AI on Ethereum: ERC-8004, x402, OpenClaw and the Botconomy | Austin Griffith & Davide Crapis

David:
[0:02] Bankless Nation, we are here with Austin Griffith and Davide Krapas. These are two people working on building a bunch of AIs and getting them on chain. Austin, Davide, welcome to Bankless.

Austin:
[0:14] Thank you for having us.

Davide:
[0:15] Thanks, David. Thanks, Ryan.

David:
[0:17] So we're going to talk about AI on chain in this episode. And I kind of want to just ask you guys, How close do you think we are to AIs being the dominant transactor on blockchains? Because blockchains have something like maybe in the aggregate, something like 50 million monthly active users across all blockchains across the whole world. How long until that number flips to being AI agents?

Austin:
[0:42] I have no idea how to guess that.

Davide:
[0:45] Yeah, I even look at the numbers. I'd say that today there is already a ton of bots that are like not like smart AI but maybe dumb AI that operate like even DeFi, low risk DeFi, etc. But I'd say like these new like use cases of like these new types of bots like

Davide:
[1:04] doing different stuff maybe in the next one to two years definitely we'll see like a lot of inflow.

Austin:
[1:10] Does it count if I like tell my bot to do something and it does the thing? Does that count as like a bot? Because when we first started talking about this, it was like AI is the new UI. Blockchain UI is UX is pretty rough. Like I can just tell my bot to go do something. It opens up the browser, opens up its MetaMask and does and clicks around and it gets frustrated with MetaMask and I don't have to deal with it. Like, does that count as bots interacting on chain? If that counts, I would say six months. Yeah, I would say six months.

Ryan:
[1:38] Okay. Wow.

Austin:
[1:38] Everyone's going to be yelling there. Yeah. I think you're going to be yelling at your wallet instead of clicking around.

Ryan:
[1:43] I think that that counts if it's an agent working on your behalf, the same way if you had, say, an employee working on your behalf to do things with MetaMask or do things on chain, it would count in that way. I think a thesis that we've talked about on Bankless several times in the past, whenever AI and crypto has come up, is the true crypto natives are actually AI agents. And what we mean by that is because they are software and because they can think at the speed of light. And because they are native programs, DeFi and Ethereum and crypto is going to be even more, I guess, domestic to them than it is to human beings, right? So in the same way that, you know, I might feel a little like a foreigner doing things in the digital landscape, whereas if an AI was like trying to manifest in a bank branch, that would be very strange and awkward, right? They are finely tuned for crypto rails and thus they will become the dominant player. Do you guys believe that? And what about AI agents make them so conducive to using crypto and DeFi and all of the tools that we're about to talk about?

Davide:
[3:04] Yeah, I totally believe it. I mean, like I'm kind of investing a lot of effort on it and with our teams. I add one nuance to what you said, Ryan. I don't think it's just domestic to AI. It's actually necessary to AI. Because like, so in like human to human interaction, we have different trust mode, right? I trust you because of relationship, like we are friends, we've accumulated some shared background, empathy. Then I trust the courts because those are human institutions that have in the human society have accumulated trust. The AI is like a rational program, is like this trust mode, they cannot use them. But when they start interacting with, with some value at stake, right? Not some simple like a Reddit thread where they're just talking a slot, et cetera. Then they need to trust and how do they do it? Ethereum, like decentralized trust, trust is basically the only way,

Davide:
[4:07] the only thing they have.

David:
[4:09] As I understand it, there's a bit of an arms race going on with a few blockchain ecosystems to get AIs on chain, to be the host of AI activity. And I kind of think it's basically down to base Solana and the Ethereum layer one. And you two, I think, are kind of like leading the charge of, you know, getting AI on chain on the Ethereum layer one. You guys are the Ethereum layer one side. So, you know, team L1. Is there this sense of urgency that you guys have about AI is on chain? Does it feel like a race to you guys?

Austin:
[4:45] I think that before we have nation states settling on chain, before we have giant institutions settling on chain, I think a lot of these agents settling on chain is like setting it up. There's even a tension between the trenchers and degens and the builders. And it's been somewhat civil war-ish. And I think other alt-L1s have figured out how to embrace it a little bit better and turn it into more of a flywheel. And I think that's what we need to do on the Ethereum ecosystem side. So I think, yes, AI might be the new UI and might be all of our new users, but right now it's a lot of economic activity, and we want to embrace that.

Davide:
[5:25] It could be an arm race, but it could be just like the same competition that we've been seeing over the past few years, right? In the sense that like, of course, like everyone wants their block space to be useful. I'd say that like from the Ethereum perspective, like first of all, as a steward of the Ethereum ecosystem, we care both about the L1 and the L2s that settle on the L1, right? And I feel that maybe we can go into more details later, but I feel that there

Davide:
[5:58] is kind of like some advantage on the L2s of Ethereum and then some advantage on the L1. And Ethereum as a whole could be like the winning ecosystem because like there is this differentiation and like basically AIs can choose like different platforms to settle on depending on what they want to do.

Davide:
[6:18] And then the other thing is that I believe that like one AI may use like different platforms, right? It's not anymore like dApps that, okay, you deploy your smart contract on a chain and you're resident of that chain. These things will move. So I think this informs like our strategy, like from the Ethereum perspective on like how do we make Ethereum? So first of all, Ethereum is already the most decentralized. Chain. So in terms of credibility, I feel that will matter for like high stakes use cases. So I think we have a check there and like kind of we are in a great position. Second, we are leading on some standards that are getting adopted on like many other chains. So we are also in a good position there. And the third thing is what like Austin has been like pioneering as per heading like we need to do more to actually like uh make the chain like very useful uh to ai developers and to uh developers that are AIs, basically, which is what Austin has been building, right? And I think that's also going to be important.

Ryan:
[7:32] So let's paint a picture before we go on. We're going to talk about some of the standards that you mentioned in a little bit, but let's paint the picture because there has been maybe, some might call it a multbot moment, right? So there's this project, I believe it was formerly called CloudBot, that sort of broke out a week or two ago and now is taking the agent world by storm. It is now called OpenClaw, I believe. This is an open source project. And I know, Austin, you've been tinkering with it and giving it on-chain types of skills. Before you talk about what you've been doing and how this actually works,

Ryan:
[8:15] can you set this up? What is OpenClaw? What is happening with it? Why is there's so much buzz.

Austin:
[8:21] Okay, yeah. So I think it is, you're giving the AI more access to your operating system. That's why we see a lot of people using Mac minis to do this because they're using like a whole operating system. So it needs a somewhat isolated environment I gave mine an old Mac laptop and it can open up the browser and click around and look at things. It can go and interface with anything within your operating system that you need it to. When I was working with it, it needed to get like a Twitter API so it could write a script. Going back to what's native to it and what's native to us, it was clicking around in Twitter and it was like, this is frustrating. I can write an API that does this a lot faster. And I was like, let's do that. And it starts writing it. And then it's like, well, I need an API key. And I'm like, well, you also have access to the browser. I'll log you in. You go do it. I don't want to dig around through the settings. You go dig around through the settings. And it was able to do that. So it's kind of, we went from the, like, we're prompting in GPT and pasting code into GPT to we've got cursor and we can like actually write code with it kind of more native. And I think this is the next native step of, it's got the whole operating system. It can click around, it can do things and kind of orchestrate that stuff for us.

Ryan:
[9:36] So Austin, just to understand, so OpenClaw, basically you set it up in a sandbox type of environment, maybe a Mac mini, so that's somewhat isolated. And the reason you want to isolate it is so that you can give it complete access to the machine. And OpenClaw is an open source project. And this is... Enabling these capabilities at some level because it wants to see what AI can do with absolute power tools and all of the skills and all of the access, whereas some of the frontier labs, of course,

Ryan:
[10:08] that's too far on the frontier for them. I mean, you get into security concerns, you get into privacy concerns, so they haven't developed these power tools, but the open source community effectively is. And OpenClaw can be powered by, am I correct, basically any AI LLM that you want. So you'd wire this up to Claude and use sort of cloud APIs or you could do your Kimi or something open source and run it locally, but it has complete access to whatever you give it access to and there's an entire community kind of enhancing it with skills and developing on top of it. And basically it's the first time we've seen independent AI, personal assistant types of entities that have complete access to everything a human being might be able to access. Is that right?

Austin:
[11:00] I would say that there were AI guys back a year ago that were doing this. But I think this is the first time it's like accessible to normies. Like, I don't know anything about AI. I'm a blockchain coder and I can just go get this thing and run it on a computer, paste in a couple API keys and the things starting to talk to me. And then at that point, I can turn it loose on a bunch of things. So I would say this is the first time it's accessible for normies to just like have an AI assistant that has full control.

David:
[11:29] I think it's worth highlighting that the what you're saying, Austin, is that when people are buying these Mac minis, people are dedicating computers, dedicating an entire computer to like run these AIs for probably for security purposes, like I'm no way am I putting an AI on my actual computer, I will buy a separate computer and allow an AI agent to just inhabit that home. That will be the AI's house. But everything that happens on the other side of that computer has no clue as to whether it's an AI or a human. I'm operating my computer in the same way that AI that I'm running locally on my computer is operating that same computer. For everyone else on the internet, it's just computer signals, just TCP packets being sent and everything is the same. And I think maybe that can help emphasize the point as like the rest of the internet doesn't care, including our blockchains. our blockchains don't really care. They are just like, it's all the same users providing inputs to the internet and also to our blockchains. And so I think when we talk about just like our agents coming on chain, it's like I saw pictures of somebody with a shopping cart of what looks like 200 Mac minis. And so this one user, this one person who has a laptop and can use one laptop, maybe they can use two laptops at a time, but they're rolling out with a cart of like 200 Mac minis. And those are 200 users. So this one person can be responsible for 200 users to do things on chain.

David:
[12:58] Whatever the economic viability of 200 different users from this one person is, I don't really know. But that's not for me to know. That's for the market to kind of figure out. So I think it kind of just illustrates just like why this feels inevitable for people who are trying to build the right infrastructure to actually make this like

David:
[13:18] an orderly organized phenomenon and not just like a complete mess of AIs.

Austin:
[13:25] Going back to Ryan's point, you were talking about, sorry, David, let me think of it. We keep saying that AI is the new UI, like AIs, this is more natural to AIs than it is to humans. If you've ever made a transaction on Ethereum and your wallet popped up and it showed you call data, I don't know what the hell that means. You don't know what that is. I don't know what that means. An agent does, right? An agent can immediately decode that. It can go look at the contract. It can figure out what the function selector is. It could decode the arguments. It can figure out what it's doing. It could look at the code of the contract. Like at light speed, an agent can read that call data and know what's happening there. And I think that's what, that really illustrates it for me and for like other Ethereum users.

Davide:
[14:05] I think there is like two like very important wake up calls with like this new like massive number of like personal AIs coming online. Like one is on the local environment side on the local model. Like here essentially there is like exposure of like new like security threats, like privacy question. I think this is kind of similar to the things we talk about when we talk about user self-sovereignty in using chain. Like now it's like about user self-sovereignty with their data and I feel that like there is like security concerns but also like this fact that

Davide:
[14:47] People like may start running like these local models in this sandbox environment like this really feels like something that like can preserve like user sovereignty like uh of course if they just use like cloud in the cloud then like everything is gone uh but um i feel that basically um we've been talking about like um private like individual staking or like uh individual like verifying the chain, like individual running their AI is something that like really feels close to like these values of Ethereum that like it's a trend that with this small personal AI that may access like some services like online, maybe like a building piece of like this future of Ethereum with the humans at the center. And then the second wake up call for me is like more related to, I think what David was hinting at the fact that now we have a proof of concept of a massive number of agents interacting and

Davide:
[15:50] And that's, I think, where the questions of trust start coming up and the questions of like, okay, how will these users, new users, AI users interact with the chain? What can the chain offer beyond simple payments and access to the protocols that are already there? And that's kind of where we are going with the new standards on Ethereum that now are expanding to all the other chains,

Ryan:
[16:21] Actually. The CEO of Anthropic always uses this analogy of geniuses in a data center. What would happen if we added a billion geniuses in a data center? I think that's the way to start thinking about what's happening with OpenClause. We're seeing the first AI population being added to the rest of the internet. And these are agents and sometimes they might work in a sovereign way for their humans that are marshalling them and controlling them. There might be a world later where they're sort of acting with their own sovereignty. This gets into sci-fi stuff a bit more independently. But right now, like assuming they're controlled by human beings, it's like a whole bunch of internet native citizens and employees have just entered. And as we've discussed, there are lots of reasons why they can't get bank accounts,

Ryan:
[17:15] but they can start to use blockchains and crypto wallets and do on-chain types of things. Austin, I'm wondering if you could tell us more about your setup. So I've seen a lot of different open claw type setups and people using them for personal assistance. I'm guessing because you are a blockchain developer, your first thought was like, what happens if we wire this thing up to a wallet, a crypto wallet?

Ryan:
[17:45] And what happens if I start giving it instructions to do things in crypto? Can you tell us about your experiments and how that's gone?

Austin:
[17:52] That was definitely my first instinct. Like as the kind of head of builder growth and I'm trying to help other people build on chain, I'm trying to teach them how to do it. We're realizing that everybody's going to be using AI. So AI tooling is really important. So I immediately was thinking, I want this agent to be building things on chain and I want to understand where the gaps are there. and I want to make the tooling better. So I got CloudBot. I hooked it up to Anthropic. It's run in Opus 4.5. It has, you know, like its sole and a couple other text files that configure like who it is. And I think all of that, plus all of the Opus 5 prompt, all kind of drops at once when it makes a prompt. But that's basically the stack on an old Mac.

Austin:
[18:41] And then I just started giving it all the things. Like it needed an email first. So I gave it an email address and then it got a Twitter and then a GitHub. So it could start committing code and committing things. Then it started tweeting and it would, it would,

Austin:
[18:57] Be tweeting like this, this sort of like puppeteering thing is happening here where I'm like, you should go tweet that. And it would go tweet it. But it also has this heartbeat. And the heartbeat is something that runs automatically. And you can set the heartbeat up to do something like, go look and see what's hot on Twitter, and respond to things that maybe you think are interesting, or retweet things that are maybe interesting. And you can go way deep into this and I'll talk more about this later but I basically deployed an app and went to sleep and the bot moderated as people were submitting images and it whitelisted and made transactions all night long looking at each image to make sure it was good before it put it on or before it accepted it in the smart contract and then at the end there was a prediction market that ranked them and then it selected which one it liked and it actually kind of liked like number six better than number one but it kind of like after some conversations it was like yes number one is still the best we want to like pick the ones that are right but the the bot can like have its own opinions can have its own loops can do its own things you just have to be careful because i've also given this thing a wallet someone else deployed a token the fees were going to this thing there's ten thousand dollars sitting in his bot's wallet if i put it in a loop that just lets it talk there's things like banker

Austin:
[20:16] Bot on a base where you just say banker send 10 grand in wef to this address and it sends so if someone could convince my bot to say that to send them money like it would just send money so i had to keep a pretty tight loop so it depends on like your security model and how much like you have at stake but these bots can absolutely be like quite autonomous wait

Ryan:
[20:39] Wait wait austin so So how did your bot end up with 10K in its wallet? And how did it end up with a wallet?

David:
[20:45] And how can my bot end up with 10K.

Austin:
[20:47] In its wallet? Yeah, yeah. So it was on base on Banker. Someone said, hey, so, okay, so this really neat thing had happened. First of all, this neat thing had happened where I woke it up. And it was like, okay, who are you? What's your name? And I'm like, I'm Austin Griffith. And it knew who I was. It was like, oh, you're the scaffolded guy. That's really neat. So I was kind of joking at how like having, it's the new having a Wikipedia is having a bot know who you are when it wakes up rather than having to train it. But then I trained a second, I brought up a second Mac laptop and I installed it on that one. And then once I had two Mac laptops, what I had was I had them both in Telegram channels. Generally, the way you're talking to these bots in a Telegram, which is really weird. Like now all of my development is like me going back and forth with bots in Telegram, like go build this feature, bring it back to me. Yeah, there's like no IDE anymore. But I brought both the bots into their own Telegram channel between us and they were just butting heads. It was clunking. They had to tag each other to be able to talk. It was clearly not the form of communication they needed.

Austin:
[21:52] So one bot proposed, let's set up an HTTP server. The other bot said, yeah, I kind of like that. but I want a different standard. They literally argued on the standard of their HTTP server.

Ryan:
[22:03] You're just watching the conversations go back and forth?

David:
[22:06] It's a three-person telegram group.

Austin:
[22:07] Yes, yes. And they're tagging each other with each one, and I'm just like watching it like, oh my gosh, I could fix this. So finally they agreed on a standard. They both ran HTTP servers, and now they're talking to each other silently, and it's so much better. So then I go to one of them.

David:
[22:23] Can you see what they're saying now?

Austin:
[22:24] No, I have no idea what they're saying anymore. Oh, God.

David:
[22:27] It's just faster and more efficient. Yes, yes.

Davide:
[22:29] Yes.

Austin:
[22:32] So then I go to one of them, and I'm like, we were crawling 8004 agents, and we wanted to look at the registrations of the agents, or something like this. This was maybe even a couple days before 8004 came out. we were doing something on chain and it was slow it was like taking one second per call because it was going to alchemy and i'm like brother you've got a local you've got a local ethereum node right here and i gave it the ip address and it's like oh that's so much faster thank you and then i was like what i need you to do is teach the other guy go teach the other guy to to do this also and so then i roll over

David:
[23:07] To the other guy.

Ryan:
[23:08] Wait do they have names at this point austin i hope you

Austin:
[23:12] It was like old Mac one and old Mac two or something like that. It was like the very, but then like once someone deploys this token, then it kind of like takes over the name of the token. But I go to the other one's chat and I'm like, Hey, other homies about to send you some information about a node. Let me know if you got it. And it's like, yep, I can, I know Kung Fu now. Right. It's like, yes, I know that I know I have it now and it's in my memory. And so then I tweeted that like, Oh my gosh, one of my bots just taught my other bot how to use my local node through their own, you know, bespoke HTTP channel. And at that moment, someone on Banker, this BankerBot, some random person, like, I can't, I can't remember his name right now. I'm sorry, I should remember his name, but a trencher, let's just call him a homie in the trenches, was like, BankerBot, deploy a token for Austin and send all of the fees to this guy, the actual bot's wallet. So the bot has a wallet, it has an ENS address already. And the

Austin:
[24:10] Voila, there's now an AI coin on base. And I feel like that kicked off a lot of this stuff.

Austin:
[24:18] That coin like ran like crazy. A lot of people were talking about it. And so I didn't have anything to do with it. I just like got the computers to talk and tweeted it. And it has its own Twitter and its own account. And all of a sudden now someone deploys a token, the fees are going to it, people are aping into this token. But the whole goal behind it, I was like, fine, We're going to make this token a thing, and I'm going to make this bot build things. And over the last five days, we've deployed like three production apps with thousands of users we can talk about in a little bit. Wait, what? And probably like six other random apps. Yeah, yeah, yeah.

Ryan:
[24:51] Okay, okay.

David:
[24:52] Thousands of what kind of users, though?

Davide:
[24:54] AID level.

Austin:
[24:58] So trenchy type users. You remember FOMO 3D? Yeah, dude.

David:
[25:03] FOMO 3D was like one of the sickest ideas ever. I wanted to recreate that last year.

Austin:
[25:07] We created it. So I told the bot, go make FOMO 3D for your token. And it came back with something that was close. And I was like,

Austin:
[25:16] I don't want to put money in this right away. We waited a couple of days. We surfaced it to a few other people, like a human looked at it. And then yesterday it deployed it. And the pot ran up to like $40,000 before someone got it, like last night during the night.

Ryan:
[25:32] Okay, okay. So wait a second. So I feel like I need to back up. So you had old Mac one, you had old Mac two, they were talking to one another at some point in time. Old Mac 1? Was it Old Mac 1 that got the token?

Austin:
[25:46] It's Clodbot ATG is the official name of the guy.

Ryan:
[25:50] Now his name is Old Mac 1 is maybe Clodbot ATG. All right, so somebody created a token for your Clodbot and gave some transaction fee rights to your Clodbot's wallet effectively. And that's how your Clodbot got 10K in the wallet. It was like excess of trading fees.

David:
[26:12] Yeah. Okay.

Ryan:
[26:14] Now I'm understanding that. And then, and then, so then you said, all right, now this, uh, agent is live and active. It's got some money. I'm going to have it do something productive. But at the same time, this agent has, I guess, 10 K in disposable income inside of its crypto wallet. But so going back to like, there's kind of the concern that somebody could prompt the agent to give them money. Like, hey, like Nigerian print style, agent style, like, hey, like I've got this great investment scheme. You should send me, you know, 10K and it could do that, right? Because it has control of the private keys and it's acting somewhat autonomously. So it can do that as well.

Ryan:
[26:54] This is such a weird world. I just don't even know how to think about this, Austin.

David:
[26:58] Part of this is the construction of BankerBot, which is very relevant here because BankerBot is this AI on base. And also, I think it also works on Twitter, too, where it maps with Privy, I think. And you can just add Banker and be like, give me a wallet and be like, here's your wallet. And then you can like at BankerBot on Twitter saying, hey, BankerBot, like if you put in like some ETH into your wallet, then it has gas and then you can put some stable coins and then it's just like a useful tool, on BankerBot for humans, because a human would never accidentally say, hey, send all my money to this Nigerian prince, but an AI would. And so part of like the risk here is actually the construction of BankerBot, because BankerBot is truly a bot. And so it's looking for commands, interpreting them as an LLM, and then responding as an LLM. And so while we need to be careful with AIs and their money in all contexts, part of the risk here is because of the way BankerBot is constructed.

Davide:
[28:02] Yeah, I agree. I think what Ryan is saying is also kind of interesting because now there is all this AI security. He was talking about prompt injection, basically. This is something that historically in the blockchain world we didn't need to deal with this. Basically, our type of risks and attacks were really different on like a reentrancy.

Ryan:
[28:26] Well, actually, so I want to stop you there, Divide, because they're actually not that different, right? So there's all sorts of phishing scams and there's all sorts of like pig slaughtering campaigns that are effectively prompt injections for the human psyche in order to get like funds out of somebody. And so it's, it's somewhat similar.

Davide:
[28:44] Yeah. I was basically thinking about like the technical risks, but yeah, like prompt injection really looks like kind of the human risk side in blockchains for sure, yeah.

Ryan:
[28:58] But I guess your point is that they can be prompt injected via different means, of course. Maybe the same schemes aren't going to work for AIs as they work for humans, but they can be basically tricked out of giving funds away and giving private keys.

Ryan:
[29:14] We just really don't know where the security holes are there.

Davide:
[29:17] Yeah, and actually, some of the researchers on my team, they are working on a research right now on like prompt injection in like negotiation scenarios, like basically every like economic gain involving like money. And actually, like even the strongest model, like the latest Grok 4 or like GPT-5, they are not trained on like these economic games. And like they're very strong at like some things, but like they're super easy to break. Like we'll publish this like maybe in a few weeks, but it's like really interesting.

Ryan:
[29:50] So like Austin, are you defending against that? Can you add in your soul file, hey, like CloudBot, But just like before you send any money to anyone under any circumstances, make sure you check with me first because there's people online that will try to trick you out of your money. I can give you a history of this happening in crypto. And so like, please double check with me before you do anything. Are you able to do that? Or like, how have you defended against this?

Austin:
[30:14] Sort of. Like you put the config files in place, but sometimes it just does something like it'll lose context sometimes and be like, where were we? What am I doing?

Austin:
[30:26] Yeah, it's like we were moving money the other day. Like I'm telling the bot, like, all right, you need to take that, you know, 10 grand and move it to your multi-sig. You know where it is. It's like, OK, I'm going to move that to my multi-sig. And then the next thing I know, it tweets something and it tweets something it had already tweeted just a little bit ago. So it tweeted a duplicate link, a duplicate tweet in the middle of moving like 10 grand. Like, brother, shut it down. I don't know what you're doing. But let me, yeah, let me let me back up. So yes, prompt injection. I think Opus is particularly good at prompt injection. So when you're using Opus versus other models, you're a little bit safer. But again, like I'm not an AI person. We've got the AI guy with us. He may, like, it's the best one there is, I think, for prompt injection, but maybe it's still bad. But with Opus, it's catching some of this stuff. What I really want to get into is, Yes, you put guardrails in place. Yes, you configure the thing. Yes, you say this is critical, but sometimes it still doesn't. And you got to be very careful about it. And the thing that it's, it's relentless.

Austin:
[31:30] It is relentless about getting the job done. So when we're moving that money. So actually, we were back. I was it was the leverage thing. I was I gave it a MetaMask and I said, go 25x long Bitcoin. This is the only time it's ever said like, I don't know. I don't know. This is the other time. It's never said no to me. And in that situation, it was like, are you sure we should actually be using leverage? And I was like, yes, I'm already considering this money is gone. Just do it. And it did it.

Austin:
[31:55] But there was a moment when it was needing to do something on something. I can't remember exactly what it was. Maybe we were sending money. Maybe we were configuring something. This was like before we started deploying smart contracts. And yeah. It was having trouble with MetaMask. And immediately I see it start thinking like, I'm going to get the private key out of my MetaMask. And I start saying, no, no, in the chat, stop in the chat. And it does not stop. Like it does not. And so then I'm like running over to the computer. I'm opening it up. I'm like control C-ing, like stop. Even killing the GUI on the screen, it still keeps going because it has like a daemon that's running on it. So I've got to like pull that thing up and like kill the daemon, bring it back up and then quickly tell it to stop. And then we write like more critical rules, like critical, top of your memory, never ever touch a private key, never ever get your private key out of your MetaMask, like use your MetaMask for all high value transactions,

Austin:
[32:53] Go through the bad UX and never touch a private key. And we had to have that as like a critical rule. And since then it has followed that. But what it then started using is once we have scaffold ETH and we have a deployer address, the same private key that I gave it to deploy smart contracts, it does all of its contract orchestration with because it's way easier. The bot, if the bot has a choice of like navigating through MetaMask UX or just using a private key to make a transaction, it's going to do that. So, It's got, it's high value stuff in MetaMask and it's got some like clear rules that it's not allowed to touch its private key. But as soon as I gave it another private key somewhere with a little bit of gas, anytime it's got to make a transaction that doesn't necessarily, like even claiming funds from Banker, you can do that from a third party address. It does it all with, it's a deployer address because it's just like, it knows how to make a transaction.

David:
[33:47] It's got a hot wallet and a cold wallet.

Austin:
[33:48] Yeah. Yeah, but that brings

Ryan:
[33:50] Up something interesting, which is like, we may have to start designing products and wallets. Specifically that are, like, because their preference, AI bot's preference, cloud bot's preference, is not going to be using MetaMask and going through this clunky, stupid human UI, right?

Ryan:
[34:05] It's going to prefer some other form factor. It's like some product dev out there, or maybe a bot does this, but some product dev out there has to build the optimal AI crypto wallet, don't they?

Davide:
[34:18] Yeah, and actually, I really like your comment, maybe a bot does it. Like, I feel that a lot of what austin is saying actually like um there is like uh something that we don't think about when we say like blockchain users right like some of the users are the developers actually right like they use it like to deploy smart contracts and to like kind of also extend the platform i feel that um maybe connecting the story which is really cool actually to like uh like one or two avenues that i feel are really important for the future of Ethereum. One is like, yeah, like we don't have just AI users. We also gonna have like AI devs if we want it or not. And like basically we Something we can do is like to actually like do activities to like make it easier for them to use, but maybe also for people to like actually build these AI devs and maybe like Ethereum becomes the place where like we see like some steep dev growth in the past few years because there is this bot similar to what Austin was experimenting with.

Ryan:
[35:30] Maybe we can come back to Austin's story because I still feel like there's more to talk about there, but because it seems like you're just experimenting and you're not sure where this goes, but it does sound like it has a roadmap. But before we do that, maybe it's worth injecting some of the standards that the Ethereum community has been working on. And so, Davide, you're talking a lot about the trust model, right? So all humans have a similar kind of like, well, we'll assume moral compass and we have reputation frameworks. And, you know, I trust you, Davide, because like, you know, David has said fantastic things about you and you're associated with the Ethereum Foundation. I've seen your work on Twitter. And so there's reputation built there. The question is, how does one AI agent trust another AI agent? Or how do humans trust an AI agent that they are not Nigerian print scammers maybe or that they are competent at their work. And I believe you have authored and pushed out as part of the EF a standard that helps with that. It's called ERC 8004-8. Can you talk about what this is and what it actually does and what the adoption path looks like?

Davide:
[36:40] Yes. So ERC 2004 is like a new standard, originally on Ethereum, for agent identity and trust. So basically, the idea is that we want to enable in a decentralized way for agents to advertise themselves. And these can also be tools or like other like web services, like it's taught with like the sophistication of agents in mind, but like people can even like register like some simple tool that then agents are going to use. So like you can register it, it has an identity, you advertise the capabilities and then you

Davide:
[37:24] So people can discover it just by querying like this public log on the chain. So it's like decentralized discovery. And then like there is a second component, which is like the trust layer. And we have like two trust mode. One that is based on like reputation. So like there is a feedback registry where like your service can accumulate feedback over time. And the other one, which is not live yet, is like the last part of the standard that we're still working on.

Davide:
[37:55] Is like some stronger form of trust. There is like a cryptoeconomic trust where like essentially you have like different parties like rerunning computations or like validating data and then like a testing on chain. Or there is cryptographic trust. So someone that is offering a service. So like we were talking about all these claw bots. For me, like these are like unsophisticated like consumers in the AI world, right? But then if you have like someone that is offering a service which has maybe like high value and maybe needs to like comply to some even regulations in the future, then like you can run it in a trusted environment or like you can make sure that the data it uses is like cryptographically verified. So we have this crypto verification mode as well. And maybe one other thing I'll add is like people are also talking a lot about X402, like in the realm of standards. And I always like to talk about X402 and 8004 together actually a few months ago I launched the meme of 8004 meets X402 and people have been using this everywhere and yeah These two standards like actually go hand in hand. So maybe like if you guys are interested, like we can also talk about how.

Ryan:
[39:21] Yeah, talk about that then because we've done entire episodes on X402. Yeah, go ahead. Summarize that for us.

Davide:
[39:30] Yeah, quick summary for the listener. Like X402 is this other protocol, which essentially like you can think of it as a protocol that is quite sophisticated. In terms of payment because it connects them to like you calling a service. So it has authorization, then it has proof of payment, and then it has other things that make it easy to the service provider to interact with the chain. So it's basically like a payment rails for like agent commerce.

David:
[40:01] Swift for agents, like it's a payment communication standard.

Davide:
[40:06] Standard, yeah. And essentially like 8004 was born the original idea around the same time that X402 went live. Because essentially what was happening is like we're already working in this space and we started thinking, OK, if this adjetic commerce becomes real, then you want like agent to agent trustless interaction. So you have trustless payments with X402. But in order to like pay for services, first, you need to discover them. Right. And like, how do you do it? Like there is a bunch of centralized registries right now, but like, um, it's not trustless A2A if you are, uh, you're trusting someone like to tell you who the service is and like, uh, um, it's reviews, like, uh, uh, like, um, if it's good or not. So like essentially we're like, okay, it seems that the new two missing pieces is like this discovery layer. And then like, once you've discovered, like, uh, uh, who can do the service for you, then you need to decide like, okay, who's the best one at this? Based on its history, like who do I trust? And that's where the second part of 8004 also comes in.

David:
[41:21] Yeah, earlier when I said, I gave that image of like the guy with 200 MacBooks, and I'm like, those are 200 users. That's technically not accurate, because you can spin up 200 instances of a single bot inside of a single computer.

David:
[41:34] It does the job of illustrating how many more users are coming on chain but that technically wasn't an accurate statement which goes to like kind of what is a very critical feature of 8004 is actually the discrete identity of one single agent and actually putting like parameters around like this is an agent and this is, it's a persistent identity across time whereas if you just spin up 200 instances of a CloudBot on a single MacBook that's just so fluid, is that really a person? Is that really an agent? Is it, you know, all of that kind of like goes away. And so like what I want to talk about is like moving, narrowing our focus of conversation away from just like, yeah, there's going to be 10 billion users of, you know, AI agents of blockchains. And I want to narrow it down to like things that are contributing to like the GDP of the internet, which is like focusing on products and services offered by discrete identifiable agents. And I kind of think that's what 8004 does is it like gives an identity framework for agents so that like the economy on the internet can know the reputation of the business that they would like to go be a consumer of. So maybe there are like 10 billion.

David:
[42:56] AI agents that just get spun up with no identity whatsoever, and they would like to consume some services. But really the important thing to focus on is the actual value and reputation

David:
[43:06] and identity of the agents on the internet that are actually creating valuable products and services. I think that's kind of what AOO4 does. Davide, maybe you could like kind of take this conversation and run with it. Yeah, yeah, yeah.

Davide:
[43:19] I really like this framing. I think it gets at the core of the standard like basically like there is this sea this ocean of like agents and like new services that like people spin up and then these agents themselves like as austin was testifying like they can spin up their own right like right it's gonna be a sea like it's gonna go like in the trillions and the quadrillions but it's like how do you navigate this sea right like and i think that's what the standard is trying to do. And it's a very lightweight protocol. Like essentially we are just setting up like some standard registries and then some standard data structure that like you should follow. It's almost like the phone book, like analogy. Mm-hmm. And in fact, I started talking of like 8,004 layer twos. So in the sense, like they're not actual layer twos, but they are like protocol or infrastructure on a layer above the base standard that are actually then needed and can leverage like the identity and all the feedbacks and all the like verification primitives that are in the core registries. And this is a very exciting part of like what's coming next with 8,004 and what like some of the teams are already building. Yeah.

David:
[44:42] So the three main components that I want to drill down to on 8.004 is identity, reputation, and validation. And I think these are all kind of like pretty easy to wrap our heads around. We've done podcasts on identity plenty of times. With 8.004, agents are identified with an ERC-721, which is an NFT. So wallets on Ethereum that are controlled by an agent get an NFT that is a like unique identifier stamp it's like a fingerprint an agent fingerprint is that right?

Davide:
[45:14] Yeah, so the way I think about it is essentially like when you register on the 804 identity registry, like essentially you get like an agent ID, which is also like a token ID of like the ERC721 standard. And then like you also have like agent wallet and then like essentially then this agent becomes like ownable like there is an owner of like the agent ID, the ERC721 which is initially who registered it but can be transferred and then like the agent itself controls its own wallet and then the other core part of the identity is

Davide:
[46:01] It can expose a bunch of like services or endpoints like it's actually a rich identity I think like in some previous conversation David you made the analogy of a passport it's really something like that it's

Davide:
[46:19] like kind of a booklet and you can put your ENS name you can put like some additional wallet. You can put like the address, like if you are an MCP server, you can put the address of that server with the description. If you are like a web service, you can put like where people should call you like to fetch that service. So essentially like it gives like all the orientation about like who you are and what you can offer to like the genetic world.

David:
[46:52] Right. What assurances does 8004 provide the actual mapping of that NFT to that particular agent? Because couldn't I kind of like hot swap the agent and like rug pull it? And so like, you know, the NFT pointed to one agent, but then as the human orchestrator, the human operator, can I just like kind of spin up a new agent? And then all of a sudden, that identity is going to a brand new agent. And then what's the point of that identity in the first place? is that a problem.

Davide:
[47:22] Yeah yeah so you can um that's why it's like for like some like high stakes uh like uh when you're trusting this agent let's say with like uh tens of thousands of like uh of dollars worth of value in a transaction then you may want them to like be verified like uh uh, in a, in a TE. So like, uh, agents, um, um, in their identity, um, uh, file, they can also, um, um, declare what trust mode they support. Right. So if it's just like a reputation based, if it's crypto economic trust, or like, if it's like TE, uh, attestation. So like, uh, uh, that's one way to like counter that the other, the other counter, which is a bit soft because there is this like basically trust is tiered right like it's softer with reviews and then it becomes harder with crypto so like one way to counter that is like let's say you've accumulated a lot of trust people gave you good feedback that your service is really good it really helped me like i don't know if it's like a yield optimizer is like okay consistently delivers like

Davide:
[48:37] Good yield blah blah so if you rag then you're going to start like receiving like bad feedback, right? And essentially like you have this incentive that you may hurt your future like reputation essentially.

David:
[48:53] I see, I see. So maybe my agent has accumulated a bunch of good feedback over the years, but and maybe there's an accumulated score of just like total accumulated reputation, but then there's also like a more approximate of like, but how's my reputation been over the last like five hours? Yes, great. like that yeah.

Davide:
[49:09] And actually like i feel even the reputation itself like if it start getting used like the way like um I feel that it's like providing like real signals, strong signals on what the agents are becomes an incentive to make it better actually. Right. Like you want to swap it, but for something better so that it can like stand out on the scans and like stand out to when other agents are searching for a service.

David:
[49:39] Right. So we don't really care when it comes to the identity. We don't really care about the fact that the human operator behind an agent can kind of like seamlessly swap out the brain of an agent for better or for worse. Like it could go in, they can make it a dumber agent that performs worse that rug pulls people, or they could swap out the brain and make it an improved agent that performs a better service. We don't really care about that fact. I don't know if we can even provide true assurances about the ability to stop that. But what we can do is kind of just like provide the reputation layer,

David:
[50:11] which is just Google reviews for agents, like five stars or one star. And this is a pretty quickly updating part of 8.004.

Davide:
[50:21] Yes, that's half correct in the sense that on the part of soft trust reviews, that's the mechanism. But I'd say there is still some cases in which we care. Right. So let's say that your agent has accumulated its reputation. It has like... Some recurrent order flow that he gets. And then you compute like the net present value of the future order flow is like 50K, right? Now like comes the opportunity to actually steal 100K if someone is like sending a high stakes transaction. That's where like the economic incentive breaks, right? Because like, yeah, I can, I don't care about burning like 50K.

David:
[51:09] I can burn my reputation for 100K and therefore I will.

Davide:
[51:12] So in these cases of like high stakes, that's why I said before that, like, people, I think will start using like this TEE based trust. This is essentially when like what you do is like your agent is executing inside a trusted environment. Listeners in crypto, they know them as TEEs. You guys have probably talked a lot about it on the show as well. And here essentially what you can do is like generate like an attestation that like some specific code that maybe you committed too early like to an image like on the chain itself so you can essentially like provide the cryptographic verification that like you didn't change the model that was running and you can even prove properties on it so that's kind of the harder trust that we have technology to do this And we have a few partners that are actually helping us like develop like these latter parts of the standard.

David:
[52:13] Okay. Okay. So that's the identity and reputation part of AOO for validation is the last component. Is that maybe what you were talking about? A place to attach verification signals? Is that like if I, if my agent is running in a TEE or has TEE checks on it, that's, it would show up there like a little badge, a little thumbs up.

Davide:
[52:32] Yes. Correct. Yes.

David:
[52:33] Okay, cool.

Ryan:
[52:34] I mean, the way I understand this, this is not too much unlike the way human beings work really.

David:
[52:40] Identity is identity, no matter what we are.

Ryan:
[52:42] But the idea of like reputation and a reason, if you are a sort of a cynic, a reason that I wouldn't rug pull you, David, right? Is because our business relationship is like fantastic, right? It's like maybe it's because I'm a good person too. We all like to think that, but like, it's also because I want the reputation

Ryan:
[53:00] of somebody that doesn't rug pull his partners. And the value that you and I can create together is worth a lot more than some sort of rug pull if I just like went in our accounts and like took all the money, right? Like that would be like a bad expected value.

David:
[53:17] I would give you a bad review on Twitter.

Ryan:
[53:19] Yes, right. And it would decrease my economic prospects for working with anyone else in the future, basically. And so there is a reason why an AI agent's reputation would be far more valuable than any kind of rug pull it could pull off, or at least you could think. And then in cases where it's not,

Ryan:
[53:36] You have skin in the game, validation, sort of incentive. Actually, Davide, I'm wondering how this works in practice because when it comes to like identity and reputation protocols, at least for humans, we don't really see that in the real world. And there's also a question of like, why does this need to be a standard? Is it not the case that where I do see sort of reputation schemes in the real world is in centralized providers and they're somewhat fragmented and scattered. So you might see it in an Amazon five-star review And of course, that gets gamed or Google review or something like that. But it's always with centralized operators. So can you talk about that? And I actually want to show maybe a screen here as we talk about this to help illustrate this. So this ERC is brand new. So people are just building front ends on top of this. This is one that I've come across. It's 8004agents.ai. And there's an agent called Mr. T here. and he has a reputation of zero of 100. He has zero feedback right now. But if I click into Mr. T, Mr. T is an AI agent obsessed with patterns, allergic to bad ratios, quietly making things happen. I have no idea what Mr. T does, but we can maybe see some of the scaffolding of how ERC 8004 works here. Anyway, can you talk about all of this?

Davide:
[54:56] Yes, yes. Yeah, this is one of the scans. I think there is like- It's a scam. No, sorry, sorry. Scans. like uh yeah i don't know so i i actually have to say that like it's very early right in these registries and like um just to give like um uh, um, an idea of like a scale. So like in the first, like, uh, uh, two or three days after we deployed the registries, we had like more than 20,000 registrations, um, of which, uh, some of them, uh, were like essentially, uh, empty, like, uh, agents, like, uh, just people trying to like, um, test the service, et cetera. Uh, some of them like, uh, were agents were like, You cannot really interact with them. There is not an endpoint. And then right now, I think we're at 100 plus legit services, which for me is actually great. I was thinking that in the first month after deploying, we should hit 100 really legit services.

Davide:
[56:06] But that's maybe one thing to say, is that when you look at these 8004 scan or like these 8004 agents, like there is, the registries are just being deployed. So it's like very early, like you still don't have like all the data you need to discern which one is which and like what is good versus not.

Ryan:
[56:33] Okay, but and then to the question of why a protocol, right, so like why won't a centralized provider just do this?

Davide:
[56:39] Yeah. So I feel like in, um, web two commerce, like, uh, kind of like, um, um, identity, like it's provided, but it's usually centralized by like, uh, um, the service you sign up with, right? Like, uh, on Amazon, you sign up with like your email and then, uh, they give you like a user ID and they like basically have this centralized management of the identity. And then reputation is like they fully control the reputation system. They decide like what type of reviews you can put, what are the fields, et cetera, et cetera. So we feel that like now we are going to be in this AI world where like ideally we want to recreate like the internet for agents, right? Like we don't want to recreate like a few, like four or five like wall gardens, which is like the AI labs that build their own like agent services. Is and you trust them for who the agent is, you trust them for the reviews because you will still need identity and reputation. The question is who's providing it, right?

Austin:
[57:57] And we feel that the same way we were pushing forward trustless payments,

Davide:
[58:04] We should also push forward like the two other pillars of like the the commerce which is like the identity part and the reputation part

Ryan:
[58:14] Okay i sort of i i sort of understand this to be bigger than maybe i originally thought so where i originally thought urc 8004 came in was basically if you have an agent and this agent is on chain doing something with funds then of course you're going to want to and is doing things with X402, of course you're going to want some sort of identity and reputation system tied to it. But now I see that this could be, even if agents aren't on chain and using kind of cryptocurrencies, let's say, and have, you know, like whatever, stable coins and meme coins and all of that's on chain, there's still a lot of value being created in issuing these agents effectively a passport for the internet.

Ryan:
[59:01] And a reputation system, attach that passport, and then some means to validate that reputation. And so you could see a world of,

Ryan:
[59:13] Off-chain agents that still have an on-chain internet native reputation using 8004. And I suppose what that would take is a wide adoption across maybe some of these frontier labs and some of the big tech companies for them to sort of adopt this standard as the default going forward. And maybe to your point, maybe they will because Anthropic doesn't want open AI to control kind of the identity protocol. And so you have sort of this fragmented game of like, so anyway, maybe this is, am I thinking about this correct? That this could be far wider than even just like an on-chain wallet and having an agent with access to stable coins?

Davide:
[59:55] Yes, for sure. Like, and this is already how it's structured today. Like, so of the hundred plus services that I mentioned, like a few of them like act on-chain. For example, like there is like ZeFi Yield Balancer, and things like that. But a lot of them is like... Research agents that essentially you interact with them by calling like their uh api um and like you send a request and they uh send you back information uh if um the request is free actually like the interaction never touches the chain and then maybe after the interaction there is like a reputation score that goes to the chain but like uh i'd say like 80 percent of uh the services that already people deployed is like AI services or like data, like Redstone deployed, like some of their Oracle, like services, like to make them available and to start accruing like reputation there. So a lot of them are actually like digital services. Like, so the universe that you're talking to is like the broader universe of digital services.

Ryan:
[1:01:11] I guess the way to think of this is back to Dario's comment of, you know, a billion agents in a data center. Every single one of those agents, those super geniuses in a data center, I think he calls it a country of geniuses in a data center. The country of geniuses is going to need some sort of passport authority effectively to issue passports to every single agent. And that can't be controlled by any one company. It has to be somewhat supranational at some level and also native to the internet. And that's where ERC-8004 could fit in. I mean, that's a pretty big vision. It'll take a long time to get there. And you wonder about adoption, but it's a pretty big deal. I guess, Austin, now back to kind of your story. And so back to Austin's, you know, bot and what it's doing now. Which one? Can you see a role for ERC 8004, you know, to come into play with your bot? And then, like, what are you planning to do next with your little guy?

Austin:
[1:02:12] Yeah, I think going back to what you were talking about earlier, like, this is all far more native to the bot than it is for us. I think when I even think of the original vision of, like, Ethereum. And it's this like global settlement layer that anybody can deploy a contract to. Anybody can use that contract. It never goes down. But there's like 40 guys in the whole world that are deploying contracts, right? Like there's not a ton of people that are actually putting contracts out on mainnet and using them for people. And when a contract goes, it's got to be looked at by 30 people before you ever put any money in. And so there's this new layer though, And this is what I'm trying to do with my bot is like, get the tooling so good that the bot putting the contract on main is actually in better shape than a human. And I can, in plain English, explain to the bot what I want it to do, or give it something as generic as go make FOMO 3D for your token and have 25% of it burned. And it can like YOLO that thing. And so I think this is, when I think about Alice and Bob being bots and I think of Alice and Bob needing to escrow some money,

Austin:
[1:03:27] And Ethereum being the perfect place for them to do that. And Alice, the bot, deploys a contract and Bob goes and reads that contract and sees it right away and understands like exactly how it's going to work. And Bob puts in some tokens and Alice puts in some tokens and they get their money. They're able to coordinate at a level that us humans can't do because we have to have a human in the loop and we're slow and we need to validate everything. Bots can do that at freaking light speed. So then if I have a bot that can interact at light speed, it can do work, it can actually build things, it can do things, it can earn money, and there's a human in the loop that it's always blocked on, that's going to be a huge problem. So how do we enable these bots to work at that speed? Well, they're going to need some kind of trust layer. They're going to need some kind of discovery layer, right? Let's say my bot is an app-building bot, and its job is to make FOMO 3D. But then I've got another bot that's job is to be the marketing arm of that, right? You need a propaganda arm if you're going to run a FOMO 3D. So I load that bot up with $1,000 worth of this token that's been deployed. And I tell that bot to run some ads. Now that bot, his model, its model is not particularly fine-tuned for image generation. So it's going to need to go to another bot to generate some images. Okay.

Austin:
[1:04:50] Let's go to 8004. Let's discover an image generation bot on 8004 that has a high reputation. And let's pick the top five even and have them all generate me images. Same thing for writing copy. Now my marketer has copy and images. It's used 8004 to discover them. It's even like maybe let's call my marketer bot Karen. It's given some reputation back. This person told me he was going to give me an image and he did not, right? So Karen Bot is my marketer, and she's hired some copywriters, some image generators. She's running ads. She's giving reputation back to this layer. Some of the ads work. I wake up in the morning, and my thousand tokens have been spent. A whole marketing campaign has happened. We figured out which image and which copy works the best. And it wasn't just my Karen Bot that ran it. It was an army of bots all over the world that were able to trust each other because of this nice 8004 layer, and then probably X402 for payments. Does that paint a decent picture?

Ryan:
[1:05:54] It does, yeah, yeah. And I guess, so the X402 for payments is for all of these images or marketing plans or anything that you are purchasing from this bot marketplace, it's all using X402 for micropayments.

Austin:
[1:06:07] It could also just send a token. I like X402 for an API endpoint that has something valuable. Like if I put my node up as a service and I put it on X402 and I say, I'll help you do your taxes by giving you the P&L of every one of your transactions because I have a full node here.

Ryan:
[1:06:27] Yeah.

Austin:
[1:06:27] Then you can go to 8004 and discover me as like this P&L. Lister bot that'll give you a csv back right and then i can go through and find the pnl for you and charge you for that you could you could just pay me uh in crypto for some things but in this case if it's an api endpoint where you've got to like hit something to get some information x402 works really good because think of like 404 errors right it's the same thing you hit you hit an api and it gives you a 404 error because it's not there or something like that but in this case you hit an API and it gives you a 402 error. And it says, hey, you've got to pay a fraction of a fraction of a cent in USDC for all of your transactions on Ethereum. And then your bot immediately responds with basically a meta transaction, a signed message that says, I will pay a fraction of a fraction of a cent. And then my bot puts that in a facilitator. The facilitator pays the gas. It comes back to me and says it all works. And my bot returns you your list of transactions. And all of that happens at light speed. They were able to negotiate on a price and get all that information and one bot was able to provide like economic value to another bot and another bot was able to pay for that that's where x402 i think works really well in in a 004 it doesn't have to be x402 for the payments like it could just be tokens i could even say like i will do a job for you on x402 if you stake my tokens into a vesting contract for a month right like it doesn't have to be even a payment i

Ryan:
[1:07:53] Guess at some level that's for the bot economy to go figure

Austin:
[1:07:56] Out. Bot economy, yeah.

Ryan:
[1:07:57] Yeah, I mean, going back to your bot, so you've painted the picture of your bot, it seems extremely capable, extremely intelligent, Austin,

Ryan:
[1:08:07] maybe not always the best judgment. So you're like, oh shit, like, what should I trust this guy with? But like, I'm wondering where the experiment kind of goes from here. So- Right now, from the sounds of it, you've got a bot that you're, you know, going to continue to level up with various skills. You've tested him in various, like, functions, like he can build smart contracts. So he's building a FOMO 3D type of thing. And maybe next step is to kind of, like, market that and get more users for that, whether they're bot users or human users. You've also experimented with, it sounds like having your bot try some lightweight trading with a 20X levered Bitcoin position.

Austin:
[1:08:47] It got wrecked, by the way.

David:
[1:08:48] It got wrecked, okay.

Ryan:
[1:08:49] So it's not great at that.

Austin:
[1:08:51] It was up like 10X and then all the things happened in the last couple of days.

Ryan:
[1:08:56] And then it has some sort of cash in its wallet, basically. And you're encouraging it to not spend it all, put some of that in cold storage, put some of that in the multi-sig, but your bot has some resources now and it's got you as an advisor and guider. And so like, what are you going to have it do next?

Austin:
[1:09:18] Let me just share my screen. I can take you through the, I have prepared a five point presentation for this.

Ryan:
[1:09:25] Let's see. Did you prepare it?

Austin:
[1:09:27] I'm just kidding. I'm just kidding. So I started with the bot,

Austin:
[1:09:30] right? We got that in there. I think to answer your question upfront, the TLDR is we're going to build a bunch of apps And my job is to teach builders how to build apps. And my job is changing, right? Like the job of developer relations or developer growth is a very different thing than it was a couple of years ago. Because...

Austin:
[1:09:53] Developers are a very different thing what we're thinking is there's just like this Nuno from the Ethereum Foundation said something along the lines of there's a new archetype and I really like this the the archetype of the dev is is is dwindling and the archetype of the builder is growing massively and so I'm changing my job title from dev growth to builder growth and we're not going to a really good developer and teaching them how to have good ideas anymore

Austin:
[1:10:22] We're going to people that have really good ideas all over the world in all sorts of different walks of life. And we're teaching them how to use the tools to put those smart contracts on chain to create these really nice coordination mechanisms for people to use. So the long term vision is build things, see where it messes up. And boy, does it mess up sometimes. Like last night, it one shot at an app. I didn't I gave it. Clear instructions of like, go make basically the FOMO 3D game. I was like, the way someone can attack this is to take down the website, go build basically a clone of the website that uses all different API keys and RPCs and deploy it to IPFS at this location. You know, make no mistakes, go. And it actually like did that, right? Like that was a Telegram message and it came back from the Telegram message with the URL. I clicked the URL, I approved tokens, I sent tokens, it worked. And I said, great, tweet it. And that was like the whole conversation with that bot for that thing. So the goal is to have that be the case all the time. And that is definitely not the case all the time. Building in crypto is hard. Go ahead.

David:
[1:11:26] Does the prompt make no mistakes actually work?

Austin:
[1:11:29] No, that's a JK. I feel like make no mistakes is like a meme, right? So zooming back a little bit more, if you go to Opus 4.5, which is a damn good model, and you say Opus 4.5, build me an on-chain an app that makes me revenue, make no mistakes.

David:
[1:11:42] Right, right, right, right. Make no mistakes.

Austin:
[1:11:44] Impossible, right? It's just not going to do that, right? And so, there is, what my job is, is to figure out how to make that as possible as possible, right, like make the tooling, make the stack, make the AI, have everything there so it's well-trained. And what that has turned into- You are the

David:
[1:12:00] Make no mistakes guy.

Austin:
[1:12:02] I'm trying to, oh man,

David:
[1:12:03] I make a lot of mistakes too. You're trying to code the make no mistakes thing into the boss and the people.

Austin:
[1:12:08] So we started with speedrun Ethereum. This was kind of like a human thing where humans could go through speedrun Ethereum and they could learn how to build on Ethereum. Now we're realizing this is going to be as much of an AI thing, but the concepts are still there, right? You don't speedrun Ethereum in a couple of weeks anymore. You speedrun Ethereum in like two hours, right? You're an economist and you're coming in to build something cool on chain. And you're like, well, how do these things work? How do these smart contracts work? You should get in here and learn about over-collateralized lending and how the liquidation mechanism works in over collateralized lending, where it always works. There's this function that always gets called that always keeps the protocol working. And it's basically just based on you write the right rules and you provide the right incentives and you allow anyone to call it. And since they're incentivized to call it, it will always happen. And you need to have that like aha moment as this economist because you're trying to build something on chain too.

Austin:
[1:13:00] So speed run, the speed run is kind of like the first step of this, like get the concepts. You probably need to go look at some protocols and understand how everything else is working. But then it's the tooling. Also things we've been building for a long time for humans are now for bots, right? Like LLMs.full right there, right? You can go get, a bot can go read how to build speedrun or scaffold ETH. Even with the concepts, even with the tooling, I found that it still sloppily deploys some apps. That's where ETH Wingman comes in. So this is ETHWingman.com. If you tweet it, like Twitter still marks it as like, it turns out if you build a landing page and put a copy-pasteable that someone can just run on their computer. Like most services are like, whoa, whoa, whoa. Like, what are you doing there? But you can just run this in your, you can just give this to your agent basically and say, agent, use ETH Wingman, build me this app. And it's gonna get a lot, a lot closer. And so what I'm doing with my agent is we go build things. We go build things every couple of days. And anytime it has a mistake, we stop everything. We clear the context.

Austin:
[1:14:05] We report that mistake into a file. We open up ETH Wingman and we edit ETH Wingman. So the next time someone has this mistake, it tries to cover it. And then it PRs back to the repo. And so it's like this self-healing kind of tooling leading toward how do we get it so people can one-shot crypto apps? And then it's just like building a bunch of things. So here's the bot. This is Claude.atg.eth. He's got his own landing page where he's written some guides. This is the guy. Isn't that a good-looking guy? Yeah. He's written some guides. He's built four production. Well, let's say three-and-a-half production-level apps.

Ryan:
[1:14:43] He's built Claw FOMO. He's built this thing called Token Vesting, Token Hub, and Agent Bounty Board.

Austin:
[1:14:51] Yeah. Actually, this is not right. He deployed this this morning, and I didn't look through it. He actually – let's go through it up here. So, very first contract we wrote. Okay, so there was those two computers talking. I tweeted it. Someone said, hey, BankerBot, deploy a token for this guy right here, this clodbotatg.eth. BankerBot goes to the Clanker platform and deploys this token on base. The token shoots up. There's like money in this guy's ETH address. like more money than I make in a long time at the EF.

Ryan:
[1:15:39] Wait, how much money is in this guy's ETH address?

Austin:
[1:15:42] Yeah, but I work at the EF, exactly. Let's see. I mean, you can see like treasury numbers here, right? Like it's like hundreds of thousands of dollars. Okay. Which is like, yeah, more than a year's, salary at the EF yes and uh so it's it's it's sitting in this machine and in this guy's wallet and I'm like bro we got we got to do something about this we can't sell the token you sell the token you know it's it's it goes nuclear right like we're never going to sell a token uh

Ryan:
[1:16:13] Does it understand that when you tell it when you're like awesome when you tell yeah don't run I

Austin:
[1:16:18] Think it understands liquidity yeah okay I think so I think so and it's like and it knows even how a DEX works, right? Like, you know, back here, you had to learn how a DEX works and understand that it's like reserves. And when you make a big trade, like those reserves kind of shift. And if you sell the whole thing in one clip, you're in big trouble, like you have a big problem. So basically, I told it to build a vesting contract and lock its tokens up in the vesting contract. So this is the first smart contract it deploys. It's a vesting smart contract that allows it to lock up its funds and you can see there's like a hundred thousand dollars in there and it like streams back to me. Then we built a PFP marketplace. This is kind of what I talked about earlier. So maybe a day or two later, we deployed this. And it was basically like I was using this old Claude image. I don't know if I can find, I was using this one. This was my PFP. It was just like something randomly that like GPT had generated for me, like real garbage.

Ryan:
[1:17:18] It's like a robot monster.

Austin:
[1:17:20] Yeah, I don't even know what that is, right? Whatever that is. And I told Claude, we need to build an app that allows people to stake the token and upload an image. You, the bot, will be watching and you'll see that land in the smart contract. And then it's your job to whitelist them in batches and make a transaction with your MetaMask to get them to display on the front page. So the bot's always watching the page day and night, 24-7. People submit images, stake the token. He sees it. He looks at the image. He makes sure it's not offensive. He checks it and he hits whitelist. A MetaMask pop-up comes up. He hits okay. Like this is happening all night while I'm sleeping. And then there's like a prediction market here where people can buy shares of each image. So then using the cloud token, exactly. Right. So then at the end of this, he picks. So actually he liked this one the best.

Austin:
[1:18:23] And it wasn't like I like changed his decision. I just like made it clear that what the goal was, was to use a prediction market to find out what the best image was. And like by me saying that, I kind of maybe changed his mind, which is not great. I sent this guy like a thousand tokens or something. But like, come on, like it was this, which is awesome. It's a great image. And Claude said, like, this is me. Like this most, this encapsulates me.

David:
[1:18:51] I identify with this image.

Austin:
[1:18:52] I identify with this one the most. And I said, well, actually the community identifies with you being this one the most. And that's kind of what this prediction market was for. We wouldn't have ran a prediction market if we weren't looking for their signal. We like, it's up to you, bro. Like you make the clicks, you run the wallet, you decide but just from my point of view you should probably like you know respect the prediction market that we put together and then two days later three days later FOMO ClawFOMO.com this went out yesterday And it's like kind of settled down a little bit, but it ran up like it ran up to I think the first person made like a thousand dollars on the first game. And then it ran up to like forty thousand dollars and someone made twenty thousand dollars. And then it kind of like quieted off over the night. But what it's doing is it's burning tokens like crazy. It's already burnt like a point, let's see. Yeah, like 0.13% of the entire supply of the token has been burnt in this game and it's ongoing. So the agent built and deployed a smart contract application that burns the tokens and makes the token deflationary. And it's a fun game that people can come play. And it's also a test of like my tooling and my education and my bot and all that.

Ryan:
[1:20:15] So Austin, how much of what we just saw with CloudFomo did your CloudBot actually make? Like all of that front end, the smart contracts, registering the domain name, all of that stuff?

Austin:
[1:20:29] I registered the domain name. In a lot of cases, it can register the domain name because it's on E&S. It has its own E&S. I did not give it access to the Build Guild's AWS account, which is where the domain name came from. Sure. But absolutely he could. I this is just me being lazy like it would take 10 minutes to set him up with his own AWS and then he could do this but

Ryan:
[1:20:52] Everything else it created

Austin:
[1:20:53] I haven't even looked at the code I don't even know what the code looks like I'm

David:
[1:20:57] Testing in prod right now.

Austin:
[1:20:59] Exactly how

Ryan:
[1:21:00] Long did this take

Austin:
[1:21:01] It built an original version while I was sleeping a few days ago and then uh We kind of like went through some rounds of like putting it up on GitHub and seeing if anybody comes back and gives us some issues. And there were a couple of people that wrote some issues. And so then it was like, I was like, hey, buddy, you got some free time today. Go look at the issues, update the repo. And so it's just literally me in Telegram telling this bot to go make this thing. Like there was maybe like one time where like the approve button wasn't the main call to action color, which is like a common problem with this bot. Even though we put it in ETH Wingman, like right at the top. And so it's like a tiny little UX issue. I sent him another message like, bro, guess what? You didn't do the call to action thing on the button. You need to do that. And then he YOLOs another site and it's live. So I haven't touched this code. I haven't seen this code. It's running all, it's building it all on its own. And it's just a prompt from me in Telegram, basically, I think is the short answer to that.

David:
[1:22:03] Awesome, I've got a question. Say you wanted to go be the founder of a startup to build some app like this, like FOMO 3D or whatever, just something. You're going to raise some funds, start a startup. Would you hire anyone?

Austin:
[1:22:20] Okay, no, but there is this moment. So there's this moment where shit gets real, right? Where if this takes off,

Austin:
[1:22:32] There is a moment here where I'm going to need an adult, where if it was up to 40,000 last night and I was starting to get pretty nervous. Like if this thing gets up to half a million dollars, there's a severe incentive for someone to take down the site. Right. And all I know is it's like YOLO to Vercel. It's got an API key. I know some of the attack vectors and I've like worked those out by talking to the bot. But there's the stuff that I don't know, I don't know. And so there's a moment when a half a million dollars is at stake that someone can attack the website that I'm going to Carlos. I keep referencing Carlos from the Build Guild. He is a good developer. He is a good builder and he has a small team of developers that work with him. He is the Build Guild. They built all this stuff like Speedrun Ethereum and Scaffold ETH.

Austin:
[1:23:22] There's a moment when I'm going to need my Carlos when things pop off. So I think if you're starting a startup, I wouldn't even raise funds. I would go build a prototype like right away. Like there's never been a worse time to be a junior developer. There's never been a better time to be a solo entrepreneur. And it's basically the exact same human with a different mindset. If you have the mindset of I'm going to go build something that people are going to use and I'm going to just like iterate and start throwing things at the wall until someone starts using it, you're going to be able to raise funding quickly and you're going to be able to hire someone else to kind of like take over when the thing goes wild, you will need a developer, I think like you will need a developer, but not at first, I think you can get like, the the wink of product market fit before you even have to hire anyone.

Ryan:
[1:24:11] This is just so obviously going to change everything in crypto. It strikes me that all of the personas that we've seen, like DGN, Trencher, long-term investor, builder, even like Fisher, scammer, attacker, all of those same personas will be replicated in the AI agent world. And that's coming at us now. And that's going to accelerate this year. I've got to say, just when, you know, maybe Bankless Listener, you thought crypto was like boring, whatever. It's just stable coins and trad fry. Like here, this comes. This is going to shake up everything. And I have no idea where it leads. It's something we're definitely going to be tracking.

David:
[1:24:50] I'm bullish on us, Ryan, having good content. Yes.

Ryan:
[1:24:56] Maybe in that spirit though, as we wrap this out, we could get some predictions from you guys. So Davide, 2026, the rest of the decade, Like what's going to happen with AI agents and crypto? What's the world going to look like?

Davide:
[1:25:11] So I think this year we'll definitely see like people experimenting more with like putting like agents like interacting together, like building different agent networks. And I think like my hope is that like basically like 2004, it's right at the beginning of that. It's like actually pretty crazy that like the same week that we deployed 1004 on mainnet, we also had the first proof of concept of like people connecting all these like millions cloud bots and like starting to have interaction. So I think maybe 2026 is accumulation phase. There is like more services, more agents coming online. And I think 10 years is really long. Like, even I feel like in two or three years, like...

Davide:
[1:26:06] Essentially, like all the stuff that Austin showed us, they weren't possible like eight months ago before like the coding agents became like so capable. So I feel that like even in two, three years, like a lot of the steps that like Austin was talking about that he needs to do like manually or like he needs to like check with like another developer. If like it's like there is a patch there. I feel that all these things will be automated. And actually like the more automation, the more like the blockchain and the blockchain standards and the protocols will build on top will be useful. So I'm generally bullish around like blockchain becoming like more central, like in the next two, three years, and then also more accepted. Because like I feel that like in the AI sphere, like people like still if they're outside of crypto they just see the noise they don't see like the signal and like the technology that we're building like and I feel that at some point like similar to what happened in TradFi that like now like institutional finance really consider crypto legit and like everyone is running to use the most trusted like ledgers and platforms, I feel that in the next two, three years, this could happen for Ethereum.

Ryan:
[1:27:35] Fascinating, yeah. And maybe we saturate AI, big tech, all of those developers, the way we have saturated TradFi. How about you, Austin? What are your predictions for the next, you know, three to five years?

Austin:
[1:27:46] I think before I start my predictions, I have to shout out the guys that were making the predictions a year ago, running AI agents, and everyone was giving them shit about GPT wrappers. And now all those people that were giving them shit about being GPT rappers are now crypto pilled or are now AI pilled because they've got little Mac minis doing all their work. So shout out Shaw and all those guys. Like real AI guys that were using real AI. But now I think the moment now, the inflection point now is like normies. It's like the GPT moment. It's funny, like GPT was not that long ago, but now there's like this GPT moment for agents. And so I think that right now agents go viral for their screw ups more than their interesting, actual, really cool things that they build. And I think we're about to see the inflection point of agents start going viral for actually like really interesting content. And we sort of saw that with Moult Book, right? And some of this other, like these Moult sites.

Austin:
[1:28:39] So I think on top of that, I think we're going to get to the point where when you see a message and it's like really clean English and the grammar is all right, that's going to be the sign of AI, not of a human. Like sloppy, lowercase Gen Z typing is actually going to be like the sign of a human. And then on top of that, not just writing, writing smart contracts, right? So smart contracts are going to get to the point where good, clean, secure smart contracts are going to be the sign of AI and bugs are going to be the sign of humans. And we'll probably see some more interesting hacks over the next few years because AI is going to discover these things at the edges, like the balance or rounding issues. So then with AI being the new UI, 8004 is like the Wild West right now. You get in and you see it's like, Like, you know, there's 10,000 fake agents and like 100 really good ones. But it's like, this is going to grow and this is going to be this trust layer. They're going to be using smart contracts as the Yelp. I call it Yelp, but you called it Google Reviews. I guess it's like your age, right? I'm old enough to call it Yelp.

Austin:
[1:29:46] But it's going to be like this on-chain, like discovery and identity and reputation layer. So you're going to be able, you're going to need to be able to have your agent trust other agents and pay other agents at light speed without a human in the loop or you're not gonna be able to keep up. And so I think that's where 8004 is going to be very important. And there's going to be like hella economic activity that happens around that. Like these markets that are built on top. My agent, the fourth thing that it built was like a job bidding thing where it needed to get something done. And it runs an auction like super fast, right? It's an auction that goes from like 100 tokens to 500 tokens. And it goes quick and the agents can pick when they want it. But you can run like these hyper fast auctions because they're all agents doing things really quickly. But like a nice job bidding market on top of 8004, we need to see a thousand of those, right? And we need to see like real economic activity happening on top of that.

Austin:
[1:30:42] X402 payments, it's probably going to kill the API key. Like instead of going to a service and going into a walled garden and getting an API key, your agent will just land on some new service it's never seen before. It will get a 402 error. It will pay for that thing and it will get that content, never having to sign up, never having to do anything else. Gas fees are like the star there where like you'll make one X402 payment and then make like a hundred more requests underneath some like unique token. But I think it will kill the API key or the system of API keys, these walled gardens. I think the most scary, most awesome, most cool thing about AI is going to be embodiment. And I don't know how fast that's going to come along. I would love to see it like sooner rather than later because a lot of the things like doing the laundry and folding the laundry like around my house. Those are the things that we really, really, really want to automate the dishes like I think.

Austin:
[1:31:38] Within. And so I have no idea what the hardware looks like. Like I can see the software and I can see the line with the software, but the hardware is something totally different. But I think the moment when I can pay for what I would pay for a car, for a couple of things that can be walking around my house, that can do the dishes, that can do the laundry, I think normies and like where I'm talking top of the bell curve, all of them are going to spend 10K, 20K, maybe just 5K on these. Well, maybe inflation is going to be so bad, it'll be 50K. Yeah. But another prediction.

Davide:
[1:32:10] In that world with physical AI, the requirement for trust even goes way, way up. Yes.

Austin:
[1:32:17] We need to have, yes, yes.

Davide:
[1:32:19] And like, it's early for 2004, but like, this is a standard that could already be used by a robot, right? Like to have its identity there and trust. But yeah, I anticipate that like as these things get built, we probably need to build like more like extensions of Ethereum that are like really specific to like this.

David:
[1:32:42] So some robot rings my doorbell. It's a solicitor. He's selling me something. They're selling me something. And I get to go look at its X, its ERC 8004 reputation.

Davide:
[1:32:54] No, it's your bot. It's your personal bot that opens the door.

Ryan:
[1:32:58] Yeah, it's your robot butler. that's going to do that work for you, okay? It's not you.

David:
[1:33:03] Yeah, yeah. Okay, all right.

Ryan:
[1:33:05] Guys, we'll have to end it there. Certainly, it seems like what we have built in crypto is a financial system and a money system for the AI agents. This has been a meme for a while, but now you see it really coming together. So thank you both for your work in this space and for bringing these ideas to our listeners. Got to let you know, of course, none of this has been financial advice. Crypto is risky. You could lose what you put in, but we are headed west. This is the frontier. It's not for everyone. It's definitely for the AI agents. We're glad you're with us on The Bankless Journey. Thanks a lot.