# Verifiable Data

- Channel: [New Foundation](https://streameth.org/new_foundation)
- Date: 2024-07-18
- Duration: 38:09
- Topics: blockchain, verifiable data, AI, decentralized, ethereum
- Watch: https://streameth.org/watch/66998677237940f4dfc82062

## Description

Cryptographically secured information ensures authenticity and trust in distributed networks.  Kamesh Elangovan (OpenLedger), Josh Benaron (Irys), Branimir Rakic (OriginTrail) The video features a panel of experts discussing their respective projects and the importance of verifiable data in the blockchain and AI industries. Josh, the founder and CEO of Iris, introduces his project as a scaling solution for Rweave, aimed at increasing transaction capacity to support more projects. Branimir, a founder and CTO of Origin12, talks about their decentralized knowledge graph technology that has applications beyond supply chains and is now implemented across various blockchains. Kamesh, a core contributor to OpenLedger, shares their recent funding success and describes OpenLedger as a decentralized data infrastructure for the AI space inspired by Web2 technologies. The conversation touches on the evolution from Web2 to Web3, the need for verifiable data in AI, and how decentralization can prevent centralized control over AI advancements. The discussion also explores how verifiable data can enable incentives for maintaining up-to-date foundational models in AI, contrasting with proprietary models like ChatGPT. Overall, the panelists emphasize their commitment to enhancing data intelligence through blockchain technology.

## Transcript

Let's start with the introduction. I think maybe I will be the last person to introduce and then like why can't you start? Let's start. So I'm Josh, founder and CEO of Iris. Iris is, I mean, a few ways to look at it, but generally Iris is an L2 for Arweave or a scaling solution for Arweave. two for our weaver or scaling solution for our weave our weave is the permanent storage chain and essentially we act to scale that and um our focus for the last few years has been essentially um taking our way from doing 10 000 transactions a day to doing kind of millions millions a day and onboarding hundreds of hundreds of projects as it kind of relates to verifiable data um i mean storage networks generally kind of tie very heavily into the topic. I'm sure I'll expand on that in the next kind of few minutes. All right, I guess I'm holding it right. So hi, everybody. I'm Brandimir. I'm one of the founders and CTO behind Origin 12, decentralized knowledge graph. What we're building is quite unique. So it's a combination of blockchain and knowledge graph technology. And it's been in use actually quite heavily for many years now. So we launched in 2018 on Ethereum. Actually, previously, we were building the same technology, which is open source and decentralized. Since 2018, we were building it as a closed source centralized startup, if you will. We were successfully deploying it even before. We actually found the need in decentralization through use cases. So we didn't start in Web 3.0, we started in tangible projects in Web 2.2 where we were primarily focused in supply chain. So maybe you've heard of Origin 12 before through all of the implementation in supply chain, but actually it's not just supply chain centric. Rather, knowledge graph technology is actually a kind of symbolic AI. The previous panel, one of the panelists mentioned that actually. So we consider the decentralized knowledge graph a neurosymbolic AI system. And one of the panelists mentioned that actually. So we consider the decentralized knowledge graph a neurosymbolic AI system. And one of the key features it brings is called decentralized retrieval augmented generation, which we can dive into a little bit more, especially when it comes to verifiable data. And yeah, today Origin Trail is on many different chains. So Gnosis, Base, Polkadot, historically was also on more like I said started on Ethereum so still big part of what we do is all on Ethereum and it's very much standard based so we'll get into that as well so happy that you guys are here and I guess please introduce yourself as well. please introduce yourself as well. Yeah, so I'm Kamesh, one of the core contributors of OpenLedger. We recently raised money from Polychain and Borderless. Just we done an announcement one week back. So what we do, we are a decentralized data layer and complete data infrastructure for AI space. So we got inspiration from a scale AI in Web2, which deals with the data processing of data and the end-to-end data pipeline activities. So we want to be an infrastructure which can support the entire data for AI space. So that's all about us. which can support the entire data for AISP. So that's all about us. So this is supposed to be a free-flowing conversation. We don't have a moderator. So I guess each one of us can just go ahead, and I guess the topic is verifiable data. So you can go if you want to throw in anything. Maybe we can start, like, why we get into this data business, right? So what kind of, we always have a, myself I always have inspiration towards, you know, doing more intelligence using the data, even from the web too. In the early 2005, 2006, right, so we tried to build a lot of shopping intelligence and commerce intelligence and everything. So data is always powerful, right, from initial stages of web 2, and now we see web 3, and then now we see AI space. It is full of data. So myself, I always want to be in a data space. So that's the journey. Now we are into data where why we need verifiable data. So whether we really need verifiable database on eigen so he was telling even before eigen he he want to build a you know decent laser database which can be verifiable that you know that is the whole idea so eigen is trying to build more like trying to bring bringing crowd cloud to crypto so that is what the overall ideology of high gain so we are trying to become more like a relational data of high gain that is a way we started and then we getting a lot of traction because generally People don't want to store data in a decentralized database because it is not cost efficient And also it is not very much needed Right especially you can give a ZQ proof for that why you want to store in a decentralized database much needed especially you can give a ZQ proof for that why you want to store in a decentralized database that kind of exceptions we're getting but we're getting a lot of pull from a companies they were trying to they need a you know verifiable data so I feel like there is a clear you know roadmap and use cases for a verifiable data inIS space more than a historical data historical data right now I know lot of enterprise companies able to do great job by storing in a centralized database itself by giving a zk proof or whatever that is the way generally starting I feel this AIS space right now it is going to really take over every industry right now, even from the mobile or signal, whatever we do, it is going to be AI. So we don't want to give a complete power of AI to centralized companies. their own LLMs but they don't want to try to they don't want to train or they don't want to use chat GPT so because they don't want to so give the data to that party company so so they everyone try to take your open models and try to build by just training their own data and try to keep it further firm right that is the way it works so I feel like verifiable data playing a very much important role in terms of what kind of data we really used to build this LLM. So for example, what is the source of the data? Who contributed the data? Whether we can incentivize them? So that is the way we can really grow this foundation models to up to date. Because Charged JPD have an incentive mechanism because everyone is paying for it. But open models will not have an up to date or there is no much synergies or people does not have that kind of incentivizing towards that. So what we are trying to do is to, if we are able to build a verifiable data, so we can build more ecosystem and incentivizing mechanism around that. So that will really drive and build this ecosystem also, I believe. Yeah. The question was why we're in the world of data, right? I mean, my background is in database internals specifically. So working on the inside of databases, basically. And then I kind of got into crypto and then into the storage world. Mostly because, I mean, yes, crypto and then into the storage world um mostly because i mean yes it's there are similarities to the the database world um but also uh fundamentally like my my view is that uh the end game for all of the storage networks and their visions are actually all the same which is that whichever one wins becomes this canonical record of history for humanity and it kind of disrupted disrupted this idea that that history would be only ridden by winners and you know back then i wanted this is like back in like 2020 i wanted to contribute to this in some way so that was kind of the original reason i got into got into data but when it ties into to verifiable data um once you spend some time in the in the world of storage networks you kind of start to realize that uh kind of a couple things one One is that storage alone is not kind of self-sustainable. And you can't really, my view is you can't really build a whole network around that. And then two is that fundamentally the unlock for a lot of the storage networks is actually verifiability. And it's the ability to create, I mean, talking in the abstract sense, to essentially create systems that can make decisions based on data. And not even kind of, I'll come to the AI side in a second, but just thinking about something like a deepened network. Essentially, a lot of these deepened networks generate a bunch of data. They define a bunch of monetization rules around that. So they say, okay, print tokens based on, you know, some work that's been done, some data that's been generated, etc. So verifiable data essentially enables you to move this logic into a smart contract and define some of these more interesting rules on-chain. And then another mental model there is that basically it's expanding what can be done on-chain by enabling verifiable data because you can start to move more and more on-chain. On the AI side, it's very much the same thing, right? It's like, okay, if you think about AI agents, all of those kind of topics, it's really just about things, systems, whatever, generating outputs, storing it, and then making decisions based on those things. It's roughly the way this works. So again, that's very dependent on verifiable data, because if you have two systems talking to each other, you need to have some verifiable way for them to talk to each other. You need to make sure that you provide guarantees that things are stored, etc. You need to define rules and smart contracts based on that. Yes, it applies to Deepin, it applies to AI, it applies to a bunch of different spaces. So I got into the space for one original reason, but over time developed conviction around data as a primitive within Web3. That sounds very interesting. I'm actually curious to dive into a little bit more of asking you a couple of questions on that. When it comes to verifiable data, I have a kind of feeling there's... So there's the pre-AI world, and then there's the AI world, like you said. And then there's this whole concept of what verifiability really... We mean by it, depending on the context. So in the, let's say, pre-AI boom with LLMs, verifiable data was already very important. So we've seen that a lot with enterprises that we've been helping build on top of Origin Trail for many years. For example, U.S. retailers such as Walmart, Target, Home Depot, through an association called SCAN, Supplier Compliance Audit Network, they've been exchanging verifiable security audits on factories based on something called verifiable credentials, data standard, and essentially doing that in a privacy-preserving way through the decentralized knowledge graph. This has been going on for many years. Prior to any AI, what they were really looking for is how they can vet for certain companies all the way in China, which is where basically all of them source their goods. What you buy in Walmart and in Target kind of all comes from the same factories. So for them, it was solving a problem of trust in the supply chain in this way. And not just them, it's been other companies such as Swiss Railways, who are using Origin Trail to track trains and goods and parts of trains. And essentially, they have a distributed network. So the train doesn't go only in Switzerland. It goes all the way to Belgium or France or wherever. And they have an inherent decentralized system that needs to connect. Not the decentralized in the sense that we consider it with all the trust benefits or really primitives, but rather kind of a physically decentralized system that is hard to track. And for them, verifiable data is really important because they drive all of the business decisions based on that. So verifiability was a really big thing even before the boom. And now with the boom, with AI, I'm sure, like you guys all know about the problems of hallucination and things like that, really AI becomes the new UI. People are interacting with more and more AI all the time, even when they don't know about it. And the info they get back, some percentage of that is not correct just for hallucination problems. And the problem might be getting even bigger because as AI keeps consuming all of the public data in existence, essentially starts consuming its own outputs. And this problem is called model collapse. So with model collapse, actually, we've seen posts from guys at XAI, at Grok, actually having to really invest energy into cleaning out data before training their model. And there's public, there's quite a lot of tweets about it, actually. So long story short, the problem is getting bigger, and it's more out there. Verifiable data really is becoming an essential need, in my opinion. And it's not just in our crypto world, where we consider it somehow an input to a smart contract, rather for anybody, really. The problem is so big that, in my opinion, it's actually a perfect fit for Web3. And that's where we need to come and provide the verifiability that our technology can bring. Obviously, what does that mean? Do I put something on chain and that becomes true? Obviously not. See, the AI is already fighting me. So yeah, garbage in, garbage out. We didn't solve it yet just by putting a hash on chain or using a decentralized knowledge graph or storage network or whatever. So we need to get better at it. And I think what we just heard here about is the deep end approach where actually we get to remain owners of our data in the sense of private data so we don't feed it to chat GPT like default, like you said said because it is by definition valuable and find ways to connect it together which is particularly what we're interested in how we can connect different statements from different places and actually increase verifiability by having it somehow sensibly connect in a neurosymbolic, I think that's the way to go. And in terms of verifiability, there's many layers to it, but definitely what we're doing here, all of us, is I think becoming a fundamental layer of the internet on top of the seven usual OSI layers, or at the bottom of it, the way we can observe it in many different ways. But anyway, I'd really like to hear what you guys think about what is verifiable data for you from the perspective of, let's say, garbage in, garbage out problem. And I guess you were the last in cycle, though. Maybe I'll try to put it this way. So what kind of use cases are important using verifiable data? Any real-time use cases that we can think? Without verifiable data, what will happen? Like with verifiable data, what will happen? One simple problem which I see, if it is a separate, take an example of a trading model. You need a trading model to do the trading on behalf of you. a trading model right you need a trading model to do the trading on behalf of you right so you whatever inputs we give it to the trading model needs a very favorable data because it is not a educational platform anymore right if it is educational platform it's fine so we don't want to really get a very favorable data for that right so when it's come to the real time applications which we deal with the finance or addition decision making apps then verifiable data plays a crucial role so what i feel a verifiable data plays a major role in ai space by giving on-chain data verifiable on-chain data to llms on models will create an interesting use cases so which can because it is not like a web 2 web 2 even if you have a ll model it can't log into the bank account it can't transfer the money or it can't do the trading on your stock market you know right there is a lot of challenges on that when it comes to crypto it is completely permissionless right so it can if you give feed the necessary data and then everything is so everything is smart content driven. So we don't want to really stop or do anything. It can do for us. In this case, data plays a crucial role. If it is someone, human is sitting in a midpoint and trying to control every step of the data, then that case is different. Now, if you give this data and then bot do it on his own it won't depend on anyone right so it's very important to make sure that we give a verifiable data for llm models especially some of the use cases like trading or any financial you know use cases let's say if i want to book you know flights to Dubai right if it is less than thousand dollars right so if that is the case right yeah you can't you it can definitely chat GPT is having a lot of drivers where you can you know use the tat party APS and then you can trigger it so you imagine if you have a social consensus and you know four five members are confirming that this is the cost, then it can trigger and do the activities on behalf of you. So it is much safer and better. So those kind of use cases will naturally evolve. Maybe I like to just pick your brain, try to understand what are the use cases you think. Verifiable data is very much important in the AI space. What are the use cases you think verifiable data is very much important in the AI space? So I really categorize verifiable data in two ways. You can probably even abstract it to one. But I see it as one basic provenance-related thing, so just understanding where data comes from, who created things, that kind of information. I see that mostly more as like a human security thing. It's like people just want to be able to verify this information. Similar to how when you're in a browser, you see the little padlock, you see the little lock in the top left corner, which kind of, no one actually knows what it means, but people are happy that it's there. And people know when, people, you know, Chrome flags when it's not there. So I think there's that side. And then the other side that I think is more interesting and can be more of a business opportunity as well is around, again, decision-making. I'm just going to be a broken record. But really, it's all about can two non-connected things be able to talk to each other and verify things permissionlessly and i really think fundamentally from verifiable data the things that mostly matter it's about knowing if things are stored where things are stored and then and and who and probably who published it and it's a mixture of that information um and you know crypto is just a way to facilitate most of most of that so um the the use case is verifiable data. Yes, on the provenance side, it's really about, okay, well, everything on Twitter will probably be posted as a hash on chain or something like that, that more provenance use case. I mean, in the shorter term, Adobe is working on kind of similar provenance standards, but I think on a 10, 15-year time horizon, most of the stuff will move on-chain to some greater or less degree. And then on the verifiable data side, on the actual proof side, which I think is, again, the more interesting side, it's very, very hard to pin down use cases because, again, really the way my mental model for this is if there's any system which is generating tons of data and defining monetization rules based off that, which I could list off a ton, but I use Deepin and AI as examples just because they are the easiest to understand, I think, in the short term. But yeah, any time there's a system that generates data, defines monetization rules roughly based on that, that's where verifiable data becomes really useful. I think that's a very interesting way to put it. I think provenance is definitely a big part of it. It's kind of like almost like a must-have, I think, if we are going to talk about really verifiable data. So who the issuer of some information was and how can you verify that indeed it comes in that shape that it was issued and when and so forth. So like classic things, timestamping, it, more detailed way than just a lock. But essentially it's up that alley, absolutely. For example, the Walmart and Target system that they use to exchange and other retailers. used exchange and other retailers. What's interesting from use case perspective for me is what we mentioned is really the growing adoption of something called retrieval augmented generation. So the ability that obviously LLMs are going to be prevalent everywhere and the ability to help them actually build their results, be that a Gentic activity or just chatbot or something of that sort based on some form of verifiable data. I think it's a very broad set of use cases here. So think of AI assistance, think of, not just that, think of search engines, for example. So traditionally, RAG sort of subsumes the notion of search. This retrieval comes really from retrieving the relevant information from some knowledge-based database, whatever you want to call it. So Google does that for many years, right? And they're using things like Google PageRank and all kinds of things that are under the hood way before AI to sort of sort the relevant content up on the highest of the list. And of course, then they monetize that through ads and then they sell you the top space, which is what we're trying to avoid because that system obviously has shown quite a lot of flaws and it's not the way we want to go with Web3. But ultimately the retrieval, or rather information retrieval as a branch is not the way we want to go with Web3, but ultimately the retrieval, or rather information retrieval as a branch is not going away. So ultimately using those techniques, such as RAG or decentralized RAG, because everybody's building their own chatbot RAG type of thing right now, as you mentioned. So every company, and they don't want to give it to ChatGPT for obvious reasons. So what is a world like where we have a lot of different RAG systems all over the place? Another disconnected system that has a huge potential. So think of it as we're here at this new internet conference, right? So the world pre-modem. Very cool, yeah, you could do things on your computer, but that's like what the world pre-DRAG is with just RAG, because you can only query your knowledge base. And of course, I understand the reasons might be, yeah, I kind of curate my knowledge base, I take care of it. Yeah, that all makes sense. It's kind of obvious. But the potential of network effects is incredible. Just as Modem brought this basically through Metcalfe's law, the ability to connect different computers and build the internet in the first place, and that's why we're here today, I believe that's going to happen with AI and RAG. So we're going to have LLMs supported by verifiable data. Some form of RAG is going to be underneath, but prior to that, the retrieval step is actually going to be underneath, but prior to that, the retrieval step is actually going to be decentralized. So maybe I query, I don't know, Josh's knowledge bank, and under the hood it actually finds that and says this information comes from Josh, signed by Josh's key, and verified by a few other people. And then I can actually see, aha, this statement has certain power. So I can use PageRank to sort if, I don't know, a certain statement is true or not. Truth, obviously, not having an algorithm for it, but having the ability to rank things is very important. And so I think the use cases are actually very wide and the verifiable data is almost becoming a necessity because the more junk is going to get created. By the way, there is a point in time, which we maybe have crossed, we don't know, where there's going to be more LLM-generated data on the web than non-LLM-generated data. And you tell me if this data is verifiable. And then LLM start training on that. So it's an exponential problem and it's hitting us all over the place. We have elections coming in the US. We have all of these major events. And they all basically might be quite endangered by, you know, we had deep fakes before and fake news. What happens now? So fundamentally, I think where we need to go as an industry is we need to offer a solution to AI, really, providing root in the verifiable data. And I see a hand up. So I guess we can take questions. And if you don't have a mic, I'm happy to walk to you and just bring it to you. Can you hear me? I think the guys on the stream can't hear you. I'd like to stand up. Thanks. Go ahead. We can just share this one. Okay, thanks. You're talking a lot about the sort of verifiability of data here, but in a decentralized AI system, in particular a D-RAG system, do you need or do we need some kind of standard, some kind of standard of knowledge? And if we do, how would that look? And what is it that we're using right now to quantify knowledge or value knowledge? The truth is the sort of the mecca that we're all trying to get to, right, is that knowledge is truth, and that's the most valuable knowledge. But if truth is difficult to pin down and sometimes changes, what does a standard or a fitness function of knowledge look like in a decentralized system? I think it's a great question. Oh, okay. I think I'm too loud now. And so layered, I'm going to try to approach it from a couple of elements. We've obviously not solved the problem yet. One part of it is definitely standardization. And what we've discovered building, especially adopting the technology, which is really the key driver for us at Origin Trail, is we even go ahead and we sometimes are too radical and say that we don't think something is technology if it's not being used. So we put it in the hands as many companies and a lot of them really demanded some form of standardization. At the same time as a builder, if you pick something as a standard, it just makes your life easier if you're at that point where you can do it. So for example, Origin Trail is based on a lot of W3C standards, such as RDF, Resource Description Framework, SPARQL, which is the query language for knowledge graphs, which has also a federated querying protocol embedded in it, but also things like Verifiable Credentials, Decentralized Identifiers. I think kind of a bundle of standards like that, and probably not limited to those. I've heard that you guys are also doing something very similar in that direction. It's absolutely necessary because obviously the systems need some way to understand each other. But more importantly, I think they need to match statements. So one great thing about RDF or kind of graphs is they take this... One great thing about RDF or kind of graphs is they take this, the way when Google invented knowledge graphs, they said it's no longer strings, it's about things. So that we don't have like data, but rather we move to the layer of knowledge. So a thing would be a person, a building, or an on-chain account, or a transaction. And those in relations actually form basically the graph. So we have two things with the relation, one or more relations in between. So if we codify in such a symbolic way the world, we can really build this total historical repository, as was previously mentioned, in one place. And if we describe it in such a way then if my system and your system observe the same things at let's say different places let's say that's railways and then in french railways see something and then french railways see the same train two hours later passing i'm not sure two hours is enough but you get my point so you can observe and you can match certain statements um because you have drilled it down to the symbolic level of description. And then obviously identifiers come into play. We need to somehow identify things. Though identity in Web2 world actually has a lot of things that were already built out. So what does truth mean though? On top of that, that's a very tricky question. And I think the important part of what we're building here in the Web3 space is that we empower whoever the one consuming the information is, that they can decide what the truth is. So open algorithms, open data in that sense. And then you have an algorithm that you pick and say, okay, for me, this is my worldview. And from this worldview, I'm going to basically determine the truth, rather than having Google or ChatGPT determined for me. Hope that answers it. But you guys want to pitch in on the same question? So, yeah, so I think the flow of data in AI, right, so we need some kind of standard for it, right? As of now, what we are trying to do is to work with the data labeling company like sapien and then like you know inference companies like rituals and like we are a data layer right so we are trying to form an association or a formation of how to flow the data in a space right so we are trying to working on that we are open to work with people who are you know exploring new stuff on a space trying to have a standard because i'm able to see there is a huge impact uh that happened before by setting up a standard for a security token right so where you know for security token they created a separate uh you know we all know erc20 they created a trc20 or yeah so yes yes T 20 or something like that so so the idea is to create a standard and then like work with everyone to make sure there is a flow of data is going to be on this standard and this is the way that you can verify the data and everything so we are really working on that part and then like you know the important other part is on this verifiable data part I'm coming back to the same topic of yesterday I'm reading a article about this story protocol was talking about the various I know who IP royalties and our other things work right a is all about like a lot of creating of a deep fix quite easy right giving royalty to the person right who is the actual creator of uh take a music director or any celebrity right so he can think on his own and come up with his own video content recently also i heard some songs uh which is created by is became so viral and finally right the shop if you have to uh i mean spotify have to take it down right so the idea there is no any kind of standard as of now so they trying to figure out the problem of you know the future is that IP is a trillion dollar in a couple of trillion dollar market and how to because of the a boom the IP and royalties is going out of the way right now right the creators are not going to get the benefit the same thing happened the initial stages of evolution of web 2 also the lot of pirated data there is no any possibility of tracking and everything right slowly the governance body came and then like you know it became ott and then there is again royalties and everything is flowing again back so i think it's quite interesting things what they do so why i'm telling about story protocol it is all about ips also it's a kind of a data right as a verifiable data is quite important to this space right so so one of the biggest use cases is what i see with the story protocol which working on the verifiable data other than what we talk about the on-chain and historical data that everyone talking about but especially the interesting part is on the ips and songs and videos who are creators celebrities and everybody need royalties around that right that is the way the world working right now so to facilitate that i believe that we store other than historical data this verifiable data is going to play a big role for a space for example if you create a content what kind of data it used to generate this content who contributed that right or we have to give a royalties for it can be able to map everything and incentivize them so if you able to make it happen then this will really create a then we can work with the ai instead of every country is going against to a nowadays side because uh it can generate a stone common control rate by getting this kind of layer because everyone wants to make money right so the idea of celebrities to get they want to get they don't lose the law royalties what they get so everyone want to get that royalty and whatever because they are the creator so how to facilitate that be using this blockchain technology is going to play a big role especially in the verifiability that using this blockchain technology is going to play a big role, especially in the verifiability. So I feel like this is the biggest use case is what I understood in this space. It feels like a bootstrapping problem. How do you verify things that are already not verified? Does that make sense? Yeah, so the idea is to create a standard, as you were telling me. We have to create the standard and try to get all the data. They're creating a layer one where they are trying to get pull all the data and try to put all this information and try to make it verifiable using you know some kind of training data if you imagine the training data is available right so you can verify the training data then they can figure out like royalties who associated with this content and everything so it is in quite initial stage but this is going to be a big challenge because nowadays even like we all know that we have started using celebrity voices for the advertisements right people are not even uh asking permission from any of the celebrities to use their voice or modulations and everything right see initial stages of it but it is going to be a big problem i think if you go into someone who is going to address this problem deeper and deeper, I think they are going to have a billion dollar market cap is what I heard. All right, so do we have any more time or are we one more? Yeah, you wanted to pitch in? You guys definitely asked that. Yeah, got it. Yeah, I mean, any other questions maybe from the audience? We have maybe room for one, I guess. Maybe we can tell you what are the billion-dollar companies that we can build using data? Yeah, go ahead. That is not all we need right so what are the 10 billion dollar companies that we can build on using verifiable data other than this uh historical on-chain data right there i you know hundreds of companies doing the same right no point in keep trying this historical data and trying to you know doing some kind of wallet and addressing the wallet and figuring out the history and everything right we are trying to play the same again and again so when we try to move something from web 2 to web 3 especially this story protocol is quite impressive right they're trying to address a couple of trillion dollar market i think that's why as16c invested on them again i heard that they again going to lead a big round again also I feel like you know the people are trying to figure out a new problems right instead of trying to play only around the on-chain data and verifiable verifiable is important but what kind of data you are going to bring to verifiability yesterday I was talking to in the Polychain capital Olaf and then he was talking about you know when someone is having so much of followers on twitter right if he can adjust do some kind of attestation of it and then try to give some kind of verifiability for him in web3 or you know putting putting it on chain and try to give some kind of privileges for him right so those kind of use cases he was very bullish about and then like you know and try to use chat gpt for everything maybe like the world is going to be chat here after right whatever we need either we will just chat our voice right so in chat if i want to move the asset from arbitrum to optimism and then buy so and so and then like you know sell so and so side if you can able to do everything on chat then we are getting out from the problem of you know because i am trying to buy five million dollar worth of bitcoin all right because began was crashed couple of days back right i'm not able to do it right it take three to four hours for me to get this trade done in a in a decent lay system right i am able to only bought i bought just 300k right it's because of user experience lack of liquidity lack of liquidity happens because of poor user experience i think if you able to fix using this llm by putting all the verifiable data into the llm i think that will create a lot of billion dollar we don't want binance right so if you can just all you need is just type it and get it done what you want right so i think that will create a bigger use case i think i believe that binance will launch the own hellum that just people can just tell hey i want to buy so and so coin and so and so price and sell so on so side it become so easy. Then everybody will become more traders, they can pull in more and more new users, I think. Those kind of use cases we should really think, and that will create a lot of billion dollar companies, I believe. Alright, sounds very cool. I think we're out of time, and we're mega late, so the next panel should start. We have a short break between this one and the next one. Alright, thanks everybody.