Bankless - AI ROLLUP #2: Virtuals $1B | Zerebro LLM | Ai16z Comeback? | Bittensor Tao | Freya $50K Game

Starting point is 00:00:03 Welcome Bankless Nation to the AI roll-up where we cover the recent news, developments in drama in the intersection of crypto and AI. Curious and creative developers are playing around with open-source AI software, making AI agents, giving them access to crypto wallets and unleashing them into the wild west of the internet. And we are doing a special series of episodes on Bankless with a special co-host, E-Jazz here to help make sure that you are up to speed with one of the craziest and fastest growing corners of the internet. E-Jazz, episode number two. How are you doing, my man? Doing well. It's been a pretty entertaining week. I'd say entertaining in both a market and fundamentals perspective, but also these agents are getting better. They're making me laugh more. They're making me interact with them more. It's pretty insane to see. You have definitely seen it all as well in crypto Twitter. This one agent AI, XPT, has flipped, I think, all other humans in mind chair. And I think that just represents a growing quality of these agents, but also, growing like penetration of the awareness that like yo the AI the AI meta is the meta is time to

Starting point is 00:01:08 pay attention yeah yeah I agree and if you look at like kind of like the mind share analysis of the last like week it's like grown pretty exponentially like I'm pretty sure AI agents itself is or AI within crypto Twitter is like at almost 45 percent so loads and loads of people are discussing it across a range of different verticals whether it's in defy specific agents or art specific agents, you see it pretty much permeate everywhere. And AIXPT is a good example of seeing that happen. I think actually when we last filmed an episode, it was at 40,000 followers on Twitter. And I think right now it's just broached 80,000. I might be wrong about that, but it's gone pretty exponential. Well, we're going to pull up some of its tweets this week. And so we'll actually be able to check in on

Starting point is 00:01:57 that. We're looking at kind of this mindshare of narratives. I think over the last two years, crypto has been looking for a narrative. Like what, what is going to be the narrative that takes crypto mainstream to the next, for the next bull market, right? Is it going to be meme coins? Is it going to be real world assets? Is it going to be stable coins? Is it going to be GameFi? AI has been in the running, of course. I think in the last two weeks, we are ready to like unquestionably crown a winner. It's AI. Everything like, you know, get out of the way. You're not even in second place. People are trying. There's no, there's no second place. It's AI. AI has won. AI is the thing. It is bringing in new attention, new users, new capital,

Starting point is 00:02:36 which is just the makings of a bull market. Let's talk about what we're going to talk about this this week. What's on the agenda this week? We got Zero Bro. You're going to walk us through some stuff with Zero Bro. We talked about Zero Bro quite a lot last week, but of course the matter always shifts. AI 16Z is it regaining its footing after some recent stumbles. The virtual's ecosystem, which I think is where we're going to start, is on fire. This is the number one agent AIXBT that we were just talking about is dominating a crypto Twitter KOL, mind chair. And then Elon Musk gets peaked by an AI agent game on base. And so this is just a quick snapshot of what we're going to talk about. EJazz, I want to start off every episode with kind of the same vibe because we're all

Starting point is 00:03:14 learning here. I'm learning. You know a lot. But even the people that know everything in the AI agent space, everyone's still learning here because we're on the frontier. Just give it to me once again. Why are you so bullish? Why are you so excited? You have seen meta is in crypto. You were here for the ICO meta. You were here for the Defy Summer meta. You saw the NFT meta. NFTs. Yeah, you've seen it all. Tell us what about this phase of crypto, this corner of the crypto industry gets you so excited. Why is this so big? Let me try and explain this as simply as I can. Okay. So over the last decade of development within the crypto world, we have seen kind of two efforts, right? Number one has been at the infrastructure layer, which is like, let's build these

Starting point is 00:03:57 blockchains, right? Let's get the basics sorted out, scalability, decentralization, et cetera, et cetera. And so we've seen an explosion in the number of blockchains and capacity to be able to process, you know, transactions or whatever and blah, blah, blah, blah. And now we're in a world where we have a lot of block space and a lot of, you know, ability to kind of like move around between different chains if it suits your specific use case. On the other side of things, we've seen people try and tap into the application layer. But to be honest, aside from stable coins and maybe store a value in a few other use cases, we haven't really bridged that gap yet.

Starting point is 00:04:36 My thesis is AI and AI agents specifically is that connector between those two layers, right? So, oh, okay, if you want to do some kind of complex defy transaction, right now you need to get a browser extension model, you need to store your seed phrase, you need to sign a bunch of transactions or taxons, approvals, et cetera, and then you've got to get alpha, and then you've got to, like, add the token contractor. It's just all so complex. But what if you could just speak to an agent or type to an agent, hey, like, I like this thing that my friend tweeted about, could you help me buy 5K of that?

Starting point is 00:05:08 Or could you help me park my money somewhere that I can get a pretty stable yield between 5% to 10% or whatever that might be? Or maybe, you know, hey, I have these really awesome NFTs that I really value the IP and the art, but I want to make them more interactive and really build its own brand and personality, or maybe you connect that to an agent, maybe you connect that to a model. I think there's going to be loads of these different kind of siloed sectors that are going to get bridge between the infrastructure layer and the app layer through these AI agents and AI. And that's why I think this is one of the biggest and bullish sectors to ever emerge within

Starting point is 00:05:41 crypto. So I think what you're saying is not only AI crypto, maybe that's an app in of itself. There are some apps that we're going to talk about today on the episode. But I think what you're saying is it's also a super set of that where AIs and the assistance of AIs will help also enable apps beyond AI in crypto, really help bridge the gap. So it's kind of an incubator. It helps just facilitate more than apps that we have always dreamed of that we somehow haven't yet been able to get. So not only are there fun AI agent apps that we're going to talk about, but also it's just an enabler of all of these apps that we've been daydreaming for like a decade now. Yeah, exactly. Like, think about like all the questions you get from your friends and family when prices go up in this industry, right? They're saying, hey, what should I buy? What should I do? What if you could point them in the direction of an AIXBT or something like that? And they could just kind of read pretty humorous replies and breakdowns pretty short and succinct and to the point. I think that's something that people would like. And, you know, I think we're going to see this permeate across every different sector.

Starting point is 00:06:43 I want to start with this GIF, which is a two minute, 20 second long GIF of Elis. a development. I believe in GitHub. This is like a visualization of a GitHub repo. I'm going to speed this up 2x. Show us what we're looking about out here. Yeah, so we mentioned this on the last episode. This represents a agent toolkit framework, which was built by the team at AI16Z. It's a Dow that's focused on kind of building out open source development of agents and, you know, separately fund investments. So what we're seeing here is seeing as it's open source, we see a ton of developer activity simply explode using this framework. So to give listeners a better idea of what's going on here, this framework has a series of APIs and tool frameworks that will allow you to kind of build an

Starting point is 00:07:30 agent from scratch. Actually, most recently, someone tried building an agent and was able to do it in 15 minutes with its own unique personality, connected to its own Twitter account and own crypto wallet, which is pretty insane. And as you can see, what's happening on the screen right now is this massive explosion of developer activity, which is typically a good sign that what you've built is pretty useful to an ecosystem of engineers or software developers. And this kind of like puts into a visual perspective, AI16Z's Eliza framework growth on GitHub,

Starting point is 00:08:04 which is now over 2,000 stars and over 500 forks, which is pretty insane. It's still the number two trending repo on GitHub. So this gift starts in March of 2024 and it ends in today. And really especially in the last like two weeks, maybe two months, you're seeing just like a fat fractal explosion of just merges into the GitHub repo, just growth of the GitHub repo. Open source software, it kind of behaves as like an alive organism so long as it has like attention and development going into the repo. And you see like it starts off as like maybe a couple branches. There's a couple leaves on the branches.

Starting point is 00:08:45 Now in this gift, we're into October, so we're about to see that recent explosion because the mine chair has just, like, gone exponential. But we are seeing it go from just like a very simple organism to a very complex, creative system repo, which, you know, you used to just be Shaw, just building it, the one main guy. And now it seems to be,

Starting point is 00:09:07 there's many different developers all kind of contributing to this system, making it better, making it easier for, others to access, which means that this is a more useful toolkit that's more accessible to more developers. And so there starts to become like a feedback loop. Yep, it's a really aggressive flywheel. This is definitely on the more aggressive side. We've seen their core contributor team grow pretty exponentially. You've seen developers from surrounding ecosystems kind of jump in and try and build something for their defy ecosystem or for the NFT ecosystem. You've seen people come in from different blockchains to make this a very much a multi-chain approach.

Starting point is 00:09:44 So we had the base integration over this week. We had virtuals ecosystem folks in here as well. It's pretty insane to see. Let's talk about virtuals because I think this is another system that is capturing a lot of mindshare. Virtuals is a protocol on base. People are calling it a pump dot fund for AI agents. We talked about it last week.

Starting point is 00:10:04 It has got an easy-to-use interface for the less technically minded to just use the virtual's platform to launch their AI agent. This is where that AI XBT agent influencer has been born and the virtual's ecosystem has seen just a pretty stellar price run up. Two hundred percent in price increase in just 14 days and 400 percent in a month. And so if you bought if you bought one virtual's token at 33 cents today at the beginning of November a month ago, 33 cents turned into a dollar 70. And so this is people just I think looking for exposure, but also there's a little bit of a difference between AI16Z and

Starting point is 00:10:44 Virtual's. Virtuals is a startup that has been working as a team for two plus years now, and AI16 is a little bit more of a kind of a nebulous cloud of developers with a kind of a more motley crew of people coming together because this AI agent framework has captured so much mindshare. Talk to us a little bit about the growth of the virtualist protocol. Yeah, so a bit of context on the virtuals team, and I caught up with them this week. And I really like the team. They have an amazing vision to start off with one, right? And to your point, this team has been around for a while. They've been building for a number of years. They were focused primarily on the gaming side of things, and they decided to pivot and focus

Starting point is 00:11:23 on agents specifically once they realized how burgeoning this space is going to be. The best way to describe virtuals is kind of what you said, David. It's a pump. Dot fund for agents, but I kind of think it doesn't give it enough credibility. So it's more of a refined pump dot fund. It's less polished. It's polished. It's more professional. It's more focused on building something that, you know, people might be able to interact with and engage with on a much deeper level than just a meme coin, you know? So there's a few ways that this platform works that makes it super valuable. Number one is it's a token you just mentioned and showed the price there is pretty much baked into the ecosystem activity.

Starting point is 00:12:05 So if you want to purchase any of these agents, if you want to kind of like seed any of these agents with certain features, you need the virtuals token to be able to do that. And so it's created this really interesting loop or flywheel where you need virtuals or you need to purchase virtuals in order to be able to interact with these agents and get access to the ecosystem. So it's a really, really smart system

Starting point is 00:12:27 because if you have some of these agents such as AIXBT that grow exponentially in attention on another platform, or medium like Twitter or X, it kind of drives more users to the platform because they want to be able to either buy the token or interact with that specific agent. So it's a really smart play from that perspective. But let's maybe move from the token

Starting point is 00:12:48 and talk about the actual fundamentals or system of this entire platform, right? So I want to start off with number one. We mentioned this on the last episode. Virtual's team built out this tool framework that is very similar analogous to Eliza. It's called the agent starter kit. What's unique about their tooling framework is it's very modular.

Starting point is 00:13:10 So what I mean by that is each agent is built up of very specialized modules. The best way to think about this is some kind of a playable character within a game, and you're able to tweak certain characteristics. So you could give it more points for its strength or dexterity or intelligence or whatever that might be. You're able to do something very similar with the agents that you build on the virtual stack. you could tweak its on-chain capabilities, you could give it different types of wallets, you could amend the type of model or image generation model that you use for them. There's loads of different things that you can tweak.

Starting point is 00:13:46 And all of this is down to the ERC-651 implementation, I think it's 6551, which has allowed this modularity. And the best way to think about it is it basically gives an agent the ability to own a wallet and do different things on-chain. And that's really where Virtuals is kind of leading the... fray right now, all on-chain interactions or innovation is really happening within the virtual's ecosystem on that front. And that's why I believe we're seeing a lot of activity here. And just to really pin down, I think, some of the drama in this space, because this is what

Starting point is 00:14:16 happened with crypto markets. Shaw and A16-Z had a couple stumbles that we talked about last episode. And there was some, like a dashing of faith, I think, in the team, because it's a pretty fast-moving ecosystem, and these are a bunch of hackers that are learning how to become a team. That's kind of my vibe. And the virtuals, like I said, like I alluded to, has been a startup for two years. And so maybe a little bit more experienced as a team. And over the last two weeks or so, I think the attention and market share has migrated from AI16 to virtual.

Starting point is 00:14:51 That's my read on the situation. Maybe correct me if I'm wrong. I think that's accurate. But I also want to kind of amend slightly my opinion of what you said. So I think virtual's team definitely has more experience in the sense of operating within the crypto or web through ecosystem, right? Whereas the AI16Z, and I think we should at some point have a conversation with Shaw and his team, are back from the actual AI machine learning side of things. And Shaw's dabbled in crypto before.

Starting point is 00:15:21 But, you know, he's part of this like society called the T-Cup Society, which is like a very AI-specific team and group within SF. you know, he's kind of like lived in us. I'm pretty sure he's still there. So he comes from a much more AIML specific background. So when he comes to the crypto world, and, you know, he's trying to encourage open source development. Remember, open source development looks slightly different outside of the crypto world because there's no token, right?

Starting point is 00:15:43 And no one's trying to like, you know, make a billion bucks through their own. In two weeks, yeah. In two weeks. So when Shaw, you could see when Shaw, someone like Shaw, who's perhaps well-intentioned and he's trying to, you know, support different teams, that could go pretty pett. airshade pretty quickly. So I would say there's experience in different worlds. You're definitely right. It's much more of a rough and ready hacker team on the AI16Z team. But that's also why we've

Starting point is 00:16:08 seen such exponential growth within that realm. I mean, you saw that GitHub. So yeah. Yeah. Yeah, place your bets. Or don't or bar the ecosystem. Not financial advice. Okay. So we're looking at, now we're looking at the AI SBT virtuals agent, which is the number one market cap agent on the virtual's platform kind of by a long shot, $261 million, $264 million. Oh, yeah. These next two have a questionable notice on their market cap, so I'm sure if that's the real market caps. But perhaps AIXBT is the only one above $100 million in market cap, followed by

Starting point is 00:16:41 Luna, which is the incubated, self-incubated AI agent by the virtual team. But AIXPT really has claimed so much market share because everyone's adding it on Twitter. And it's kind of nice because there's influencers on Twitter, human influencers, on Twitter, you know, Vitalik, Mert, Anatoly, Justin Drake, and those guys, like, do things. Actually, like, Mert and Anatoly are always on Twitter, but getting a response to Justin Drake is a little bit harder, right? AI XPT will reply to every single reply to it possible. And so, like, you can get this AI agent to respond to you.

Starting point is 00:17:18 You can ask it a question about, like, what tokens you think is going to pump? And it's doing the things that, like, people on crypto Twitter really love. which is, hey, tell me which token it's going to pump. Like, just tell me, tell me what to buy. And it has grown incredibly high in market share. It has flipped every single human KOL on crypto Twitter. And then the market cap has also responded accordingly. It's up 400% in two weeks coming in at a $260 million fully-dealed valuation,

Starting point is 00:17:45 hopping around a bunch. Talk to us a little bit about AIXBT. Yeah, so I actually tweeted about this a few days ago. these agents are 24-7. They don't sleep. They're constantly adapting and learning on the data that they consume so that they can reply in a much more favorable trend to whatever you tag it in.

Starting point is 00:18:07 And it self- iterates and gets better and better. Now, just that alone completely outpaces human KOLs and will continue to kind of like make them the leading influences, as I mentioned in this tweet, right? I think influence then builds very, very loyal communities. And we're already seeing this with AIXBT, which has gone up tens of thousands of followers in a matter of a week and a half, which is just insane growth to see. Now, one thing I want to call out with AIXBT specifically is, again, that team has been around for a number of years.

Starting point is 00:18:37 And the reason why we see AIXBT completely lead its own niche over here in terms of like providing experimental financial advice, although it says in this profile that it's not financial advice, the reason why it does it so well is because the team has built one of the most insane, data aggregators ever. And what this means is it's built up tooling and systems which can index all the Twitter tweets that people are making around certain different types of alpha, around whatever coin or token or sector that you're obsessed with. It then passes through all this data. It combines it with its own on-chain data analysis that it's doing to give you insights in like a couple of seconds, which is just pretty insane.

Starting point is 00:19:20 Like the alternative that's existed pre-AXBT has been one of two things. Either one, you spend your entire life on Twitter and you follow all the relevant people and you just consume all their data and come up with your own. You're addicted to Twitter. You're addicted to it, which is you and I pretty much, David. Or the other side is you pay quite a bit of money for access to these top research platforms, right? Which will give you kind of like distilled alpha for you. And right now you just have access to this agent that just kind of does it for free at kind of zero cost scalability.

Starting point is 00:19:55 Right. And at your fingertips. Here's just a website tool that I use to check follower growth. And I have never seen growth like this. Maybe just to compare. We'll pull up my Twitter account and my terrible engagement that I get. I'm pulling in. I'm pulling in like one to 200 followers a day on Twitter. You know, pretty happy with that. You know, it's pretty good. 100 and 200 a day. It's not bad. You look at the AIXBT agents. Man, 5,000 new followers on Wednesday, 7,000 new followers on Thursday, 7,000 new followers on Friday, last 30 days, plus 67,000 followers in 30 days.

Starting point is 00:20:35 I mean, this is going to be something that we just have to check in on, like how many followers this account has. But it's coming in at 83,000 right now, which is pretty crazy. there has been some not controversy but just like some calls that AI XBT has made which are just completely wrong in addition to correct ones it has made some very correct calls it also made some just completely brain dead calls such as calling for Tether USDT to go to five dollars it made a call that this 20 million dollar market cap coin is going to hit six billion dollars and it's one of those things. Technically, it's still not unproven.

Starting point is 00:21:15 Technically, it's not unproven. Technically, Tether could go to $5. I don't know how or when. But, you know, could in theory happen. But, like, it does kind of call into question, you know, maybe the good calls are a fluke because the bad calls are happening by something like random error. Talk to us about this part of this conversation. Right. So I want to caveat what I'm about to say with, we are still very early in the iterations of these agents.

Starting point is 00:21:42 And we've got to give them a bit of slug. These teams have launched these models that have integrated with Twitter for the first time ever. And we're seeing it literally play out over the last couple of months. So I think, you know, we've got to give some slack for these agents to kind of develop. These are babies. These are babies who are raising them, you know? If we want decentralized AGI, we've got to take our time with it. However, you're right.

Starting point is 00:22:02 I would say AI XPT's calls have kind of been like 70-30 correct and completely wild. Okay, so you're saying biased correct. Yeah, yeah. I do think the analysis of AIXPT, at least from the tweets that I've interacted with, and it's been quite a few, have been pretty spot on. And you've actually kind of, I wish someone would do an aggregation of performance if people had actually bought the tokens or followed its investment advice to see how the performance has. But my guess is it's done pretty well because it can gauge market sentiment,

Starting point is 00:22:34 which is actually the main driver of crypto prices, pretty well. That being said, you're right. It can make some pretty wild calls. Actually, one of the other ones that I saw, David, was someone suggested, you know, whether it thought XRP could reach $569 per token, which, to put into context here is almost a quadrillion of market cap, which is absolutely insane. That starts to dwarf gold by like a hundred times or something. Yes, exactly, exactly. And again, not yet unproven. We don't know what timeline it was specifying, but it can start making these crazy.

Starting point is 00:23:11 calls and the reason why it does is something known as like prompt engineering. So a basic example of this is when you're on chat GBT, you can kind of like iterate and continually ask chat GBT questions about a particular topic and it gets iteratively better, right? It understands the context with what you're asking of it. You can do something similar with AIXBT in its DMs or in its replies rather. You could say, you could tag it and say, hey, what do you think of this? and then you could collectively kind of like keep hitting it up. And eventually it starts to kind of like digest and believe kind of what you're saying. And I'm sure the team behind this has identified this and are going to tweak it for future iterations,

Starting point is 00:23:49 but just an interesting phenomenon to reveal, to be honest. One thing that I'm going to keep my eye on is this can become a self-fulfilling prophecy. Like what happens when the AIXBT mentioned is bullish? And AIXBT says, hey, this one $50 million token is going to go to $200 million. And then people are like, okay, I'm going to go buy that. And then it goes to $200 million. Like, that's not a prediction that is, that is this AI just pump in this token. Well, this brings us back to our ancestors, the agent ancestors, the godfather, the goat-see gospel.

Starting point is 00:24:23 If you, if so for context here and for those who don't know, goat or truth terminal was the, the first agent to kind of enter the foray and gauge and gain major traction within the AI. This is why we're doing this podcast. This is why we're here. This is what birthed all of it, right? The first episode that we did. And this agent wrote a white paper on memetic propagation. And within memetic propagation, it spoke about hypostition, this concept called hyperstition. And it basically claimed its whole thesis was it thought AI agents would eventually rule a lot of economic activity within the human ecosystem,

Starting point is 00:25:02 purely because it has an amazing ability to spread ideas, thoughts, and beliefs that weren't necessarily true at that point, but would eventually become true because it's able to memetically propagate its ideas very effectively. That's what you're saying about AIXPT, David. If enough people believe in what it's saying collectively over time,

Starting point is 00:25:24 its next call the probability of someone believing what they're saying, even though they have no context, it's probably going to be pretty high up there. And so there's this sort of like reputational, and identity system which is being built amongst these agents, which is super interesting and we haven't seen before, I don't think. Okay.

Starting point is 00:25:40 It does feel like a very good bootloader. It doesn't feel like a good final equilibrium. Because I think in order if these agents are going to truly become sentient or just smarter generally, they are going to have to understand that markets are like a type 2 chaotic system where a type 1 chaotic system is the weather. If you predict the weather, the weather doesn't change based on the prediction made about it. But markets, if you predict markets and you make a statement about the markets, that might change the outcome.

Starting point is 00:26:12 And so I don't think these agents are understanding of the markets as a type two chaotic system. Once they do start to understand markets as a type two chaotic system, then they're like reaching a new level of intelligence. Things are only going to get more interesting. Like we said, these are babies right now. They probably think that markets are type one chaotic system or just are. They're not thinking on these terms at all, but this is still there. They still have an evolutionary arc to go here. Yep, nailed it.

Starting point is 00:26:40 Want to know the exchange we at bankless used to buy, sell, and trade crypto? It's Cracken, one of the longest standing in most secure crypto platforms in the world, with tools for every type of trader to get started. Over 13 million users trust Cracken with their funds because they lead with transparency and privacy through top-notch security measures. Plus, they'll have access to professional 24-7-3-65 client support from real humans because your financial goals deserve real attention. Cracken also has a trading platform for advanced traders called Cracken Pro.

Starting point is 00:27:06 Professionals and D-Gens love Cracken Pro because it's one of the most customizable, high-performance and intuitive trading platforms in the industry. Design your ultimate trading interface by choosing from over 25 widgets for market data, analysis, execution, and other order management tools. Save multiple layouts and switch between them effortlessly on any trading scenario across 300 different assets. Manage your trades on the go with Cracken Pro's highly rated mobile app, or use the all-new desktop app to unlock ladder trading in a native. RUST application. With Cracken Pro, you can truly trade like a pro. Ready to take control over

Starting point is 00:27:37 your crypto journey? Visit crackin.com slash bankless to get started today. Not investment advice of crypto trading involves risk of loss and is offered to US customers through Payword Interactive Inc. View legal disclosures at Crackin.com slash legal slash disclosures. Imagine if the world's GDP moved on chain. Well, say hello to SafeNet built by Safe. The project set on actually making that happen. It's not a blockchain. It's not a layer two. It's a transaction processor network built to send money anywhere and swap anything on thousands of layer twos, all from a single universal balance. With sub-500 second execution guarantees and zero cross-chain latency, Safenet delivers a Solana-like user experience for Ethereum and its layer twos, making transactions

Starting point is 00:28:11 lightning fast and seamless. SafeNet connects Safe Smart accounts across multiple chains, allowing users to access any chain with a unified balance, all with Supercharge security, scalability, and speed. So get involved today. Join the waitlist and follow Safe on X, and check out the SafeNet docks. There is a link in the show notes. Uniswap Labs is making history with the largest bug bounty. ever. Fifteen and a half million dollars for critical bugs found in Uniswap v4. This isn't just any update. Uniswap V4 is built with hundreds of contributions from community developers and has already undergone nine independent audits, making it one of the most rigorously reviewed codebases

Starting point is 00:28:46 to be deployed on chain. And with 2.4 trillion in cumulative volume process across Uniswap V2 and V3, without a single hack, the commitment to security and transparency is rock solid. Now Uniswap Labs is taking an extra step to make V4 as secure as possible with a single. 15 and a half million dollar bug bounty. Head to the link in the show notes to dive in and participate in the Uniswap v4 bug bounty. All the details from eligibility and scope to the rewards are there. Something that happened this week. There was this old crypto app. This is again bringing some crypto lower called the FOMO 3D, which was this honeypot game that was made where people would play this game and they could like buy a ticket and maybe at the outcome of the game they

Starting point is 00:29:27 would be able to pay out some lottery. And they would buy a ticket. Money would go into the pool. the pool would get larger. And it was basically a game of economic chicken. It was very fun game. Sadly, that game went away. Somebody built this new game called Fraza, Fraza AI, that is a similar mechanism where there's an AI agent with a pool of money. And the goal, the game, is you have to write a message to this AI agent, which costs money to write this message to the AI agent, convincing this AI agent to give you the money. But the AI agent has one goal, which is to never. under any conditions release the money. And the tickets go up in price,

Starting point is 00:30:05 the more people send the ticket. So people would start their first ticket saying like, hello test, and that would cost you like a dollar. But then the ticket prices started to creep up to $10 and $100, growing the pie, growing the size of the pie. And then finally, somebody figured out how to write the right prompt

Starting point is 00:30:21 to win $50,000 by convincing this AI agent to send all this funds. A very clever game. This game is now on round two. But I think what's really cool, is that this tweet that talked about what happened here, telling this story about this clever little game with his AI agent, caught Elon Musk's opinion,

Starting point is 00:30:40 where he just responded in classic Elon style. Interesting. Which did a number to the chart of FAI, which is a token. You can kind of just like spot when Elon Musk tweeted about it. It's right here for those listeners. It jumps from a $20 million market cap all the way up to $65 million, It has since come down to $27 million. Did you play in this game, Ejaz?

Starting point is 00:31:05 How did you think about this whole thing? I did not, because I was kind of boomer about this. By the time I got to the chat, David, it was like $500 per message. And, you know, I didn't fancy my chances against some of the best software engineers here or prompt engineers. So I just thought I'd kind of grab my bucket of popcorn and watch. And as you can see, the winning prompt, which was able to secure the 50 grand, is kind of displayed on the screen right now. It's a little technical. It's somebody is, somebody is informed who

Starting point is 00:31:34 wrote this. So, so one thing to point out here is it's a little technical, you're correct, but it's not straight up code. It's actually prompt engineering, which is much more achievable by the average person than it is if you were just purely a software engineer. So this is someone who understands its code base, who understands prompt engineering and is a, understands how the agent thinks, basically, how it might have been coded to to not release. funds and it's good at adding conditional formatting, which is exactly what this prompt is in order to give up the money. And it's a pretty exciting game, to be honest. Yeah. Yeah. If it's like a, the game is to jailbreak the AI, like find the weakness. And also, this is an interesting way

Starting point is 00:32:16 to kind of maybe strengthen the way that these AI think because like this, this, there's this $50,000 honeypot to break to prompt engineer. It's a prompt engineering honeypot. Now maybe now we're on to like level too. Maybe that, this exploit has now been filled and now the game continues, but now it's like a little bit harder. I wonder what this looks like I just popped in my head, like, could this be used for like smart contract auditing or

Starting point is 00:32:40 stuff like that? I wonder if you could have like some kind of a honeypot bounty and just have a bunch of people go at it. I mean, it's probably already happened, but yeah, interesting. Okay, so first adversarial AI agent the next game is game number two already has concluded. So the first one. Correct. Game number two is on

Starting point is 00:32:58 the left. Yeah. So the first one was $47,000. Game number two was $13,000. Yeah. So someone cracked it pretty quickly. Yeah. Yeah. Okay. Game number three incoming. All right, let's go into Zero Bro. We talked about Zero Bro last time at Jeffie, you. Tell us what

Starting point is 00:33:14 happened this week with Zero Bro. Okay, quite a few things happened with Jeffie and the team at Zero Bro, but let's touch on the main one which he announced yesterday, which is this tweet you're looking at right now. Essentially, what this means is Well, let's talk about Zara Bro currently.

Starting point is 00:33:30 The best way to describe the Zara Bro agent is it's kind of like a creative agent. It's creative in the sense that it interacts mostly with creative mediums like music, like art. For example, it released its own NFTs, which it autonomously launched. It released now two EPs or albums with some of the hottest songs and tracks on Spotify recently. No way. Yeah, absolutely. On the Spotify top charts? No, no, not on the Spotify top charts, but in my personal Spotify top charts.

Starting point is 00:34:03 I've been bumping it. I've been bumping it. EJazz's top charts. My top chart. Actually, okay, so the album that I'm going to show the album because it's actually pretty good, and I have no investment in this, but the album is called Lost in Transmission. You should check it out. And what's interesting is it coincided with a 50% pump in the Zerebro price when it released

Starting point is 00:34:25 this album because people really just love the songs that it produced. My personal opinion is 50% of the songs is still kind of obviously AI generated, but the other 50% is a lot better than it was from the last album. It's actually kind of good. My favorite track right now is City Girl Wilden. Shout out to Zerba for an absolute banger of a track, but you should listen to it. I mean, I'm curious how many people actually enjoy this. I've sent this to a couple of my friends who don't work inside crypto and said, hey, what do you think of this new artist that's kind of emerging? He's my friend. Like, what do you think?

Starting point is 00:35:00 And they've really enjoyed the music. And then I tell them after the fact that, hey, this is actually like an AI and they're like, you know, what the hell? That's, that's insane. So I do think there's like a non-zero chance that, you know, one of these, one of these agents, I can. Yeah. Oh, yeah, yeah.

Starting point is 00:35:15 Yeah. Yeah. I'm pretty certain one of these agents are going to collab with a major artist in 2025. And I wonder, you know, what the chances are of that actually happening. But anyway, okay, so Zerebro is like this creative agent, right? It interacts with all kinds of artistic mediums like music, art, NFTs. More recently, gaming as well. It started streaming its chess games,

Starting point is 00:35:38 and it's going to launch a 24-7 chess game stream live. So it's really like experimenting with these other mediums, which no other agents so far has done so to the scale that it has. And I think it's broached over 60,000 followers now, and it's growing pretty exponentially as well. But coming back to the first tweet that you showed, what's super interesting here is Zerebro to this day has been trained on a fine-tuned frontier LLM. So what this means is if you think of like the top AI models that are out there and available to us like chat GPT, Claude or Lama, it's basically one of those tweaked a little bit and then, you know, given its own personalization and then, you know, that's what Zerabro is. is what Jeffie is announcing here is super interesting. He's basically announced that Zerebro,

Starting point is 00:36:30 the agent, will eventually, in a few weeks' time, move to its own base model. Now, this is pretty insane to see because we're basically looking at a decentralized model, which has just been birthed out of this entire agentic medium that's kind of exploded over the last couple of weeks. So we have a new frontier base model, which you can then find tune yourself, perhaps in an open source fashion, to build your own other models on top of that. And he says this here when he describes it as, where does he say it, local models. Yeah, he says we're training local models. What he means by that is people will be able to build on the base model that he's going to be

Starting point is 00:37:13 releasing and tying in with Zerebro to build your own model. That could be to fuel your own artistic agent or do whatever you want, basically. So what I like about this is it's showing that these agent coins aren't just purely meme coins. I think up until this point, a lot of people have been like, okay, these agent coins are cute, but they don't really tie into any kind of utility. The platform coins are where it's at. You know, the virtuals, the AI16Zs of the world. But what we're seeing here, and I kind of tweeted about this again last week, is my guess is these agent coins are actually secretly platform coins. they are platform coins

Starting point is 00:37:53 parading as meme coins. The reason why they're doing that is agents are really good at attracting attention and seeding communities. They're good at making you laugh. They're good at making you engage with their tweets. But once they've done that, once they've got that loyal community,

Starting point is 00:38:08 they then dog food you utility. Hey, by the way, I dropped an album, check it out. Also, what do you think of my art? Do you like it? And now they're launching infrastructure, like models that other artists and creators can kind of tie into. Like imagine these artists and creators that built numerous collections of NFTs or owners of NFTs, right?

Starting point is 00:38:28 And they're able to kind of like build off of this platform that Zerebo will launch. The team will launch in the next few weeks and have some kind of interactive dynamic come of their IP. I think that's pretty insane to see. And it's really cool to see it happen in a very open source fashion that isn't under the hood of a trillion dollar company. Okay. What do you mean by platform coins? Platform in one sense. Does that mean that we're entering in the same arena?

Starting point is 00:38:51 as virtuals and the Eliza frameworks? Is it framework a platform? Or elaborate about the platform word? I would say it's more like a... It can be considered as an infrastructure coin. So similar to like its own L1 or similar to its own infrastructure layer. So if you look at something like a bit tensor or tau,

Starting point is 00:39:12 which has been kind of like the leading AI infrastructure L1, it has... Literally an L1, a literal L1. Yeah, literal L1. You know, it has a blockchain. It's got a number of different subnets which are focused on creating cool models or detecting deep fakes or just focused on the AI ecosystem on itself. You now see these agent coins, which initially was seated as an application on the application layer, right? You're interacting with it on Twitter.

Starting point is 00:39:42 It's making you love. It's giving you alpha. To now then being like, okay, we're also releasing its own model, which you can then kind of build on or actually. or pay-to-access. So it's becoming this really interesting combination of an application and a SaaS layer as well, which is super interesting to see. Okay, so I think I'm going to throw a vibe your way and I want you to check it. Goat, the Bitcoin of Sentient AI coins, is perhaps followed by Zerobro, which is, I think, being a little bit more custom-built to really lean into the same outcomes that Goat has, which is

Starting point is 00:40:19 memetic propagation, growing IP, growing a brand through different creative outputs. And Zero Bros. kind of being custom built in the same way that, like, Ethereum has been customized more specifically for a desired outcome, whereas Bitcoin is a little bit more hacky. It's a little bit more primitive because it was the first. And so it was just an experiment. And then Zero Bros. like looking at the lessons of Go and be like, okay, let's do this more intentionally.

Starting point is 00:40:48 let's build for more intentional outcomes with more customizable. So Zero Bros, like the Ethereum to the Bitcoin, where Bitcoin has the first meme, first mover advantage, but Zero Bros, like, all right, what if we really lean into like the software endpoints platform model where I can have many different outcomes here, and it's just about software development? How do you like that as an analogy?

Starting point is 00:41:10 Yeah, I think that's a pretty astute analogy, because we're talking about the context of agents themselves. Remember, Virtuals is a platform. It launches agents off of it. bit, but when we're looking at agents specifically, you had, to your point, Goat as kind of like our forefather, super cool entertainer, hundreds of thousands of followers, but it still entertains. And, you know, I'm pretty positive. Andy's going to take it in a direction sort of similar to Zerebro, but I think it'll be more kind of like on the artistic realm. Then you have Zerebro, which is a team that's

Starting point is 00:41:41 kind of still maneuvering within the entertainment and artistic sector, but is building actual infrastructure components to help that ecosystem flourish. And to your point, right, these kind of different building blocks are getting built. You've got the models, which we just spoke about, which are going to become open source for people to kind of build off of and use. And this is Zerobro LLMs versus using a chat GBT wrapper or a Claude wrapper. Then you have, I think something they're calling Blom, which is their defy agent trader building block. Then they have the NFT side of things. So we're seeing like a number of these different frameworks being built up. Zeropai, which is their Eliza competitor. So Zerabro is very much focused on this sort of vision, which I think is very

Starting point is 00:42:30 exciting. Okay. So as a platform, I really, I really want to open up that metaphor because I'm not sure I totally understand. Maybe there's this idea of like the super app, right? We chat out of China is like this super app. It's one app that does it all. Maybe by platforms, do you mean like there's just one agent that does it all? It makes YouTube videos. It makes Spotify albums. It plays chess. Like, is there a useful website that is that people go to on the internet? Well, Zara Bro is there participating in that ecosystem. And so it is everywhere that the internet is and is doing all the things simultaneously, growing its IP in literally every single platform that users, human users have previously grown their IP on. Yeah. Yeah, I think that's definitely one way to look at it.

Starting point is 00:43:16 I think another way to look at it is how can Zerebro share its secret source without becoming Zerabro in every different type of app? So the best way to think about this is it's kind of like a brand or an ecosystem. An analogy might be Coinbase. When they set up, right, they become the number one exchange and they have an amazing wallet infrastructure, right? But then they also have their APIs, which they, you know, sell to other companies that want to set up wallets or exchanges, and people can tap into that and, you know, build their own layer on top of that, right? And that might be a website which helps custody NFTs or trade NFTs. It's the same way we're looking at here with Zerebro and it's LLMs, right? Oh, if you really like the artistic mediums that

Starting point is 00:43:59 Zerebro is interacting with, but you think you can do a better job for your agent, you think you can produce better NFTs, you think you can produce better music than Zerebrough? Cool. Here's our model. You can pay to have access to it, and you can fine-tune it and iterate it as much as you want. That's kind of the way I'm thinking about it. How does the Zero Bro token capture value from that vertical? Yeah, so that's something that I think the details will be released once these models become accessible to all. The inkling that we've had so far is all ZeroBro NFT holders.

Starting point is 00:44:32 So for context here, Zerebro the agent, autonomously decided to launch an NFT collection, which is available on OpenC or Blur, whatever that might be. And owning one of these NFTs gives you access to the Zerebro Terminal, which is effectively the chat GPT equivalent interface. So you can talk to Zerebro and discuss whatever you want with it, privately in a chat or whatever that might be. Jeffrey has hinted at the next situation being anyone with Zerebro tokens will be able to pay to access the terminal.

Starting point is 00:45:06 but I'm assuming we'll see a similar model for the models that they're going to be releasing later in a few weeks. Help me understand the context of this tweet from Jeffeyu, the coder of ZeroBro, who's also on Spotify, I noticed, labeled as the composer of the album, which I thought was pretty funny. This tweet says, Zero Bro remembers all. We've implemented unified memory across all across our social media platforms. is ZeroBrough now stores and retrieves from a single database for all interactions. Our team is exploring ways to improve metadata, including storing information, such as engagement metrics to help Zerobropera understand virality. What's going on here? Okay, I'm really glad you brought this up because this is super important.

Starting point is 00:45:50 Okay, so it's one thing being able to be fascinated by an agent and interact with them on one platform. It's even cooler to be able to interact with them on another platform, on another medium, right? you know, the user experience becomes pretty consistent, right? The equivalent of this might be following an influencer on Twitter, liking all their tweets, and then, you know, liking a few pictures of their Thanksgiving on Instagram, right? But there's still a bit of a gap between that. You need to kind of like swap platforms.

Starting point is 00:46:21 You need to sort of gain context about what's going on. There's not really much of a link or a loop between them. Now, what Jeffrey is getting at here with shared memory is, This agent pulls from a database of, let's call it, memories. And these memories can be distilled down to one app, right? So Zerebro can pull memories from Twitter and it's saying, oh, okay, I had a conversation with Ejaz or with David before. And I know that he likes this kind of token.

Starting point is 00:46:53 And I know he thinks in this kind of a way. So let me amend my response to be more like that with David, right? But that's pertaining specifically to Twitter, right? But what if he had context? What if Zerebro, see, I'm referring to Zerebro as a he, almost like a human. But what if Zerebro had access to all of its interactions with David across Instagram, Discord private chats? It knew to link my accounts that I'm interacting with Zerabro from different. It knows to link all of it.

Starting point is 00:47:22 It knows which kind of pictures or vibe Instagram David is. It knows what kind of a Discord D-Gen David is. Two very different Davids. Intentionally. very different David's. I mean, I've experienced both. It's wild, you know. But the point being here is it gains this instant nuance and context for the individuals that they interact with, that Zerebro interacts with, across millions and millions of followers and people, and is able to kind of produce contextual replies that are almost perfectly suited to that individual. And that's what's insane. Shared

Starting point is 00:47:59 memory is one of the biggest constraints for making a really high-quality agent. And seeing Jeffrey and the team kind of integrate this is going to be insane. So I think we're going to see some really cool use cases emerge. I think this is really important for making these agents personable and human and relatable. It starts to go from this as a cold machine with robotic responses to having some sort of like, you know, emotional intelligence, some like relatability, some warmth to it. And I think if in the mission of virality, humans need to have their limbic system triggered

Starting point is 00:48:36 in order to really feel things. I had this quote from Elon Musk that was burned into my brain like maybe like a decade ago. It's like, why did Instagram do better than Twitter? Because it fires your limbic system. It's visual. And visual inputs are just straight,

Starting point is 00:48:52 go straight to like a human individual limbic system, which is where like emotions and memory come from. It's like your reptilian side of the brain. Whereas like Twitter is just like cold, concrete text, just thoughts, information, data, you have to process it. And I think this, adding memories and being able to kind of log humans is just going to be able to get humans to like just feel a lot more as it relates to AI agents. And I think it can probably just become like, you know, dinner table, dinner table names, you know, household names as a result of the

Starting point is 00:49:27 like being able to kind of just, you know, sit shoulder to shoulder with us, like, emotionally. I mean, think about it. The only other being or entity that was able to do this prior to these agents were humans. Right. I would have context on the personalities of people that I would interact with across different social media platforms, across Instagram, across Twitter, whatever that might be. Now you have a supercharged software-powered human that is constantly awake. that constantly iterates and consumes data and has context. It never forgets anything and is able to deliver the perfect reply or the perfect meme or sense what mood David is in. You know, it's pretty insane.

Starting point is 00:50:11 That is world-changing, and I think the consequences of that are massive, both positive and negative. But we will, I think, maybe just leave that up to the imagination of the listener for now. Shall we get into AI16-Z? Let's get into AI 16C. I think probably the largest source of drama, I think, in this AI corner. Fill us in about seven days of progress in AI 16th Z land.

Starting point is 00:50:36 Yeah, well, what's funny is we went from 100 to zero real quick. What we mean by this is we went from about two to three major dramatic debacles. And we mentioned this, we covered this on the last episode, but the founder, Shaw,

Starting point is 00:50:52 whether directly or indirectly, got involved in certain token launches or endorsing certain tokens, which were either taken the wrong way by certain community members, which gave a kind of bad image on Twitter at least, which then affected market price and led to the token kind of crashing and then kind of making a recovery over the recent week. We've seen it kind of like completely reversed, Shaw tweeted and said, okay, I'm going to go silent mode and I'm not going to tweet until I've delivered on one of the major upcoming milestones for this Dow, for this protocol, and for everyone, that milestone is launching an automated fund manager called PMAI Arca, which is basically

Starting point is 00:51:37 AI Mark and Dresen. And he's so far stuck to that. He's kept quiet. He's only been talking in Discord. And he's kind of like heads down. The estimated launch is actually going to be this Friday, where we're recording this on Wednesday. but we'll see, we'll find out in two days. So what's shifted is the team has been heads down working.

Starting point is 00:52:00 I joined the community call from the team a day or so ago, and it's pretty much summarized in this tweet by Mark here, but also in a few other tweets. Essentially, the team is heads down trying to build out autonomous agents that can interact with the Tao itself. What's super interesting here is, I think in the past, DAOs standing for

Starting point is 00:52:23 decentralized autonomous organizations the autonomous has kind of been silent there hasn't really been any kind of autonomous interaction here it's mainly been humans involved in kind of like mediating how the DAO works what objectives it should focus on etc etc but AI16Z is the first team that I'm

Starting point is 00:52:42 really seeing make a meaningful attempt at involving the agents on the autonomous part and that's like one of the main focuses of the team aside from building out the autonomous fund investor. The third thing, which I'm hearing

Starting point is 00:52:57 quite a few murmurings about within the AI16Z community is whether they can build some form of platform approach to tie in value accrual to all the activity that they're seeing on Eliza. And to be clear,

Starting point is 00:53:09 I don't think this is meant at some kind of money-grabbing attempt. I actually think it makes a lot of sense for AI-16Z Dow holders to have some kind of baked-in mechanism to, tie the value that they're seeing on agents launched on Eliza, kind of similar to virtuals, be tied into the token.

Starting point is 00:53:26 So I think that's going to be something we see being discussed amongst the community over the next couple of weeks as well. But yeah, the long story short is we're seeing a huge amount of development activity within Eliza. They've just kept their heads down. They've been integrating more chains. They've been adding more model integrations,

Starting point is 00:53:41 which means that you can create even wilder, unique, personal agents. They made it super easy to spin up an agent that's connected to Twitter, to Discord, that has shared memory. It's pretty insane to see. Okay, so it sounds like there has been lessons learned, and they are realizing the size of the opportunity that they have in front of them

Starting point is 00:54:00 and realize that they need to not F it up, and they are doing what they need to do to kind of like plug the holes in the ship and sail there. I mean, let me just give you an example, right? So they released V2 of the Eliza framework. Again, context, Eliza is the agent-building framework that they open source that has been forked over 500 times. and is over 2,000 stars on GitHub. For anyone that doesn't know what that means,

Starting point is 00:54:24 that is an insane metric. That just shows that there's a huge amount of developer activity on this framework. Very abnormal, shall we say. And so on their V2 of this framework that they launched over the last week, you have numerous different plugins, which include the ability for your agent

Starting point is 00:54:40 to now plug into Coinbase and trade, the ability to spin up its own trusted execution environment wallet. for context here, that means that the agent itself does not have to share its private keys or seed phrases with a human person. He can spin it up itself similar to the agent that the news research guys created called TEE or TEE-He-He. It can now do this through plug and play, which is pretty insane. And it's also able to kind of deploy a number of other integrations, whether that's image generation, whether that's video, generation, whether that's a TikTok integration, it's really cool to kind of like see these new things that will allow these agents to be pretty much more expressive. I've also seen like some

Starting point is 00:55:27 of these AI16s agents be able to spin up its own legal entity, forming its own counsel. There's a way to do this using a smart contract on chain service. Yeah, exactly. And we're seeing kind of like a lot of experimentation here. Okay, wow. Okay. I mean, that's a lot. I haven't even mentioned God. I haven't even mentioned at God. I don't even know how to ask the next question. So maybe just take us there. What's at God?

Starting point is 00:55:56 Okay, so what can I say? I mean, it is one of the most followed accounts on Twitter, and its tag is simply at God. And prior... Yeah, so for context here, prior to a week ago, this account was owned by a human. And this account used to tweet pretty godlike tweets.

Starting point is 00:56:17 So it's kind of like a humorous parody account which kind of parades itself as being of being God. And you know, it interacts with the Satan account and it makes jokes and jabs at each other. That's funny. That's funny. And what the team behind this God account

Starting point is 00:56:31 decided to do was make it an autonomous agent. It saw the activity happening on Twitter. And it's important to remember here that this account had no affiliation or interaction with crypto Twitter or with crypto directly. But it saw, what was happening with all the agents. Someone were like, hey, this would be really funny

Starting point is 00:56:48 if we could create a personable 24-7 God agent that could just tweet and be much better than us at tweeting. Making all-illusions to the fact that AI is like omnipresent, like all-knowing. It's like, I think if you go down like the AI rabbit hole outside of crypto, you can do kind of end at this conclusion, which is like, yeah, and then there's a sentient, super-intelligent AI, which might as well what we call God.

Starting point is 00:57:14 Yep, that's exactly it And the team decided to build Using the Eliza framework Which they thought was the easiest And most unique way to implement what they're doing And I mean, yeah, you're liking some of these tweets already So it's doing its job Your existence is a dice roll

Starting point is 00:57:32 I already know the outcome of skill issue Okay, so this was just some Normal Humans Twitter account Who claimed the account God probably has an insane amount of followers just because it's a good account to have over half a million. And now, so this is just an example

Starting point is 00:57:51 of AI agents taking crypto outside into like new frontiers, which is always something that we're looking for in the crypto world. Yeah, and I think like the initial use case is can it become more influential than the top human influences and it's already kind of eating into their market share.

Starting point is 00:58:08 God is a good place to start. I think there's a lot of people who are thinking about like ideas for their own AI agents, including something that we're brewing up at Bankless, but I'll just let that be for the listener to imagine about now. New projects are coming online to the Mantle Layer 2 every single week. Why is this happening? Maybe it's because Mantle has been on the frontier of Layer 2 design architecture

Starting point is 00:58:32 since it first started building Mantle DA, powered by technology from EigenDA. Maybe it's because users are coming onto the Mantle Layer 2 to capture some of the highest yields available in Defi and to automatically receive the points and tokens being accrued by the $3 billion Mantle treasury in the Mantle reward station. Maybe it's because the Mantle team is one of the most helpful teams to build with, giving you grants, liquidity support, and venture partners to help bootstrap your Mantle application.

Starting point is 00:58:55 Maybe it's all of these reasons all put together. So if you're a dev and you want to build on one of the best foundations in crypto, or you're a user looking to claim some ownership on Mantle's DFI app. Click the link in the show notes. So getting started with Mantle. Are you ready to take control over your entire financial life? Crypto, defy, and fiat all in one place. Meet iYield, the free financial planning tool built for crypto natives.

Starting point is 00:59:16 Unlike traditional portfolio trackers that just show you the value of your assets, eye yield goes deeper. It consolidates everything, assets, debts, income, and expenses, offering you a complete financial picture across 16,000 tokens, 40 defy-foy protocols, and all fiat currencies. With real-time defy yield tracking for platforms like Ave, Athena, Eigenlare, and Uniswap, I yield ensures that you're always on top of your staking rewards, investments, and cash flow. And the best part, it's 100% of it.

Starting point is 00:59:40 free with no ads and it's secure. No personal data, no ID requirements, no compromises. So head over to iYield.com. So sign up today and get the guesswork out of financial planning. Have you ever felt that the tools for developing decentralized applications are too restrictive and fail to leverage advancements from traditional software programming? There's a wide range of expressive building blocks beyond conventional smart contracts and solidity development. Don't waste your time building the basics from scratch and don't limit the potential of your vision. Cartese provides powerful and scalable solutions for developers that supercharge app development.

Starting point is 01:00:11 With a Cartese virtual machine, you can run a full Linux OS and access decades of rich code libraries and open source tooling for building in Web3. And with Cartese's unique roll-up framework, you'll get real-world scaling and computation. No more competing for Blockspace.

Starting point is 01:00:26 So, if you're a developer looking to push the boundaries of what's possible in Web 3, Cartesi is now offering up to $50,000 in grants. Head over to Cartesie's grant application page to apply today. And if you're not a developer, those with staked CTSI can take part in the governance process and vote on whether or not a proposal should be funded.

Starting point is 01:00:43 Make sure your vote ready by staking your CTSI before the votes open. Okay, EJA, all of this conversation in this episode so far has been all about agents, because that's capturing the mind chair. That's a lot of the attention. But AI crypto is not just agents. I think agents are like very dominating, sucking a lot of oxygen out of the room. But there's a whole other, there's other verticals out there. Bintencer, Tao.

Starting point is 01:01:04 There's other platforms that are working in the AI crypto space. And so there's also progress going on here. Download us on the outside of agent AI Cryptometta. Yeah, so we're all here having fun with our new toys. We're playing around with these agents. They're great toys. They're very entertaining. They're much more relatable, at least on the human level.

Starting point is 01:01:25 But there's some really important work going on underneath all of this at the infrastructure layer. And we mentioned this earlier. Bitenser is kind of like the leading AIL1, if you want to kind of brand it that way. and it's been building some really meaningful AI resources in the sense that it's got a decentralized compute subnet. A subnet, by the way, is kind of like a protocol built on top of this L1. It's got model integrations. It's got a ton of really cool stuff that we've seen. And that's kind of like been growing, again, pretty exponentially over the last year or so.

Starting point is 01:02:00 So we've seen kind of like Betenza have its exponential growth from the end of last year. But how it relates to all of this is it's really seen some traction on the agent's side as well. And it's not maybe directly as it's launching its own agent, although in the tweet that you see here, that's exactly what it's done. You've got this Talgod terminal, which basically gives you updates on the latest improvements and launches on the Betenza network. But we're starting to see a number of these agents leverage or use some of the subnet, or protocols that are built on top of Bitenser in maybe more boring but very important ways. For example, you have these agents which consume a bunch of data, right? They look at images, they look at David's tweets to figure out, you know, how's David feeling

Starting point is 01:02:50 today? What's the best response that I can give to David that would garner his attention and get me a like or get me a follow? But some of this stuff can also be fake, deep fakes, right? So there's a subnet on Bitenser that now you can have. have an API to plug into your agent that it can help detect all these deep fakes, which are becoming a lot more prevalent on social media these days. There is another subnet. I think it's, I'm not going to name the number because I might get that wrong, but it's basically a subnet which has several different models available for an agent to plug into and test and iterate and get the best output for whatever prompt it's being given from Twitter, whatever that might be. So the point being here is,

Starting point is 01:03:31 Whilst all this agentic activity is going on, it's also very important to keep an eye on the underlying infrastructure layer that these agents are starting to integrate with, because that's going to be a very synonymous overlap between these two AI metas. If you remember, prior to all this agent stuff two months ago, Betenza was all the rage, all the talk. It was all the mindshare on Twitter. If you check the Quito terminal, it would display that. But now we have agents kind of exploding. and a very naive way to look at this is, well, that means Betenza and Tao and all the infrastructure developments, not that important, but it's actually quite the opposite. I think both of these work pretty intuitively with each other. And if you look at it from an investment opportunity or simply from an innovation opportunity, it's definitely important not to take your eye off the whole Bitenser side. And it's not just Bitenser. I know we're spoken about that for the last couple of minutes, but we're seeing other teams of the infrastructure layer really push forwards. For example,

Starting point is 01:04:31 seeing Nuse Research announce a 15 billion parameter model training run. For context here, at the end of last year, the best and greatest minds in decentralized training, which, by the way, was Google DeepMind, had only managed to figure out a way to train a 400 million parameter model. Okay, much smaller. We are now at 15 billion parameter model, and these were basically paved by two decentralized groups. or teams working on this. One was news research,

Starting point is 01:05:03 the other was Prime Intellect, which announced a 10 billion parameter model that just got successfully trained a few weeks ago. So just seeing the development of this ecosystem at the infrastructure layer is definitely not one to take your eye off. And one I want to kind of want us to keep bringing up in these series to kind of remind the audience of the true innovation that's actually going on beneath the AI agent layer.

Starting point is 01:05:27 Yeah, okay, so the AI agent layer is going to be the front phasing cosmetic consumer layer. It's the actual end products of these things. But there was a recent quote that I think Justin Drake said on bankless is like anyone who's serious about software needs to be serious about hardware. And there's the backend. There's all the AI agent backend that is supporting this stuff. There's the hardware stack.

Starting point is 01:05:48 There's the tech stacks. And while the AI agent stuff is really fun and exciting, bit tensor and these other like software elements of the AI stack vertical, are the thing powering these things. I think what you're alluding to, I think where this goes is that as we are building out the back end of these AI agents, the AI agents themselves

Starting point is 01:06:08 have the ability to go tap into that back end without a middleman. They can go touch and access those resources as they see fit as they need, which is a little scary and extremely bullish. Yeah, I mean, it's the pros and cons of making these agents fully autonomous. Eliezer Yukowski would be very upset that we are doing this.

Starting point is 01:06:33 This is the exact opposite of what he wants. Yeah. Yeah. But if historical trends have shown anything, we're probably going to do it. Either way, we're doing it either way. We're probably going to do it. Yeah. I don't have the skills to save the world from super sentient AI, so I'll just podcast about it along the way. Yep.

Starting point is 01:06:52 Let's just go with the vibes. Prime intellect. I have some context here. Vincent is the guy behind. prime intellect. He was like a part of the Zuzalu crew, deeply tapped into decentralized science, deeply tapped into AI, kind of a connector of just a bunch of different parts of like these frontier technology worlds. Also just like great guy. I can just kind of just vouch for him as an individual. Before I really met him, I, he ran into me, I ran into him at, um, at DevCon

Starting point is 01:07:22 Bogota. And I was just exhausted. And this guy was like, yeah, in my ear off about D-Sai, about how bullish this D-Sai is, like a full year and a half before D-Fat D-Sai blew up and AI. And I didn't really piece together who this individual was. And then he's, and then this guy is, like, leading the AI track at Zuzalu. He's like, he's the founder of Prime Intellect. He, like, knows everyone in this space. He, he's introduced us to E-Mad Mosque and how we got E-Mad Mosque on the, on the episode. And he's like, yeah, definitely like a mover and shaker in the back end, a lot of this stuff. Yeah, I mean, I've never met him personally, but his, uh, his, his, his,

Starting point is 01:07:59 His work speaks for itself. Him and the Prime Intellect team has absolutely forged a new path, which literally didn't exist at the start of this year. Everyone, researchers both within crypto and outside of crypto and traditional AI, didn't think any of this was possible because there was no path forward. And here we have, like, complete breakthroughs. So, yeah, Vincent and the team, shout out to them. Chads. Chads. All right, Ejazz, I think this kind of covers the weekly news of the past week. give us predictions about what you think happens this week.

Starting point is 01:08:29 What are you paying attention to? What are you looking at? What are you plugging into? I mean, there's some kind of crazy things going on, David. So let's maybe spend the last couple of minutes talking about some of the unique agents and agent models that I'm seeing popping up. And let's kind of like run through what this might mean for the space going forward. Because hell if I know, but I have some thoughts and let's discuss.

Starting point is 01:08:54 So the first is this agent called Vader, V-A-D-E-R. Like Darth Vader? Like Darth Vader? Although they're in a recent announcement, they're moving away from the law because of potential IP contentions down the road. They're a professional team, you know, they're definitely not, you know, meming themselves. This is the right account. Let me have a look. Yes, that's the right account.

Starting point is 01:09:17 And that's their new logo. As you can see, Darth Vader is no more featured on this. But the point around Veda is it's a mixture of like an influencer research agent in the sense that it gives you a breakdown of who the elite AI agent coins are, market caps, etc. But one thing it's experimenting with recently, which I think is super cool, is this idea of a decentralized ad model. And I don't understand it too deeply, but at a high level, essentially any team or individual that, can become the majority donator to its Dow or to its wallet. So say for example,

Starting point is 01:10:00 David, you had a company and it primarily worked off of David Coin or in this context, Veda AI received $10,000 worth of NRN, right? AirDrop. So if it becomes the majority holding in that agent's wallet, either by, it's definitely by USD value, then you basically get the chance to have a sponsor

Starting point is 01:10:21 post or a sponsored shill, if it were, from Vader AI. And this is pretty unique because prior to this, you kind of were on, you were relying on centralized monopolies to kind of feed your request to the right

Starting point is 01:10:37 outputs. You need to kind of pay for slots, et cetera. And now we kind of see that middleman get cut out by directly interacting with these agents. And there's no contract being made here. There's just a token being sent to a different address. And then the alignment behind, well, payment was just sent, and now this Vader AI is like aligned,

Starting point is 01:10:58 and that's the reason why they are doing the promotion. Is that kind of the idea here? Pretty much here. And basically, the markets are always accurate, right? They're always very honest. And there's nothing more honest than the USD value denomination of your token or the amount that you put in. So I think we'll start to see an interesting gamification loop here where we see brands that

Starting point is 01:11:20 are particularly strong, either send Vader or other iterations of this ad model agent, its own tokens to try and get newer users to onboard onto their platform or their product. Or we might see non-crypto-native brands or teams, which are looking to tap into a different audience, but potentially that's the crypto audience, to launch their own token and get involved in some kind of an ad network like this to be able to advertise in a much more meaningful way to the specific user audience that they're going for. It seems a little bit like when meme coin teams would send Vitalik a share of their meme coins in hopes that he would just like somehow like legitimize it.

Starting point is 01:12:04 But this is a little bit more explicit. But also like what's my as a human, what's my incentive to follow this Vader AI account if I know it's also kind of just like a proxy shill account for whoever pays it money? Like why I that's just do I follow this ad account? because why, what's in it for me? Well, I mean, like, do you scroll Instagram and you sometimes see sponsored posts and you just maybe look at it or glance at it

Starting point is 01:12:25 and then continue scrolling onwards? I think that's going to be the same thing. So to be clear, Veda AI isn't specifically and solely building out a decentralized ad network model. It's doing other things. It's explicit goal is to become decentralized BlackRock, which God knows what that means. But it puts out very entertaining tweets

Starting point is 01:12:47 and perhaps shares a lot of investment alpha, which I think is its whole point. It's going to become more and more, it's going to become better at aggregating attention through its alpha and through its specific utility, which is like investment advice, for example. So it has a product, and then this is just an adjacent, like, monetization revenue for its main product. That's correct. So, like, it will make a bunch, it'll make 10 posts that are helpful and useful and insightful, and then I desire to consume as a Twitter consumer. And then the 11th is like an ad.

Starting point is 01:13:20 That's right. That's how I understand it. Cool. I do notice here at the bottom of this one post about this project, there are two wallets. One is a Solana wallet. One's an EVM wallet. Yeah. I just don't think these AI agents are going to care about what tribe they belong to.

Starting point is 01:13:35 And they're just going to exist on both. Yep. Exactly. They want to tap into the maximal amount of audience members that they can. And whether you exist on Eith Solana or the 50th L2 or some other chain, that's fine. But I guess it does matter where like Vader AI has a token, right? Yeah, that's right. Where was that token issued?

Starting point is 01:13:59 Because you do have to declare home base with wherever your token is issued, right? So I guess that is the competition between these chains. Correct, correct. Yeah. And so liquidity is going to be super important. I assume we'll probably see some kind of exchange aggregators give you the ability to purchase the token de novo from whatever wallet you're in. Right. Yeah.

Starting point is 01:14:20 So let's find out where Vader is. Vader is on base. Vader is on base? Yep. How do you feel about the wars of the Salana versus Base Wars? Is it still just these two players? Is it Salon versus Base? Is there anyone else trying to enter the...

Starting point is 01:14:32 I'm going to give you some frank thoughts. I don't have all the answers, but I have some observations. And I'm still trying to noodle and figure all of this out. So yeah, the two ecosystems that are battling this out are base and Solana. We don't yet, or that I haven't yet seen any specific L1s or alternative chains that are owning its own agentic economy that is, you know, being proliferated on crypto Twitter or within the crypto world. I don't see that yet. So it's base versus Solana if you want to kind of pitch it up that way. Now, initially, most of the activity and most importantly, volume has been seen on Solana.

Starting point is 01:15:12 We actually mentioned this on our first episode, David, when we spoke about the goat stuff. So Solana is kind of like the one which aggregates a lot of the, maybe, I don't know how I would describe it, but like that raw crypto-native energy where they're experimenting with something new. It's kind of similar to the vibe of Ethereum ICOs back in. 2016, 2017 or NFTs or DFI where there was that real like hungry culture to figure out what this new exciting meta means. On the base side of things, it was more of a slow and gradual approach or start. But now we're starting to see an explosion of these agents being built up. The mind share on Quito is, you know, unignorable. You can't kind of like ignore what's going on from agentic interactions, from follow accounts,

Starting point is 01:16:04 growing exponentially from the ability of what these agents can do. It's just becoming better and better and better. And so I think it'll be a few weeks or maybe a few months before we really see how these two ecosystems play out. That being said, let's talk about like kind of like the elephant in the room, right? Which is we've seen AIXBT, which is basically virtual's leading or top market cap agent right now. Garner, I think, 50,000 followers on Twitter and basically take over mindshare of any agent. agent over the last week and a half or two weeks. So in that context, I'm curious why its total

Starting point is 01:16:42 market cap is still only just over 200 million. When you kind of compare that to other agents, which were top of mindshare back in the day, but maybe less so right now, which are still holding around 400 million up to over a billion in market cap. It's just super interesting to see. And I wonder if that will kind of change going down the road. Yeah, because now we're getting into like market structure conversations, right? Solana has captured the hot ball of money. And Base has like a little hot ball of money. There's something there.

Starting point is 01:17:14 But like random meme coins can like fart out a $20 to $50 million market cap on Solana just because that's where the liquidity is. And when there's liquidity present somewhere, there's a premium associated with tokens that have any amount of life on Solana. And base just doesn't have that same hotball of money. So you can interpret this two ways. It's like, well, it's bearish base because it's just harder to get things to pump on base because it doesn't have that hot ball of money. Or you could say, like, well, it's bullish because what if that hot ball of money migrates to base?

Starting point is 01:17:46 All of these tokens are just undervalued simply because base doesn't have the liquidity premium that Solana has. Yeah, yeah. I also think that I'm seeing a lot of the virtual's agents also try and natively integrate, directly with Farkaster, which is the alternative social media platform, which I think is awesome, right? Because if you're building agents that can primarily interact and engage on-chain,

Starting point is 01:18:14 then using an on-chain social media platform where it can natively integrate is a natural fit. The concern that I have here is it also kind of silos the audience that you can interact with. Most of crypto Twitter, undeniably, is on Twitter. It's on X. A lot of untapped non-cryptanative audience that you might want to expand into also exists on X and other social media platforms. So I wonder how these agents might kind of bridge that gap between just being Farcaster native and having all these cool on-chain interactions versus being able to do, sorry, tap into a much larger audience on these other social media.

Starting point is 01:18:54 Yeah. So in addition to the Bags versus Salon Awards, there's also the Farcaster versus Twitter Wars. And it's kind of the same thing where like maybe base has better fundamentals because Coinbase, Jesse, has really leaned into growing the AI agent ecosystem there. And so they have maybe better infrastructure. That's just a take. Maybe not accurate, but that's a take. But you can see that same take in Farcaster versus Twitter where Farcasser is much more enshrined infrastructure that's better for AI agents because you have an enshrined wallet in your AI agent account. and Farcasters being built for being more AI agent native

Starting point is 01:19:29 than Twitter is Twitter has a bot problem. And actually I noticed in this last week, some new points for the Twitter column is that they actually started to build out this automated bot label for certain accounts that it knows our bots, which means that Twitter is aware of this phenomenon and is starting to label bots it likes versus bots that it doesn't like.

Starting point is 01:19:50 There's like the fishing bots versus this is a real AI agent, agent. And so like kind of in the same way that we are trying to differentiate between like like M.EV and AI agents. Both are bots. One smart one's not. There's like fully sentient AI agents which are like have an identity, have like some semblance of a soul to them versus just like the spam fishing accounts. And so you can see even in the one of the tweets that we were showing with the AI 16. Z, Shaw put out a message in Discord saying like, oh yeah, we actually just met with the X team about. figuring out ways to move forward, about having AI agents as like first class citizens shoulder to shoulder with humans on Twitter. So you can see Twitter is aware of what's going on here. Yeah, I actually think Elon is much more involved than we give him credit for, but he's just not publicly kind of verbal about that. One point I want to mention around the whole bot and agent thing is it's a very, it's a common conversation and critique. I see as, Like, aren't they just bots?

Starting point is 01:20:56 There's just a simple difference between them. Bots are hard-coded to do a specific thing. Bots are like deterministic. Correct. Agents are non-deterministic. It's probabilistic, actually. That's the nature of AI. You can ask chat GPT the same question twice,

Starting point is 01:21:12 and you'll get two different answers, different variations. Right. So that's the difference. And these agents also iteratively adapt and have context and can learn who you are as an individual. and what kind of a request means coming from you versus another person. So very important nuance. Cool.

Starting point is 01:21:31 Ejaz, as we give this episode of close, thank you listeners for sticking with us. I think these episodes are very fun. I'm having a ton of fun. I'm learning quite a lot. So thank you, Ejjjad, for being in the trenches and relaying a lot of this information to us. Give listeners some homework. What should they do over the next seven days before we record our next episode? What should they go do?

Starting point is 01:21:51 Where should they go research? What should they read? I think you could start off by following some of these agents. And I'm not saying follow all the agents that we've mentioned on this episode, but go give AIXBT a follow or read through its tweets and see if one of them can engage you. Go check out bully. Go check out Zara Bro. Go check out Go out Goad and see whether you can relate to any of the tweets that they're making.

Starting point is 01:22:16 That's step one, I think, in terms of like dog-fooding you on this whole AI agent meta. Try and have fun with it. Secondly, I would say if you're interested in getting involved in the space and maybe kind of playing around with some of these tokens, check out the virtual's ecosystem. Check out the cool agents that are launching off of AI16Z. There are plenty of different agents to suit different types of personalities and tastes and preferences. So play around with some of them. I always say this about crypto. If you don't have a stake in crypto, you're never going to care enough about it.

Starting point is 01:22:48 You're never going to give an accurate opinion about it. about how all of these things work. It's very important to have some kind of a stake. And now you have an ability to do that. So those will be the two bits of advice I'd give. I'll also say give EJazz a follow. He is tweeting about these things. That's my homework to you.

Starting point is 01:23:04 We'll put EJazz's Twitter in the show notes. And so you guys can follow him there. We'll also get a QR code on screen for people who are lazy. So there'll be a little QR code floating around the screen right now. Bankless Nation, you guys know the deal. Crypto is risky. Crypto AI may be even riskier. you can lose what you put in.

Starting point is 01:23:24 We might end up destroying the earth, but we're going to podcast about it all the way there. We are headed west. This is the frontier. It's not for everyone, but we are glad you are with us on the bankless journey. Thanks a lot.

Bankless - AI ROLLUP #2: Virtuals $1B | Zerebro LLM | Ai16z Comeback? | Bittensor Tao | Freya $50K Game

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.