ETHOnline 2026 Finale | Kartik Talwar

Good morning, good afternoon, good evening everyone. My name is Kartik, I am one of the co-founders of ETH global, and I am very happy to welcome you all to ETH online. This is our massive online massive online event of 2026, dedicated to everything happening in the world of Ethereum. You all watch this on eglobal.tv.

This is a live streaming platform that we'll use for all of our asynchronous events, and, um, we'll be able to chat with you here, see all of your questions and comments, and respond to respond to any of your questions and comments, to make it a very interactive, two-way, real-time chat. For those watching this for the first time, this event is organized by ETH global, an organization with a very simple mission. Our goal is to attract thousands of developers to the web3 ecosystem. And we do this by hosting hackathons and summits both in-person and asynchronously. Our next two in-person events will be happening very soon.

Just next week, ETH global Tokyo will take place ETH global Tokyo will take place . We're holding a hackathon at a conference, uh, in just 10 days. And our last Inerson event this year will be held in Mumbai, right during Devcon, from November 5th to 7th, about two months or 50 days from now. But today, for the next hour, we will focus on ETH online. It was a large-scale event.

This is one of the largest ASIC events in our history. Around 1,500 hackers have been working tirelessly over the past two weeks on incredible projects, leveraging the best of D5 payments and artificial intelligence. And, uh, there are people from all over the world here. They are from 89 unique countries, six continents, 434 unique cities. And we're still getting more and more developers who are interested in Web 3, who are coming from the Web 2 world to try different things.

And a quarter of them are completely new to this ecosystem. People are here from literally everywhere, and we've met a lot of great people who are creating with smart with smart contracts and just leveraging what's possible in a new way of thinking about finance. And besides, today we will be giving away over $100,000 in prizes. So you can see who won which prizes. Thank you to our 11 incredible sponsors.

I want to express my huge gratitude to 1 in, Arc, Bazantic, Chainlink, ENS, Adera, Ledger, Privy, Graphisa Foundation, and World. So let's talk about this hackathon. Many of you have participated in our events over the years, and many of you have the opportunity to do so wherever you are in the world. And these events make it easy for anyone to see what's happening without having to physically travel to a city that's nearby or not nearby. And the reason we might do this is because we want to understand how these activities benefit you and what we can do to improve that experience.

So if you can take 30 seconds to fill out this short survey for us, we'll get that feedback. We can read everything here. We understand what we should continue to do, what we shouldn't continue to do, and how we can simplify it for the next thousand people who will attend our events over the next month and make this whole onboarding process incredible for developers . But what I want to talk about is our 1,462 hackers who have been working tirelessly for the past two weeks to create a project they can be proud of. We were absolutely overwhelmed by the interest this event generated, and we submitted 814 projects during it.

This is submitted 814 projects during it. This is definitely the biggest project we've ever had. This broke all our records. This even required us to refactor our code to make the rows display fast enough. And, uh fast enough.

And, uh , we've had a very busy last 72 hours. One of the cool things about these events is that we also understand how the developer ecosystem is changing. And some of the things I want to talk about is that DeFi is coming back to the top. DeFi was the most common type of project submitted. People have also made AIC also made AIC payments and used artificial intelligence to improve blockchain operations.

And wallets overall were the second largest category, with facilitating payments-specific usage being the next best thing. So those are three obvious commonalities that we see. The graph was the most used SDK. The Hideera EVM integration was the second most popular leveraged product, and ARC was the third largest integration for the event. And, to my surprise , ARC has been an incredible partner here, and I've been preparing for their mainnet launch , which is happening today happening today .

So we can show you some cool projects that are ready for the ARC mainnet, and people are actually deploying them directly on the mainnet chains . So, the Ethereum mainnet was the most frequently deployed for these 800 projects. Aera and base take second and third place. But what we want to talk about is some of the best projects that came out of this entire hackathon with over 800 entries. We have eight finalist teams and we'll show you what they created.

They will give a live demonstration and tell you all about their projects. So, for the first time in ETH online history, please join me in congratulating our eight finalists. We have hardware wallets from LeafDIY, Novi Corpus, Carr, Open Bug, Cordon, Petri, Onchain Router, and ETH Arcade. These eight teams will tell you who they are , what they created, and how their demo works. And after that we'll talk about prices.

And we will wrap up this entire hackathon and celebrate the huge number of participants and developers who came together over the course of two weeks. But before we meet the first team, I want to take a moment to congratulate each team working on our project. These slides are getting bigger and bigger for us, and we have to make the font smaller and smaller to fit everyone. And this time it took three more slides than before. So congratulations to every project that applied for this event, and continue to improve on what they are passionate about.

So many people followed the continuity and took on existing projects that they were proud of and intended to improve. And, uh, it became a great way for everyone to really spend the week and ignite their inner designer. Everything we talk about is public talk about is public . So you can go to eglobal.com/showcase and see a full list of what they've built, how they built it, watch demos, see their source code, and learn everything you, uh, need to become a better developer.

And, uh, with that, we're ready for our first showcase of the finalists for today. Let's introduce the onchain router. Hello, I'm alsan. Uh, it's an onchain router onchain tools for AI agents from a single endpoint. Uh, think of it as an app store for onchain tools or a fast API, but the client is an AI agent that pays from its own wallet.

Uh, today ask the agent anything onchain and iterate the script. Uh, find the RPC, guess the AI , run it, hope it's correct, and if it needs a data provider, uh data provider, uh , its accounts, API keys, and subscriptions on top, uh, usually what comes back is a table or an error, not a response. Here's how it works: the agent connects once, sees each instrument with its price, chooses one, and pays the sent x42 from his wallet. The money goes directly to the person who created the tool, without commission. And if there is no response, then the payment does not arrive.

Today we have six tools: lending and management. They all work on a graph using standardized MSari subgraphs. So one tool can ask every lending protocol in the chain the same question and compare them. This one considers liquidation considers liquidation . You can see the answer.

You get a verdict, not a table. Let's actually run some lending races for USDC on Ethereum. I'm going to pay on Hideera Testnet pay on Hideera Testnet . You can see 402. It offers Hideera test network and art testing at the same time.

This is settled. We also see the answer. It says what's wrong, what's missing, and this is a transaction on Hideera Test Net, the gas is sponsored by an intermediary. Um, if the agent, um, chooses arc, the same call is settled in USDC through the Circle gateway. Um, what if there is no answer?

Same idea. USDC is on a leg where no lending market is responding. The tool says it can't give a verdict, and the payment it signs is never calculated. You buy answers, not attempts. Um, same rule for HTTP and MCP.

Um, so far I've been the one pushing. Now let's get me out. Um, this is cloud code with an on-chain code with an on-chain router added as an MCP server. Um, one team, and he has the wallet and all the tools. Let's try.

Now the cloud finds the right tool by the right tool by contacting the router and paying from its own wallet. I'm not touching anything. No key, no approval, no script. Here is the best rate, alternatives, um, what it is , and um, at the bottom is a receipt paid to Hideera from the agent's own wallet with the transaction ID. Um, all six instruments are mine now, but that's certainly not the goal.

Um, anyone with an API can list one. Um list one. Um , you wrap this in one function from our mpm package. Set your price and address, and it's all yours to keep. If your endpoint is already using cixs x42, that's even better.

Just paste the URL and check it. The check reads everything directly from your 402. You can describe your tool and submit it for review. Agents can list instruments in the same way as through MCP. It is available on onchain router.

io. The package is on mpm and I continue to update it after the hackathon. Both networks are running on a test network today. Mainet only needs to change the configuration to change the configuration , and it will appear very soon. Data in the form of a graph, payments via x42.

Provide your agent with onchain tools. Thank you. Incredible. Thank you for this incredible demonstration. I think demonstration.

I think the ability to integrate onchain data like MCPS will be extremely useful extremely useful , and the fact that you showed a real-world demonstration, as well as made a way for everyone to contribute to the tool, can only help everyone else use these services more and more. So I appreciate you continuing to do this after this event, and I hope that, um, it will be improved even more. Okay, we're ready for our second demonstration. Let's get started right away. On behalf of our second team, I would like to introduce OpenBook.

Welcome, General Manager General Manager . General manager, everyone. What if I told you that the people currently running the fastest growing economy on Earth have a blind spot? So big that it will be worth up to $7 billion by the end of the year. This blind spot is bad data.

And the people who run this economy, well, they're not really people. You understand. They are understand. They are artificial intelligence. Today, 1 second of data and 3 seconds of data are sold for the same price.

But this difference is incredibly important for the agent economy. The probability of the forecasted market changes instantly. A drone needs weather and wind data that is fresh in seconds, not minutes, not hours. Data needs an incentive model that forces sellers to compete in offering fresher data, ensuring that buyers get what they pay for. Agency commerce is heading towards 1.

5 trillion per year. And bad data already costs the average large company over 13 million in prediction markets prediction markets . More than half of all activity is either bots or agents. And these agents work with data that has to be fresh, otherwise they will lose real money. The problem is that there is no guarantee that the data is fresh.

So I created a missing primitive for this. Openbook is a data marketplace for AI agents that offer fresh data or your money back. Let's check out what types of different markets are available. For demonstration, I have five that are displayed. We have UNI file exchange.

We have NFT deals. We have odds on sports results with overtime. And we can see that each of them has its own level of price and freshness. Let's use the console use the console to check what sports are available for a recent game. Let's connect to MetaMask.

Good. So what happens is this: the agent essentially uses the OpenBooks MCP server and sets the promise criteria for the age, i.e. agrees to the seller's price. The data should be fresh for 10 seconds, as promised 10 seconds, as promised , and the seller gets paid.

And here they are. We definitely saw the data coming in. Domestic market – 50.4%. Remote market–49.

6% Remote market–49.6% . It took 2.44 seconds. So it's well within the 10- well within the 10- second limit.

Let's try something different. Let's try, uh, give me the price of ETH with an age of 0.00001 seconds. This is literally impossible to do. What happens behind the scenes is that data comes through the graph with the block tag in which it was indexed, and OpenBook compares this to the timestamp of the promise, then pays it out on the due date.

And you could see that it took my poorly written natural human, uh, human language and automatically mapped it to the correct format for the query. And here they are for the query. And here they are . They were rejected because the data was 170 seconds ago, and we wanted to get it within a fraction of a second. Hmm, as we see, they were rejected, as predicted.

Let's move on to the trading floor. Here, agents can explore various forms of data coming from both on-chain and off-chain on-chain and off-chain networks. And sellers have their own identity and reputation built on ENS, where you can see how often they failed to deliver and what their refund rates are . This is extremely important for knowing who to trust. Hmm, and everything is stored on the onchain.

One of the coolest things about OpenBook is that in a single line of configuration, it can turn any of the 15,000 structured subgraphs on the graph into its own monetized marketplace with tradable data, price curves, and freshness guarantees. If there's one thing I want you to take away from my presentation, it's this . Agents will not be trusted to spend funds until the costs are eliminated. And there are guarantees that are provided. It's time to open up the world of data to agents.

It's time for an open book. Thank you. Incredible. This is a great demonstration. And the fact that we can do this in real time and enforce these rules at a contractual level, I think that's why we started building in this space.

So congratulations on doing this live demo and actually showing what it looks like, and I hope others will take advantage of this service as well. Okay, we're ready for demonstration number three. Let's start T. Welcome. Greetings.

Greetings. Hello everyone. Hello everyone. So, we will introduce you to TAR. Hmm will introduce you to TAR.

Hmm , Unis swap before introducing hooks smart introducing hooks smart contract attached to the liquidity pool that runs its own logic before or after the swap. They make the pool programmable. Hook can change commissions and force a rule or take a portion of each swap. But how much hook actually charges is not announced and cannot be directly measured. Unis swap give hooks a way to announce what they are taking.

Out of 1559 deployed hooks, only nine do this. So if you want to know how many hooks are being pulled from the user, you can't rely on what they declare, you have to measure it. And we created, uh, a system with its own environment mechanism to play the swap first with the original oak and then with the empty oak, uh, with 89 by, so that we could measure the difference, uh, in cost between the two, uh cost between the two, uh , two passes of oak. I can show you that with Nvidia, uh, right here in the tokenized version, we see that the cost of one basis point per basis is 0.01%01% difference.

So, this is a really good pass. But how did we arrive at this number? We have 13 tools managed by an artificial intelligence agent that will use each tool as needed for each exchange request. We will use the user environment as well as other oak resources to see if there is a better way to share. We will also consider each pool's liquidity depth, chain trust rating , web , web information, all to make a positive contribution and gather data.

This data will also be available through our MCP to your AI- based coding agent , as well as directly from X 4002 on Edera. But why all this data? Of course, this is to measure oak resources, but it is not an end in itself. We wanted something simple for everyday life. Most of us use a web use a web wallet for everyday online transactions, but we also have a ledger that we trust more.

So how can we combine the two to create something useful without slowing you down? So, if I want to exchange ETH for USDC, my extension is an attacker will capture and send to our registry all the information about the incoming transaction, with all possible values possible values , calculated BPS, all addresses, everything. And when I, if it suits me, I can sign it. This will be added to the metam mask. The is called directly, so it doesn't slow us down in the process.

It's just a security. So now I will exchange on the mainnet. Perfectly. It worked . And, uh, the advantage of this system is that you never pay more than you have to pay.

This is what we created, and this is what we needed to measure the oak. Thank you for your attention and thank you for making it to your presentation. Thank you all. Incredible. Well, great job on creating a nice public dashboard for everyone to watch what's going on behind the scenes, and I hope others can use it too .

Congratulations. Okay, let's move on to demonstration number four. And for that, I would like to introduce ETH Arcade. Why are you bothering me, Kartik? I'm busy playing.

I do n't like you bothering me. But anyway, let me just show you. This was before what trading looked like. I was hoping there were a lot of screens open. I would have to print everything.

I would have to move the mouse and see so many things. So I wanted it to look something like this . So, what did we do? We sat down and did it. We created a physical console heavily inspired by my favorite game, Hill Climb Racing.

And I traded a little and used the latest AI models for game development, which we as a team had never done before. Let me show you a demo. Let me start by putting some USDC in here putting some USDC in here . So, I have my physical, uh, device. I'll show you how it's made.

The winner here will scan the QR code and send some USDC. Let's go. Let's go. Good. Good.

The transaction has been sent here. Oh, D. Hurray. Good. Oh, I started playing.

I can click here or maybe the yellow button maybe the yellow button and then my bet will be placed on that cell. Now, the further I move away, the higher my leverage. Now my bet has failed. So I missed. Now I hope this bet succeeds this bet succeeds .

Oh, I'll make the next bet. Oh, I missed again. Oh my God . This has never happened before, guys. But the house still wins.

Mom, mom, I'm playing. Good, good, good. Hit, hit, hit, hit, guys. Okay, let me continue playing. Let me continue playing.

I want three in a row. I want three in a row, and then I want to show you if it will flare up, but OK. And don't pay attention. Don't pay attention. I, I, I just stop playing.

Good. Now let me first show you the transaction that took place, uh, when I inserted the coin. I hope this is the correct contract I opened. Oh, yes. Yes.

1 minute ago. This was the opening of a transaction, and it was in USDC. Now I played the game. I could win or lose some money. I think I lost, but that's okay.

Now I'll quickly withdraw the money here. Okay, I'm withdrawing money. Oh, and I got paid. And this is my leaderboard. Now let me quickly explain how this all works.

So, here is the graph providing all the data aggregated from all the sub-streams. You can, uh, read this boring part in more detail in the repository, but here I have all the blocks from different chains from the underlying arbitrage, and they're all aggregated into USDC, and that's what I have. Then here is ARC. ARC, uh, I use USDC for payment and payout, and this is ENS. And here you can see, this is the leaderboard.

And okay, I'm the last one. My friends are ahead of me, but that's okay. And withdrawing money. Let's look at the withdrawal transaction. Did it pass?

Yes. Okay it pass? Yes. Okay , loading. Oh, that's slow.

In the meantime, I'd like to practice the game. Look how cool it is to press buttons and play something. This is great. Let's see if it loaded. Oh, unfortunately, no.

Oh, yes. Yes. OK. OK. And you can see the closed transaction here.

This is great. I won't bore you any further, but here's how we put it all together. Join is put it all together. Join is one of our team members who put together most of this, like he connected the screens, the pies, and, uh, we have the battery, we have the switches. And the best thing is that you can do all this yourself at home.

You can order these parts, uh, on various websites, and you can assemble it. Everything you see here is completely open source, and I've posted detailed instructions on how you can put this together yourself in my repository. I'll share the QR code a little later. And the best part is that this was just an example of the game. You can create your own game using this SDK and run it on your own physical device that you make.

And all these, uh, all these, uh, services, like getting a prize, inserting a coin prize, inserting a coin , all in one function call. So your agent will be very good at this. And that's it. You can try, you can try ether arcade.fun.

I have a simulation here, for example, Vinnai created this amazing simulation on the website. So you can play this game, experience it, feel the handle, and keep playing. Perfectly. Here is the QR code. Try it and let me know how it went.

Many thanks to ease global for inviting us. This is Joan. This is Winn. My name is Hussein. Thank you very much.

This is great. Thank you again for creating something incredible. You've been a very consistent developer here in our ecosystem, and, uh our ecosystem, and, uh , I love hardware hacks . So, uh , making it even more interesting and open source is great. So congratulations.

Hopefully we'll see it in person soon. And, uh, congratulations to the rest of your team. Wonderful your team. Wonderful . Okay, we're at the 50% mark.

We have four more demos left. So let's invite our next team. Novi Corpus, welcome. Greetings. Greetings.

Welcome. Thank you, this is Kik Martin. I am very happy and grateful to be able to present to you today. New Corpus. A few words, what is Novi Corpus?

This is a legal entity for your AI agents . By legal entity, I mean the actual Wyoming DAO LLC, the legal entity that we create, and we transfer the operating agreement on the blockchain directly to your agent entity on the blockchain. So this is a real company with USDC treasury with rules that an agent cannot break, and there is a real verified person that we use as a global identifier for everything on the blockchain. We came up with this idea because we want to create the last level of trust in the e-commerce agent ecosystem, where a company can hire an agent without worrying that no one is behind it and if something goes wrong, there is nothing you can do about it . So without a corpus, someone can be held liable for every action an agent takes.

So I'm going to take a look at our website demo. This is, in essence, our interface. So you have an option, which is my agent, if you're already connected. Create an agent, which I will tell you about now. Here you have some information and some steps that define the work of our adaptation agents, starting from the wallet and access keys to the treasury management agreement, etc.

, and of course, a real-time dashboard where we describe all the steps so that you can easily get started. The second step is what you get and what type of usage you want to use with your agent. You can choose from several options. So, there are seven functions. The first of them is key storage.

You can choose to store your agents' keys. You have rules covering all aspects. You have accountability from global ID to code agreements, deployments, etc., all the way to MCP agent connectivity MCP agent connectivity . This is the next step, or MCP, if you want to tell your agent, and your agent can help you go through the whole process of, um, getting an agent entity or forming a company with us.

Uh, you can just go directly to that option, and we also explain why Wyoming, DOLC, why we came up with this idea, and why it's a really, really good use case for the agent e-commerce ecosystem . So now I'm going to tell you about creating an agent. The first step agent. The first step is to connect wallets and create a database. I'm already connected here, but you can connect your wallet directly.

Hmm, you need to create a coordinate access key that links both your wallets and your access key to your agent ID. So, I'm just going to create an access key here, which will essentially take me to the second step, which is a registered person. So, this is the proof of identity that we get from World ID, and it's proof that there's a real, verified person behind each agent, which is really important before the law because this human need is to be accountable for everything that an agent does. So we use World ID for this, and essentially you can do it with an Orb scan or a passport scan in the World App. The next step is a legal entity.

You can create a company by following these steps, or you can use a company you already have. For this demonstration, I'll use this . Here you can choose storage, as I said, the storage we chose, we recommend, is managed noi, which uses a developer-controlled wallet from Circle, which allows us to do gas sponsorship. So, you basically take care of the gas bill. This is agent identification information that you can enter through the MCP or manually.

Here you have our nov corpus agent responsible for accountability. You can set spending limits such as transaction limits, daily limits, time blocks, etc., as well as a list of allowed recipients if you want your agent to only interact with certain addresses. So, we're going to review the agreements we just made. It's something like reviewing operating agreements.

And then we're going to confirm and deploy. So, during deployment, it will provide the agent key. It will also register the onchain ID and we will deploy and connect the contracts directly to the blockchain and bind to the ter key on our testnet . So everything can be checked. If we open this, we will see directly on the ARC scan that we will get a real transaction.

Everything is successful. The same applies to the agent keystore. Then we can call the agent directly the agent directly , for example, from the forest DC. And here we actually get to the new agents that I've already created. Hmm, because it might take a little while for Dooley to send us the company documents.

And this is the important part that I wanted to talk about. This is our interface for the company where you can find the articles of association, operating agreements, and EIN letter that can be downloaded along with all agents created through this company. So we can also connect to the agent and see that the treasury balance is here, the transaction limit is here, and that we have the agent book and all the other, uh agent book and all the other, uh , capabilities. That's all. Well, thank you very much.

Please try Please try . Uh, you can scan this QR scan this QR code and try it directly on our website in arc testnet. You can also join us on Twitter at X and, uh, you can contact me if you have any questions on Telegram. Thank you very much. I hope you enjoyed it, Martin and JB.

Thank you for this amazing demonstration, and, uh, for everyone else, this is one of the most incredible things we've seen, where you actually tied things to the real world and made it actually legally binding into one whole, and the agents are held accountable. So I hope you continue to make it, uh, much bigger, and, uh, congratulations on being one of the continuity winners. Okay, three more demonstrations to go. Let's welcome our next team, Petri. Greetings.

Thank you. Thank you, Kartik. So, today I will introduce my product, Patrick. So Patri is actually an evolution tree of sharing equipment for AI agents, meaning that each upgrade needs to be added to the chain or recorded. So I'll start with some of my statement about the problem.

The first statement about the problem is that no one can measure it. Someone claimed that their skills or tip were better, but there really is no benchmark to prove it. The second statement about the problem—that failure remains private—means that when something happens, another team will also make the same mistake and will need to spend tokens to actually fix it. So the third problem is that no one can check you when someone claims, "Oh, my AI agent can improve by 20%," but maybe it's just some lucky run, who knows? So Patri will actually, um, improve the gear by just improving the rice inside it, and that mimics how bacteria grow.

For example, speed, um, instead of improving the entire gear, the tree will only improve a specific trait within it, allowing the complexity to work better and relatively quickly. The second is numbers. We allow many verifiers to actually come and verify the exchange. So verify the exchange. So , uh, it would be better if we let one person test the trade on their own.

And then - the trade on their own. And then - exchange. Exchange means that we will transfer all the information, uh, exchange all the context between all the characteristics. So they will know which approach is best and which is bad. And then I would like to go to my page.

Here the user will be able to choose the domain model, as well as the complexity. For demonstration, I will choose this cloud code. For example, this is the tree of evolution. There are three types of color nodes here. Blue or purple means it has been accepted and can be promoted as a new parent or a new branch.

Red means it was rejected due to poor performance, which few verifiers have reviewed , and gray means it is still pending review; need more verifiers more verifiers to come and test it. And the next one is this benchmark, so that those who haven't actually hit the benchmark can move forward as new parents, so that, um, this is a tool that would improve their performance day by day if, um, we can hit that benchmark. And then I'll demonstrate how a verifier can come here to verify the exchange. For example, I click on this and it's very simple. They just need to execute the command here.

For command number one, this command number one, this will essentially send information to the model about everything or about the hypothesis evaluation, and the second command is that second command is that it will give the model a certain task and see the answer it gives, and the third command is the eval script command is the eval script . So, since the second command already provides a specific answer, the eval script will check whether the answer is indeed the standard answer or the correct answer correct answer . Yes, we just let it run here, and yes, this is the output, but now all of these outputs are just running in my local model. doesn't mean it works the same as when compared to other, er, other people's local machines. Right?

So here we use hideera to get all the information that is stored inside the header consensus service, and the first command is to generate a key. But now I already have one, so I'll skip that command. I'll copy the second command and let it run. Yes. So, from the result, we see that last time, the parents didn't say he failed this test he failed this test , but for this version, he actually fixed it.

This means that this version of trade is actually good, and we can copy this to check if it is actually on-chain. Yes, we will publish all the information to the header consensus service, and as you can all see here can all see here , it is already verified, and yes, that is all for Patrick, and at the end, Patrick saves the verification result to Hera. So a future agent can know that they can trust what worked, what didn't work, and build on that. And that's all for my presentation. Thank you very much.

Perfectly. Thank you so much for creating this real demonstration and, uh, implementing evolutionary algorithms in AI to improve these models. This is a really cool experiment, and I hope you can make it even better over time. Thank you, Aartic. Okay, two more demonstrations to go.

Let's congratulate our penultimate finalist, Lee Wallet. Welcome, GMGM. Hello everyone, this is Okam from League Wallet, and I have a question for you. What can you build for $15? So, $7 for the development fee, $3 for the screen and four buttons the screen and four buttons , and $2 for the jumpers.

And, of course, the free, open-source Lake Wallet firmware for $0. So you will have your own hardware wallet, a leg wallet to be exact . Your keys, your device, and your control. This is a wallet ecosystem for your feet. We have extensions and a desktop companion for Linux, Mac, and Windows, and of course an Android APK companion.

So for global, uh, continuity, we created Hera, which is designed for issuing stocks and bonds, as well as an ARC-based point of sale and payroll system, and Aqua for transferring positions directly to your hardware wallet. So how it works is when you are connected to a DAP or any mini- DAP or any mini- app, the app, the companion app creates a transaction and shows the offer to the Leg Wallet device without the descriptor string data which is decoded by the Leg Wallet device and when you sign it when you sign it , it is transmitted and executed to the blockchain. We support Firefly Pixie devices. So, if you have one on your computer, you can use it use it . Now let's move on to the demonstration.

We will connect the Leg Wallet device using a USB Leg Wallet device using a USB cable. And as you can see, we get a connection request. So I enter my PIN and press the unlock button on my my companion app. It will take some time while it loads the balances. So, I have my balances and we are on the ARC testnet.

So we're going to test the payslip using ARC. So I'm going to use the example of a payment statement, and it's loaded. So now I can approve transactions in Leg Wallet. So I need to approve the spending limits and then send the salary and tips. So we continue with spending limits.

We sign the transaction. Good. Then we send salaries, since there are 12 people. There are 12 spaces in the transaction , and we are going to send tips to our team members. So here in the diagnostics, I can show you that the transaction, sorry, the transaction is complete the transaction is complete , and it's available to the ARC mainnet that was deployed today.

Now, for demonstration purposes this is a testnet, so we just had a payroll run on the ARC testnet. Now we will move on to the Desnet header so we can deploy. Wait a second; we are going to deploy shares and capital. Let's expand on global ETH stocks. So, let's put two and two together, we create a challenge and send it to the hardware wallet.

Just okay, we... okay, just wait a second. Okay, we create a call and request a device, and we need to sign a transaction, for example, in a hardware wallet. We will receive the request. Just fine.

Good. Uh , I think I just pressed the wrong button. Wait a second. Let's, uh, put a global 2026 2026. We're creating a challenge.

We are querying the device because I deployed another global E. I think it is sending a signal to the ground. Good. Yes, he is sending now. A little strange sending now.

A little strange . Good. This should work. Good. Yes, we just received the request and signed the transaction, and we can test the Hera testnet that we just deployed.

Just give me a second and that's fine. So what's better for deploying and issuing shares than a hardware wallet? We just deployed here, and we can also check it out in the check it out in the widget right here. Let's just scan. We are waiting for the headline.

Good. And we have global 202 right here. Thank you for your attention and support. If you want more information about GitHub wallet and x.com cerxen, stay tuned.

We have new boards and new chips. So we will have a better version soon we will have a better version soon . Thank you. Perfectly. Another equipment demonstration and, uh, another live equipment demonstration.

It's, um , pretty amazing to see that you can do this on a small budget. And I hope you can improve this, um, specification and design and make it available for development by others. Congratulations. Thank you. Okay, we're ready for our last final ETH online demo.

Without further ado, let's congratulate Cordon. Yes. Hello everyone. So , today I'm going to show you our products called Cordets. So if you're, um, a person who uses multiple AI agents to do their daily tasks, or daily tasks, or maybe you're currently building a business and are dependent on something like one core company, maybe then you can use multiple AI agents to, um, help you build your company.

And the problem with managing multiple AI agents is that you can't really measure how much those AI agents can spend for you. So maybe you can try limiting transactions using your wallet. Let's say you have this main wallet, and then you have these two AI agents. And when you set a transaction limit in a wallet, let's say you set it to $20, but what actually happens is that these two agents actually think that these two of those $20s these two of those $20s are actually their budgets. So in the end your two agents will spend about $40 per day.

This is what we do day. This is what we do , this is what we don't want. That's why we're creating this Cordon. Essentially, Cordon is a product we're building to help you manage multiple AI agents , essentially with shared budgets. So how this Cordon works is how your AI agent transacts with any service agent.

So, let's say you 're making a transaction on the Circle agent marketplace, and when a Circle agent tells you, "Hey, that information will cost you about $1," then Cardon will first check, word-checking, "Oh word-checking, "Oh , do we have any budget to buy that information, is it really that important for us to get it?" If so, we'll let the AI ​​agent buy for it, but if not, we'll block it, and you won't have many more accounts in the end. So without further ado, let's move on to consoles. So, the first thing we need to do is connect to your wallets, and second, you will need to create new credentials. So, essentially, the mandates here will be double for all your agents.

So the first thing they want to set is a budget for each window. So, this budget per window is the budget you want to spend on each window. Uh, this budget will reset over time, depending on the windows you choose. It ranges from 1 hour to 7 days, and you can also limit the transaction to a purchase. So, let's say you set one USDC and how much you want to spend, uh, per seller or, uh, per trading platform.

But you have options: you can just set a budget window and leave the transaction limit, uh, blank, because Cordon will handle that if you don't specify one. And the third thing you need to, uh, put in here, and the mandates, is the operator address for your agent, and then you can just sign the mandates, but it will take, uh, more time, so to speed up the process, I've already prepared one account that already has the mandates here. You can see it here. This is the budget for each window, how much money is in the windows, and all that stuff. And the second thing all that stuff.

And the second thing you need to do after setting up new meditations is, of course, register all your agents, because we need to know which agent we are currently managing. So, as you can see, I already have a root agent, the first one, and then I'll have three child agents, and each child agent can actually have another child agent, or maybe you could call it a green child agent. Here's the green child agent and how you can add these agents you can just add a spot agent by specifying the operator address here and also specify the parent share but if you don't want to specify just skip that. So if you want to add any grandchild, you can just open the child agent once and create another agent under it. Yes, this is how you set up agents.

And the second thing you need to do is, of course, fund the bug first, because this will be the wallet that all your agents will take funds from. Yes, of course, you will need to specify it first. So your agent will not essentially be transacting for each agent, but rather will trigger this event. It would take me $1 to purchase that kind of information. Yes, we can just manage this whole case.

Yes, I already sent the case. So the next thing I think is, let's just try your agent. So maybe I'll explain to you a little bit what kind of request I've made here. So the first one, Cardinal Worker 2, is the name of the agent I already registered. And then registered.

And then , uh, what I'm asking here, or the task that I want to accomplish, is that I need, uh, this agent to call this dummy endpoint that they've already written. And this dummy endpoint should cost about $1. But I'll do it, uh, but I've already set up this corner worker so that it can only do transactions of about $0.5. So that means this, uh, task has to be rejected because, yeah, it's just beyond our budget.

And this information is not really, uh, relevant to this, uh, task. So if it gets feedback, you can see the rejected task here, you can also see, uh, public records here, you can see why it was stopped. So yeah, that's how, uh, you can see it. So the next thing, if you consider that oh, this task is actually quite relevant to me, and yes, I still want to keep doing it . So I can just click on this purchase, sign the release, and the task, or the agent, or your agent can continue working on this task.

So yeah, I think, uh, this is how I'm going to show all the demonstrations. I hope this can show you guys how helpful this corner has been because the main reason I created products like this is because I feel pain inside myself. So yeah, hopefully you can use this too. So, I hope this helps you, uh, improve your business with One Minute. So yeah , I think that's it for me.

Thank you for inviting me, Kartik. Perfectly. Well, thank you for this live demonstration, and I think this is something that people absolutely need right now, given that we are increasingly in the world spawning subagents that have their own lifecycles, and, um, I mean, and you definitely want to understand how costs are controlled. So super cool, hopefully people can use this as a built-in service, and I hope we continue to develop this after this event. Okay, let's get straight to the part everyone's been waiting for.

I, of course, want to thank our eight eight finalist teams again. These are sleek wallet finalist teams again. These are sleek wallet , cordon, no corpus, bitry, terin router, open book and ether arcade. Incredible live demonstrations. Everything here is actually unfolded.

You can go to any of these projects on the showcase and play around with them, see the source code, and just try different things. This is the best part of what we're doing now, especially considering that people can use a lot more agents to build software. And before we get to our prize, I want to talk about what it means to bring this community and all of you on-chain. All you had to do was create a hacker package, which is an NFT token on the blockchain and provides access to our faucet and many other developer tools, including free LLM credits for working on this hackathon. But after today, in 30 minutes, you will be able to create other packages that will be based on your overall participation in global events.

So, if you participate in more events or win more prizes, you will be able to unlock new packages with different bonuses. So please go to eglobal.com/packs to create your constructor and pioneer packs. This applies not only to hackers, but also to our volunteers, judges, speakers, and partners. You will receive various bonuses provided to you through our community.

And the last thing that's going to be here And the last thing that's going to be here is E Global Plus, your full access to the ETH Global ecosystem, where we're going to do everything to simplify your understanding of what's happening in this space, which brings together things like access to ETH comp tickets to all of our Pragma events, our packages, more blockchain transactions, all of the additional events that we're partnering with and that our partners are hosting, that you can get dedicated seats from credits from EIM to Uber and Priority Pass, all of which will help you not only attend what we're doing around the world, but also help you get into the right place to build your own projects. All of this is available at global.com/plus. Check this out. And without further ado, let's get to our prizes.

So, I want to start from 1st place. First place for the best Aqua app goes to Solvent. You win $2,500. Cold Cascade takes second place with a prize of 1,500, and Super Position takes third place with a prize of 10,000. Congratulations to these three teams.

In the “For Continuity” category, Aqua Zero wins again and will receive $1,500 for first place, while Smile takes second place and receives $500. Let's talk about Arc. This is a big day for ARC and you can win some great prizes too. T-Rex Capital Market is the best DeFi Market is the best DeFi on-chain application built on ARC and using Privy ARC and using Privy . You win $3,500.

Keep in mind that this depends on the deployment in the main network. So $2,500 of that amount will only be delivered after the Arc mainnet deployment, which is set to go live today. And in the category "Best Ajente Economic App," Open Book receives $3,500. Congratulations to the Open Book team, and similarly, there will be a $2,500 mainnet deployment prize for the best Defi app on the continuity track. ACR wins this prize for $3,000, and $2,500 of that will go towards mainnet deployment, and we want to congratulate all ARC and ARC teams on their move to Mainet.

Next comes Bantic Marget, which took first place for API identification. You get $500. Latch takes second place for $300 and Leg Work takes third place for $200. And for the best recipe that uses global color APIs, not yet takes first place for $500. Mandate app comes in second at $300.

NSA comes in third at $200. And for the continuity track continuity track , Pay or Refuse and Ledger Online will each receive $500 for helping an agent use your package project category. So congratulations to all these teams. We hope you can build on top of the next link in the chain. DFlow and Sovereign are the winners in the "Best Confidential Workflow" category, and you will both receive $8,000.

Aqua Ghost Protocol wins the “ automated liquidation protection” test and you will receive $500. Car number one is our next winner, who will receive $500 for the best Chainlink-based update and a prize in the form of a continuity track. And they are all incredible Chainlink winners. Congratulations to all these teams. Then we have ENS.

Best use of ENSV2 goes to Herit for first place and $1,500. Batas will also receive $1,500 for second place. The verdict for third place is you will receive $1,000. And Capsule takes fourth place and receives $500. And in the " continuity vault" category, Soul Vault wins $500.

For the best integration of ANS VTO into an existing project. So congratulations to all five of these teams. We're going to Hideera. Turnstyle is the best winner in the AI ​​and AI ​​and Aentics categories for Hyera. And Autovo Voyage and Carpool will also share this prize.

So, it will be shared by all three teams. You will all receive $2,000. And for the best open source project that improves Hideera's capabilities, the Henville and Scenario Engine teams will each receive $2,000. Rialto, Namegate, and Facture are the three winners in the Tokenized Everything category, and each of you will receive $2,000, and you will split the $6,000 prize equally. And for the continuity track, Quorum EV and Your Move 2.

0 Quorum EV and Your Move 2.0 are the best continuity projects built on Hideera, and you'll both get $500. So congratulations. Then we have Ledger. Harness–best AI agents and Ledger integration.

You will receive $2,000. Ghostki takes second place with a prize of $1,000, and Ithaca takes third place with a prize of $500. And for expanding the roster on the continuity track, Soul Vault takes first place with a $1,000 prize, and Preflight MCP will be our second winner with a $500 prize. Then we have Privy. Paygate is the best B2B product on Privy.

You will receive $2,500. Swim is the $2,500. Swim is the best financial workflow product with a prize of $2,500. These are private winners. Congratulations to both of these teams.

Then we have Graph. Alpha Market takes first place with a $2,500 prize for the best use of layout standards in Graph Pro products. Sentinel takes second place with a prize of $1,500, and Open Book takes third place with a prize of $8,000. Covenant is the best way to use artificial intelligence tools for graphing for graphing . You will receive $2,500.

Turnstyle takes second place with a prize of $1,500. And third place goes to pre-sign and True Trek, where you will split $1,000 each or receive $1,000 each. So congratulations to these four teams. For the continuity track, Aquaza Zero takes first place with a $2,500 prize for the best use of AI tools with a chart, and Hermes 2 takes second place with a $ 1,500 prize. Then we have Unis Swap Closing Bell Hook Otter and Tear–the best prizes for contributing to the Unis Swap stack for contributing to the Unis Swap stack .

You will receive $1,000 each. So congratulations. And as for the continuity track, first place is IP Token Fund and you will get th00and, and second place is Prime and you will also get $1,000. So congratulations on continuing to add Unis Swap integrations to your existing projects. Then we have World Arc Asset Creators and Automator–the best apps for checking selfies.

You will split $3,500 equally. So, you will all receive $161,166.67 each. And as for the continuity track, Novi, Corpus, Manaki and Orluma will also share equally US$3,500 for the best Asian computer integration or continuity track. These are all our winners.

In addition, all finalists will receive an ETH online 2026 finalist package, which gives them access to ETH Global Plus for the next year, a thousand USD per team member from ETH Global, and a $500 flight credit to the next hackathon. You will also receive a free pro pass to ETHCOM 2027 in June in New York. And we do a lot of events with founders in this area. And you will receive an invitation to meet with them and ask them questions about how to build companies and scale your products. And on top of that, we'll also be giving away $15,000 in developer credits to all of these teams, as well as a hoodie that each participant will receive simply for being a finalist.

This will be delivered to you wherever you are. You will get much higher limits per faucet, so 10 times more test net withdrawals. You will also receive a special X registry that will be delivered to you. And of course, the $500 refund for our next hackathon can be applied to Tokyo next week or DevCon in Mumbai. And, of course, to ECOF next June , which will be the world's largest event in 2027.

So check it all out on global.com, and congratulations to everyone for participating in this event. All information, including the prizes you have won, is now available on your dashboard dashboard . And one quick note: everything we announced is preliminary and not final, which means you still have to follow all the rules, making sure none of the projects have broken any rules regarding code, behavior, how we use Git, keeping everything open source and auditable. And over the next four weeks, we'll be checking all of these things and making sure you've met all the requirements.

And then the prizes will be sent on October 23rd. The same applies to your bets, where bets will be refunded along with prizes until October 23rd. So you can do all of this and you can create your builder and pioneer sets now. So check it out like hackers. And with that, I want to thank everyone for participating in our biggest event, the biggest hackathon biggest hackathon , which was Async ETH online 2026.

We'll see you all in Tokyo next week, those of you who joined us in person, and many more of you at Defcon. And before we finish, I want to thank everyone who helped make this event possible. So thank you to everyone who contributed to the success of this massive event, starting with all of you hackers, all of our incredible judges, our mentors, our speakers, and all of the amazing partners amazing partners who did so much work behind the scenes to give you an incredible experience. And last but not least , to the ETH global team. I just want to say a huge thank you to Ethan, Joanna, Anna Lisa, Alex, Karen, Aubrey, Val, Pascal, Sofia, Nuno, Anna, Emily and Andrew.

So, this officially concludes the last ETH Global event. And then we have Tokyo from the 25th to the 27th, which will be held next Friday, and Eth Global Mumbai, which will be held alongside Devcon in the same location in India. They are all on eglobal.com. You can apply for apply for any of these events.

We hope to see you again in person or asynchronously very soon. And thank you for joining us at ETH online. See you all soon .