Pushing It to The Edge: The Future of Crypto Applications - TOKEN2049 Dubai 2025

How is everybody doing today? Hello. How's people doing today? Woo. Awesome.

Well, we're doing a fireside chat. Uh really excited to have David here with me. Uh I'm excited about what David is building, Pipe, because I kind of think it represents a new era of crypto. uh as we've kind of seen in the crypto markets, I think a lot of kind of the narrative games that have kind of plagued crypto for some time are kind of slowly coming to an end. And now what I'm excited about is kind of the next era of crypto of applications that are using crypto rails but actually enabling better products.

And I I think David and what he's pioneering here at Pipe represents kind of this new wave. I'll let David do a quick intro, but he has a very quick or interesting background uh and very storied career also in kind of the web two industry. Yeah. So, so we're building Pipe Network. It's a content delivery network essentially to compete with companies, you know, like Cloudflare, uh CloudFront, Okami.

Uh previous built video transcoding, media conditioning software. uh had a really nice exit to Amazon Web Services in 2015. Sold a company called Elemental. So basically had a career of streaming video and we saw some like really fascinating math of like bandwidth delay product and some ways to like optimize the latency for video delivery and lots of other use cases, you know, from high frequency trading to, you know, inference on chat bots and things like that. And we were building out this content delivery network using like hyper local nodes.

So it's basically in the neighborhood and we're seeing like in our early versions like 10 milliseconds of latency. So it's really bizarre. It's like already about 80% faster than Okami today. Um just a bunch of engineers building like video technology. So let's break that down because I I think your background is a little unique.

Uh from the venture side, what we see is a lot of crypto natives trying to go deep in kind of other fields. Sometimes that works, sometimes that does not. But having come from a very engineering heavy background, you've built applications at scale. What got you uniquely excited about kind of the power of blockchain uh within specifically kind of content delivery networks? Yeah, I think the the power of blockchain for us is like being able to pull together this like super massive node set.

So like right now in our our DevNet stage, we've got over 200,000 pop nodes. Just really amazing people contributing bandwidth right now. And essentially if you think about like an AWS hyperscaler, they're about 350 miles from most users. And if you look at Okami or these large CDN companies, they're about 150 miles from the user. And these POP nodes that we've got deployed, on average, they're about 50 to 20 miles from the user.

So it's makes the latency so low that you can actually fully utilize the bandwidth capacity on these home connections. From our test, it's like video starts instantly. There's like no delay. There's no buffering. and blockchain made this like coordination effort like I couldn't imagine sending out 200,000 wire transfers a week or you know especially if we want to try to pay these nodes like instantly you know per minute then be possible with traditional like banking systems and trying to coordinate this effort of like you know we're in the US so like we have nodes all around the world and people just join permissionlessly uh they contribute the bandwidth so it's very fascinating how blockchain kind of enabled something that would have been nearly impossible.

I think if we tried to launch this in a web two aspect, we would have needed billions of dollars of upfront capex to probably not even achieve where we're at today. I I think the thing that gets me so excited is becauc is really around the coordination aspect as you mentioned because crypto acts as this global coordination layer that can be used to incentivize people to spin up this infrastructure as you're mentioning and it's just something that theoretically you could do in web two uh as David was highlighting But doing individual wires for all these contributors is like such a hassle. Uh you you do remove a lot of that friction uh when you move to blockchain rails and I I think the interesting thing is as that friction comes down it enables to build build net new products because the scale uh just vastly increases. uh it it's very very hard to deploy the amount of nodes and infrastructure that would need to be achieved to kind of get latency down to what you guys have done. And so the title of this talk is really along the lines of the future of crypto applications where the infrastructure has been laid but we're moving up the stack to actually build useful applications and I think it really evolves around this hyper localization.

Can you dive a little bit deeper into why kind of this hyper localization or more so like edge compute or edge latency really matters? Yeah, so there's a lot of trends outside of video. I mean video is currently like 80% of the global CDN TAM which is about 25 billion. So we essentially highlighted video workflows but very quickly half of the inquiries today are are AI inference. So it's it's very fascinating, you know, trying to cache these objects of JSON and so forth.

And the hyper localization is fascinating. Like we we talked to many people before we even wrote the first line of code. And they said, "Oh, you're going to need 800 900 million dollars of capital to deploy these servers." And um you know, the blockchain coordination brought that down to basically nothing. You know, these node runners are like first shareholders in a company.

It's it's really amazing that you can come in and like participate so early in a company like that. And I think there's going to be a ton more of like deepines kind of utilizing these like hyper local hypers scale. I think for example like hyper local advertising would be a next you know benefit of something like this to for the next entrepreneur wants to go down this route. I think for the non-engineers in the audience, uh, latency may be a little bit of an abstract concept, uh, around like peen times, jitter. Uh, maybe in kind of more layman's terms, would you be able to kind of speak to the people in the audience or the people that are watching the video why it's so important to really optimize those things?

Yeah. So a lot of most of tech measures latency considerably in all aspects. So for example any shopping site if it hits a 100 milliseconds of latency they'll see about 80% abandonment of the shopping. So they'll lose for example at Amazon they'll lose millions of dollars per minute if latency spikes. And it does occasionally still.

So bringing the latency much lower, it just improves the user experience considerably. Right now if you if you go and you watch any type of YouTube, uh no matter what you're watching, you'll eventually encounter like some buffering and things like that. That's all due to network congestion and latency issues. And the hyper local nodes down into like neighborhood level removes this like latency aspect so you can utilize the full bandwidth at your home. A lot of people have like one gigabit connections of of fiber and things now in uh certain neighborhoods and they go to YouTube and it's a little bit faster but it's not that much because they're still reaching out like a couple hundred miles away for those video objects.

Yeah. I I think the the kind of meme more recently has kind of been uh reduce latency, increase bandwidth. Yeah. And I I think that's kind of been a rallying cry because it actually enables net new applications. Uh as you kind of expand the scope or create a larger sandbox so to speak for engineers to play with, they can build that new applications or uh in terms of video increase video quality from 1080p to 4K eventually 8K and beyond.

How do you kind of see the product perspective of like what pipe is doing and even kind of hyper localization? Yeah, I think I think today like even if you are paying for an ultra premium service like Netflix, they struggle to deliver 4K video uh most of the time it's downgraded considerably and that's essentially based on like a lot of bandwidth mathematics. So I think it will open up a new new era for faster more rich content. You know right now if you actually watch anything in like 12 16k the the quality is just so beautiful like you can see just the the micro expressions on on people's faces and it's really horrible that that's not available over streaming today. And you know, we we were watching uh just a couple months ago, we were watching this um UFC fight and Netflix and Netflix's massive infrastructure and this like little boxing max just crashed Netflix and they had had no ability to resolve it.

like they were maxed out at infrastructure and I know it was there's some old stories of like we launched uh live streaming in Europe uh couple years ago with Amazon video and we had the same issue like we didn't expect so many people on Christmas day would be watching uh you know Christmas movies and uh we had saturated all of the available bandwidth we had leased from the content delivery networks we're like three of us sitting in Starbucks on Christmas and in Seattle and just like calling up every CDN like we'll pay whatever whatever it takes and unfortunately everything had been leased out at that time. So I think one of the really unique things that inspired us like from Salana they have this like priority fee market. So essentially we added that in and if you're in like a dire situation or your application really needs super ultra low latency like high frequency you can pay like more per bite and have like priority bandwidth. So we think that will probably open up a whole sloth of like new innovations that we can't even perceive right now. Uh, one thing I would love to double click on just as you mentioned kind of the Netflix and obviously your pastime at Amazon and AWS is really just the level of scale for these things.

I think uh oftent times crypto as large as the market cap is and uh there are still products that have really lackluster from a scale perspective or still emerging. How maybe like going back a little bit more into engineering and like the nitty-gritty of things, can you quantify some of the level of scale that it really takes to run something like AWS or Netflix at some of these global scale? Yeah. So it takes tremendous amount of servers and like for example it's fascinating the amount of nodes we've got at this point without even turning on incentivization and if you look at content delivery networks in particular so like Amazon's uh cloudfront system it can scale up to about 200,000 nodes the same as Okami they have about 250,000 and here we are like pre-incentive And we've got the same capacity as these 25 year old companies with lower latency. So blockchain and I think this uh this new term to me like DPN it's it can unlock tremendous innovation if like you can come and build like ultra fast software.

So like building the CDN tech is not like easy at all because like we try to serve the data through our software path in like three milliseconds. So we try to add zero delay in this and you know there there's no uh AI is going to build that today unfortunately. Yeah. Maybe you talk about that orchestration layer because I think this edge computing obviously sounds well in practice. You can use blockchain to coordinate.

You incentivize people to put up nodes across the world and that does result in lower latency which potentially creates a better product but that software is extremely hard to build. It's no e no easy feat. A lot of people have kind of attempted this. Talk about that like orchestration layer and kind of making sure that you can build and scale to hundreds of thousands or even millions of nodes. Yeah, it's uh essentially it's what we had done we'd been trained to do at Amazon is handle infinite scalability.

You know that we serviced every major major broadcaster on earth without any any type of issue. So essentially it's just fundamental like low latency distributed systems engineering. And it's it's very amazing that we can achieve what we've built so far with this system. I think um it's this deep end is going to open up so much more innovation. And to that like one of the things that I really appreciated about you was almost the skepticism around blockchain.

Yeah. And I it honestly excites me today because I I would love to see more kind of traditional web 2 engineering. They've spent say a decade cutting their teeth on these very high performance systems uh from engineers like Google Meta like building very high performance systems. What would your kind of like words of wisdoms be to those crypto curious but skeptic kind of like web two engineers? Yeah, I think for us like we we talked to a bunch of principal engineers at Amazon and Google before we wrote the first line of code and kind of vetted like the mathematics.

You know, there's a ton of math for whatever sector you're engineering in. And and this one in particular is uh the magic formula is called bandwidth delay product and optimizing for the latency essentially will invalidate the existing CDN market. So like just find the niche that you want and find the mathematics on how to optimize that and you can easily have a 10 20x you know performance improvement on existing web two companies. I think defense is going to open up some like fascinating like solo founders with like billions of dollars of revenue very soon. I I think within the deepening category overall there's been a lot of kind of excitement about obviously continuing to lean into what crypto does well at which is coordination.

Uh but historically it's been hard for some of those founders to uh bring in demand. And I I think that why I'm excited particularly about pipe and kind of the future of kind of edge computing and pushing things to the edge is because of the net better product that that can create. Uh and I think as soon as you build a product that's not only 2x better, 3x or 5x, it's significantly better where net new applications kind of emerge, you open up the new world of possibilities. And you mentioned I mean piping hyper optimize for video delivery but there now there's kind of net new applications that are kind of surprising you like uh distributed inference or just inference requests. Talk about kind of like some of the more novel things that are kind of surprising you with what you've been seeing on the pipe side.

Yeah, the I think the AI piece is is very fascinating to us because it came out of nowhere and they have like very low latency demands. So if you're using like a a chatbot and you want like instant feedback, they need that in like 50 milliseconds or below. So just really quick so it feels like you're talking to a human or uh kind of interactive robot. And I think the other side, we we spoke with some guys in uh Singapore, they're building uh these really cool robots for like food delivery and things like that. And they were immensely latency sensitive and um there was no existing technology out there for for this.

So like they they had to like manually drive these bots and and now with our our pipe network system, uh it's fully automated the robot delivery. So I think kind of robot to robot, it definitely opens up some new new innovation using the public internet where a lot of these companies would have to go and you know build out wireless infrastructure or at least private fiber and just have tremendous upfront capex. Yeah, the capex is expensive. It kind of now shifts a little bit more to opex with uh tokens which I think is interesting. And the other part of about that is with kind of the token or global coordination network of the blockchain itself, you can almost do things like Uber surge pricing where a lot of these kind of traditional companies have to spend tens or hundreds of millions of dollars to expand city by city.

But with blockchain kind of powering it on the back end and acting as that coordination layer, you can almost say like this region is underserved. If you want extra rewards, please go here and set it up. And I I think that's kind of a continuously underexplored opportunity, but something that's uniquely uh I think I I'm excited to watch as you guys continue to kind of like push on that front. Yeah, exactly. So when we call up our previous network of you know these large streaming companies our opening is like give us a list of your worst performing cities and we'll go pay geographic bounties on that bandwidth and we will prove the latency improves and yeah this like geographic uh bounties on blockchain is fascinating.

I don't think we've ever seen like a onchain bounty not become fulfilled. It's it is remarkable. I I think in your kind of perspective, what what has been perhaps the toughest transition about kind of coming from the traditional web 2 engineering world to blockchain? I think for us like crypto's all over the place of like the products there's been been a lot of really great projects we've providing some infrastructure too on Salana some deepin projects and I think the the stigma is really not the best unfortunately at the moment you know we go back and talk to directors Amazon video and they say you know don't don't mention blockchain you know, don't mention tokens, just focus on the product. And that's that's what we do.

Like we built the the product and then at the end we started thinking about, you know, tokens, rewards, all these like crypto pieces. But yeah, I think hope hopefully deepen really helps push this forward and cleans it up. I definitely agree. I I think it's fairly early on the deep end side, but these net new products that are actually better products are the thing that I think are exciting. Um, in terms of you mentioned kind of Salana itself, how do you envision things like Pi potentially working with blockchains in a future?

Obviously, you can cut down on latency uh p push things to the edge, but what does that result for the underlying chains themselves? Yeah, it's uh it's very amazing. So, like we really like Salana. I think it's it's a great engineering culture. um got got connected with Anatoli very early on and um just talked a lot of like tech raptor encoding you know compression algorithms and felt like it was a a very fascinating blockchain ecosystem and essentially it does have a lot of problems I think one of the the big problems it has right now is like a read layer so it's processing just transactions like crazy just um you know hundreds of thousands a minute and essentially trying to read that that back is is very contentious off of the node set like you're utilizing so much bandwidth and CPU that should be used for processing these transactions.

So you know one of the things that it's not not a company priority but it's very cool like experiment is like building out a decentralized read layer for Salana and I think it it opens up a lot more really unique blockchain innovations. One of the things we've talked to folks there about is like a kind of a stateless validator. So you could you could potentially spin up like hundreds of thousands of Salana validators, keep the blockchain state on pipe and these uh validators perform microtransactions to get the latest block and create the validity proof and push that out and essentially it decentralizes Salana even much more by decentralizing the read rare at this like ultra low latency. And I think if if you look at most products out there like they're in a like best effort to return the data to you and uh that's uh from my entrepreneurship that's not really practical for businesses. Yeah.

Best suffer doesn't really cut it. Uh it's kind of like you make the better product that product uh returns the data very quickly or you're getting cut. Yeah. Um I'm curious um broadly how you kind of think about uh kind of the content delivery space overall. I think if you're obviously heavy into engineering, have worked at these large tech companies, uh this is a more understood problem, but for in perhaps more layman terms, how would you want the audience or the people watching to kind of understand this?

Are there any kind of comparable companies in the web 2 world or even in the blockchain world that you kind of feel like are potentially uh aiming or have been kind of misguided even in the past about like trying to attempt what kind of pipe and edge computing or edge latency is doing. Yeah. So like in in Silicon Valley, it's very common. Startup wants to come in and create like a more like a content creatorowned. So there's no middleman.

One of the issues that these companies traditionally hit if they get like a h 100,000 users, they'll have $300,000 uh CDN bill. So like they go bankrupt in 90 days typically if they start scaling. So it's it's very difficult for a company to come in to like rich streaming media and be able to participate without either tremendous venture backing or uh you know some some really strange edge. And right now you know there there's a huge opportunity to reinvent uh content delivery. you know, YouTube took off and then 2016 it flattened out the rewards for the content creators.

And I think coupling some like really crazy innovative technology like pipe and probably some others like live pier for video transcoding, you can build some like very unique disruptive technologies today. the the aspect or kind of like I think I'm also excited about uh we we've talked about video transmission, we've talked about inference, but I feel like what you guys are enabling on the backend infrastructure and pushing things closer to the edge almost potentially unlocks this like crater economyish where people have been trying to do this for a while with blockchain rails but haven't really had all the pieces. Uh obviously the the L1 itself or the blockchain is acting as that global coordination but they really haven't had the high performance enough on the back end to kind of make something like a decentralized YouTube to a certain scale. How are you thinking about kind of like craters or like net new uh applications uh on the creator side? Yeah, I think I think because like pipe is utilizing latent bandwidth.

So it's just out there idle you for for an entrepreneur and even traditional companies the pricing is just fundamentally different from traditional CDNs. So it opens the realm for like a lot more experimenting of of building out these rich media apps and you know rewarding the people directly. Um you know I think you get you get so such a big cut from the middleman on most of these apps even even newer ones that are you know very pacific of like cooking and uh all these kind of narrow apps out there today. It's um it it it's not sustainable, you know, at this point. Basically, I think um I appreciate what Pipe is doing.

I I think broader one of my big takeaways from this conversation is that a new era of kind of applications are emerging within blockchains where they're building net better applications than were previously possible in web 2 and I think things like pipe are really on the forefront and enabling those better applications and the the product is actually building a better product. the blockchain is almost an implementation detail to that and so I'm very excited about what you're doing. I I think you're really pioneer something unique here. Is there any alpha that you could potentially share with the audience uh today? Any exciting things or how they can continue to follow along with what you're doing?

Yeah, I think you know kind of follow Pipe Network on uh Twitter X.com. Uh currently we're in the DevNet phase and uh basically next week we'll be launching incentivized test net to help reward these node operators. There's a lot of nice demand so it's time to kind of upgrade the system and uh hopefully everyone can uh run a node. Awesome.

Well, thank you everybody. Thank you David. Appreciate it. Cheers.