Shelby Deep Dive: Bringing Decentralized Hot Storage to Crypto | Pranav Raval

Transcript

Beyond the trading and beyond the payment, uh we also want to support a lot of decentralized innovation and that requires decentralized infrastructure that can actually bring value to the apps which are building it and then have an economy where the revenue that is being created is not being centralized into few big tech companies but is actually being decentralized. >> Hey everyone, quick disclaimer before we get into today's episode. Nothing said on bell curve is a recommendation to buy or sell securities or tokens. This podcast is forformational purposes only and the views expressed by anyone on the show are solely our opinions, not financial advice. Our guests and I may hold positions in the companies, funds or projects discussed.

All right, everyone. Welcome back to another episode of Bell Curve. Uh today I'm joined by Prrenav Ravvel of Apttos fame. Prrenav, welcome. >> Thank you very much. Great to be with you. Yeah, excellent. Uh, so we've got a lot of ground to cover today. Uh, and we're going to be talking a little bit about uh, some of the infrastructure that you're in. Your title is head of infrastructure at AppSlabs, but you are also leading all things Shelby, which is this exciting joint venture that you've been doing with, uh, Jump. Uh, so I want to get uh all into that, but maybe even before we get into the weeds of what you're building and we're going to pick it apart layer by layer. Um, I would love to just understand maybe zoom out for a second and understand how you were looking at the L1 landscape today. You

know, because when you look out at the world, uh, in a lot of ways for L1's, this is a great time. You've got a whole bunch of institutional interests in the United States. You know, we've got the political and regulatory regimes changing. You see a whole bunch of new assets and attention and all that stuff. uh but at the same time you know the L1 space there's always been a lot of investment there are many different L1's out there and you're starting to see maybe you know the beginnings of uh you know some network effects uh taking off and I just want to understand you know how does uh an L1 like Optos today think about positioning you know what markets are you interested in talk to me about how you just see the whole landscape >> yeah no absolutely I think this is a

super exciting time to be building an L1 and especially L1 that is ready for real world demand right So since very beginning like apta started at meta right our our our origins were with the DM Libra project and the goal was to build an L1 that can scale for the global user base. So we have been building for this demand for quite some time and we are super excited that finally we have regulatory clarity. Uh there are lots of new and interesting use cases being built and APOS itself is extremely focused on building um uh the global trading engine. Uh that's that's our goal. Um Aing our CEO he put out a big uh vision statement at the beginning

of the year and we have been head down building the global trading engine uh that is going to bring all sorts of assets on chain and make the trading infrastructure 24/7 uh with many new innovative assets coming on chain. So that is number one that is our big priority and and we are headon building that at the same time the stable coin uh and payments uh is is absolutely a critical use case that is gaining lot of traction and momentum and at DM Libra that was our beginning that was again that is what we were born to build and you can see there is lot of stable coin inflow there's stable coins uh both two big names USDC and USD and

bunch of and other stable coins are already on Apptos and so we are also very well positioned to have um very high throughput payment networks being built on Aptos. So that's those are the two key stated objectives. And the third one which probably we'll go in lot of details today is build a global decentralized infrastructure layer that can compete or even be better on some dimensions with current centralized clouds because beyond the trading and beyond the payment uh we also want to support a lot of decentralized innovation and that requires decentralized infrastructure that can

actually bring value to the apps which are building it and then have an economy where the revenue that is being created is not being centralized into few big tech companies but is actually being decentralized. So the third big initiative we have is uh we build the decentralized infrastructure um that is high performance and can operate at at a scale that all the big tech companies operate at. >> Got it. Um yeah and it's so that makes so much sense uh just given your h past at um meta right I mean everything in in web two you know some of these massive you know tech giants do everything you know very famously to scale uh and I think that kind of stands out with with

the vision here and you know we were talking a little bit you know before we got on the call of you know and this will lead into Shelby and why it's interesting but you know kind of of this vision of combining some aspects of what if historic Berkeley been relegated to the realms of kind of Wall Street uh finance you know financing capital formation but then also what has been you know kind of relegated to Silicon Valley which is value creation and creating a product that people really love and I remember in when I was getting into crypto in 2017 2018 people had all these debates is it Silicon Valley or is it Wall Street uh and you know our discussion before we we got on just made me laugh because those two things are actually highly

related. Uh so maybe you know could you talk a little bit about kind of this vision around you know um you know combining or putting much more closely together you know the financing and the value creation aspect on chain. >> Yeah absolutely I would love to. All right so let's let's talk about the Shelby economy flywheel that we are building and and the purpose why we are building Shelby. So if you look at web 3 and the state of the union right now, we have nailed the digital currency use case where we have uh uh we started with BTC and now we have Bitcoin, now we have stables. We also nail the dig tokenized assets from memecoin to RWAS. But when you really think about it, when you squint hard, this is um actually all about value exchange. In fact, even if

you want to be a little bit critical, there is also a lot about value extraction with me and arbitrage and everything, right? Um so yes while while we want to build an amazing value exchange engine because the participation in exchanging value also brings lot of economic benefit to everybody uh globally. Um we want to also add a slightly different counterpart to that with Shelby. We want to have decentralized hot storage tier where you can store serve and process data at scale. And once you put compute next to Shelby, you can analyze, train and infer from that data. So you can your data gets converted to information.

Your data gets converted to intelligence with that with that intelligence and information. Uh the D apps that you can build on top of it would be rivaling in revenue the applications we see already that that is being built. Um, you can have streaming, you can have ads, you can have social applications because now it's all backed and powered by data. The same engine that powers all the big tech revenue will be now available to everybody in the decentralized economy. And so this is the value creation engine. So when you put the value creation engine next to value exchange engine, uh I would want you to think about um a D app that is building decentralized streaming um with user

generated content, say YouTube, uh decentralized YouTube, that could now actually have tens of millions of users and hundreds of millions of dollars in revenue. When that now asset gets listed on the trading engine, it generates capital. that mirrors the IPO or large scale IPOs that we have seen generate already and that capital that is raised kind of gets in gets invested back into decentralized innovation and so this is the decentralized digital economy flywheel that we are waiting we are looking to create with Shelby and that is the endgame right for decentralized digital economy and so that primarily

motivates us to build the missing piece of infra infrastructures in the digital economy, decentralized digital economy. >> Okay. So, I have a lot to say about that, but before we get into the details, can you just describe how does the Shelby product enable something like a decentralized YouTube like just talk just for actually for folks who are less technically inclined just give us an overview of what storage is, right? Why you know the kind of tagline of how Shelby is described as this, you know, hot storage network. So, why is it so important to put you know this on chain, right? How does this enable what you're describing?

>> Yes, totally. I think um the key fundamental premise that uh Shelby brings is incentivize reads. So when you store your data in a decentralized way, anybody who now wants to access the data, they need to pay for that data, right? So this this does two things. One is that your data is being served really really fast because the reads have to be paid for and all the participants in the network are now incentivized to serve out the data as fast as possible.

>> And the second thing is that it also kind of brings the data that is worth paying for right so that brings high value data uh in the decentralized storage. So if you look at the current state of the affairs for all the decentralized storage systems that are existing right now the there is lot of data that is just kind of backup and cold storage and archival. So now users who are building or there's a D app that is building a YouTube like uh streaming application they can start they can start incentivizing their users to say hey put your content here when we serve out the data you are going to get paid then we we are then at the same time the data and the analytics that can be processed

and can be understood from the data will also allow the D apps to start having um ads marketplace places. So we are already talking to some decentralized ad ad marketplaces that can help bring the content and the relevant ads together >> and so now the D apps that is building is building its own revenue stream. the users which are building their bringing their their uh videos on decentralized storage they are also building uh they're also g making money and so every all the participants so unlike a big tech like if you think about the take rate for YouTube there's about 45 to 50% um uh any any streaming platform takes 45 to 50% take rate versus in this the

the new economy the take rate could be much much lower >> a lot of the revenue can be passed on to the users and the and the D apps and the users and all the participants start uh participating and start getting a big slice of the revenue that is being generated. >> Okay. So with that in mind, so I think also that there's something to be said for I mean you know just like if you go back and look at the web and all of the different successive layers of computing that needed to evolve in order to make it happen even something like YouTube right which by the way fun history about YouTube that was actually first a uh basically an early predecessor for uh like Hinge or Tinder uh before they you

know pivoted to YouTube uh the YouTube that we all know today and it took uh it took quite a long time for them to actually become uh you know profitable and to decipher a business model. But also think about all the different like the bandwidth constraints that needed to happen. All the different stuff that you could do on a web app uh social algorithms first instead of uh you know looking at videos that your friends like. There were so many different things that needed to happen. Um and these kind of successive waves of infrastructure that we all today take for granted. And so help me what I want to dig into now is like help me understand exactly the architecture of uh Shelby uh and then we can get into maybe even getting a little bit more concrete in terms of how it enables this

this >> Yeah, totally. I think even going back to when you talk about history of YouTube, even talking about history of internet like 1998 where where Google was uh incorporated, I believe they they collected or they indexed about 25 to 50 million web pages and they put page rank algorithm on top of it and since that like trillions of dollars of revenue has been made from data and algorithms, right? And so, uh, I think this is this is where if you want to start making something like YouTube more successful or more like decentralized YouTube more successful, you need um you need an engine that can allow you to process all of this data, run your AL algorithms on it, build kind of user

centric uh feed and recommendation and then kind of match the match the ads and users in a way that is still privacy preserving uh because you don't now that uh data is owned by the users uh users can allow access and users can selectively kind of share certain parameters that they want to use for the kind of ads they want to see right so I think there is there is a lot more that is uh that is going to happen in a decentralized economy that looks similar on the surface but behind the scenes underneath it gives a lot of power back to the users um so Yeah. So u now coming to the architecture Shelby architecture um

there are six pillars um of of Shelby architecture that uh differentiates it Shelby from the general decentralized storage uh that is there right now and let me let me bring that uh bring that slide. So let's talk about the six pillars. So first as I may as I mentioned uh the paid reads um Shelby incentivizes every single read. So data that is stored in Shelby every time you try to read the data you have to pay for it. And we have special payment channels being built so that uh there is no impact of making that payment on the performance of the read right so where your data reads

would be seamless. The second one is uh dedicated network. Again, this is a foundational change uh from how everything is being built uh every infrastructure is being built um uh in in in um uh in all the other storage networks because they all go through public internet and they're extremely bandwidth constrained. For Shelby, we are going to use a dedicated fiber network with uh very high bandwidth. So all the nodes that participate in Shelby network, they'll all be connected through probably 100 Gbps network to begin with and probably a terabit uh uh network uh soon after.

This this way all the nodes in Shelby network will be able to move lot of data around super fast. Um the the la the third one is battle tested software and and jokingly Jump and Apptos team talked about how long it took them the last time in meta and jump when we put out a new hot storage protocol and meta it took us like couple of years and similar for jump and so we rather than kind of inventing everything from scratch we are going to use some of the battle tested software from Jump's existing stack right so that allows us to go to market really really fast and it brings the reliability and performance that is required at scale.

Um whenever you build a decentralized storage system um you have to figure out how you are going to take your data split it into many many chunks and spread it across the network and you do that so that your system is resistant to failure. Right? If some nodes fail, some disks fail, uh even a rack fails, um and all of this is very common at scale, you do not lose your data. Um at the same time you really have to find the right balance of repair bandwidth and kind of your robustness of your data. You can you can overly replicate the data many many ways

or you can have very small amount of replication and then you end up paying a lot of repair bandwidths and so we are using an encoding scheme called clay codes which gives us this optimum kind of uh balance between those two. Um, in a decentralized storage system, you also need to make sure that your data is available because this is not you're not putting it in your kind of one data center and just make sure that you're watching your machines like it's it's running globally all across uh many many different data centers and so then auditing scheme needs to be extremely efficient for and it also needs to be incentive compatible. So you have to constantly audit to make sure that the

file that all the nodes are saying they have stored actually exists. Uh otherwise you run the risk that there is a node that says the file exists and they have it. When you ask them for a proof they'll actually go to another node and hey give me that file. I'll give you the pass on the proof and I'll just you know reap the benefits. So we have a very interesting auditing scheme uh that uh that prevents from all of that from happening while it is extremely efficient and of course definitely you need an extremely fast reliable and scalable coordination layer because when you're building a highly scalable storage system there are many many files you need to keep track of everything you

need to keep track of all the payments and you need to keep track of all their auditing results and placement and this is where appto shines through, right? You know, you talked you asked me about the beginning like what is Apptos thinking about because Aptos is highly scalable and highly performant. Apptos can support um a workload of this scale seamlessly and also our gas costs are very cheap right so um it does not add to the cost of the storage protocol itself. So are you thinking about Absel and Prennav as you know this is one question that I had for you uh you know in preparing for this episode is you know do you is Shelby kind of an what is the relationship exactly with Aptos like is

this exclusively kind of another layer of infrastructure which is going to live kind of permanently and exclusively you know within Aptos the protocol you know if this uh if this was successful enough you know as a hot storage layer for Aptos would you serve other L1s like how do you view the positioning here. >> An excellent question. So while Shelby is being coordinated on Aptos, which means all your files that you're storing will be managed and coordinated on Aptos, the usage of Shelby is already built for any chain and any stack. We also want to break away from the silos that we have created in in uh in like you know or maybe even the cults that we

have created, right? that you just kind of go all in on that only single stack. Shelby should be able to serve any stack. We already have multiple L1s who have partnered with us. Uh story protocol um is going to be using Shelby u near and Shelby announced a partnership and we are talking to two or three other L1s and L2s to bring their data heavy workloads and data heavy D apps that they're building uh on Shelby. And so Aptos will be the coordination layer which should be seamless and kind of somewhat hidden from all the um all the consumers of Shelby which want to kind of read and write data from Shelby.

>> Okay. And help me understand too you know I'm just thinking of let's say I'm trying to build the next YouTube right or let's even say there are all these workloads that are happening right now. There are lots of like exchanges maybe are a good example where they have they're hybrid onchain offchain and many of the workloads just happen mostly in an off-chain environment where computation is cheaper and then maybe you post something like a proof on chain right like that's kind of the and the way you know tell me if I'm misunderstanding this but the way that I view this is you're moving more uh more of the workload especially on the memory side of things onchain um right and appto is extremely you know this is a really great use case Raptos in the sense of it's very fast, it's very cheap

and it could accommodate those additional workloads but what is the benefit from the perspective of an app builder you're say? >> Um yeah so the biggest benefit that I see is that your data becomes integrated with your onchain workflows. So you are not bringing the large data onchain but there is a very seamless integration. I'll give you an example uh of how this all plays out. There is uh one very large Hollywood studio which is very interested in building uh decentralized streaming platform on Shelby. And you know I was also a bit surprised when I first heard this uh requirement and yes there there are this monopolies and middlemen and all of that

uh that that is actually a problem. One thing I realized that when you think about any content, you have lots of um contracts which talk about royalties and u IP and and and rights and so there is on one hand you have rights then you have content and then you have royalties and then how everything pays out and how everything there is a waterfall of payment that you need to manage. >> Yeah. Yeah. every step of the way is a different system right now. And when you bring it all together, including the streaming and including the payment for streaming, you can automate that entire workflow, right? You you stream the data out, you get the payment for the stream

data out. That payment automatically then goes on chain and says, "Here's my smart contract. Here's the waterfall of payment. This is how everything should get paid out." and everything settles like somewhat real time versus right now it takes months and months to settle those accounts, right? So this whole efficiency of seamless integration from your app to your data to payments and everything just being coordinated seamlessly is a huge value addition for anybody who's looking to build the app and that is a data hungry app. Okay, that makes a ton of and that uh I can completely see the benefit of that as well from the like this is one thing that I'm just kind of trying to think through in terms of what are some of the big changes uh that stable coins uh might bring about and yeah if you have

kind of an asset and then you've got you know the dollar payment it just makes everything much easier to do in one environment you can do things like real-time streaming right or something like that which would make no sense in uh you know kind of this segregated onchain and off-chain environment >> yeah and and it also brings new um new experiences and new payment models like um you know uh you might want to do uh rather than like everything works on subscription right now, right? You can you might want to just do you know pay use and you can start having now microp payments even if you are watching a short clip you can charge for it right and so all of the new payment rails can enable new experiences uh that don't exist right now and kind of everything is either ads or subscription so so I'm

also super excited about that right bringing all of this on chain and enabling new payment models to enable new experiences >> okay so you know let's go back to I was looking at that that slide I've got the white paper pulled up here as well. Um, so you know, these are kind of the the architectural pillars here of, you know, paid reads, dedicated fiber network, battle tested software stack, uh, efficient coding scheme, uh, and incentive compatible auditing, uh, all built on AP toss. Talk to me about this dedicated fiber network. Where is this coming from?

>> Um, that's, uh, that's a pretty good uh, uh, pretty good question. Um this is something that uh that has been holding the industry back so far because everything we are building is just so bandwidth constrained right and so finally we have industry has realized that uh we need more dedicated bandwidth and lot of it and double zero uh is is building that's what I was thinking >> taking the first step yep first step in this direction and so we are also working closely with double zero uh to have dedicated bandwidth for Shelby and and we believe this is just a start right this is uh these are the two foundational layers once you have dedicated bandwidth and then you have uh

high value data uh then these are all the pieces that you can start stacking on top of on top of these two foundational pillars next we are going to bring compute GPU and CPU and then there is a uh in in in Aptos we are also building eventdriven architecture um in some sort of You can call it reactive blockchain. So you can build a lot of workflows. So now you can have a workflow engine. You have your data, you have your compute, you have your bandwidth and this is where you can start actually in earnest start competing with the existing uh cloud providers and uh and build a decentralized cloud. Mhm. So how do you think about like when I think about the

cloud providers or even something like maybe we can talk about uh you know CUDA something like CUDA built around Nvidia like part of you know the lock in that these guys have um is not just you know the compute that they're able to offer very cheaply but around all of the software that they give away for free right like you have entire you know SDKs libraries for like dedicated use cases and that's a huge you know Different people have different strategies, right? Uh, you know, CUDA kind of gives away the software, but everything in Nvidia is closed source and so it locks you into the hardware.

Um, versus, you know, uh, you know, some other strategies, uh, which is, you know, more about, uh, charging for the actual software. So, how do you think about, okay, outside of enabling this, what else are you going to need to do? um you know maybe selfishly from the perspective of providing lock in for Aptos but outside of just providing the ability to do this like you know are you investing in kind of you know SDKs and you know you know use cases and stuff like that to build around here or talk to me about that layer. Yes, absolutely. I think uh what one thing you'll see and and we can uh we can talk about it is that when Shelby devet comes out, we are going to have a bunch of use cases and bunch of SDKs um uh that we are building uh that go with each of these use cases.

uh we haven't we are not building on compute yet but we are going to start building compute soon uh next shall we and that's the beauty of uh decentralized infrastructure that you can also have Nvidia hardware and you can also have uh any anything else you can have your AMD hardware next to it and you can have all of these libraries as long as you can run in um like Nvidia also supports running um uh uh supports hardware and then they have I think built-in support for trusted execution environment uh with GP CPUs and so once you have verified compute uh you can leverage all the libraries which are there they just run in a trusted context right so we can run that we can offer the new other hardware um it it depends

on what users want to use right so I think we see a future where all of these can become seamlessly come online can connect to the same network uh so that their ability to process uh data and move data around uh and kind of ingest test data and process data and read and write data uh is uh is pretty high. And so um we can offer all of these um uh all of the systems and then SDKs and APIs that make development extremely simple. Uh one of our goals with Shelby is to actually have a user go from not knowing about Shelby to building on Shelby in less than half an hour. like

just just go there, you see the example, you write it and like, "Yep, it makes sense. Here's how read and write data. Here's how make payments. Here's everything works." And just seamlessly just start and start building. >> Okay. So, walk me through uh you know on the architecture side uh the different layers that exist here. Um if I'm understanding this correct, there's a client uh SDK, RPC node layer, storage provider, and coordination layer. So I know that's uh you know we're in danger of getting deeply into the weeds here but just walk me through like overall kind of the different layers of the stack.

>> Absolutely. I would love to. So then um the six pillars um we have uh we have kind of few few horizontal crosscutting layers. Um so client SDK is where um you um you take the data you split it into these chunks. Um then we have uh RPC nodes which act as a gateway to everything Shelby. So RPC nodes are in charge of um receiving your data making sure data gets written or read uh from uh from the nodes and RPC nodes are also um kind of they can offer new and innovative payment models like for example we also see some RPC nodes can say hey actually even if you want to pay

for Shelby in credit cards I I'm happy to accept credit card I don't mind that right so of course the default would be stables but uh there would be RP RPC nodes that can provide any kind of payment mode. And so then the last one is SP storage provider nodes which are kind of akin to you know you can think about validators on a blockchain. These are going to be storage optimized large nodes with hundreds of terabytes of storage potentially pabytes of storage per node uh globally located in uh data centers where we have connectivity uh and uh they will be storing data.

they'll get paid for storing data. RPC node will get paid for the reads they're serving out. Uh and client SDK makes everything very seamless and integrable in your in your D app. And so this is where I when I mentioned that we want to go from not knowing about Shelby to building on Shelby in 30 minutes. This is where very simple easy to client SDK will come in play. >> Got it. Um okay. I really that that's super interesting. The ability to pay with credit cards. That's uh yeah that's something I had really thought of but that's very cool.

>> Yeah and I and this is where okay I'm I'm I'm going to get slightly excited but when we look at the current pricing model of the current clouds um there is a storage uh price and there is egress price. So let's take Amazon S3 for example. Uh their storage is about uh you will pay $23 per terabyte per month and the egress would be anywhere between it depends on your egress monthly egress bandwidth that you buy but anywhere is like 70 to 80 even $90 per terabyte per month. So as you

read more and more data you know there's this general complaint that hey my cloud bill suddenly adds up. Like everybody's just always surprised like oh my cloud wheel is so high because they're like oh I'm just paying $23 per terabyte per month and then once you start using your data your bill starts accumulating when you started looking into what's cost that much um we realized two things one is that um the egress cost in itself is not that high one is that they want to create this data mode and lock in right so data is um is a kind of key asset they want to lock into their own infrastructure hey why do you want to take out the data just use it with our engines and use it internally for for other use case. And the second thing is that there is a very high markup on

egress. And so when we want we are looking at and we haven't finalized the Shelby cost model yet but even from storage point of view and egress point of view there are opportunities to make it while keeping the performance characteristics the same make the cost cheaper right and so we are excited that I we are completely expecting that there are going to be large web two clients who who would be very interested in using Shelby as well just for the economic benefit. that it'll bring and so that is where the whole you know hell hey there should be a credit card payment model as well to onboard users and this is where we feel like a lot of this mass on boarding uh will happen

because they will use Shelby they'll not even realize it's being coordinated on a blockchain and they'll be super interested like oh what else does it offer right and so slowly we are going to we think that there will be a one more u path through which we are going to onboard new users and kind of new enter and enter enterprises onto onto decentralized infrastruure And is there something about the decentralized architecture which allows you to uh like for egress to be structurally cheaper or is that just a plan to undercut their current centralized cloud pricing models? >> It's it's primarily that like if you when you are when you're actually serving out data there are like two three major um there are three major uh cost factors. One is that the machines that you use to serve out the data. Um the

second thing is um the internal bandwidths to move data because it's in storage stored somewhere. You read the data out. You bring it to the edge of the RPC node to serve it out. And the third is like you pay the ISP uh to to buy the bandwidth to serve it out. Um when we put it all together and add it up, uh it's it's still um it's it's still significantly cheaper than what uh the cloud big clouds are offering. And so yeah, that's just a very high margin. uh on on the US >> and PR why walk me through the logic of why make Shelby its own layer versus you know you're a head of infrastructure at APOS lab so why not build this into the APOS protocol itself

>> yeah I think um it it primarily goes into um how albums and blockchains work right so let's take an example of apttos it's um it's a network of 150 nodes And in that network um you get each node replicates all of its storage right because it's fundamentally it's a ledger and all 150 nodes need to agree on the state of >> the ledger at any given point right and so the data gets replicated 150 times.

So the that amount of replication is only justifiable for something that is um where you are storing of high value but small data assets like your um it's a ledger so your accounting for example that should happen on chain because you need all the 150 nodes to agree uh on apttos but when you start talking about large data sets and potentially large compute you You can think about a blockchain or an L1 that can support it, but then you have to replicate it as many times, right? And and so that just becomes just adds a cost kind of explodes at that point and it just becomes very very uh

expensive to operate and use. So I think um uh just building storage layer which is adjacent to Aptos and maintaining all of its metadata on Aptos right to coordinate and make sure the data integrity is fine. But the actual blobs, they don't need to be on chain. The blobs can be offchain as long as you can maintain its integrity and make sure that all the metadata is preserved, it's coordinated, it's paid for, it it essentially serves the same purpose, >> replicating it as many times. >> And Pam, is there a differential performance for Shelby here on applications which are built on Apptos versus let's say Ethereum or Salana? So I'm just thinking about Okay. No uh we expect all chains, all stacks to use um

uh use uh Shelby seamlessly. We have apptos as you know we have crosschain accounts. So you can just use that account um and and you can read write for it. Reads have as as I mentioned reads have kind of you know many different flavors. you can directly pay the RPC nodes in in in the stables of your chain and when you write you have to use the crosschain account uh and and write but that should all be very seamless from performance to kind of ease of development everything we want to Shelby has been designed from kind of day one to be any chain any stack. H.

So, help me understand and maybe as we start to, you know, kind of wind down here, I almost I almost want to, you know, go go back and forth with you a little bit on some I have a pet idea that I really like. Um, so I've been very into this idea of content coins and underwriting. Uh, so, you know, been in media for a long time. I see the trend, individuals over uh companies and brands. And, you know, there's some some of these streamers. I don't know if you know a guy named I show speed. You get a $600 million deal from Twitch. And so one of the things that I like about um one of the things that I like about the idea of content coins is you get easier access to financing. Like you can just take a coin, you can dump it into a

protocol. And if you actually one of the benefits of like onchain social uh would be all that information is just out there and you could use that to underwrite a loan easily. So I'm just thinking is this like a workload that could work in the future? Let's say in this crazy uh world that I'm presenting here, which very few people agree with me on, and I've talked to a lot of people about it, but let's say um let's say, you know, I am a creator and I'm on this uh decentralized version of like a Twitch or YouTube or something like that. >> And all of my analytics live in Shelby, right? Like my whole history, the entire memory of all my videos, yada yada.

>> Um so let's say I want to leverage that to go get a loan. So a theoretically over on Ethereum could leverage that data say hey uh this person is creating content you know their engagement is growing twice as they have twice as much engagement and is growing at 1.5 the rate of the average creator I know what that means from uh that means historically in terms of cash flow and therefore I can just do this loan it can all just happen pretty autonomously on chain very easily is that type of thing >> you might have just unlocked the missing piece that I have been thinking about the creator economy This is this is amazing. The reason and and I'll I'll first step back and then come back to your question. Um the the creator tokens have just got a

bad reputation, right? Because of a lot of people just you know launch creator tokens. They don't back it by anything substantial. There has been like you know some examples of you know pump and dump or ruck pools and whatnot. And so I think this one of the re one of the key audience or developer audience uh for us is creators. We are actually talking to some of the very big creators uh who are interested in participating in the new creator economy and they're also very excited about it but they're kind of scared of just putting a creator token, right? And so now if the token is you first build a history, right? First build out your

reputation on chain, your history onchain, the proof of revenue on chain backed by all the data that is built on Shelby. You have all the analytics available, you have all the history available and then you can launch a creator token. Now that is very well backed by all of your data, your history uh and you can have very clear terms of like what does it even mean to own that token. So that is the one part and the other part is how now taking it to the D5 world and decentralized finance and getting loan against some collateral. I think that's absolutely possible as long as you can prove what you know what is the value you know you are putting it in as collateral and with Shelby that is absolutely easy to prove because all of

your history and everything would be uh would be readily available for for anybody to verify. So I think that's that's a you know your question was slightly different but that has triggered a very different thought process for me uh to to kind of build out a path for large creators to come on chain um on on Shelby and and and have a very compelling story of why they should build out this history and um reputation so that they can leverage it in many different ways.

>> Yeah. I this so historically there actually are I've thought about this actually an enormous amount and there are you can look at like YouTube and uh Tik Tok today that actually have these um you know creator uh actually in the beginning days of YouTube they found out they actually gave loans to creators because they're like hey you were you know they kind of found a bunch of people in the early days that were creating content then they drop off and you know they would actually go to them and say hey why you stop creating content it's like dude I'm I'm a barista at Starbucks I got to work uh you know, like, well, if we were to give you $20,000 that get paid back from, you know, the ads that we serve on your platform, would that be interesting to you? You quit your job, go do this

full-time kind of thing. And now, I think the value propos I firmly believe the value proposition for how you're going to get creators to do this is, you know, if you're even like a middling creator, you know, try going to a bank and telling them applying for a mortgage while running a crypto podcast. Good luck. Even if you're making a ton of money, even if you're making a lot of money doing it, >> but if you were to do it on chain and if you were able to I this is this is stolen valor because this idea actually comes to me from uh Jacob from Zora who if you're able to, you know, put the cash flows on chain very close to the coins >> and you're able to drive value, that's great. But having an open model where

you're creating this history and track record, you know, to get funding and a loan, you know, and it's not siloed. That's the key. It's not siloed to Tik Tok or YouTube. So there all this control over you. Yeah. >> Yeah. It's super powerful. Um >> Oh, totally. And you can give selectively give access to different people for your data. And with the new regulatory environment, I think uh there is there is just a lot of appetite to look into, you know, how crypto is adding value and how crypto backed assets uh should be. Uh I think there was a there was a regulation that uh that allows kind of lenders to start considering crypto assets uh in in in your mortgage application which which which was missing and you know all of us have kind of suffered one way or the

other because of that. So I think this is this is the right time to start thinking about this. So yeah, this this is a great idea and I >> I think we should we should uh we should uh brainstorm more after the podcast. >> Yeah, I think so too. I've uh yeah, in case you can't tell, I've thought about this an enormous amount, so you've tickled my funny bone. Um, this is uh this is really exciting. Uh I'm certainly hopeful that, you know, what we've talked about here right at the end definitely gets built. But I'm sure in the interimm there's going to be a lot of really interesting stuff. Maybe, you know, in closing here, could you just talk to us, you know, if you're someone, obviously the audience for this podcast a little bit more crypto oriented. So, you know, obviously, yeah, very exciting

to hear that a big Hollywood studio is building, but you know, maybe uh help make it a little bit more real for let's say I'm building a borrow application on Ethereum or some kind of onchain asset management on Salon or a new launchpad like how could um you know, where are some of the early traction that you're seeing with maybe more cryptonative use case uh use cases maybe to just uh you know >> Yeah, I think there's there's a generally a need for high performance data availability layers uh in itself like uh the ones that exist um you you post your data but they don't get read uh enough. Uh so I think that's that's something what we are seeing. We are seeing the history of uh chain being stored uh on Shelby. There are people are reaching out. Um lot of uh uh also

just understanding your defy strategies and kind of people are trying to automate some of your trading systems and right now they all they have to go by is like the tip of the chain and now there is kind of you know historical data you can process it um um and and can build more automate more onchain trading strategies. Um the real big unlock I'm seeing though is actually happening in decentralized AI. Uh and I'm super stoked about it because there there are many use cases like across the spectrum like what I'm seeing is that u um first of all AI data marketplaces um like if you go to

hugging phase you put your data there there's no monetization model uh for you. you just kind of have to give it out for free. Uh you pay like monthly subscription fee. So hugging face makes money. uh but you don't and and so we have like some creators have reached out to us and say hey we capture this amazing nature videography and there's so much demand I just don't know how to monetize it right and so where should I put it and it's like oh shall be it would be the ideal place because everywhere else what you do end up doing is like you have this massive upload and download which are very slow and so when you take the data out then you start processing it by the time there is a newer version on there you have to kind of manage multiple versions it becomes very messy versus Shelby

because of its performance you can just buy and just point at it and say hey just start training your model using this data. So that's one area. The second one is we are seeing um this agentic AI revolution the the revolution that's coming uh or the promise of agentic AI um all the agents have to manage a lot of context and what's called agent memory and then the context needs to be stored and kind of be portable from one model to another even when you interact like for example when you're interacting with um >> totally sticky with GBT yeah it knows >> absolutely right and so that So making that context portable is another very interesting use case. There is also MCPS

and rag based uh kind of augmented generation or or just something that enhances your model output. You can have API based uh kind of payments and then the data can be stored in a decentralized way. There's a decentralized API and you kind of you know make uh uh make money out of it. And then there are further advanced use cases of privacy preserving AI and decentralized AI. But I I I am personally super stoked about those use cases um uh that that kind of now start coming on chain and so it's Shelby's ability to kind of attract more use cases on chain than just only serving the cryptonative use cases.

>> Very cool. Well, if anyone uh you know web 2 or web 3 or otherwise is listening and wants to learn more uh get in touch with you, the work that you're doing at Shelby, what's the best way to do that? >> Um yeah, go to shelby.xyz. Uh there is also um uh there's also a link if you want to build anything interesting put your idea there. Um and one thing I would say is uh we have um some big announcements coming up with experience which is October 15th and 16th in NYC. Um so uh if you're around if anybody's around you're welcome to join and there is um maybe I can share the code as well. Use the code shelby and see what kind of discount you get on your ticket.

Uh it's it will be worth it. It will be worth checking out. >> Amazing. All right. A little bit of alpha right at the end there, Prenov. We love it. Uh thank you so much. Uh appreciate you coming on and uh yeah, we'll have to do it again soon. >> Thank you for having me. It was a great chat. Take care.