From Theory to Practice: Ethereum's ZK Evolution w/ Kartin Wong (Hyper Oracle)

**Speaker A:** Foreign. **Speaker B:** Hello, and welcome back to the Strange Water Podcast. Thank you for joining us for another awesome conversation. We've spent a lot of time on this show talking about ZK cryptography. We've gone deep on implementation and have been far out with how ZK is going to be the backbone of an automated Internet of Things world. And we spent a lot of time thinking about the conceptual purpose of zk. In searching for this conceptual purpose, we keep coming back to the idea that ZK is a technology that allows us to project computation across different computing environments. This is particularly important in the context of blockchains, which can be thought of as computing environments that sacrifice performance for credible neutrality and trustlessness. Before zk, only the most basic computation was possible in the extremely limited blockchain environment. But with zk, we can use high performance modern computers and then send the answer and a proof into the blockchain environment. If the proof verifies, then we can treat the result like it was computed within the trustless blockchain environment itself. And so that's the basic idea of how ZK interacts with blockchain and a glimpse at how it will forever transform how we use it. But the gap between that idea and an integrated automated distributed system is very wide. Fortunately, we have the perfect guest to help us see the future. Carton Wong is the co founder of hyperoacle, a company taking the very latest ZK technology and actually building out the infrastructure that drastically changes what the world computer is capable of. Now, based on the name, of course, you know that Hyper Oracle is capable of delivering prices and other data into a blockchain environment. But just wait until you hear about ZK automation or ZK machine learning, and you'll quickly see that Hyper Oracle is about so much more. It's about bringing modern computation and with it modern capabilities and modern adoption to the world computer. One more thing before we begin. Please do not take financial advice from this or any podcast. Ethereum will change the world one day, but you can easily lose all of your money between now and then. All right, ladies and gentlemen, Carton Wong Karten. Thank you so much for joining us on the Strange Water podcast. **Speaker A:** Yeah, hi, nice to meet you. **Speaker B:** Rhett, of course. So before we get started talking about ZK and about Hyper Oracle, I'm a big believer that the most important part of every conversation are the people in it. So with that being said, can you tell us a little bit about who you are, how you found crypto, and I guess once you found crypto, why did you not run away. **Speaker A:** Yeah, so I'm Carton and I'm the founder of like Hyper Oracle which is like a programmable ZK Oracle protocol. So basically if you empower this on chain Oracle will empower smart contract to perform more computation and have accessibility for more data. So about more computation we can enable smart contract to do statistic computation, AI inferencing and for data we can have Internet data and on chain historical data for smart contract and we use CKP to compress the data and computation and then get the reserve and approve to the smart contract. So that's what we do. I actually got into crypto quite long ago in high school and I co authored like one of the Python library, Python bitcoin mining library with my high school teacher. And then we are kind of dumb back then we don't think it's real money. So what we do actually we didn't use the script, we test the script and got some bitcoin it was like 50 or 25 bitcoin per block like back then 2011. But all we do basically sell the script for 10 USD and then that's like how we do. So our client got really rich but we are kind of like broke still. And then so so after I got some money I go to like us for school and and then I study astronomy and physics back to the college and then after I graduate I just go to Google and go become a software engineer and go to TikTok to become a tech lead on like the tech side for the privacy department and then I come back to grip toe because like obviously the bull market and also like the ZTA always have a plan to you know cut the department of piracy preserve and then because like they just spent bunch budget on the privacy team but they find out American government no actually they don't really care about that actually because like like a political pick for them. So like so just cut the budget no matter how good they do it doesn't matter really. And like so I just like okay then I should go to a sump like industry that actually care about this like cryptography technology and stuff. And then so I go back to crypto and back then 2021 that's when I decided to come back to crypto and then I check how many solution architecture and layer one, layer two like bunch architecture solution and then I find out it's like this like a red market, I don't want to join the dogfight amount of layer two so alternative layer one so that I just like begin to think if there's many layer one, layer two, all of them will need Oracle. Right. Like so, so I want to well position my business and then like I just like see Oracle, it would be a good industry and begin to do research on Oracle and I find out like majority, I would say 99% Oracle. Like basically it's a fork of chainlink. So that I find out like it's like that's why they always like claim themselves a chainlink killer, but they never is because they're just a fork. Right. There's no innovation there, why you win the market, no sense. And then I see like CK technology is getting more and more sophisticated and now is the time that we can use CKP to do like approving on the Oracle industry. So that I see this two cycle is overlapping each other once a business cycle. There's like a blue ocean for Oracle. And also the technology is facing a paradigm shifting so that we will be able to use the new technology to do Oracle. So these two factors together made me put down the decision that I have to do a company now to start a new protocol. I mean not a company because the crypto industry is a protocol now to revolutionize the Oracle industry. **Speaker B:** Yeah. So we'll talk about a lot about what you're building and the opportunity that's so clearly ahead of us. But while we're still in your background. So you said that you were working at Google and TikTok, right? Were you working on cryptography? **Speaker A:** Yeah, kind of. It's like trusted hardware. So basically in Google it was like it's a trusted hardware. Basically you need to do like formal verification and do like TE stuff and CKP stuff for the hardware and for the cloud computation. That's like the job in Google. And later on the TikTok is like Eco trusted computed department and then it's for the TikTok cloud. TikTok have like cloud but they never, just never public it. Just like another big tech company that their own cloud department. That's what I did. **Speaker B:** I guess you've already a little bit answered this, but I would love if you could expand on this. I think for most of us who kind of came to cryptography organically or at least through crypto, we have this magic moment that happened in the last two or three years where we've all realized that zk, while it has application to privacy, is like not really about privacy. It's much more about like being able to. Yeah, succinctness and being able to compress and then project computation in and out of blockchains. I guess my question for you is while you're working on all of this stuff at Google and TikTok is are these paradigms and these visions of what cryptography are for, like really clear? And do you think that these big companies are kind of ahead of us on understanding how ZK is going to. **Speaker A:** Not really. **Speaker B:** No. Okay. Okay. **Speaker A:** Yeah, yeah. The Web2 perspective for ZK, right. Like it's really interesting dynamic in like this Web2 company, they want to do ZK and TE stuff actually because they want to do two things. One is like because normally in Google and TikTok they call it Trusted compute. Why called trusted compute is because like it's a cloud computation. You have a client the client will worry about. Like let's say if your client is like the competitor of one of the Google's product and then you worry about that if you use Google cloud platform to run your computation, Google will steal your stuff and mess up your computation. And so the Google cloud platform need to prove to the client that they are executing a computation privately. Even Google doesn't know what the heck is going on inside the computation. That's the first one to preserve the privacy for the computation job for their client. And because some of them, some of the client will be some bank, right? They have a multiple different bank and you don't know if Google invest on some bank and you would prefer another bank over another and steal some information for another bank, right? So you will want to Google to prove that like they are running executing the computation fairly without snooping your data and computation process. That's the first one privacy. And the second one is that basically they want to tell the user that all their computation are safe and nobody can snoop, can backdoor their program database and stuff like that. Yeah. Another one is for security. It's just like the two main purpose for doing the ZK cryptography te in Web2 company but from long times ago why ZK tech was being invented, right? It's because back to 1980 there was a time that the computer was super slow in some military devices, right? So they will want that if you have a commander center that can have perform more complex compute and send a command to a certain super slow computer and then they will execute and decode the stuff like verify the information quickly. That's the original purpose for doing zk. To prove to a super slow machine that you finish a super complex computation, right. For commander center you got a cluster of computer, you run all the computation. And now we decide to go to this place. And then like you need to send this information with the proof into like a super shitty devices that the soldier have in their own side. So to confirm that it's like not a fault message from the enemy and stuff like that, it's like you need to prove the super slow machine. But like, because like, but this situation kind of got weird because like later the computer, the PC or like the private devices just develop so fast. This technology no longer, there's no super slow computer in the world for a while until the blockchain. Right. The blockchain essentially is like a super slow computer. So that we can use the succinus to complex compute and then have the results and the proof to the blockchain. **Speaker B:** Yeah. I mean there's something super interesting about how Moore's Law really took a lot of the hard interesting problems that we were working working on and just made them irrelevant when we thought forever. But it turns out that like all of that really cool, like I mean the more you know about zk, the more it literally is magic and the more you're like hiding numbers behind curves and like doing all this crazy stuff and it's just, it's like totally. I don't know if weird is the right word, but just interesting how we found ourselves like Moore's law, took care of the computational differential problem and then 30 years later we decided to reintroduce it. **Speaker A:** Yeah, exactly. And also in web2zk, you don't really need to care about the performance that much because the verifier is not a blockchain. The verifier essentially it's like another server, super beefy server. So you don't need to worry about the proof size, you don't have to worry about how much the efficiency of the polynomial and stuff don't need to worry about that. So that's why in Web two they doesn't really lean on the technology. It doesn't really develop anything until crypto is like. **Speaker B:** Yeah. And I just recently had a conversation with someone about kind of the more fuzzy thoughts around ZK and security and open source verifiability. Right? Let's say for 99% of the applications that you and I use on a daily basis on our computer, like we, they're probably open source, we probably could go find the source code, we probably could download it, we probably could build it ourselves and we could check it against the hash that the website put up there and make sure that everything's fine. But like who, who does that? **Speaker A:** Nobody. MD5 Whatever happened. That's why like the supply chain attack is like a thing in the open source software. Like you can just like poison some open source library and then all the library they rely on, that library, they will be wrecked. **Speaker B:** Yeah, yeah. And it just, I think a question that like we all need to really wrestle with is like how much of this like decentralization, permissionless stuff is like us really just like masturbating each other on like, like thematic things that we care about versus things that actually matter. **Speaker A:** Yeah, I would say the ZKP today is really efficient already compared with like 10 years ago. It's not even the same era. So we can of course use this new tag to apply on morphing, not just Oracle and also like bunch new kind of stuff like transaction for modification. A bunch technology can be applied on. **Speaker B:** Yeah, well I think there's always a really important thing to talk about in technology where it's like getting 100x or thousandx. Efficiency improvements aren't just important for the technology, but they make the technology applicable to fundamentally different. Yeah, fundamentally different, exactly. And so I think let's pause this part of the conversation. We'll return to it once we understand more about what Hyper Oracle is and does. But, but to suffice to say it's like very clear that we're now, we have just in the last year or so hit the point where ZK has hit its first thousand x improvement and we can start to perceive applications and we're still have at least one or two or three more thousand x improvements to come until it's real technology but with the background as it is now and especially where it was when you were starting Hyper Oracle. Can you talk a little bit about what, what did you see and like what were the problems that you were seeing and before you even got to the solution being I need to build this protocol, like what were the things there that needed solutions? **Speaker A:** So before I built Hyper Oracle, like and also I feel like smart contract really need something new and not enterprise have a huge attack vector. So before Hyper Oracle I think all the Oracle services is pretty much like fully centralized. There's no way they're not centralized because the traditional Oracle mechanism that basically introduced by chainlink is basically everybody just woe to a smart contract. Right. So if you have 900 people, woe for smart contract. The gas fee is unscalable. The gas fee would be like per minute it will be a few thousand dollars per year, will be like $60 million, something like that. That's like basically it's like unscalable. So which will make all the provider try to make things aggregation happen in their own server. But when the hacker had their own server aggregator server, it basically had every voting. So it's like it's super dangerous to have that. It's like what's the point to using that? Why not just have a server just directly report to a smart contract. Right. So I see the problem is that even the D app still rely on a centralized server provider. So I was thinking about can we put like Oracle system fully on chain, make it become protocol and then to make it safer, faster and cheaper to use. And then so that's like the problem exists before Hyper Oracle and now through like the most sophisticated like ZK technology we have and also the open source library, open source just recently called opml that will enable like any AI model and any on chain historical data and Internet data into the smart contract and do the computation inside the Oracle. But the Oracle is on chain because the Oracle trigger is on the smart contract and they settle in the smart contract. It's like when you start a transaction from UA account, it's like that kind of experience and the whole network is fully decentralized. And then it's like no, we don't even want run our own node. And so that was like let people have like the incentivized mechanism along with like the layer one or layer two so that every time they run this node they will just run the hyper Oracle node together so that they would be able to perform more value to the smart contract. **Speaker B:** So let's talk about like the actual construction here. So we have. Can you actually just walk through all the pieces starting with like what's, what did you need to deploy on chain when actually let's just walk through the lifecycle of like let's say the most basic Oracle process is I want to know the price of eth. Like what, how would I interact with Hyper Oracle? **Speaker A:** Yeah, that is like super simple one. This is like a price fee Oracle usage. So basically you in Hyper Oracle, transitional Oracle. Let me just use transition Oracle as an example. Transition Oracle basically you just call the Oracle provider and then like I say I want this and then make it in your image so that like right, so that I can have that in my subcontract. But in hyperacle it's different. The reason why we call them programmable is because you just program it. You just in ZK graph and just say hey, I want this data from Binance, I want the token price of Ethereum from Binance, from combat called okx and then you just write it, you don't need to call us at all. And then you just write it and then with JavaScript assembly script, super simple. And then you deploy this job, made this job become an NFT and then put it in Ethereum and then you put down a bounty for whoever execute this job for you on chain. After they pass the verification they will be able to have the reward. Let's say I put down like 0.1 eth per 100 block. Whoever can provide me this data per 100 block they would be rewarded 0.1 eth. And that is something mechanism like this. And then you probably as like operator, Oracle operator, you see this job on chain and then you begin to execute the proving from the OKs, Binance and Coinbase and then provide this data with the proof and submit to the verifier smart contract. If it passed the verification it will be rewarded and you do it for 100 block. You reward it like it's 0.1 eth and stuff like that. **Speaker B:** So the, the fundamental idea behind Hyper Oracle is that instead of this kind of the current dynamic where like we have our Oracle contract and that's great if the data that you want already is in it and if not you need to like hope you know somebody important. Yeah, hope you know somebody important enough to get your data in. But what Hyper Oracle does is flip that paradigm on its head and say all Hyper Oracle is doing is providing a like smart contract that, that accepts basically a request with bounties and then it's up to like the Hyper Oracle network to look at those requests. If any individual Hyper Oracle operator says like oh that's actually pretty interesting for me, I'm willing to do that, then they go, they run it. **Speaker A:** It's more like economy economy reason. Right? Because if the cost for you to run it is like cheaper than, way cheaper than your reward, most likely you run it. And also the whole process is on like the layer one, layer two, it's nothing called Hyper Oracle network exists. The whole thing happened on chain, the communication. **Speaker B:** Got it. And I think there's a couple interesting things under that. But really what you're saying is that there's not a Hyper Oracle separate blockchain that also needs to be maintained in parallel to the chains in which you're actually using the Oracle. **Speaker A:** Yeah, you don't need a chain to do that. So you start the Oracle core inside whatever chain you want to settle the result. Let's say if you want to do on Polygon? Just start the call on Polygon, the result be settled on Polygon. If you want to do it on Ethereum, you start a call on Ethereum and then we start settle on Ethereum. **Speaker B:** So I think let's jump from a super simple thing like price feed to something that's super complex and interesting. So I saw a couple examples of I guess like hypothetical ZK D haps that you could build using Hyper Oracle. Could you just pick one of those, explain us to what it is and then I kind of want to talk through like the specifics of how it works. So like the requester, are they actually writing out the lines of code that they want executed or. And then all the lifecycle. So can you pick a. What's an interesting application on Hyper Oracle? **Speaker A:** Yeah, we got a lot of applications but like most recent one is like something called Zkammo. It basically is like I see the potential of this application can change the whole defi dynamic because as we know AMM is one of the most important application in defi, right? But the problem for AMM is that for such a long time the market making efficiency for AMM is quite low, let's say because it needs a lot S times Y equal to K. It's like straight line. It's not even like proactive market making. So every time you use DAX you will need to pay for the alpha for arbitrust trader who is like try to balance the value between the token pair in Uniswap V2 and Uniswap V3. And so you will actually have the worst rate than when you trade in the DAX than when you trade in the centralized exchange. So people will prefer to use centralized exchange because they have proactive market making. You need to pay for the alpha for arbitrage trader in centralized exchange. And so the problem is essentially the problem is because the market making efficiencies for traditional DAX is limited. The reason why it's limited is because traditional smart contract cannot perform complex computation in it. Right? So now we have ZK Oracle. How we can resolve this problem? How about let's move the market making strategy and computation into the ZK Oracle. Now in the smart contract just need to have a verifier and then the verifier will be able to verify the computation from the ZK Oracle. So let's say I can change the market making strategy in the ZK Oracle from X times Y equals K into some like fancy AI model. We can make AI model as a market maker or we can just do a linear regression for market making, other spread computation and stuff like that. And then no matter what way, you will make it way more efficient for the market making and then that so that people would prefer to use DAX more than centralized exchange possibly in the future because you have a good market making. **Speaker B:** And I think a way to visualize this for those of us who aren't actually building this stuff is uniswap versus curve. Right. Curve has a fancier, more nuanced curve of exchange. But the problem with that is it requires the EVM to do more computation and be that the guest fee is higher. Exactly right. And so I think what you're trying to describe by the ZK AMM is you can jam pack it with 10 times as much computation to get more. **Speaker A:** And then how it works. **Speaker B:** Yeah, how does it work? **Speaker A:** Oh yeah. So basically how it works is that you basically put down the model of whatever the competition you want as a market maker as like ZK graph and deploy it as an NFT into the blockchain and your SMART contract basically your Dex smart contract basically run along with the NFT smart contract and then every time you want to have a computation code, let's say somebody want to swap in the current block high and then I submit the request and then like security Oracle compile because they see the bounty in the NFT and then submit the reserve to the verifier to your smart contract. And your smart contract do a verification if it pass the verification and then it will execute the order. Basically it's like that. **Speaker B:** And this is going to be tough without charts to really get this done. But I want to visualize this. So the idea in this example is you have the ZK AMM and that's a pretty simple smart contract that's just holding the assets and has basically a pointer that says like hey, this is the NFT which we all know an NFT is really just a smart contract. So the smart contract is going to have have the code that it wants to have executed and the bounty that it's willing to pay. And so the the ZK AMM will say hey, like let's on every block I want somebody to like this bounty is active. Yeah, all you need to do to win this bounty is perform this computation. And then I guess like the magic of the ZK is that we can verify because we have the NFT smart contract and therefore the code and we have the solution and the ZK proof provided by the Hyper Oracle operator with those three pieces of Information we can zk prove on chain that the computation was run, the result is valid and because of that we can release the bounty to the Hyper Oracle operator. **Speaker A:** Exactly. **Speaker B:** Yeah, yeah, that like that's super cool and it's interesting and I think like the one thing that I'm like a little bit curious about is at least today and we're all hoping for an Ethereum that fixes this, right? But at least today we have this problem of the fundamental difference between EOAS and smart contract accounts. And what's cool about the way you just described this Hyper Oracle system is it's super modular, it's super flexible, you can build whatever you want. But it does sound like you're adding all these different points where external actors need to create external actions. That may be. So can you talk a little bit about when you're comparing Hyper Oracle construction versus a traditional chainlink forked Oracle, how do these systems compare in terms of the amount of external actors that are required and the amount of EOA actions that are needed? **Speaker A:** Yeah, it's free aspect mainly. The first one is like decentralization, right? Because now any Oracle provider they no longer need to go through the aggregated services like channeling redstone and dissimilar type of Oracle services so they can just directly communicate with the smart contract because you no longer need the aggregator server anymore. So you don't need to go through the enterprise Oracle provider first and then aggregate the smart contract. You can just directly send to smart contract. That's the first one to make it more decentralized so everybody can participate and join. And second one is safer. Actually the reason why it's safer because every computation have the proof. So even the Oracle node is hacked, right? Because like the even you hack so like you can your proof didn't match the result. You still cannot pass the verifier for the smart contract. So the worst case in here is not. It's the worst case that you just don't have data and doesn't have the computation result and the smart contract won't lose anything for it it and that's like the security part and the third one is like the programmability. So you no longer because the operator, right, they just download the NFT basically check the NFT data and then run the program and then according to the NFT and then submit the result back to the smart contract. The whole process is like permissionless. So I should phrase that as like permissionlessly programmable. It looks like nobody have used this term permissionlessly programmable. So that anyone can just. If you want to use secure code to perform something for smart contract, you don't need to ask anyone to do it. So you can just write a program and deploy it, put down the reward and people will do it for you. It's like the concept we propose, I call it like proof of Oracle work. It's like proof of oracle work is just like you do the job, generate the proof, but it doesn't generate consensus between the nodes. **Speaker B:** The thing that jumps out to me as like scary about permissionless programmability is you know, just when you're running other people's code, your machine's at risk. Right. And oh yeah, you are. **Speaker A:** Right. **Speaker B:** Yeah. So can you talk a little bit about like what let's say I'm interested in being a hyper Oracle operator. Like what do I need to be worried about just basically running arbitrary code in exchange for, you know, small crypto bounties. **Speaker A:** Yeah, so like small crypto bounty, you just need to compute, do a calculation between your cost and you'll be raw. Right. If it's positive you should do it, if it's negative, just don't do it. So we will provide a calculator because the same logic as a GAS fee. Right. So basically if the reason why you don't need to worry about someone have like a torsion or virus deployed in a smart contract because the GAS fee for running that is quite really high. So we have something like that to in case. So the cost computation is I guess you computation is same thing, it's just like how many bytecode you're executing in the whole workflow. So in case people just write while true and then never end and wreck the computer of the operators machine, we will have that kind of analyzer for our no operator so that we can dodge that kind of like filter out all the problem code from the TK graph. **Speaker B:** Got it. So sorry, let me repeat and tell me if I got this right, that your solution to this is the same solution that basically we have in blockchain or within the EVM which is to simulate the transact or simulate the computation. Like look out how much work that actually costs and then provide that to the operator beforehand to allow them to make kind of a coherent decision. And then it becomes like yeah, like things could change in that moment and like it is what it is. But like the idea is we just try to like front as much of the economics as possible and then let the actors decide. **Speaker A:** Yeah, yeah, yeah. If they see it as negative and also have like A kind of suspicious code but they still run it. It's their own problem. Right. You cannot really stop that. Yeah. **Speaker B:** What about do I need to worry about things like somebody putting in code into a Hyper Oracle bounty and then running that code and suddenly there's something malicious in there and they've gained root access to my machine? **Speaker A:** Yeah, that would be like a security problem into concern to be honest. It's like a good question, really good question. Because I think about this question for a while and right now we don't really have a perfect solution for it. But running the whole VM image inside isolated environment is like our suggestion because you never know what's going on inside. But if you just create a virtual machine, even they got a real assets doesn't really matter. **Speaker B:** So I guess my next question is, with that being said, what kind of computing environment is available to people that are creating Hyper Oracle bounties? Are they writing Python code? Are they writing what kind of code is it? **Speaker A:** They just need to write assembly script. Assembly script basically is JavaScript that compile into awesome binary code. So just JavaScript is the same syntax with JavaScript. It's pretty simple, Every programmer know it basically. **Speaker B:** So like yeah, so you're, you're basically saying if we simplify everything we've talked to, talked about down to you write your programs in JavaScript and then a Hyper Oracle like node will run it for you. That's you've essentially brought JavaScript to the blockchain. **Speaker A:** Yeah, basically every like Ethereum developer know how to use assembly script anyway today because they need to set up the subgraph node right where they use the graph for investor for their own protocol. They need to set up something like that. So I would say majority of senior blockchain developers already know how to set up this kind of thing. **Speaker B:** Got it. And so I think for me we can talk all day about how this enables and gets us to the multichain future, which I think we all believe that the multichain future still has Ethereum as the important chain but and then all the L2s and things outside of that. I think if you have any like interesting insights or any ways that Hyper Oracle is going to solve that problem, I'd love to talk about it. But the other thing Hyper Oracle is like clearly here to solve is the bridge from the EVM into outside environments, whether that's the Internet or real world assets or that kind of thing. So before we get to that, is there anything we should talk about in particular about what Hyper Oracle is really going to do to supercharge the roll up centric roadmap. **Speaker A:** Yeah, Hyperacle is free to use, right? So every roll up basically just come with free organic Oracle services that they don't pay for any subscription fee and then which also use their token as like a nifty gas metric. So I would say every roll up will like that and they can empower them a lot. **Speaker B:** Yeah, cool. Yeah. And I think especially through this Bounty system, what's cool is that it enables anyone to have this functionality without really freaking out about your token and those kind of economic pieces. So big fan of the system you're building. So let's talk about the again, what types of things Hyper Oracle allows in regards to connecting the blockchain to the real world. So I think the big thing is you're adding an Internet connection to the world computer, you are giving connectivity. So can you talk a little bit about either the visions that you guys are having at Hyper Oracle or some of the projects that are already starting to be built on how this technology will bring us out of of casino land and into really supporting the financial system? **Speaker A:** Yeah, so there's a many way, right? So I would definitely say like AI will be the game changing technology here because like when people talk about AI times crypto, it's like kind of no use case and book high bullshit. But in our development and when I think about use case we find out many think that might actually bring us massive adoption, meaningful application. Because right now, just like bringing more data into a smart contract is not enough because you need strong computation, you need a cpu, strong CPU to process this amount of the data. Right. But if your computation ability is quite low, even you have all the data in the world, it's just meaningless for you. And so we also have the computation part in Hyper Oracle that will empower smart contract to do more computation, help them to digest all this data before they receive it so that we can see many interesting applications like AI gcft. Basically that will make the whole creator economy moody to blockchain so that every inference word for the AIGC model will be able to become a financial asset. Because in let's say five to six years from now, majority of the art in the world will be produced by AI. Right. That's just really clear trend. But before when we develop NFT technology, obviously nobody really expect that will happen that quick. So FT is more about creator economy and the artists got the loyalty fee from like the future trading and all kind of stuff. Right. We can actually parallel migrate the whole Business business model from the NFT to the AI. So the AI itself become nft and then all the inferencing or all the artwork that created by the AI model will be also the sub NFT for that AI NFT model. And this kind of like a new innovation. We actually help to boom up the technology for open source AI. Because right now like open source AI does not really have like a sustainable way for their business model. The only way for majority AI company to make money become profitable is that they close source for their model and then provide the API, right? And charge 20 bucks a month, something like that. And it's like basically blocking the knowledge. Got like propagation, right? The knowledge propagation. So we by deploying AI gcnft, they open source the AI model but still make huge amount of profit through the AI model. So it can help the AI industry to evolve faster and quicker and more open to the world. That's like the first one and then the second one is that the lending protocol. Let's say you can basically if somebody put down collateral through some organization, the bank, right? I put down my house as collateral in the bank and I want to land something on chain. I want to get some USDC because I put down the collateral. And then also the model can analyze to help to do the fraud detection to see if this collateral is real. It's like a credit card, right? It's like so you do the fraud detection so that people on chain will be able to land the stuff to another people. Easier to increase the capital efficiency for the world. So that's the changing I'm seeing because for massive adoption for blockchain technology, you need to use blockchain to do more things, right? It's the same logic that like why smartphone and iPhone is like dominating the world now you have more smartphone than people. That's because like you can use phone to do a lot of stuff, right? And then the reason why you can use phone to do a lot stuff is because like your phone have a stronger and stronger CPU and GPU every year. So like so that you'll be able to have it to do many applications that was impossible before. And so as a smart contract, if you want to use smart contract for more stuff, you will need to like, like have smart contract to perform stronger computation and more data access. **Speaker B:** Man, you said a lot of like really interesting stuff there. I think like first and foremost I. So I came to crypto like through the lens of finance and I came here into DeFi first and like for my first like I joined in 2021, and for my first, let's say, year here, I thought, like, the point of all of this was a better financial system. Like, we're going to replace everything. And, you know, like, just to cut to the chase, like, I found zk and like, that's when I realized that this really is the story about distributed computing and not about finance. And if anything, I think, like, while all the finance talk and the money talk has been great for, like, bringing people like me in, I think it, like, really clouds what we're doing here. And so, you know, my big insight was we all can kind of recognize that NFTs were this crazy fad that went from nothing to, like, billions of dollars and then came back down. And maybe bored apes or cryptopunks will be worth money like that again, but we all know that that's not coming back. And that's okay though, right? Because, like, this, it's like a different. **Speaker A:** Cycle of different assets. Exactly. **Speaker B:** And, like, what did we take from it? We took these, like, core primitives, like, NFTs. Like, ERC. There's a few of them. Right. But, like, we took these core primitives and, like, exactly, like you're saying in just the basic structure of Hyper Oracle, like, we're putting them to use. I just love that, like, these bounties and the code that you're running are NFTs. Like, that's exactly how I think that what we took away from the NFT boom really matters without NFTs actually mattering. And so, like, I so am on board with you there. And. And I know, I am so confident that crypto and AI are part of the same story. Like, crypto is about scarcity and it's about provability and identity and, and just slowing things down, while AI is about speeding things up and about, like, abundance and creativity and chaos. And, like, I know that those two. **Speaker A:** I think they both is, like, a complementary, like, relationship. **Speaker B:** Yeah, for sure. And I think on this show we've talked to Jason Morton from East Ezekiel and we've talked to Dan Shore from Modulus Labs. Like, there's. We've had a couple conversations about AI and ZK and how they come together. And again, like, I don't really know yet, but, like, I can totally imagine that, like, maybe the point of Blockchain is just to give the AI that we're building identity and, like, elevate it to the level where it can actually change things in the world. And, like, it's, again, these conversations and exactly what you Described that helped me start to really piece together how we get there. **Speaker A:** Yeah, how we get there, it's fast. We already enable any AI model on Ethereum and any L2, literally any. And then it's just like right now all the piece of the technology is there, it just needs somebody to assemble them together and we are assembling it. So I think next year like they're going to have a boom of AI application on chain and then we will see. **Speaker B:** Okay man, I let you get to the big brain part a little too early but I want to take us back to Hyper Oracle and how it actually works. So one of the things that I'm thinking about in this AI empowered world where at the end of the day we actually do need people to be running the like plugging in these computers and like servicing them and like downloading the software and like opting into the system. So like, like I said, huge fan of the the Bounty system and like how that makes much more sense than like the direct like token based system. But how can you talk through like how you are thinking about bringing Hyper Oracle operators to the table and like building a community of people that like want to participate in this kind of work? **Speaker A:** Yeah, I think it's just like we need to have a really good tooling for the operator so that they can avoid to be attacked and they will avoid malicious software and stuff like that. And then beside that I think the most important is that we made the whole mechanism smoother and people will easier to participate. You can just have a phone and then review what job is happening on chain and then even your phone can do a proof right and that you can contribute computational resources on your phone and earn some money through your phone. So everybody be able to contribute to the network and then work to assemble the world computer. **Speaker B:** Yeah man, again I love the world computer metaphor and what you're talking about really does bring it to life a little bit more. Do you have any concerns both on just the realities of building a a community but just even on a theoretical level, do you have any concerns about centralization at the Hyper Oracle layer? Like imagine a world in which T mobile is the only person willing to do this. Like is that a problem? **Speaker A:** Good question. It's not a problem at all because Hyper Oracle is fully open source and everybody can download the image and run it without permission at all. Hyperaco, it's not a blockchain but it's like a protocol, it's like a way how you communicate with blockchain smart contract so that decentralization definitely is not an option at all for us. Basically the whole design, the no element that is centralized because when you talk about centralization, it's like talking about permission. A wireless operation. Only certain people can do certain thing, right. But in the whole protocol design, everybody can deploy any computation, including maybe dangerous and suspicious computation into mainly AS NFT and then another people, anyone in the world will be able to download any ckgraph NFT to their own computer and then run the computation and submit to any smart contract that they want to submit to. We don't really have the control for the whole process. It's fully on chain. And so that's why it's kind of like on chain Oracle, it's not enterprise Oracle. **Speaker B:** So I guess with all things ZK and all things, I guess blockchain, at the end of the day like you're relying super heavily on the cryptography. And so can you talk a little bit about the choices that you needed to make when you were designing this system? And like I'm always interested in talking to builders about what it's like implementing ZK when ZK is changing so fast. **Speaker A:** Yeah, that's true. ZK does change really fast. Really fast. I remember when I started to do the Hyper Oracle there was like a paper called hyperplonc and that was like a sum check protocol for the ZK protocol. And then it was like deeping at that time. And after six months you got folding and after another six months you got like cyclic folding and all kind of like new tech that's there. But you know what? One thing never changed is like the KDG plunk is always there like from zcash. So that's why we like. The first primary library we use to construct the Hyper Oracle is actually the Halo 2 library which is provided by Ethereum Foundation. It's like completely public good, no commercial license, you can just do anywhere you like with it. And then so that's like the magic. And that's like why open source software always got more success because it just people can do it without any concern. Even probably they slower or even like perform it's not as well as another but because it's like public good. So you just willing to even take the risk that is lower than another library and still use it. Basically we just take Halo 2 from Ethereum foundation and add our own modification with Darknesslab to make it faster and more reliable and more suitable for our use case. So basically we have like a Halo 2 version called hello 2 Hybrid which we developed and Then it's like have like some feature we need when another project might not need and that's what we do. And it's also open source, many people can use it. And also I would like to want to try another library called Plonky. It's basically developed by Polygon team and then Plunky have like it's a hash based scheme. It's different with like a hello to KZG KGG based scheme. But the price for it is like you got bigger proof but your proving time is way faster and we have still working on the experiment on try to have a faster proving scheme and to make the whole secret Oracle system faster. Although right now our benchmark is quite good. It's just like 5 to 7 seconds to do the proof for the ZK Oracle which already shorter than the layer 1 plot time which is 12 seconds. But like for many layer 2 is still not fast enough because for layer 2 right the plot time is like 2 seconds. So you even want it faster. So we still work on like different experiments to see how to make it even shorter. But right now our benchmark are really, really happy for it because it's enough for us at least deploy on the main net for Ethereum. **Speaker B:** So I want to get to latency in a second. But while we're just still on implementing ZK and ZK being an extremely fast changing technology, I think the message that you were trying to get across was that yes, while things are changing it sounds like like because the way that you're using ZK is all you need to do on chain is the verification and maybe you might need to do some smart contract upgrades over time if you like other proving systems. But it sounds like you're able to build this in a way where all of like the infrastructure kind of works regardless of the proving system and then you can kind of like modularly swap in and out like whatever is the most performant, like the best possible thing at the time. **Speaker A:** Yeah. So basically we try to make the system more modular. Right. Because technology is always improving and changing. You don't know, probably tomorrow you have a new scheme that is just outperform everyone like 10,000% faster and then you just have to use it. But you don't want to redo everything. Right. So you better to decouple every code front end and back end together and then made the decoupling really good. Essentially modulus means decouple every component in the code. **Speaker B:** So you brought up latency and I wanted to talk about that. And so I Think your first answer really answers the big question, which is how fast can we do a turnaround on a Hyper Oracle request? And how does that interact, interact with block times and like totally hear you that if today we can beat Mainnet, then we're fine and things are only going to get better from here. But I guess taking a step back and just looking at the big picture of the entire construction in the ecosystem that we're trying to build, something that I've noticed is like when we talk about zk, like we really, really power through all of the nitty gritty, which is, you know, like the amount of round trips needed, the amount of just like overhead needed in, you know, verifications when that wasn't happening before. Like I'm wondering, do you worry that like the constructions that we're building are so. Because they're decentralized and have so many different components that we're always going to have trouble competing with, like just using AWS and then if so like, what does that mean about the systems that we're building? **Speaker A:** Yeah, using AWS is fine in my opinion because it's fast and reliable machine. Right? But we always need to care about the asset mechanism, right? We need to worry about for deciding protocol. You need to think about what if AWS do not work all of a sudden or just become like some evil company, just ban everyone. So as a protocol designer, you need to like, it's just like bitcoin, right? It's fine that every miner in Bitcoin is using aws completely fine, fine. But we just need to think about the migration problem. What if bitcoin is like the AWS is no longer function or they just ban all the bitcoin miner. What will happen? What will happen is that you just need the difficulty of bitcoin mining all of a sudden from six zero down to two zero. And then any laptop computer will be able to run the bitcoin node and then just mine bitcoin out of it. So that's like a migration problem, essentially. So in Hyper Oracle we have the same constraint, like what if all the no operator is AWS and all of them just like, like not doesn't exist anymore. How to handle the migration problem, actually have a plan B for it. So basically like because the bounty is still there, right? So like actually it's easier to get a bounty because if every AWS machine is running, which means like you always have the competition, whoever got the bounty first, who will be able to settle the reward for their, for their, for their competition, right? For their wallet and then but if you don't have competition, which means like no matter how fast or how slow your computer is, you just run the job, you still got the reward for it. So like the migration is like a click is like a second once the AWS is not functioning all like hungry like little computer will be able to go jump into it and like do the computation for it. It's like similar, similar computation process of pow. The migration process could be quite similar. **Speaker B:** So I think what you said is what's important in crypto and distributed computing is not necessarily to worry about like okay, what does the system look like in good times? It's like what do the failure scenarios look like? And do we have a clean glide path to get there? **Speaker A:** Yeah, because that's the secret spice that why Bitcoin is still running for tenure more than 10 years. Right. Like a long time and then they will have a problem at 15 years and then it doesn't have problem even once because the migration design really well. So once the operator no longer operated another people can just operate it with lower price but still got the same reward economy designed to make it migrate easily. We need to think about the asset plan when talking about decentralized problem with aws. **Speaker B:** Yeah and I'm curious what you think about what about to say another thing I'm starting to see a lot of to the point that I think it might solve like the latency and the round trip issue is starting with like the Eigen layer moment and then just watching everything that's come from that it looks. It's starting to feel more so quick. Yeah, it's starting to feel more and more like all of the computers that are running all of these services are going to be one computer and they're going to be the Ethereum node. Or put a different way the Ethereum stakers have computers that are much more powerful than running an Ethereum node and they're going to use that space to be running everything from Hyper Oracle to Eigen DA to even Polygon ZKEVM sequence. **Speaker A:** It makes sense. It makes sense. Yeah. Actually it's a really good design with staking mechanism is one of the best like I would call it a defi design in the space because actually have a utility and actually can have actual usage to pay for the yield. Right. That's a really good design to be honest. But what my take for that is I actually have a dual analyze the node on Ethereum. Actually majority of the Ethereum node is quite weak. They don't even have GPU with it. So which means I went in this restaking mechanism going to AI and relating to most like with staking node won't be able to perform anything and if you can perform like a complex computation already, you won't able normally they don't care to be like if node because the reward is not that high because the reward for eth node is not based on how strong your machine is, it's based on how many ETH you stake in. Right? So like you don't really want to do it just with a super BV computer to become eth node. So majority of the Ethereum node is quite like weak node node. You probably have a CPU, probably have like 30 or 40% space left over but what you can do with it like probably some simple computation. So in our condition you definitely cannot use this kind of node to do ZKP or AI computation. Right. So that's why so far we never really use any restaking in hyper Oracle because the no majority of Ethereum node is not capable to do it at the moment. And the point for Ethereum, right, the point for Ethereum to change from pos pow to pos is that low down the barrier to operate eth node so that more people can participate, more decentralized. That's the point for have an Ethereum 2.0, right. So I don't expect this will work on ZK, Oracle and AI stuff. **Speaker B:** Yeah, I think that's a fair point. I think, you know, there's people like me, I run a node in my fiance's closet and you're right, doesn't have a GPU. Like I spent like a thousand dollars on it two years ago. **Speaker A:** Yeah, it's like a BlackBerry pirate. **Speaker B:** No, yeah, I mean it's fine. It's like an intel nook. Like yeah, it probably has like 50% of capacity left to run stuff and that's definitely, definitely not enough for AI, but it's probably enough to run the calculations for ZK amm, right? And so maybe. **Speaker A:** Yeah, probably. That's a good point. That's a good point. Yeah, it might be and absolutely that's why I value there because they basically utilize the NP space that in the Ethereum node. **Speaker B:** Well and then there's operators like T Mobile or I mean there's so many of them that are huge and actually have extra compute power. I think Eigen Lair is amazing because of the amount of conversation and energy it's put into zk. And I just love that everyone is talking about what they're going to build. I was super into the idea and then when they launched and it was, they accepted all these liquid restaking tokens and it's essentially the protocol is the idea is you don't need to issue shitcoins anymore, come give us good quality collaterals to farm shitcoins. And then now it's all like, so you're hundreds of millions of dollars of liquid staking tokens, which is basically delegated proof of stake now. So the node operators, if they screw up any of the restaking services, it's just like some rando depositor who is going to suffer. Like the whole thing has a little bit falling apart for me. **Speaker A:** Yeah, it's like conflict actually with staking mechanism is kind of conflict with the POS consensus because it's just like making it not secure anymore. But it's still so far so good. But I don't know what's. **Speaker B:** Well, so far so good. All it is is a deposit contract. **Speaker A:** Yeah, exactly. And I think the vitality of the paper talk about it. **Speaker B:** No. So, you know, I love how much energy it's brought into the space and I love how many people have started building really interesting things specifically so that they could be on eigenlayer. **Speaker A:** But it's like a layer project, right? It's like a bunch. I see I can layer like a new partnership every day is like, yeah. **Speaker B:** All right, man, we've been going for an hour. I could keep, I could keep talking all day, but I can see it's dark there. So I will wrap this up. But, but before I let you go, can you please just let the audience know where they can find you, where they can find Hyper Oracle and if they got excited about hearing about all this stuff like where can they learn more and where can they get involved? **Speaker A:** Yeah, you can definitely go to our GitHub. Just go to our website hyperoralco IO, click the GitHub button, you go to our GitHub. We have a bunch of really good open source software for everybody developer as a starting material or you just want to use it to build your smart contract. And also if you go to AI Hypero, you'll be able to see that like in the first time in history that we'll be able to perform large model in Ethereum smart contract. And also right now we support llama2, chatgpt2 and stable diffusion. You can try it all out there. And you can also have me in Twitter like X now, which my handle is like a 0x carton. And then I normally in the Bay Area. I also hang out with my friend in Stanford and Blue Bottle Coffee in Palo Alto all the time. So if you guys see me and say hi and have heard about this broadcast and then yeah, awesome man. **Speaker B:** Well, I was a cardinal too. I graduated 2013 and studied computer science. But anyway, Carton man, this is incredible. Right now there's so much going on in ZK and really essentially what that means is there's a bunch of people building ZK primitives and just saying like we're building tools to abstract away ZK so developers can focus on applications and it's pretty rare that we talk to someone who's actually using ZK to develop something useful. So man, I'm just so energized about Hyper Oracle and about the world that you're enabling and, and it's not often that you can start to see the future and like this conversation is one of those moments. So thank you my friend. Really. **Speaker A:** Yeah, I appreciate it. Appreciate the time.