic thing of like when Apple pulled everyone over from writing in C to writing in Swift for programs. Yes, Apple can do magic on the compiler level to make things quite performant, but there's also a whole skill set that's lost for people that want to build

even more performant programs. And you see this even on a network like Solana, where folks were writing a bunch of stuff in Anchor, which is this abstraction language on top of it. Rust was the default high performance for a while. Now, C is the default high performance for writing Solana programs. And there's actually some stuff that's been written in assembly. And you have to be willing to get that low level on performance optimization. And part of this is the fault of VCs, where There's a whole incentive structure in VC, which is don't worry about costs, just worry about scale. And if that is the structure you're building under, who cares if your server bill goes from

$800,000 to $8 million because you've grown 10x and growth is all that matters. The problem, though, is you see all of these companies, they hit IPO, and then suddenly they have to do a bunch of cost downing on their infrastructure, but they have no DNA and they have no expertise to do it internally. and it's a huge transformation they have to go through. And we saw my friend worked for Etsy back in, I think this was probably 2014-15. And they had done a migration to the cloud.

They had this incredible architecture set up in AWS. But Etsy's got 2 to 3% gross margin business. They were taking transaction fees on a bunch of payment processing, and some very smart person there looked and said, our EWS bill is not going down. This thing is increasing at 7 to 10% per year, as Amazon raises prices, we need to move out of the cloud. So part of his job there was orchestrating a giant migration from the cloud back to bare metal servers that were run in two different

data centers, one in New Jersey, one in, I think somewhere in Colorado, and they massively reduced their cost over the short term. But the delta there, if you look at the average price of something on AWS, it is outpacing inflation, which is not what you would expect from the Moore's Law type perspective, but it's because people are willing to pay for it. And the companies that are willing to put the work into optimizing and moving back, that is a far more effective form of cost reduction than swapping out all your customer service people with AI agents. It's the boring stuff that people get wrong. I think in the last 10 years, too, there was this software infatuation and the successful venture stories or startup stories were things like Etsy, to your point,

Shopify, just break out software applications to the most part that scaled revenue really quickly. And from an objective perspective, I think this is one of the things people actually take to without being articulated about Elon Musk is that he's one of the people that has actually taken this attention to things like hardware and systems, which especially years ago, what was really not paid much attention to by startups as an example. Yeah, I mean, we look at what were all of the sexy space companies doing, not even before SpaceX, but along that same lines, a lot of them were focused. on software systems or things that we're going to,

we're going to hand away of how we get them into orbit. Once they're in orbit, here's the skill set that this thing offers. And it was a very, very different perspective to say, look, if we're going to actually make space a real thing, we need to rethink the entire way that we built launch systems. On the business model, rethink it, on the engineering side, rethink it, take risks that the industry was historically not willing to take because the number one customer for the industry historically was the U.S. government.

apart from the DoD, it's very risk-averse. Even the DoD is pretty risk-averse. NASA would tell you they would rather spend 10 times more for something than have a 10% chance that it blows up on the launch pad. Because the PR damage will do more damage to their budget than the cost-overrun. So that whole model shifting, I really think it's a pretty good analogy for what we're doing with Double Zero

because we are shifting the entire business and cost rationale around private fiber in a place that's really only been accessible to HFT firms until now. Makes sense. I want to draw a through line. I was going to come back to an element of double zero. I wanted to pick your brain on two, which I think a through line here is these infrastructure systems, particularly the lower down you go in the stack, which gets the hardware,

they tend to become stagnant, I think, just because of how much has been built on top and to your point that we don't necessarily have a lot of people who are thinking about innovating on this front. And it's difficult. how do you look at from your perspective, the double zero perspective, creating a low-level infrastructure that can innovate as opposed to become stagnant, let's say even if you take on prominent market position in five years? So one of the things that we don't do is we don't offer any servers for people.

It is really truly just fiber and some compute resources implemented through FPGA. But this is a thing that Solana's always been very good at. And I was a very strong advocate of this way of thinking when I was at the Salana Foundation for the last four years, was that we should build things as low level as possible and with as few opinions as possible. So this is really how we're approaching double zero. We are building a series of tools that people can use to do things that I have no idea what they're going to do with. So we can do things you can't do on the public internet. Maybe it's worth getting some of this offering stack, but at the very basic level, we have fatter pipes and we have pipes that have

lower latency than the public internet. So you'll get 100 gigabits of connectivity as opposed to one, and your latency will be 30 to 70% less than you'd get over the public internet. But we can also do things like run multicast over this network, which is a technology of basically hardware acceleration of packet replication. So if you need to do something like propagate blockchain state from one machine to 2,000 machines, today you have to send out 2,000 copies of that data manually. With multicast, you send out two copies of it, and the network replicates the data in hardware acceleration on its own. And so it really is a send-it-forget-it type of model, which really changes how performance systems can operate. And it also reduces the penalty for not being the

first one that the system sends out data to. We can also do things like insert precision timing into the network. So you can do things like time-aligned packets. You can know exactly what time the packet hit the edge of the network. And so that's really useful in all sorts of financialization and trading systems as well. You can now time align when stuff comes into the network. You can make sure that every one of your validators has extremely precise clock systems. And you know, okay, this event happened here at this node, we can now reconstruct the order that something should have happened based on wall clock time, not based on something else. So there's a lot of these types of functions and services that just are not available on the public internet. I don't

know what everyone's going to build with them. Smarter people than me, you're going to come up with really awesome stuff you can build using this toolkit. That's the same thing we saw at Solana for years. I don't think anyone predicted that meme coins would be a huge thing on Solana or that we'd have tokenized stocks. The point is that they could be. You created that sandbox to your point.

I think it's a great point and in the interest of better technology to be unopinionated to really be non-verdical in a way to improve on each layer of this stack. Do you find attention because I imagine a lot of software engineers, to your point, more of a spoon-fed infrastructure solution, where they have something that probably is opinionated and that checks 10 boxes at once instead of one or two? Is that difficult to wrestle with? And relatedly, who are you selling into with that in mind? So this is one of the things about crypto that's just such a joy. We don't have that problem in crypto. The engineers who are building stuff on Solana are very happy to be getting

into the very low-level systems, and they like how much of this they're exposing. I think a real analogy here is the rise of Navidia. And at the time that Navidia launched Kuda, which is their programming framework for accessing vector processors on GPUs, it was not the cool way to do it. All of the work on compilers

and all the work on this system was really about how to make it easier to use. And Kuda's innovation was, there's actually a lot of really smart programmers out there that want to be able to direct, address things to hardware. And what they allowed people to do is write in languages they were already fairly familiar in, but add a few flags that allow you to actually address the hardware directly. And Kevin Bauer's famous example is dense matrix multiplication. It's actually very hard

to write that in a non-Kuda system. And it's a hello world example is basically how do you dense matrix multiplication on a GPU. So when we look around and we see, oh, well, everyone's just building stuff in software abstraction layers. No one really wants to address the hardware. Like you're saying, people don't want to get this low level into the stuff. What we basically find are the systems that are going to need double zero and would benefit from its performance are already thinking that way. So maybe we fast forward three or four years.

And yeah, we have to do some type of more accessible abstraction layer or we have to have a bunch of coursework for how to actually optimize infrastructure for double zero. But the place we're at today, those people find us. We don't have to go find them because they've been looking for something like this for a long time. It's almost an inadvertent benefit then. Let's take meme coin trading an example because it forces people to really prioritize high speed because how lucrative it can be, whereas there are plenty of legacy internet apps that could frankly run milliseconds slower and I don't think people would care that much.

Yeah, and that's fine. That is the big differentiator here. So a great example of this is there's a new prop AMM on Solana, Humidify, and And what it basically did is they wrote the whole AMM program basically in assembly. It's using this tiny framework. And because it's so small relative to other programs, and it requires so little compute units, which are basically sort of a unit of transaction processing on the Solana network, it pays about six times the compute-to-feet ratio that any other Defi program on Solana pays.

And it still pays incredibly little to send each transaction. But because the ratio is so high, validators are like, oh, I'm absolutely going to process these transactions first because they're really efficient and they get paid comparatively a ton. What does that mean? Well, suddenly, this prop trading AMM is the fastest AMM on Solana and they can quote half a basis point or less. We executed a trade yesterday on chain where we got 10 basis points of basically fee and slippage in this process, which is better than you can get from any OTC desk or any centralized exchange. And this was fairly significant size, too. This wasn't a $2,000 transaction or something like

that. And that type of performance, we're going to see this widening gap between traditional systems and blockchain systems because you can get this insane performance and you don't need all the compliance and security overhead. What other use cases, adjacencies, technologies, do you think double zero is most primed to improve upon to work with in the interim? I think there's a lot of the AI space that cares a lot about connectivity. We're not going to be servicing open AI or someone like that that's got billions of dollars to build their own data centers and their own high frequency fiber cables themselves. But there's a whole bunch of companies that are in between stage where they've got a technology advantage, but they don't have the expertise or

the timing to go out and build data centers and build fiber connectivity between them. News Research is a great example of this. They're usually a second customer on a lot of these big AI workloads, but they waste a bunch of time syncing model data back and forth. So if we can change that sync time to five minutes as opposed to 20 minutes, that's 15 minutes more processing that they're going to get on their models every time there is another customer in that system. I think the other area we see are CDN systems. So there's this group called Shelby, which is a storage network that's being built by a combination of APDOS and Jump, and that's going to run on double zero. And because it needs multicasting of data and it needs fast access to data and these sorts of things to be

able to do CDNs in markets that are not well served by CDNs at the moment. So I think there's a lot of these different types of applications that we'll see. Trading is absolutely one of them as well. The highest profit, most turnkey solution here is just if you're trading on a distributed system, you need to get data as quickly as you can to as many places in the world as you can. On this note, and to touch more broadly on the business, you have the fiber network that you work with, the 10 providers you mentioned, and then you guys are a coordination layer, ultimately providing a service to customers. How do you look at how value should flow between you and your 10? I don't know if we'd call them clients or providers that sit underneath. What we basically run on is a different reward model than most blockchains do that more closely resembles a proof of work.

than a proof of state system. So we have a whole blog post on the website about this is called proof of utility. But we basically use Shappley values to determine what is the incremental value each network link provides to the overall network graph. And the benchmark for this is the public internet. So if your link that you're adding does not beat the public internet, it gets no rewards. If it beats the public internet but actually doesn't connect anywhere that paying customers, in this case, blockchain validators or RPCs are operating in, then it also doesn't get any rewards. But if we do the math on this and we say like, okay, if your network link didn't exist, the

network would be 20% worse, then you receive 20% of the rewards for that given epoch of time. So we have this system where we are incentivizing competition between fiber providers to always be trying to basically steal each other's market share. And the best way to steal someone's market share is to provide a more performant infrastructure link. At the same time, validators and RPCs are the ones that are actually paying to use this service, and we're working on a percentage of revenue model, basically.

So if we're taking about 5% of block rewards and priority fees in MEP, in exchange for this higher performance system. Now, we're also not going to take that 5% unless we're making them at least 5% more profit. And that's our trade with folks is, yeah, you're going to give us some percentage of your earnings, which is going to be more than maybe you're used to paying. But in net, you're going to make more because you're able to operate on this high performance infrastructure.

And as part of the value proposition, we've talked a lot about performance, but to my understanding, there's also a privacy aspect at play here. Yeah. So privacy is, I would say more of a theoretical problem than a practical problem on the internet today. So if you are sending a trade on a blockchain right now, that data is moving over. the public internet. And it might be passing through 15 or 20 different providers as it gets to its destination. Deep packet inspection is very good at the moment. It is entirely possible that someone is holding back some of your packets arbitrarily so they can try and front run you or something along those

lines. Now, there's very little evidence of this actually happening today, but it is a real theoretical concern. This is one of the reasons that traditional financial firms don't use the public internet to trade on. It's slow. Yes, there's that. But you're also leaking data, about what you're doing. So part of the thing with double zero is we can actually provide cryptographic assurances to folks of data that moves over the network, and we can catch providers that are holding back data.

So if we see, hey, it's supposed to take 64.3 milliseconds to get from here to here, sometimes it's taking this one 67 milliseconds. What are they doing? And that sets off an alert that we can go and investigate and see if someone is selectively holding back traffic or something along those lines. We also will be able to in the future offer traffic encryption as well.

We can do this in such a way there's very, very little overhead involved in that encryption. So you're usually in this trade-off game now of, oh, I can encrypt my data, but it slows it down. And if you're on Zoom, who cares? No one's going to care about that slowing down of data. If you were trying to trade, it may become a meaningful slowdown. Understood. I did want to ask you, hitting on some of the topics we discussed before, do you and Double Zero ever have ambitions to become your own settlement layer for assets?

I don't think so. This is one of those classic things of, I don't think that's where our expertise lies. I think that double zero, we're in a much better position to simply not compete with blockchains to instead onboard all of them to double zero and have 10 networks running at a million transactions per second

and have a percent-based model that just takes a small cut of all of those in exchange for providing a ton of value-ad services. There's a lot of layer ones out there. And I think you have to have a really credible architecture or thesis decision about why yours would be better. And if all our answer is, it would be better because it's running on high performance, private fiber with hardware acceleration and technologies you can't get on the public internet. Going back to the very first thing we were talking about about non-network effect businesses in Tradfai, why would I not want the network effect of Solana and Suey and Aptos and Megaeth and Monad and Stripes L1 and Arc and all of the network effect of Solana and Sui and Aptos and Megaeth and Arch and all of the, these things operating on top of double zero, why would I ever limit our market to just people

who are willing to adopt the double zero chain? It's a very clear case for me of grow the pie as opposed to trying to take a bigger slice of it. Usually when you try and take a bigger slice of the pie, as opposed to growing the pie, you actually end up with a smaller piece of pie. You're pulling from a much smaller base. The final question I wanted to ask you, what do you look at the end state of this segment of the market being, do you think that we have a host of call it luxury fiber networks that the applications that really need them or those that are willing to pay can pay forward. Do you think we have this slow public internet, which still powers most of what we use? Where do you think we end up? This may not be like a closing topic. We might go a little

long here, but we have to think back to 2020 because 2020 informs a huge amount of this. And what we saw in 2020 was a very strong effort to de-platform certain applications. and certain ideas from the Web 2 world. And irrespective of what you think about those efforts, morally, it turned out that when a social network got kicked off of AWS and got kicked off of Cloudflare, it was extremely difficult for them to run infrastructure, let alone if they got kicked off of DNS,

which is also a centralized thing run basically out of the United States by one company. What we basically saw was we already have this world you're talking about of a two-tier infrastructure system. the folks who can pay for Cloudflare and the folks who can pay for AWS and the folks who can pay for Akamai, their infrastructure is more performant and it doesn't actually run on the public internet. It runs on this entire private system that's completely opaque that runs alongside the internet that you have to pay to access and also they can kick you off at any moment for any reason because net neutrality also, first off, doesn't apply to the US anymore, but even if it did,

does not apply to private networks. So you get into this place very quickly. where we have an internet where pretty much anyone in a few positions of power can monkey with how the whole system works, let alone the U.S. government can just say, hey, and we saw the emails, Facebook, take down these posts. Hey, YouTube, remove this video. And that is a difference in sending a letter to Facebook saying, delete this content, and sending a letter to YouTube saying delete this content, and sending a letter to Verizon. saying lock this content is the only difference there is the political will of the United States versus the political will of the Chinese Communist Party. And I don't want to hang my hat on

that differentiation in an era where free speech is kind of a little bit less than it used to be. You compare the 90s to the 2020s in terms of free speech, and we have less of it now than we used to in a lot of ways. So this is a long way of saying when we think about these types of systems over the long term, I want credibly neutral alternative to the existing communication systems, and I want basically universal systems of prioritization. So today, it is very hard to say, if I'm epic games and I want to prioritize all the Fortnite traffic, so kids can play more games and buy more skins, my bottom line revenue can go up, I basically have to go around and bribe every ISP to do that. You'd think there'd be a protocol on the internet that just says, I want to pay more

to make my data better. You can't do that today. So double zero. really over a long-term time horizon, our goal is to become the default prioritization layer for anyone anywhere that needs to send data in a timely fashion. I think the other piece we're looking at over the long term is replacing TCPIP with public key, private key encryption and verification. Right now, the only way that I know a server is correct is I'm verifying a SSL certificate, which is issued by a third-party authority. I actually don't care about routing to an IP address on a blockchain. I don't actually want to talk to this specific IP address. I want to talk to this specific cryptographic signature. So we can just cut out the entire middle layer, for lack of a

better term, and actually start routing things through public, private key cryptography. And I think that is going to be a pretty big unlock for the ability to unwind a lot of these centralized, we literally call it a trust authority. That's what a CA is. We can try and unwind some of these centralized trust components from the system and actually increase competition. At the end of the day, I'm a capitalist maxi, and I want more competition and fewer access barriers to building systems that compete with one another. To those points, do you think these base infrastructure layers are inherently monopolistic, if not pushed otherwise, to the point of, first off, higher powers will want to control, and they can do so via the base infrastructure. Secondarily, there aren't a lot of

innovators. Do you think that's just what we converge on, if not for other efforts? I think that is the story of every economic system going back as long as we've had an economy. It's like a central bank almost. Exactly, yes. There's two things we've done very intentionally to combat this. The first is it's a token project. And the beauty of a token project is no one owns Bitcoin. Yeah, there's a bunch of people who maintain it and they have more power over it. Michael Saylor has got more power over Bitcoin than I do. But at the end of the day, it is a giant consensus on how these things will work and function. And that applies to Ethereum.

That applies to Solana. You can't block someone on a Bitcoin, to your point earlier. You can't block someone on Bitcoin. Or you can block someone on Bitcoin, but it requires you getting 51% of the rest of the Bitcoin network on board with you. These systems are still built on trust, but they're built on the trust of the crowd, as opposed to the trust of an individual privileged actor in a system. You can still change something in the Ethereum ledger, but the amount of people you need to get on board to facilitate that change,

it's like amending the Constitution. It's a very high barrier, and that actually creates a ton of security. And the second component for us is this proof of utility system, where proof of stake systems can turn into an oligarchy at some point, where you just have, oh, I have 4% of slana, so therefore I earn a ton of rewards, even though I'm not doing a ton of work.

With the proof of utility system, it's very similar to Bitcoin. If you have 10% of the hash power of Bitcoin, you're going to get 10% of the new blocks, irrespective of if you have Michael Sailor bags or if you own zero Bitcoin. So we've tried to build our system in a way that it encourages long-term competition, and it also rewards folks

who are able to put in that little incremental advantage. And actually, I think the true mark of success for a system is if someone with a better idea or better technology can plug into that system permissionlessly and earn outsized rewards for outsized contribution. Yeah, I completely agree. Austin, it's been a pleasure chatting. What should people do if they want to learn more about Double Zero

or explore working with you guys in the context of their business or operation? So, www.xyZ is our main website. You can find out a bunch of information about Main Net Beta, which is launching at the end of September. You can connect a validator and run on it, as well as get in touch with us, if you have fiber to contribute or any other crazy ideas you want to explore.

Brilliant. Thank you again. Thank you. Thanks for listening to another episode of On the Brink with Castle Island. To find out more about Castle Island, visit castle island. Visit castle island.vc. To listen to all of our podcast episodes, please go to On thebrink dashpodcast.com or just click on the tab in our website. Thanks for listening.