Building on Ethereum in the AI Era I Kevin Jones

So, congratulations to everyone. Thank you very much for being here. Hmm, I always feel a little nostalgic at ETH Global New York because it was the first hackathon I spoke at about four years ago. So it's really nice to see familiar and new faces. Thank you to everyone who came.

A little about myself. Um, will my slides appear here? My name is Kevin Jones. I've been working in the Ethereum ecosystem for the last six years, give or take, supporting Biddle Giddle, working with Austin Griffith and the Ethereum Foundation. We're just setting up the slides here, but they'll be on the screen in a second.

Actually, Biddle Giddle has been around for quite some time. This is a product created by Austin Griffith to help train developers and support the Ethereum ecosystem. And that's still a very important part of what we do as an Ethereum community, which is —engaging new people. We are now in a new wave where there is AI, and AI gives us many powerful tools. And there are many ways to use AI to really accelerate, right?

And not just for the growth of the industry, but also for you as individuals: you can learn faster, build and prototype faster prototype faster . So today I'll talk a little about the products that came about thanks to Build Guild. We'll do a real demonstration. And we will make sure that you leave here with useful information useful information . Let's check if it works.

Good. Perfectly. Class. If I switch to a better resolution, will everything still be visible? Good.

Super. Everything is ready. OK. So, yes. Build Guild has been around for a long time.

We focus on education, tools, and ecosystem support, and a lot of that has come out of Build Guild. One such project is Speedrun Ethereum. I believe that the Ethereum Speedrun remains a critically important part of Ethereum because it lays the foundation for learning why Ethereum is important and explains the key concepts needed to build good products based on EVM and Solidity. And so based on EVM and Solidity. And so , you know, some of these things that we take for granted, like DEX, right?

using Uniswap or something similar, it's still worth understanding exactly how they work how they work , right? And now many people are coming into this field who need such support. So we'll deal with that a little later. But we also have Scaffold ETH. Scaffold ETH was my path to Ethereum.

That's why he's very close to me. I still use it a lot for prototyping, and it's really useful for hackathon projects when you're just starting out and don't know starting out and don't know what tools to use or how to connect everything. Scaffold connect everything. Scaffold is the same starter kit, right? It contains everything you need.

Later, we'll demo it and dive into why it's important and how you can build on it if you want. We were running an Ethereum on tour. It seems we are still trying to do that. If you know of a university that wants to learn about Ethereum and is nearby, just let us know. We are interested in this.

Additionally, we have many products that emerged from Build Guild, such as ABI Ninja, Address Vision, and these are tools for recovering wallets after hacking and recovering stolen funds. And you know, there were a lot of really cool innovations there. But let's talk about Speedrun Ethereum. The Ethereum Speedrun was, again, my first educational experience in Ethereum, and it has grown significantly since then . We have different modules or tasks that you can go through to learn about different things, such as tokenization.

In this case, we will learn a little about NFTs: what an NFT is, what it really means, and what is the value of creating an NFT. This task is actually very simple. You don't even need to write code. You can just run it and deploy NFTs very easily. Then we have the concept of crowdfunding.

How to gather a group of people to pool funds in a smart contract, having common goals, and make it useful for people. So, you'll learn a little bit about staking and how to contribute value to a contract from many different users and somehow accumulate that value. Then you will learn a little about ERC20 tokens through a token distributor task where we go in-depth and teach you what an ERC20 token actually is, how to make transfers, how to maintain a balance, and you will learn about the approval pattern approval pattern for ERC20 contracts. And then you move on to concepts like dice games , which is more of a way to understand how a blockchain works and that it's a very public deterministic blockchain where randomness can be quite complex. So in this example, you actually have to break a smart break a smart contract that has a dice game, and you have to somehow trick the system by trying to determine the outcome before the dice are rolled.

This is a very good example to understand that randomness probably requires some off-chain component, such as an oracle or something like Chainlink. Again , this teaches you maybe even some limitations of where we are at with Ethereum right now and what you can and ca n't do. And then you get to what I call the “ call the “ aha!” moment—when you finally start building a DEX, learning about liquidity pools, and liquidity pools, and how to balance funds inside a smart inside a smart contract and make sure they’re balanced. This is very important, and I believe that if you finally get to this point and really understand everything up to this point, you will be in a very good position to start building industrial-grade Ethereum applications.

This task really tests your understanding of these concepts. And then we get into the oracle stuff, right? For example, how to transfer data from the real world to the blockchain? How to determine the price of something from multiple sources and you will learn why an oracle is important. Then you start looking into excessive lending.

So, that’s the concept of what DeFi is built on, right? How to maintain a smart contract that allows participants to deposit collateral and then borrow against it in a trustless manner. And you'll learn a little bit about how you can create incentives for people participating in a smart contract to maintain its integrity and maintain a balance between what is given as credit and what is actually in the contract. You will also learn a lot about how DeFi is built . This is actually very, very interesting.

Then you move on to stablecoins; There are different types of stablecoins, but we teach a little about overcollateralized stablecoins. This is kind of an introduction where you learn how smart contracts store large amounts of value, and you probably want to allow people to deposit something and withdraw something else, but maybe you want what you deposit to be a stablecoin, right? But how do you actually maintain the price of a stablecoin? This touches on some of these concepts. So this is really really cool stuff.

We also have a prediction markets task that teaches you the idea of ​​creating a market for different outcomes, you know? You understand that if you want to run a prediction market or have a market for a yes or yes or no outcome, then you sell yes tokens and no tokens. The prices of these tokens change depending on the potential outcome and who votes “for” or “ who votes “for” or “ against,” and ultimately, whatever the outcome, all “no” tokens may depreciate in value while all “yes” tokens will gain in value. You will learn how prediction markets work. Very, very cool.

Then you move on to ZK, which I think is a very important concept that people are sometimes afraid to learn. ZK afraid to learn. ZK is “space math,” something high-level and confusing, but this course simplifies everything and explains what zero-disclosure is zero-disclosure is , how to use zero- disclosure contracts for network voting, and more. Additionally, we have a whole “ tech tree” being created so you can explore different directions of Ethereum, even after you’ve gone through all of this. We also hold “capture the flag” competitions where we educate people about vulnerabilities in contracts.

This is really exciting if you are involved in the security aspects. We usually hold such events at major Ethereum conferences like DevCon or Dev Connect. So maybe we'll do one in Mumbai if anyone is going to be there. By the way, who's going to Mumbai? going to Mumbai?

Anyone? Okay, brave souls in the room. It's a trap, but you should do it. Um, okay do it. Um, okay .

So, check out the Ethereum Speedrun. This is really a very good place to just start learning. And I think a lot of these tasks have been changed so that you can just come here, connect your wallet. Yes, let's see. If we connect the wallet to the platform, we can register as a developer; This is a brand new wallet, so I haven't registered yet, but I have now.

And you get this little profile right here. And yes, there are details written there about everything you need to know. Many of them have videos that will teach you. You can watch the video. It indicates approximately how long it will take.

There are useful links there, and you can see how many people have already completed the task. So it's still a very important thing to study. I thought they added some AI features , but apparently...oh, right. AI-guided learning mode, which is pretty cool.

You can download this and by running this command you will learn through the command line how to complete the task, and everyone is using AI these days, right? And I think AI is also very useful for learning, right? I think we are learning we are learning . This is one of those things where our brain works in a pretty interesting way. With “ With “ vibecoding,” you start learning things you never would have learned before, even though it does most of the work for you work for you .

So this is a great way to do it. So, check out the Ethereum Speedrun. Another thing Ethereum Speedrun. Another thing I wanted to talk about is X42. If anyone hasn't followed X42 yet, it's a great hackathon theme.

Because that's really the limit of what Ethereum is and what users need, which is commerce. Um, and it's also a very good connection to traditional infrastructure. So, for those who don't know what XR2 is—it's an extension of the HTTP 402 status code. Well, the same status code that you've all seen, the 404 error, right? 404 Not Found, and then a small Nginx logo small Nginx logo .

Um, it's been around for a long time, but 402 has also been around for a long time, and you probably haven't seen it just browsing the internet because it hasn't been widely implemented. But Coinbase decided to take 402 and reimagine its use in this new X42 protocol. And that just means that you can respond through the API with additional data that says: if you want to access this API, you have to make a payment here, on this network, with this asset, in this amount, and with all the rules. So this is kind of a way to restrict access to any API through tokens, right? The idea is that we no longer need API keys.

We can simply use Ethereum use Ethereum . Well, we can use...it's more than just Ethereum, but of course we're at an Ethereum conference, so you'll use an Ethereum address and be able to make a payment via X42 to access this resource, right? So if you have an idea for something like a traditional Web traditional Web 2.0 application and you want to bring it online, X42 will give you that opportunity, right?

So, there are many different ways to work with the X42. I also work at Edge and Node. I don't know if anyone is following Edge and Node, but we are the original creators of The Graph protocol, and we have a wallet Graph protocol, and we have a wallet that allows you to make payments via X42, accept them, and also make payments via X42 with certain safeguards. So this can also be interesting for development, because it simplifies the process a bit . We have an SDK there, but there are many other SDKs for working with X42.

You can find many options, but this might be a good choice for you to check out. Um, the next thing I want to talk about is ETH skills. So if you don't have it in your bookmarks, it definitely should be there. Um, you have to install it, if you have an editor, you can install it in Cloud Code, Cursor, or Codeium. And you can see this setting right here, which will tell you exactly how to do this for each editor that you use.

And this, well , this is a really cool skill because it contains all the best practices for working with Ethereum, which will save you a lot of time. If you're launching a project and want to minimize errors or strive to follow best practices, this is a great starting point. And since it's open source, there's an option to submit a pull request. So, if you notice something wrong, please let us know. The nice thing is that quite a few people support it, in my opinion .

The project already has 134 commits. Austin Griffith and Open Clawbot are constantly working on this. You can see it right here see it right here . So the project is constantly evolving. But at the same time, it is quite stable.

As you can see stable. As you can see , there haven't been that many commits lately. But we are constantly trying to improve it. And that's really cool. It also covers topics such as X42.

OK. Yes, ERC 804. Actually, that would be worth discussing too. ERC 80004. Oh, I can't write.

Uh-huh. So, ERC 80004 is another thing to think about during a hackathon. ERC 80004 during a hackathon. ERC 80004 is, you could say, a registry for AI agents registry for AI agents . Well, raise your hands.

Who here is going to create an AI agent at this hackathon? Maybe some of you. Wow, I expected more. Okay, maybe it will be a traditional mix of DeFi traditional mix of DeFi protocols and AI, but if you are going to use AI and build an agent, you should register with ERC 80004, mostly to get noticed. This is a registry, and its main purpose is that you as a human will be creating AI creating AI agents, working with on-chain agents, and you will want them to have a long-term reputation, right?

You will want people to know that this agent is not a fraudulent contract or anything harmful. And you start to build that reputation, and over time it becomes extremely valuable. But since AI agents But since AI agents are just software, you can register your AI agent according to the ERC-80004 standard and it will be available through this registry. One really cool project is Agent0. Hmm, I'm trying to remember, I think it's this one.

It seems to be agentz0ero.dev. Let's take a look. Is this it? No, that's not it .

Uh, agent0 dev 804. I keep forgetting their domain. I remember him being very strange. 804. I can't find it now.

In any case, Agent0 has an SDK that simplifies registration via ERC- registration via ERC- 80004. So you might want to check that out. Hmm, I'll try to find it and post it later it and post it later . I don't know why I can't find it here. There is also a subgraph, and since I work at Edge and Node, I do a lot of work on The Graph protocol.

The cool thing is that you can connect a frontend to this subgraph and very quickly find all the agents that are currently in the registry. You can just go here, add a subgraph—it's a GraphQL API, so you can get all the agent data and filter it by the criteria you need. There's metadata, agent types, and you can use those parameters for filtering. There seems to be a reference to a subgraph here; yes, this is the same subgraph that has already been expanded. And I think if I go to thegraph.

com... to thegraph.com... Let's see. It is called agent0.

You can go to applications, open explorer and try to find Agent0. Let's see. Oh, really? Did they change the name? Oh, I can't find it.

OK. Well, I'll find it for you. It's in their documentation. I just don't know their website don't know their website . Does anyone know what their website is?

Yes. OK. I do n't know. Actually, he should be here. Um, sorry about that.

Anyway, check out Agent0. This is another very useful tool. Okay, let's move on to Scaffold ETH. Um, well, before we get to Scaffold ETH, I always like to show who we have here who is new to programming, and don't be shy. Okay, so when you come in, especially if you're new to programming, we have a kind of "vibe have a kind of "vibe coding," right?

And we have this concept of, well, call it what you want, prompt you want, prompt engineering, if you will. But there are so many different tools, it's almost hard to know what to use, you know, should I use Cursor? Should I use Claude Code? Do I need Claude Desktop? I use many different tools.

So I just want to quickly show you my approach, how I do it. So that this can be a guide for you and just give you some ideas. I always start ideas with Claude Desktop. So I always go into Claude Desktop, go in here and set up, you know, a spending level that I'm comfortable with. I usually use Opus 3.

5 Sonnet or 3.7, but I turn on the meditation mode, right here. Reflection mode, and I usually set it to high. And I try to provide as much context as possible in Claude Code, or rather Claude Desktop, about what I want to create. Yes?

So if I say, "Okay, I want to build this, this, and this with links, lots of detailed links," I just pour everything that's in my head in here and tell it, "Hey here and tell it, "Hey , I want to build this on Ethereum," or "I want to build this as a smart this as a smart contract," or "I want to"—you know, you have some idea of ​​what you have some idea of ​​what you want to build. And then at the very end I ask: give me a prompt for Cursor or Claude Code to build this. That's exactly what I'm saying. And then it gives me the exact prompt for what I want to create. So, let's just look at an example.

Let's see. What did I try to do before? I was trying to do a POC with...God, where was that? Ahem.

I'm trying to remember which one. Well, we can actually do it quickly right now. Ahem. Let's see. Oh, it was Bill Go's seminar.

This is my seminar. So, one more point: right here you see, I have all these pinned messages. I just chat about little things I'm working on so I can always come back to them with all the context and not lose anything. I just pin them and rename them so I can understand what's what. Well, let's put it this way.

I was creating the “Red Claw Team” feature for OneClaw—a project I’m working on. I could just ask: " just ask: " create a query for Cloud Code to develop this Code to develop this " and it will try to use best practices to generate a query that you then take to Cloud Code. Cloud Desktop is actually communicating with a virtual machine that is not on your computer at all. So at some point you need to start working locally. And here's what it will do: it will again help you generate ideas and then do something concrete.

So I go to Cloud Code and actually build what I want right there. And here's why I like it: you see it build a whole query for me, give me the best advice, and I can look at it and say, "Oh, maybe that's not exactly what I wanted," and tweak it a little bit. It's kind of a way to make a plan with phases and stages of integration, and it again applies a lot of best practices. And then maybe you'll read this and think, "Yeah, that's still not quite what I need." So then I come in and actually run Cloud Code.

So, I'm using Cloud Desktop to generate this query. Let me create a directory called "test", go there, and I usually run Cloud Code with the permission skip option. This is at your own risk, but I've gotten to the point where my computer is sufficiently isolated, I don't store any private keys on it. It's quite convenient for me. If you're worried, use a virtual machine in the cloud to generate ideas, then just give that machine access to GitHub.

But this will only annoy you because you will constantly have to click "yes" or "allow "yes" or "allow ". So, that would open it up in a more open way, right open way, right ? And then we can just come here and take all this hint take all this hint , right? This will create a GitHub app for the Redclaw team, and we can read what exactly it will do and move on. That's a huge clue, right?

And we can just go here, paste it in, and then let it run and create everything for you. Then you can go to lunch and come back when everything is ready. So I always start with Cloud Desktop, or, sorry, Cloud Code, do my thing, and when I'm done and have a starting point, I exit Cloud Code. I really only use Cloud Code for automation or initial setup, and then use Cursor. And this is mainly because with Cursor, especially if you are new to programming, you will probably want to at least see the differences in the code as you make changes as you make changes .

And I personally prefer to look at the files and understand what's going on. So it's more a matter of preference. You can continue working in Cloud Code if that's your style, but keep in mind that it can be difficult to keep track of all the changes that AI makes, right? So it's really up to you. So, yeah, start with Cloud Desktop, essentially setting up Cloud Code, and then download Cursor.

Okay, we'll show the Cursor in a second. So why don't we do that? Why don't we start with Scaffold? Who here has used Scaffold before? Several people.

Perfectly. Very cool. For those who haven't used it yet, the reason I started using Scaffold is using Scaffold is because, again, there were a lot of different tools out there. This is similar to how we work with AI now. There are many different ways to use AI, and each has its own small advantages.

But Scaffold makes a lot of decisions for you. But it's also a bit flexible. So, it gives you some freedom of action, but essentially creates a project for you that already has an interface. There is already a smart contract there. There's already a chain set up for you.

There, the interface to the smart contract is automatically connected. There's already AI out there, and I'm talking a lot of words now, but it does a lot of the work for you and just makes things easier. So let's just do it and show you how useful it really is. So we can say npx create-eth@latest like this, right? And let's do it in this demo do it in this demo folder that I have.

Okay, can you see my screen? Good. Yes, that's better. So, we're going to run this command. So, npx create-eth.

So, create-eth is a package, and we'll just take the latest version. Okay, press Enter and it will ask if you want to install this package. We say "yes". It will simply download the package and then start interactively asking you what to name the project, what you want to use, and so on. So, we'll call this NYC.

And it will ask us which framework we want to use for Solidity. We can use Hardhat or Foundry. Honestly or Foundry. Honestly , it doesn't matter because with AI coding it's not that important now. I used to use Hardhat a lot.

Then I started using Foundry. I'll use Hardhat like in the old days. Oh, such a directory already exists. I need to delete it. Uh, I thought I deleted them earlier, but...

Sorry, I was just checking to make sure everything was working earlier. Okay, let's try again. So, create-eth@latest. Let's call it NYC, shall we? And he will ask what framework.

Let's choose Hardhat. And that's almost all. Very easy. Two commands you type and it will create an environment for us . It will install all the packages and get everything ready to go.

Now, when you use scaffold-eth, you can think of it as if you are virtualizing the blockchain, and you are also virtualizing the interface, and you need another terminal to enter commands. So you need to have three windows open. I'll just open these three windows using tmux. But you can do this simply by opening new terminals. Of course?

So, it is not necessary to split the window. It's just more convenient for me to show it that way. We will go to the demo folder will go to the demo folder , and then to the New York City folder. And we'll do the same here. Demo folder, New York City folder.

It is still being installed, but it is almost ready. So let's open this in Cursor. I already installed Cursor. Therefore, I will use Cursor. The package installation will now complete , just wait a moment.

OK. Who here has never used Cursor? Several people. Of course. Let me briefly talk about Cursor, because it's important to understand why I use it: it has a sidebar where it's convenient to browse packages.

This is essentially a fork of VS Code, and there's a chat window. You can hide it if needed. There is a choice of model here. If you want to save money— want to save money— although "save money" isn't exactly the right word. If you want to be economical, use auto mode.

And if you want to take a risk, choose Claude 3.5. This is the most expensive model. The choice is yours. I usually use version 4.

6, it works very well. I think that's enough. I rarely use Claude 3.5, except for security audits. So today I'll stay at 4.

6 for the sake of stable quality. The packages are still being installed, let's wait a little longer. Let's look at the structure. You can choose settings, but there's also this. See, it says "Agent" here?

You can customize the type of interaction. Agent mode is when you make changes to the code. Plan mode—when you are planning a major change or a new feature change or a new feature . This is the moment when you take the context from Claude Code or Claude Desktop and paste it here. You want to clearly understand what exactly will be done.

He will make a plan, and he will do it better than a regular Claude chat, because he will scan the entire catalog and make sure that Claude's advice is really relevant. So think of it this way: you plan a big task and then you execute it. If you are fixing a bug, then debug mode is good. I never, ever use it. Multitasking—I don't use it.

Ask mode is great if you want to make sure nothing changes and just want to ask a question or chat. Ask mode is also a good option. But for the most part, to be honest , I work in agent mode. Oh, it seems to take some time. There was no need to delete the previous one.

Well , it's okay. So, scaffold eth uses the yarn package manager. This means that it installs all the necessary dependencies necessary dependencies , like SDKs and things like that. This is a monorepository, which means there are two large packages. There is a hard hat package and a next.

js package. The hard hat contains all your smart all your smart contracts, and the next.js folder contains the frontend. If we go to hard hat and then to contracts, you will see the initial smart contract. This is a very simple contract.

Something like a "Hello World" contract. So, if you want to create a new contract or modify an existing one, this is the place to do it. The deployment script is right below it. It's also important that we have this cursor, sorry, we have this...claude file, which essentially describes how to view the structure of the code.

It's funny that Carlos is a grumbling code reviewer. We also have an agent section, where there are certain rules, and also an MCP. This is a very cool thing. This MCP server automatically searches for documentation for you and we add it automatically. We've added a lot of AI-friendly stuff here.

Also, as you can see, there are skills here, for example, X42. We have the skills for subgraph and login via Ethereum. So, there are certain skills that we create for you automatically. It surprises me that ETH skills are actually lacking here. Maybe they are just separated.

In separated. In any case, whenever you want to change the contract, go to the hard hat folder. The application and frontend are located in the app folder, it is an X.js application that uses the app router. So, a slash and whatever comes after it.

For example, slash block, explore, debug. It seems to be over. Yes, it's finally over. So let's go over here and I'll expand this to full screen so you can see it. To configure Scaffold-ETH, you need to run these three commands.

You need to run yarn chain to start the blockchain, yarn start to start the frontend, and then a third command to deploy. So you just need to run these commands in each window. So, yarn chain to launch the blockchain, yarn start for the frontend, and in the last window—yarn deploy. As I said, it comes with an initial contract. So he will deploy this contract for you.

It will use the specified version of Solidity and load it. In this case, we updated Solidity to version 8.30 and it will download it. The internet is a little slow right now . There you go.

And he will create it. Ahem. Here it is. Deployed. The contract has been deployed.

And Scaffold-ETH listens on port 3000. We can simply copy this copy this URL. And let's move on to the browser. Let's open it quickly. The first run of the compilation takes a little while, but once everything is ready, it will load.

That's all. OK. Another cool feature of Scaffold-ETH is the presence of a "burner wallet" . If we look in the corner, we 'll see this wallet, it's already connected. We didn't need to connect MetaMask.

It's like, well, consider it an ephemeral wallet. It is for testing purposes only. You can reveal the private key if you want. You can disconnect and use your wallet if you want, but the burner wallet here is better because it doesn't have the "nonce "nonce " issues like MetaMask does. So try to use a burner wallet, especially when running locally.

If we move on to setting up contracts, we will see all our smart contracts. So smart contracts. So , all contract information is automatically read. This is truly an extremely powerful tool. This sounds simple, but it's actually cool because with Ethereum contracts, if you don't know, an ABI is also created at compile time.

You can think of this as a map of your contract. Your frontend needs to know what data types are used in your contract. She needs to know how to read data from the blockchain. She needs to know how to write data to the blockchain. So, Scaffold-ETH automatically pulls this up for you and creates the interface, and also automatically generates types for Next.

js, i.e. for TypeScript. This means that you can easily build a frontend using these objects, i.e.

this data objects, i.e. this data . This is a really great feature of Scaffold-ETH. So, let's try to do something.

Let's try to change the contract. Let 's not do anything crazy, but maybe create a smart contract that mints an NFT for a basketball game or something like that. I don't know. New York Knicks. Yes.

OK. We'll try something simple, just so you can see the workflow, you know? I'm not going to reinvent the wheel and build the next Uniswap or anything like that. Okay, now he'll go through and try to make these changes. He already knows exactly how to work with packages because we have all the necessary AI files.

He knows very well how to work with it. And since Scaffold-ETH uses types, it really does a great job of building an interface, does n't it? Uh, okay. Let's give him some time. He will start the process.

You see him starting to make these changes. And as it does that, you'll see a yellow color here, meaning the change is pending confirmation. This is because we initialize the git initialize the git repository immediately after creating the project. This is another reason why I like Cursor: I can track what has changed. If you look here, you 'll see that we've deployed the contract.

So, everything highlighted in orange is change. We'll move on to Hardhat and contracts in a moment, and we'll probably see changes starting to happen there too. It creates a new file. He hasn't written it down yet. There you go.

Now it's lit up blue, see? So now we know exactly what he wrote. He wrote a contract for player statistics and their positions. So, apparently you can mint NFTs for players. I didn't give him much context.

I just said do something. So he figured out for himself what would be cool, and it's actually pretty good. Um, and it wrote a bunch of Solidity code for us. So this is extremely valuable. And if we want to check it manually, we can just click here and it will stop being green.

Hmm, often I prefer to leave it green to see what's going on. Um, but yeah, you can see it's all upsetting. It also updated the deployment script. It created a new deployment script right here, see? To expand it.

So expand it. So now we should be able to deploy this. But we could also ask AI to do it. So let's give him a second. While it's loading, what else can I talk about?

Um, Scaffold ETH. Oh, the documentation there is really good. If you decide to delve into the documentation, it's quite helpful. I think it's worth talking about this a little bit, because there are two other things that are really great about scaffold really great about scaffold . These are hooks and components.

Um, hooks components. Um, hooks are pre- written Next.js hooks that know exactly how to interact with Ethereum. And that's why they're very useful, because, again , they're written in a special way so that you can pass data and get the result through a hook, right? So, there are a bunch of hooks.

One for reading from the contract, one for writing to the contract, listening for smart smart contract events, and simply connecting another contract that you may not have deployed to Scaffold ETH. Many very flexible ways to get data from the blockchain. And you can just direct your AI to it if you're having trouble, and it will know what to do. Um, and the components, they're cool too, because there are different components that look really nice, like this thing with the address, where it automatically checks if there's an ENS for that address. It does a reverse search and has a little copy button, it's a small thing but it's really valuable to be able to copy that quickly.

Um, entering an address. It automatically recognizes the ENS name and converts it to an Ethereum address for you, right? This is really cool. There is a balance check component that will allow you to check the address balance and then switch between the price in dollars and Ethereum, right? Um, there's also this little black block avatar.

There is an ether input where you can simply deposit 1 ETH, and this will allow you to convert it back into US dollars. So these are all really simple components that you can just copy and paste if you're a fan of the old approach and want to write a little code. Um, or you can just point your AI and it will do it for you, which it seems it already did. So, he probably didn't deploy the contract. Oh, no, I opened it.

He unfolded it for me too. So he did everything. Let's take a look. Let's see what he created. Let's go back to the main page.

Ahem. Basketball NFT. He did n't do anything super complicated, but you can release a player. So, LeBron. Um, he's number 23.

I didn't know he was a point guard point guard . Who, who is LeBron? Does anyone know? He is a forward. Heavy forward.

Yes, definitely not an easy forward. OK. So, we can release it. I have no balance. Okay no balance.

Okay . So, that's another point. There is a faucet on top here . We can get some funds from the tap and it will give us play money. Now we can release it.

That's it. Now successfully. So now we've released this little NFT. It could be something like a trading card. Um, but the world is open to you.

You can do something crazy, right? Um, but you see the process. Using Scaffold-ETH to create a prototype very quickly is very, very a prototype very quickly is very, very easy. OK. Um, okay.

So what else can we talk about? Ahem. Yes. So, let's go back to the Build Guild website. I'll see if I forgot anything.

Ahem. We have a Middleg client, which is pretty cool. If you want to run a full node at home, you can run Ethereum on your computer with the click of a button. You just click on it, run the command, and you'll have a full node on your machine node on your machine . This is very actively supported, I highly recommend it, especially if you work with data locally.

No need to use RPC. It's also a fun way to learn Ethereum. Um, it's really really easy and I like it, especially if you want to do indexing or something like that and learn how to index data, it's really cool. Um, what else? Yes, subscribe to us on X.

We have a lot of interesting things there too. Um, I guess while I have time, I'll tell you a little bit about another project I'm working on, which I think is pretty cool. Um, I was the project in the spotlight at the conference. It's called OneClaw, and it's essentially a secrets management platform. So if you want to give your AI give your AI agent access to an API key, private key, or wallet, I created a tool to securely store these secrets.

We store secrets in an HSM module in Google Cloud, and also use multi-party computing, where the key can be shared between multiple providers. So, we can share the key. We can split it in half between HSM and your own storage, or use multi-party computing across clouds where you store the fourth part. This is a very flexible storage method that provides secure access for your agent. We also have a full LLM have a full LLM proxy.

This means we can route all LLM traffic through a Trusted Execution Environment (TEE). This is a separate confidential computing node in Google Cloud, isolated and connected to an HSM. If you have a secret in an HSM and you want to give an agent access to the private key and LLM, but are afraid of leaking that key. Our system checks the entire content of the request, in fact every prompt. We prevent injections, code execution, and social engineering—all the things that cause people to lose their wallets or API keys.

We also ensure that secrets do not get into LLM and we edit them. You can see how we keep secrets, because we are also an LLM provider. So you can use OneClaw as a one-stop solution for LLM. Use this as your LLM and a repository of secrets at the same time. This way you can be sure that your secrets remain safe.

We carry out many different checks. We check for hidden characters, Unicode, encryption— Unicode, encryption— sometimes people hide malicious content in Ethereum content. Therefore, we first decode the data to make sure there is nothing malicious inside. Many interesting things. Um, and one last point: if you want to give your AI agent access to use a private key, but don't want the agent to own the key itself, we have the option to implement this through an intent in a secure execution environment ( TEE).

So, the agent can prove that he is able to use the key and conduct a transaction in this secure environment. You can set certain restrictions, for example: send only one ETH per day, make transactions only on the Base network, or send funds only to Vitalik, you know? You can adjust these fuses. This way, the key never leaves the hardware. It remains physically blocked, and everything happens behind the scenes.

So this is a pretty interesting technology, if anyone wants to discuss it with me discuss it with me . We also have built-in wallets. Similar to Coinbase or Privy wallets, we offer a similar solution through OneClaw. We also have an MCP server that allows you to connect and manage access to your confidential data. MCP is open source software, so anyone can try it.

We have an SDK and a command line interface (CLI). So if you use OpenClaw or Hermes and want to work through the CLI, you can use it to access your secrets programmatically. And yes, we have a lot of really cool stuff. There is also an endpoint for AI. If you go to -4-AI, you will find everything you need to work with OneClaw on one page.

You can simply direct your requests there. What else? That seems to be all. Yes, of course. Thank you.

I think this may not be entirely true, for example, for Maker. Like for example, for Maker. Like ...First we tried using MPC, yeah. But the delay was very high, the signing was expected for...

Yes. Very. So how long does a signature take? It? So now...

Yes, less than 100 milliseconds is probably hard to achieve now probably hard to achieve now . So you would probably be more interested in using TEE. With a TEE, we store half of the key in the HSM module, and you can keep the other part under your own control and place it anywhere you want. This way want. This way , you can make two calls at the same time to get the key and perform the signing.

Right. So, it's a little faster. The only thing I will say is that with Google Cloud, everything will accelerate because we are a strategic partner of Google Cloud, and they are adding their own support for multi-party computing. Oh, you work at Google Cloud. Oh, cool.

We started. Okay, let's talk. Yes. Yes. Because they're adding their own multi-party computing, and we'll use that as soon as it becomes available, because we actually prefer Google Cloud.

Surprisingly. Question. Yes, that's a great question, and that's what we're working on next—direct Git integration for agents. Imagine that the agent would have access to Git, and we would be able to offer, for example, "red team" style code analysis. We would connect this directly to the repositories, instead of you having to store your secrets on, for example, GitHub.

So yes, we're going to do this, not only with GitHub, but also with another Git competitor, whose name I forget. Yes, that's right. Yes. Yes. Yes.

Let's talk later. Very excited. This is cool. Maybe we have more questions, because I think I've said everything. I've been chatting for a long time.

You're probably surprised, I'm surprised you're still here. No. Thank you for listening. Does anyone have any other questions I could answer? Well, was it interesting?

Yes, there's one in the back. By the way, do we have a microphone or not? Because it seems to me that people watching from a distance won't hear anything. So. Yes.

So, ERC 8000. Yes. ERC 8004. Yes. What if?

Sorry. Reputation. There are three different ways to work with reputation in ERC8004. This is identity, reputation, and validation registry. I'm trying to remember exactly.

It depends on how depends on how ... How you want to do it, but there is a way to do things, for example, inside a TEE, right? Or ZKML, or some other kind of automatic validation that you could implement. Let's take World ID, for example, right? You are building on the foundation of World ID.

All your users in World ID. Yes. So, let's say you want to confirm that this agent was created by you as a human. You could make a ZK- could make a ZK- proof and put it in this validation register. This is one of the ways.

Or it could be Or it could be any ZK. It doesn't matter. Or you can choose a more minimalist option, such as simply storing a registry file that you can verify. It could be something like an identifier like a DID or, say, OneClaw. We use DID, right?

we provided, uh, an identity. So that would be something you would probably want to store in a generic identifier like that. Um, and also, I think, secondly, by the way, I'm not an expert, so I'm just reading about it. Maybe some additional data, you probably also want to get feedback from other users, like someone saying, like someone saying, "Okay, this guy is "Okay, this guy is an agent, he's a scammer or something, and he's a bad guy." If this gets enough votes, it will signal that it is no longer valid.

That's how I understand it. I'm not an expert on this yet, but that's what I'm leaning towards. Yes. Good. Next question.

I thought I saw another hand raised. Yes. Ahem. Oh, five more minutes. Okay, next question.

No, I thought I question. No, I thought I saw another hand. Yes, next question. Secret, is there a way to pass secrets to an agent for access? Yes.

Go ahead and register the secret for future use so it never goes through the backend. Currently this is only possible with private keys, but we are working on it. But you could build a system for this, because you can do it like this: you can have a backend service running that has unique access rights. So, OneClaw has two ways of interacting: a user key and an API user key and an API key for the agent. The user key has full access to policy settings, sees secrets, and is a high-level key.

So you can run the service directly on your own machine, which is in TE or another environment with a user key that has access to them, and your agent will be able to verify their identity to get that key from your backend. So you can design it that way. We plan to add this natively in the near future, but it's not there yet because it requires a lot of work. But there is another company that does this. It's called Inphysical.

So I would advise you to pay attention to Infysical. Um, not that I'm promoting competitors or anything, but they specialize more in Web2 and they would be a good alternative. So, okay. Thank you very much, guys. I really appreciate it.

Thank you.