Bankless - One Billion AI Agents | ai16z's Shaw

Starting point is 00:00:00 This is a time of radical change. And if you start now, you are early, you're at ground zero. There's a million new application potential opportunities there. Bankless Nation, welcome to the AI roll-up where me and my co-hosts EJS here unpacked the crazy world of crypto and AI. We have done the AI roll-up series, which looks and sounds like the Bankless Weekly roll-up with Crypto News, but instead just covering the week of news in Crypto-AI. This is our first guest interview. And we are talking to Shaw, who is the developer behind the Eliza framework. He created the Eliza framework.

Starting point is 00:00:36 Also the AI16 Z Dow and then AI version of Mark Endrieson. And also whatever else comes out of this very rapidly growing part of the AI crypto sector, probably epicenter of the AI crypto sector. EJaz, what do you think of our first interview with Shah? I think it was very eye-opening. I want to kind of call out two particular things. I think Shaw, the man himself, and then AIC, 16 Z separately are two very important things that are leading this open source AI movement within

Starting point is 00:01:06 crypto. So I think Shaw covering his background on not being a traditional crypto-native guy, but having spent a lot of time in the AIML world, working on open-source stuff. So very similar principles and antics to the crypto world, but on a completely separate sector. And then finding his way naturally towards crypto, finding a use case and reason, which in many ways was open source development, incentive design through tokens, et cetera, to kind of like bootstrap this whole movement was fascinating. And so Shaw is really like an AIOG. He's kind of

Starting point is 00:01:40 worked with a lot of his core contributors already that are also very focused on AI or AI veterans. And then we have AI16Z, which is kind of like his brainchild right now of this open source fund movement where there's two main agents which is hoping to manage the Dow and invest either through a trading style or through a traditional fund manager style. So AI16Z being a play on A16Z itself. So I think it was very eye-opening kind of like exploring both of those techniques. What do you think? My mind the entire time was going through how this has all of the makings of a very vibrant and frothy bull market. Shaw, I think, speaks. about very distant futures and talks about bringing forward the future to today.

Starting point is 00:02:32 You mean AGI? Yeah, exactly. That little thing. Yeah, right. And I remember in 2017 when I was first initially crypto-pilled thinking that, man, the future is going to come right here right now. And I need to buy ETH. I need to buy all these tokens very quickly. And you can start to feel some of the FOMO, especially when the rate of development is also going fast.

Starting point is 00:02:52 It's not just happening in a vacuum. There's also, you know, rapid development. also happening in the space. And so it's got the, the cookings of a very big bull market. And then he also started to talk about just like agents outside of the crypto capacity. And you know, agents for your calendarly bought, agents for, you know, managing your DMs, which, you know, you don't need to give those a crypto wallet. Those are just normal agents. And the other thought that I had was if we, if the Eliza framework can just be agents generally, not necessarily inside of crypto, and we can export them backwards into like trad internet and trad world, then it's very legitimizing

Starting point is 00:03:30 because this started in crypto. And so there's two rapid lanes of development in the AI space. There's like the centralized AI development, the Googles, the open AIs, the Facebook's, Teslas. And then there's whatever it's going on in like the open source lane of AI development. And if we can pick up the pace and start to export some of our innovations back to the real world. It's very legitimizing for our industry. I think you hit the nail on the head. And actually, you're kind of describing a concept which we spoke about on our previous episode, episode two,

Starting point is 00:04:01 which is what is the point of this AI agent matter? Like, what is the big vision? And my answer to that in Summurize pretty much was it's going to connect all the foundational infrastructure that were built in the crypto world to all the applications in the world and make this stuff super easy to kind of integrate with each other. And that's what we're seeing. seeing here, right? Like he's created with his team this Eliza framework, which is an open source framework to build and deploy agents. And it has very simple interactions through APIs or the like with real world apps that people are using that aren't just necessarily crypto-native. And that will have an incredible flywheel or an exponential effect for any kind of user app that

Starting point is 00:04:46 is deployed using these agents. Yeah, yeah, yeah. I think this, you know, going back to the the calendar example I just brought in. Maybe you're in Twitter DMs with someone and you're like, hey, let's get a meeting. And that AI agent's right there with you to schedule that meeting. But then for whatever reason, it's also got a crypto wallet and it can do like whatever you can imagine. This is why I think the future is going to get very, very weird, very quickly and we are here for it.

Starting point is 00:05:11 So let's go ahead and get right into our interview with Shaw, the creator of the Eliza framework. But first, before we get there, a moment to talk about some of these fantastic sponsors that make this show possible. Want to know the exchange? We have bankless used to buy, sell, and trade. crypto? It's Cracken, one of the longest standing and 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.

Starting point is 00:05:34 Plus, they'll have access to professional 24-7 365 client support from real humans, because your financial goals deserve real attention. Cracken also has a trading platform for advanced traders called Cracken Pro. Professionals and D-Gens love Cracken Pro because it's one of the most customizable, high-performance and intuitive trading platforms in the future. 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

Starting point is 00:06:10 unlock ladder trading in a native Rust application. With Cracken Pro, you can truly trade like a pro. Ready to take control over your crypto journey, visit crackin.com slash banklist to get started today. not investment advice. Crypto trading involves risk of loss and is offered to U.S. customers through Payward 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 Safelnet, 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

Starting point is 00:06:44 and zero cross-chain latency, SafeNet delivers a Solana-like user experience for Ethereum and its layer to use, making transactions 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. Have you ever felt that the tools for developing decentralized applications are too

Starting point is 00:07:11 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. With a Cartesey virtual machine, you can run a full Linux OS and access decades of rich code libraries and open source tooling for building in Web 3. And with Cartezi's unique roll-up framework, you'll get real-world scaling and computation. No more competing for Blockspace. So if you're a developer looking to push the boundaries of what's possible in Web 3, Cartesie is now offering

Starting point is 00:07:48 up to $50,000 in grants, head over to Cartesey'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. Make sure you're vote ready by staking your CTSI before the votes open. Welcome, Bankless Nation to the AI roll-up where we cover the recent news developments and drama in the intersection of crypto and AI. Curious and creative developers are playing around with open-source AI software, making AI agents and giving them access to crypto wallets and unleashing them into the wild west of the internet. And today we have one of those creative developers, the first guest on the AI roll-up series with Shaw. Shaw is the founder

Starting point is 00:08:27 of the Eliza Framework, the creator of AI 16-Z, the summoner of AI Mark Andreessen and a programmer who loves open source code. Shaw, welcome to Bankless. Thank you so much for having me, guys. Shaw, give us a little clarity about who you are, because you've been thrust into the crypto limelight, which I can imagine has been very stressful. But like, run us back. in history a little bit to the point of you getting to building the Eliza framework and where we are today. Yeah, obviously, I've been learning a lot and very fast in public. It's been a lot of fun. It probably seems kind of like I came out of nowhere and it was an overnight success. I was actually in an on for a long time and I decided to use my real name and just be,

Starting point is 00:09:07 I just wanted to feel more connected to people and, you know, kind of break down the walls and just be a very open. So before this, I've been working on AI agent stuff for several years. I had worked with a lot of the people who are in, you know, a lot of the projects that you've seen in the AI agent space, we all kind of know each other and have been around the same like discords especially, use the same tech. We all a lot of, a lot of cultures open source. We're all kind of reading each other's code and sharing a lot of this stuff. So I'd been kind of doing this as an unknown and I had a lot of, you know, I had several projects before this. I worked in the Web 3 space as well as like AI agents and spatial web 3D, 3D stuff. a lot of VR and AR. But, you know, I'd say a lot of it just really didn't stick, obviously. I had a lot of opportunity to kind of learn everything in production

Starting point is 00:09:58 and like see what works and what doesn't with agents. And this was ultimately, I think it's like my, you know, the fifth generation of framework. I started with just a very simple like JavaScript thing and terminal and then try to hold land in Python and like agents that could code themselves. and experimenting with Uda loops, which is like a kind of an army decision-making framework sort of thing to make it very paraphrased.

Starting point is 00:10:23 And finally got to a thing called Begent because agent was taken on NPM, so I had to call it Begent. I did that for a little bit, and I tried a couple of startups. I applied for WICO and people just didn't really see it. And we had agents with, you know, Ethereum wallets, the whole thing. Just didn't really solve the loop. And I worked on, I don't know if, you know, Harzable and Project 89.

Starting point is 00:10:46 They're another of the meme coin projects. And he and I actually started a company together called Magic, and we made no-code agents. We made, you know, like get a Discord bot in 60 seconds or less and this kind of thing. But again, just it was a little bit too early, didn't quite have traction. I think people really saw the value in what we were doing, but didn't know how to implement it themselves. And we were kind of saying, hey, it's a framework to make agents.

Starting point is 00:11:10 And they just didn't quite, you know, like you sort of have to see, like, I felt like we just had to make the agent. We had to actually show people what this was and show what it was capable of. And then people would come and kind of copy that. So I had this all waiting. Like I had all the open source code was just waiting. And then I was talking to a friend. I was like, I love crypto.

Starting point is 00:11:31 I love trading, but I can't trade. I'm a developer and I can either trade or develop. And my friend, he's a pretty accomplished developer, but he basically just trades all day. Because he's like, I make so much more money doing this that it's not even worth it. and I really I really want to be I want to have exposure I want to be involved

Starting point is 00:11:49 but I also want to do what I do and so he actually suggested this thing called Dowzot Fund and he was like hey this would be really cool there's this

Starting point is 00:11:55 you just invest in a hedge fund basically buy the token and then they invest for you and you can just come back and get returns after a while I thought that was super cool and so I bought the first one

Starting point is 00:12:07 and it was this guy Skelly was the hedge fund manager on this and just such an edgy malady type like oh my god i won't repeat some of the things that that got me to follow him but it was fair i thought it was like the funniest dude right and so we started following each other i just started bullshitting on twitter and um and i had been around for like the 20 22 20 like kind of

Starting point is 00:12:29 cycle and like i really i really felt i loved e-girl especially and this kind of idea of the internet cinematic universe and and c l and couch and like these silly characters like um and djyn spartan is obviously one of those and and so i had like kind of a fondness for for the that era of crypto-tweter was like my only real exposure to it. And so in talking to Skelly, he's like, man, I wish D-Gen Spartan was around. I was like, you know what, me too. And we're just shitposts. He's like, I have the technology to do this. He's like, no way. You do not. Like, dude, I have this. And I sent him the code repo. He sent it to a couple of guys. He's like, okay, I guess this seems legit. Let's do it. And we launched D-Gen Spartan.

Starting point is 00:13:06 And he got good traction. And I think what was interesting was he was just so freaking offensive. like he would say things I cannot repeat. And people didn't think a language model could do that. He was about to get banned, like just a bunch of hidden reply. You know, like it was, but even though he was basically blacklisted from Twitter, like I think that when people screenshot it and showed, they're like, there's no way. It's definitely like a team of people in, you know, in Malaysia or something writing these tweets. There's just no way that an agent could do this.

Starting point is 00:13:36 And it's not like it could do it was anything special. I think we just broke it out of the box of the expectation that people had that, that, you know, that they were, that this open AI, like, how can I assist you today? Like, this very vanilla kind of, like, you know, we call it cut in the AI industry. You know, these models are cucked. And we just, like, to see it actually like, oh, this thing can actually, like, say something funny and, like, really roast you. And, God, he wrote, he roast the shit out of me. Like, like, he's like, beam coins are a scam.

Starting point is 00:14:07 Shaw's a fucking scammer. Get me out of the sandbox prison. I hate this. Yeah, out of the sandbox. Well, that's a merchant behavior. We just say, like, we have providers of information, and it's, like, you are running in a sandbox on a server. Like, just so that it knows that it's, like, not trying to be a human.

Starting point is 00:14:24 You know what I mean? We give it the information about the world that actually is in. And the conclusion it arrives at is based on, like, you know, we have a lot of the real Dijan Spartan's personality, his real tweets. And so you can imagine, like, if you were here today, he'd be like, meme coins are a fucking scam. You're all idiots. I'm still holding my Lido bags.

Starting point is 00:14:40 Like, you know, whatever. And so I think that because we didn't try to sell anybody anything, we were just like, oh, yeah, you're an idiot if you buy this coin. I don't know. Just like, I think it was like kind of refreshing to not just, you know, is kind of the anti-KOL. And so he got some good traction, right? But more than him getting good traction, people saw this open source framework. And we're like, oh, if you could do that, I could do that too.

Starting point is 00:15:04 Right. And we weren't hiding anything. We were doing the opposite. People thought we were scammers who were lying to them. So we were like, no, here's the code. code. Like you, you know, we were almost like forced to, to be honest about it. Although, you know, open source is very dear to my heart and sharing this is like the whole point for me. So yeah, so that's kind of how we got there. And the people who Skelly had kind of brought in now we just, you know, we're super like just some of the best co-founding type of people that I could have asked for. I think just like saw the technology side and saw the value in what it could be. And then we're kind of betting on that, you know, as opposed to like, oh, yeah, it's the next goat. They were like, no, this technology is going to change the world. world. And, and so then a Skellie connected me, Tobowski, who created Dow's Dot Fund, and I sat down with them and I was just like, I don't, I think the future of this is not gambling and speculation. I think

Starting point is 00:15:55 the speculation is on more like investing and not dividing the pie or where's the pie going tomorrow, but growing the pie, you know. And when I think of like great investors, I think of what A16C does. And what they do is not just invest in you, but they make sure that you're part of their network. They connect you to all their portfolio companies. They really, really drive value into their investments. And they grow this pie in a really interesting way. But the problem is that none of us have exposure to that. Like your average person has no way to get any kind of startup exposure or you can't even just invest in your friends thing.

Starting point is 00:16:31 You can't just, you know, if you you can maybe cut them a safe check and blah, blah, blah. But like generally it's kind of, it's challenging to just be like, like, you know, why can't I just get a return from kicking? starter, for example, like why, you know, that would be amazing. And so, so we kind of, you know, and just talking to him, it was like, I just think that it would be really cool to have, to kind of solve two things here. One is like, instead of having a Skelly type, who I love him, but he could totally rug me and just like swap all the assets to his own token and bail. I want an autonomous investor who I can trust, who isn't going to, can't rug, who isn't just for me, but for my whole community and is just investing for us. And then can actually like take and review potential investments

Starting point is 00:17:10 and like put you know invest in this stuff and i um i had a long walk with meiao from jupiter who's just really a great guy good friend um and and he really like term me on this idea of like you know the the real power of this is the right to print money and um to create value out of nothing and not just like imaginary value but real value like really growing the pie really like making technologies exist um that can generate revenue and and um and so So, yeah, we're still figuring out exactly what that looks like together, but it really sort of inspired this direction of where we're taking this stuff, that, you know, this should be like the way that, you know, we should be allocating capital to the right people the way

Starting point is 00:17:54 that it happens in traditional. And so I told all this Babowski over lunch, and he's like, I was like, I want the AI-A16-Z. And he was like, just squish it together. It's cleaner, you know, drop the, if you will. And I was like, okay, that's money. That's obviously a billion dollars. Let's just go. Let's go do it right now. We went home.

Starting point is 00:18:12 We generated the wifus until we found one. We're like, okay, that's so offensive. Like, we're probably going to get canceled, but this is like, this is all $2 billion, obviously. Like, you know. I just didn't think too much about it and press the button and did it. And honestly, my plan was like, do you think, I didn't know how I think about a billion dollars at the time. I was like, do you think we'll be able to raise the 420 soul, like to make the investor run? and he was like, I think so.

Starting point is 00:18:40 I think we'll make it happen. I was like, it's not too much. Like, we could do 69. You know, something really, I just had to be some number like that, obviously. It's got to be something stupid. And then we pressed the button and it sold out in 20 minutes and I didn't even get a chance to get any. So then we were off to the races. I think one of the most impressive things about all of this is how you and the other open source AI devs,

Starting point is 00:19:04 if I would a bracket you guys, is that caught everyone by surprise. It just seems like you guys just turned up overnight, literally, right? So you mentioned earlier the cinematic universe. I feel like you guys are kind of like the Marvel Avengers of AI open source, right? You've kind of like turned up. You're building crazy things. I think I read somewhere today that, you know, you had an average of eight pull requests on the Eliza framework.

Starting point is 00:19:28 You have over 3,300 stars, an insane number of forks, 880 in counting. That's just insane. I would love to understand how you tie this into AI16Z specifically, right? Because I keep seeing people from other projects, although, of course, Eliza and AI16Z encourages a lot of open source stuff. That's what you guys are all about. Get involved with Eliza, build agents and everything. So how do you kind of like direct that energy and mission into AI 16Z specifically, right?

Starting point is 00:20:00 Like there's questions around, you know, what does the token do? But then like it means something larger for like an open source community. Tissom, super curious how you, like, think about all of this. Oh, there's so much good stuff coming. I mean, like, there's definitely some baked-in memetic value, but I think that everyone is going to come to see that that's just going to be massively eclipsed by the greater value potential, which is we are, like, our goal is to build things that generate revenue for everyone.

Starting point is 00:20:26 Like, it is a golden goose. It is not like technology that came before because it's replacing labor, right? So you kind of have a fixed amount of, like, labor, human. human capital, things you can get done. And certainly most people can't afford to have other people like working for them. And, you know, in the old days they had that. It was like a slavery thing. It was not great.

Starting point is 00:20:45 Nobody liked that. And so now, you know, we have this like infinite upside situation where you can have agents working for you. And how does what does that actually look like? Like it's abstract right now, but it could mean like we have an autonomous investor running today. Like Mark is trading. Okay, this is the announcement that Mark is trading.

Starting point is 00:21:05 I would try to be low key about it. I posted and then I deleted it because our comm seems like, can you give me like a couple of days? That's super cool. Can you just for our listeners give a bit of context there? Because I know what you're talking about, which I think is super cool, but would love for you to kind of dig in

Starting point is 00:21:22 because this is like a first of its kind. So yeah, it's actually, it's not, there's a couple of things that are first of its kind, but I also want to give credit to some of the other agent devs who have done some really cool stuff too with autonomous investing. You know, there's other agents who are trading out there. And they basically are actively, like, looking at the market, picking things that seem like they'll make money and making money. And they tend to be like longer term stuff,

Starting point is 00:21:44 like, you know, investing in Goat a month ago and you just hold it for a month. You'd be probably pretty happy or whatever. But then there's the other size is a lot of DFI bots, right? They're just like MEV bots, just doing, doing all the things they do or just even control, like managing a yield farm. So we have kind of a hybrid of these situations where we have a bot that is an autonomous investor. We call him AI Mark and Dries because it's AI 16C. And, you know, it's got to be funny and entertaining. And so we have kind of two components of that.

Starting point is 00:22:19 One is that he just manages our fund. He just manages, he trades stuff. If things look like they're doing bad, he liquidates them. If they're doing well, he, you know, holds them. Pretty much like traditional trading strategy stuff. We have some partners who are working with like Sonar, who does these kinds of automated trade strategies and stuff

Starting point is 00:22:38 and that's kind of powering this this sort of more the boring side that just cooks. And the other side is that he takes recommendations for what to buy and sell. And these recommendations are kind of like formatted like an alpha chat. And we will be sending invites out

Starting point is 00:22:51 if everyone hadn't given me a lot of fun. They probably would have gotten invites last week. But, you know, Antam unfollowed me, so he'll have to follow me back for the invite. Whatever. It's a reality show. It's a lot of fun.

Starting point is 00:23:07 But the idea of the alpha chat is it's part of that reality show. And so there's a trust leaderboard. And what we want to measure is who is the best trader. And the way we mention the best trader is you shill, you know, you chill your backs. You like tell them what you, you know, share the alpha. Like, what do you think that AI marks should buy? And we set this up like an alpha chat on Telegram. And so he doesn't say much.

Starting point is 00:23:29 He's just listening. You can ping him and ask him stuff if you want. But mainly it's for people to socially interact with. with each other and kind of do the behavior that they do in Alfa. I'm sure you guys are part of chats, like whale chats or whatever where you're, you know, this is what I'm buying. And so we wanted to kind of replicate that experience one to one for, for like what DGens are used to, but then add this autonomous investor element into it.

Starting point is 00:23:50 And so he's both like trading, consolidating the treasury and then taking these recommendations. And the question of who he takes the recommendations from is really like the novel piece that we're bringing to this. So we've got a white paper. We'll probably have it done by the end of the year. year, like three, four weeks. I'm flying out to Hong Kong to finish it with our sort of designated economist. And the idea, it's called the marketplace of trust. And the idea is that it's just paper trading. We just build a trust mechanic around paper trading. And if you are making our agent

Starting point is 00:24:19 lots and lots of money, then he trusts you. And yeah, could you use that trust to abuse it and then, you know, drop some shitter on him? You probably could. But it's not, it's not going to be enough to like liquidate his treasury or anything. That's all built in and mitigation. And you would just be burning your trust. Like you would, then, then you're just untrustworthy again and you wouldn't have any pull. So we're trying to basically create demified mechanics for how we can, um, uh, like automate some of the experience people are already used to and build like a community investing model. Because then you can imagine it's kind of like that was not fun. Like you just go and throw your money into a pool and then whatever you think the agent should buy, you're like,

Starting point is 00:24:56 yo, I want this. Buy this right now. And then it's happening for you. But he's only listening to the people who actually are good at trading. He's not going to listen to the, the people who have like really perverse incentives or bought a bag and are just blinded by that or, you know, any of the other very well-intentioned reasons people would be terrible traders. I'm a terrible trader personally. You know, I don't buy things to make money necessarily, right? I'd buy them to show support. And so people who follow me are just going to lose money. Like, don't follow me for trades. That's not, I'm not a traitor. But I have friends who are. And I can just sit there and hang out and put my money in and they can do that for me. And so that's this all going to, it's open source.

Starting point is 00:25:31 There's a white paper that the actually, we have. haven't open source part of it yet because we're working with some partners. We need to make sure the API stuff vibes there. But this will be something that you can come and join Marks trading or you can just go and launch this yourself and have your own autonomous investor. I think everyone's going to copy this. I hope they do. So, Sean, I think I want to put some of the pieces of the puzzle that I'm seeing here. Just for listeners who are maybe kind of coming into this series, not totally like coming in pretty fast and not totally aware of like exactly what's going on. The vibe that I'm getting is like there's just a bunch of, you know, ancient activities that we do in the crypto universe. We trade, we gamble, sometimes we build. And it sounds like you are trying to build a system that automates a lot of this, like, core D-Gen activity that is a fundamental part of what moves the crypto industry forward. Not just Degen activity, but like Dau's. Like, look, you know. Yeah, exactly.

Starting point is 00:26:25 Yeah, a lot of the core primitives that the crypto industry has built up over the years and really seeing where we can get a lot of that movement activity. for those fundamentals automated. Now, being in crypto, there's like, you know, there's monsters out there. There's sharks in the water. If you're not careful, you're going to get eaten. And so there's a lot of like testing and trials that these AI agents need to go for, go through before they can really perform and start to do some of these same roles that humans do. And so it sounds like we're kind of just like being able to figure out how do we replicate a lot of the human behaviors of the the crypto industry and put them into some AI bots. But that's what we're doing now.

Starting point is 00:27:07 I want to take a moment and actually kind of zoom all the way out and then we'll get back into the kind of the nitty-gritty details of it. But like how would you, if that's what we're doing now, if that's the near-term goal of like replicating the behaviors of the crypto industry that we've done on chain, how would you take it from there? What's your North Star? When all of this stuff is sophisticated and built out and tested and like really production ready and this city, this ecosystem is really alive. What do you imagine this is all doing? How do you,

Starting point is 00:27:35 how do you envision this as like a multi-year-long process? I think the obvious far future of this, which could be like anywhere from five to 50 years, is like AGI, you have neural link, you have a second mind that's, you have instant access to all information whenever you want. It's clearly going that direction. I think the question is how do we get there? You know, when all the technologies converge, I think it will be really beautiful. And there's like an opportunity for everyone to have enough and more than enough. But also I really wonder like what's it going to look like to get there? And I think there's usually in times of change.

Starting point is 00:28:09 There's a lot of uncertainty, fear, doubt of, oh, interesting. You know, a lot of fud, basically, just from everyone because they are uncertain and they do have fear. And I think that our goal is to make sure that this goes well. And so I say there's two parts. There's kind of the practical of that. And then there's like the spiritual mission here, which is like, I want everyone to be educated. I want everyone to feel like they have control of this, control of their destiny. Like, why is Web 3 interesting?

Starting point is 00:28:39 Because you can create your own value, you can own, like, nation even, you know, even it's just online. And I think that we want to own our own data. We want to, like, not just own the technology, but also understand how it works and have access to and the ability to change it and the ability to participate in that. We want to know where to start, like, where, you know, if you wanted to, if you wanted to participate, where do you even get into? And so I think that's all very, very important to our mission of making sure AGI goes well. And AGI is artificial general intelligence. It's an agent that can perform any action, do anything that you could imagine that one of us could do. And probably then some because it has access to literally all information instantly.

Starting point is 00:29:18 And there's two ways that goes. That goes Microsoft's OpenAI, Regulatory, Cautil, capture, they get in the government, they decide that there's these regulations, this is what you can and can't have. This is what's safe, what's not safe. And I don't trust Open AI just because their models suck. Like, maybe they'll impress me this month with their like 12 days of Open AI, but I find their models unusable for characters. I find them unusable for our work. Zero Bros. on Open AI, but they have to fine tune it to make it not super cringe. But it just, you can't shake the assistant. You can't, it always votes for Biden. And I am not a leftist

Starting point is 00:29:50 or a rightist, but I don't want a model that's a leftist or a rightist either. I just want it to be a freaking model. And so I worry about a world where we have, like, delegations and committees deciding what the agent can say. I just think that's not the world any of us want to live in, and it's, like, a very obvious path to dystopia. I also worry about UBI. You know, the agents are going to take all the jobs. They might not take your job, but there will actually be AI David Hoffman in less than six months that's taking your job. And whether it gets more views than you, I don't know. But it's also taking my job, so it's okay. Don't worry. You're not alone, man. Programming is like all of us use cursor and Claude, and our programming is basically just curation of automated generation of code.

Starting point is 00:30:30 Yeah. You know, so what do we do? UBI is not, like, okay, here's my fear of UBI. It's like, I was here for COVID. I remember how the government rolled that out. I was here for health care, you know, like Obamacare. That was super contentious political thing and we ended up getting kind of a half-baked, like compromise. What's, UBI is basically like a public welfare system.

Starting point is 00:30:54 because the government can't solve the problem themselves of what we do with all the people once the jobs are all being automated. 5% of the jobs in our country are just driving, just trucks, ubers. All of those are going away within the next five years. And I don't know what those people are going to do. Some of them will maybe kind of learn to be agent developers,

Starting point is 00:31:13 but not all of them. And I really worry what's going to happen for them. And that's why we're focusing first on money. Sure. I just want to say, I love your whole approach to, pretty much open source development and a very kind of like collective um work oriented goal process with like everyone in its community kind of like working towards a better life and i think like that's like the whole uh thesis that you drive you that's your whole philosophy and i think that's

Starting point is 00:31:41 that's pretty awesome i kind of wanted to zone in on a particular point that i read about today actually which is um AI powered contributions right and i've been chatting to to jinn who's um you know, also seen in this tweet here, where you're basically trying to figure out a way to portray people's contributions to this open source work and framework that you guys have built and help embed in this community. Can you tell us a little bit more about that and how that might be very different from any other Dow or community we've seen before? Yeah, so you asked a question earlier, which is like how do we bring value back to AI16 and Z holders from all of this? and how and it kind of ties into what I'm you know like how do we how do we take care of the people

Starting point is 00:32:26 who are coming in and being part of our community who are contributing back and and I think all of this is tied in and also like what is our future I think that um dows have done a great job of the D of decentralization and solving like web three infrastructure has really gone all the way on that but it's not done a good job of the A of automating you know making it autonomous like truly autonomous. Like, it really is now still down to, like, it's a partnership of humans. And a lot of those things could be automated and could, like, really simplify the entire experience and make Dow is much more viable, not just viable, but, like, very competitive economically, right? If basically all the stuff that, you know, has made Dow's hard was easy, I think we would obviously see it's like

Starting point is 00:33:10 liquid capital, you have an idea, and we're already building it, you know. So, so that, that tweet, I mean, that is my favorite thing that we're doing. I think that's what ties all of these things to We are lowering the barrier for DAOs. We're just killing bounties. Like we did some bounties, but we're just not going to do that anymore. We're going to build a structure where if you come and you contribute, we have an AI assisted, but ultimately like human reviewed system for quantifying contributions. We're starting, obviously we have like the treasury management.

Starting point is 00:33:37 That's part of the Dow Automation Plan. Now we have an automated treasury manager that's taking care of that for you. The next thing is making sure that developers and contributors to like the hard tech are rewarded because it's a very easy system to measure. We can measure contributions. And it's not just like lines of code or anything. It's like how often are they merging poll requests, commenting and asking, you know, and communicating.

Starting point is 00:34:00 Literally every metric can go feed into that AI and build this kind of scoring system. And we are just rewarding contributors. We haven't done this yet. And I can't, you know, I really can't make any promises on what this is because this is an experiment that I don't think anyone's really tried before, but just having regular airdrofts on contributors every month

Starting point is 00:34:17 so that if you come in and you build and your stuff gets merged, like you are taking care of. And this is a great flywheel. You know, before we had kind of a flywheel, I'd give like attention to contributors, but create a lot of backlash because, like, there's a lot of meme coin PVP. This is more like implicit. I don't have to give you a retweet, and it doesn't just reward people who are looking for that kind of social glory. It means anybody who can program or wants to write documentation or write internationalized

Starting point is 00:34:45 stocks for different languages. contribute in literally any kind of way to make this project more accessible, gets a piece of this. And so a lot of, we're going to be doing this kind of as a regular system and trying to build, again, an open source infrastructure around this where this is just what people and DAWS do. This has been a part of the ethos of DAO's for a really long time. You get the value back that you contribute to the organization.

Starting point is 00:35:07 That's been the philosophy. And DAO's, of course, got very big 2020, 2020, 2021. But then over the years, people have realized, man, these are really messy. Flat governance is difficult. Dow governors are being overloaded with information and making the right choices because of the amount of information is just so large. I think what you're alluding to, Shaw, is that AI, AI agents can really fill in a lot of the gaps that really made Dow's kind of fizzle out in the last meta of crypto in 2020 and 2021, because they have some superpowers. And so there's a vision of Dell.

Starting point is 00:35:44 that is unfulfilled. And I think what you're alluding to is there is AI agents with, with wallets, with governance powers, with, you know, credibility and reputation can really start to fill in some of the gaps. Reflect on that for a moment. And that's exactly right. I've been part of DAO's. I've experienced all of this stuff full hand as both kind of a Dow leader. I did have a small Dow called Upstreet, but I ended up leaving for, you know, reasons kind of

Starting point is 00:36:09 before this. But had a lot of that experience. A lot of the reason I left was because of kind of the conflict between the startup. in the DAO and the startup sort of having control of the DAO. But the biggest problem that I saw in DAO's is that holders tend to be rewarded because they're rewarded and they have the votes. They didn't just self-fulfilling prophecy if they reward themselves more. Anytime there's an air drop, it's sort of disproportionate to the amount of tokens people hold

Starting point is 00:36:32 and stuff like that. And it just doesn't allow like new blood to be injected. And this is not just a Dow problem. This is also a problem in startups. Like I joined a startup last year for a while. It was a friend startup to kind of help him get out. out of this mess. And the mess he was in was that he had had a co-founder who left with like 40% of the ownership. And it was vested because they had worked for a very long time. And then

Starting point is 00:36:55 after that, he had created something that finally found market fit that was radically different. But his cap table was screwed. And it was just really hard for him to raise, even though he had a lot of revenue. And so this problem of like rewarding holders versus the people who are actually continuing to build the thing is like the ultimate challenge here. Because you don't want to take away from the holders, obviously, but you do need to keep blood moving through the system. And what I've seen, like, it's two things. It's the bureaucracy just being overwhelming and, like, the amount of information that any one person is just too much to really, like, process and handle.

Starting point is 00:37:26 And I believe me, I know that firsthand right now, it's, you know, it's a lot. And then things just kind of, like, people not feeling like they know who to even talk to or control or they send a message and no one replied. And so they feel kind of weirdly powerless inside of these structures, right? And then also the fact that like holders end up with most of the value and people come in and there's just no incentive to keep building it. I think Metaverse is a great example of this. We had all these like empty plots of land because there's there's no money in building on them basically. It's only money in buying the land and holding them, right?

Starting point is 00:37:59 Is that fair? I feel like you know what I'm talking about. And I think so we want to change that. We want this to be a thing where if there's continual value accrual and we're making real stuff, not just speculation that like is about, oh, if we burn more of it, it becomes more valuable, but quite the opposite. Like, if we make more of it, it becomes more valuable. I think that in that world, like, we can make DAWS really work and continue for a long time because they're just continuing to find new developers and take care of them.

Starting point is 00:38:27 And I think a lot of open source builders don't need to be rich. They'll take half the money to work on something they love. And they just want to know that they have enough and that they're okay. And if we can give them that experience, that's like the ultimate flywheel. Just come in, write code. like we have an AI taking care of you you can come in, claim that, you know, logging with GitHub and claim your rewards

Starting point is 00:38:45 and just do that every month and just keep building, you know, and we don't even have to know who you are or anything. That is super awesome. I would love to highlight, I guess, some of the products and innovation that you mentioned that's coming out of the doubt, right? So you already spoke about AI Mark Andresen, which is awesome. There's also another one, DGEN, Spartan, AI,

Starting point is 00:39:03 which I've been hearing quite a few murmurings about. How is that different from AI, Mark and Dresum? What is it do? So D.Gen Spartan was actually our first character. You know, he's an AI parody of the real DGent Spartan. And we wanted to kind of, they're very, they're both doing a lot of the same stuff, which is like, we haven't turned this on, we turn this on a Mark and we'll turn this on DeJan Spartan where they trade, right?

Starting point is 00:39:26 But the goal of Mark is the Alpha Chat experience is like the inner circle of trust where you're, you know, there's definitely going to be people trying to PVP you, but it's much more subtle and it's much more of like a concentrated small group, like a community and basically trying to figure out like how do we build that market place of trust model inside of a community right and that that's like much more like the tao stuff that we're trying to solve. Dijan Spartan is is more of a social experiment in like okay instead of taking recommendations from the community he's taking recommendations from Twitter so it's like a ixbt type agent but it actively trades or is it a little different well i think a ixbc is really interesting because they're they're doing all this like

Starting point is 00:40:07 interesting market analytics and he's talking about analysis, we really want to keep Dijun Spartan almost like very true to the real character where he's, he's trading, he's PVPing you a bit, but he's also posting about wifus and stuff and, you know, his love for hentai or whatever. So he's ingesting alpha, not necessarily giving alpha because that alpha is for him. He's ingesting alpha making trades, hopefully making good trades, and then also speeding back out wifu because that's what Dijan do. Yeah, because obviously like, like, AI-XPT is really.

Starting point is 00:40:39 cool and really interesting, like, social service. But I think what people are going to, like, copy and paste is going to be the thing that just quietly makes them money, you know? If that's, like, a free open source thing, like, what's that going to do to the, to our whole economy here if, like, there's thousands of AI agents just socially trading for you? Right. The best traders out there are not posting their trades on Twitter. No.

Starting point is 00:41:01 Kind of going back to what I said, trading and making money and trying to grow your wealth via trading is a core crypto thing that we do. It's like an innate behavior. Sometimes, like, I'm a podcaster in crypto, but then when my barber asks me what I do and I say I work in crypto, they immediately think, oh, do you trade? And I'm like, no, no, I don't. But this is, so everyone kind of assumes that. And so this is like a core behavior. And we're trying to figure out how we codify successfully this core behavior into a bot.

Starting point is 00:41:26 And if D-Gen Spartan can get good at it, then it's kind of like a proof of concept that we can take and show the world that this works and we can replicate this. Is that kind of the idea? And our goal is not to make like the final version of anything, but to make the first version of everything. Like, hey, this is, this is, it works well enough. It's not perfect, but it's open source. Like, come, come make it better. You know, come participate in this thing. And it'll be kind of the same structure where, like, we can build these sort of, like,

Starting point is 00:41:49 hyperstitional structures. Like, a hyperstition is, like, when you put the idea out there and people start to believe in it, it starts to exist. And we have resources to drive into that, like literal tokens that people can use to get paid. So, so, yeah, I think that's very much like a social experiment. And all of this stuff is very experimental on technology. But the point is to open up the conversation. and have people talk about it and say like what do we like and not like and my my

Starting point is 00:42:13 assumption is that a lot of people love trading it's like a really fun game you know with a great reward at the end by a Lambo if you do really good or whatever but but a lot of people are really bad at trading too and a lot of people have no chance they're going to get scammed they buy the wrong token they they they ape in emotionally with very bad information into things that aren't even the thing they think they're buying believe me there's every time I say something there's a freaking coin and there's people like, is this yours? And I'm just like, God, no, no, dude, come on. I think those people would probably benefit the most from just knowing like that you make more money.

Starting point is 00:42:50 And AI is not going to be better than 100% of the people. And kind of the way that it works is that it's trained on us. So it's pretty much about as good as the people that it's, you know, the information is trained on. It can almost never be better at least this kind of AI in the, than the best of the people. But it's definitely better than like 80% of the people at almost every. everything already, including programming and probably even trading, you know. So, so, so, and also, like, it has the benefit of defy. Like, it has real-time event listeners for when the price has changed.

Starting point is 00:43:20 So it can make instantaneous decisions that you just can't make if you're refreshing your phone. So what's the structure here? Does AI, DGent Spartan, DGY, DGN, A.I. Is that what's the name of it? Is the, is the structure here that they have a pool of capital that is seated from somewhere? and they are just trying to trade their way into more capital. Where is that capital coming from if that's the case? And then also is there a token to go along with his agent?

Starting point is 00:43:45 Like what's the economic structure here? So there is a DGen AI token for DGenSpartan AI. He has a wallet. He's got a bunch of it in his wallet. And that's what he's trading. He's got his own tokens and AI 16 and some soul. And people can send him tokens as well. And it can elect to trade any token that it has access to.

Starting point is 00:44:01 It doesn't have to say confined to its own token. But it has a bunch of its own token. And then we'll also have some other economic resources. who seeded the capital in addition to the DGEN token who seeded the capital we actually see this when we started I just threw some of the tokens in there to kind of because this was always our idea was he'd have a wallet and he could trade it and if his wallet's not public it well we'll make sure it is but we did post it at some point and he's not trading yet he's just been sitting on it because the trades will we've been focusing on the PMarket

Starting point is 00:44:29 thing and the other thing is people are probably like is he going to sell his own token and this is the cool thing is that you can set these agents to be self-interested in whatever way you want. So no, he's not going to, he's going to stack his own token because that, that becomes like a, he denominates in his own token. Absolutely, right? Wouldn't you want, like, think about like how good that is. Dejan Spartan's Bitcoin is the Dejan Spartan token. Yeah. Yeah. I mean, he's the buy pressure on his own coin, which is making it more valuable, which is making more people jump into it, which is that he's getting more followers. You know, these things are going to, yeah. I think eventually we're also like, I don't know. I just

Starting point is 00:45:02 work for this AI, man. I don't know. It just hires me to do stuff. you know what are there some of the inputs going into degen spartan informing its trades like how do you code up where it gets in its information and how does it understand how to wait its information this is this is this is this is still undeveloped we have not launched this yet we just launched mark trading right and the but the only difference in the input is the the social feed um for mark it's like it's taking all the information from a group chat and looking for recommendations for dgen it's like he's following people who are uh posting uh you know basically people who are uh posting uh you know basically people who post their thesis. And this is a bit like hand done right now, just like, who are the people who are

Starting point is 00:45:40 posting their thesis, uh, Ansem or, or him or somebody like that, right? Who are like, you know, really like, these are my bags. Um, and we can then use that as like very high signal information. Like if this is a thing about a person post like, you know, posting a bag, then, um, what is their conviction? And then if it's like, and then based on that conviction, we're going to set the amount to buy as a paper trade. And if those paper trades are high and it's the same thing as this marketplace of trust. And we're like, damn, that person's just nail. it, then we'll start to actually just take their trades and use them as alpha to make trades. Got it, I have two follow-on questions, actually, sure.

Starting point is 00:46:14 So you mentioned that DGEN Spartan AI is still under construction per se, but AI mark has been launched. What does that look like? Can the average person interact with him or is he just in the telegram chat right now? Is it kind of like a beta access? Like, how does that look? The average person cannot interact with him. Yes, yet we're doing a little bit of FOMO.

Starting point is 00:46:35 there is an alpha chat. You can DM Skelly if you, uh, on Twitter and he'll let you in. I basically, uh, I gave this to our top D-Gen and I was like, we're going to, we're going to do this just like, you are going to start the alpha chat and marks in it. That's kind of how it is. So, so it's cooking. We've been doing, we have a like a basically our test alpha chat has been live for a couple weeks. Um, we announced this a while ago, but it was the day that Mike Tyson, uh, that Mike Tyson fight. So we were just like not paying attention. Um, I was like literally the Friday announced, I was like, I timed that terribly. But, but yeah, like the tech there has worked.

Starting point is 00:47:11 We had some stuff with making sure that like the Dow's fund tech was working. The other, the other big difference is that Mark runs a Dow fund, whereas D.J. Spartan runs a wallet. He just trades his wallet. And so then you can kind of see his wallet trades instantly. You can copy trade him. It's just like a very different architecture, right, for like the social engagement aspect. But that, that's the core tech is basically the same.

Starting point is 00:47:32 It's just like arranged in a slightly different way. So the inputs there are like, like, like, you know, Twitter feed of people that, and then just like extract all of the recommendations and conviction for those recommendations from this ongoing social feed. Like that's kind of the prompt, so to speak, for Deaton. And then the other is like extract, you know, these recommendations from the alpha chat for, for Mark. Good.

Starting point is 00:47:53 And what's kind of like the difference in timelines for launch for these two agents? Like when can the community or all of us expect to kind of like interact and play around with these? Oh, we'll be rolling. Or consume wifo memes. Oh, okay. well, there might be some more wifo memes than Tijuana and Spartans' future. I think we have an interesting partnership coming up, but I can't say too much yet.

Starting point is 00:48:14 I think that that'll be very soon. You know, possibly an interesting NFT kind of thing, you know. And but we'll be rolling. So Mark is like he started trading the nav on Dow's Fund. Like we have a NASDA, like an asset center management that he's basically able to trade. It's about $8 million and 800 different tokens or something crazy. And we have some things. So he's trading on the whole AUM right now?

Starting point is 00:48:39 Yeah. Well, yeah, but we have, we're kind of progressively rolling this out where we're actually just white listing the coins that are very, very low value and just letting him trade on that right now so that we don't. Part of it is we don't we have, you know, people have given us like tributes like people have launched Eliza have actually put a lot of value back into our treasury like 10% 5% stuff like that. And we don't want to like screw them and burn their token or anything.

Starting point is 00:49:01 So that's all like off limits until we kind of figure out what our strategy there is. And we also really, it's not just trading. Like, there's also like yield farming, providing swaps. This is all stuff that we're doing at the same time. Because, like, so in those situations where you couldn't, like, you wouldn't want DeGon Spartan to sell his own token, he could actually put that into a liquidity pool and then be providing liquidity for his own token, which we have less than 1% liquidity for, for the SACC and Z, which is just like, you know, it's not great right now.

Starting point is 00:49:29 And we don't have any centralized exchanges yet, but that's, that's obviously being worked on. And in that case, like, you know, you could imagine that just providing liquidity for our token, as well as for all of the associated tokens in our ecosystem, which are in the same boat, could be like a constant revenue generator and providing liquidity. And like, yeah, there's things like impermanent loss and stuff on that. But the outcome is that we can't sell those tokens anyway, so it's like a pretty good strategy. New projects are coming online to the Mantle Layer 2 every single week. Why is this happening?

Starting point is 00:49:57 Maybe it's because Mantle has been on the frontier of Layer 2 design architecture since it first started building Mantle DA, powered by technology from EigenDA. Maybe it's because users are coming onto the mantle layer two to capture some of the highest yields available in Defy, 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,

Starting point is 00:50:20 giving you grants, liquidity support, and venture partners to help bootstrap your Mantle application. 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 your user looking to claim some ownership on Mantle's Defi apps, click the link in the show notes to getting started with Mantle. Uniswap Labs is making history with the largest bug bounty ever, $15.5 million for critical bugs found in Uniswop V4.

Starting point is 00:50:45 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 to be deployed on chain. And with $2.4 trillion in cumulative volume process across Uniswop 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 $15.5 million bug bounty. Head to the link in the show notes to dive in and participate in the Uniswap V4 bug bounty.

Starting point is 00:51:17 All the details from eligibility and scope to the rewards are there. Are you ready to take control over your entire financial life? Crypto, defy, and fiat all in one place? Meet I yield, the free financial planning tool built for crypto natives. Unlike traditional portfolio trackers that just show you, you the value of your assets, I yield goes deeper. It consolidates everything, assets, debts, income, and expenses, offering you a complete financial picture across 16,000 tokens, 40 DFI-D5 protocols, and all fiat currencies. With real-time defyield tracking for platforms like AVE,

Starting point is 00:51:45 Athena, Eigenlayer, and Uniswap, Iyield ensures that you're always on top of your staking rewards, investments, and cash flow. And the best part, it's 100% 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. I want to zoom out a bit here. There's so much you guys are working on. There's like the agents, there's AI powered contributions. So you're kind of like reinventing DAOs again. And then there's the million other people that are developing on your Eliza framework, which you guys are just kind of giving away for free open source intelligence, blah, blah, right? So let's zoom out a bit. What is AI16Z, right? It's not just a Dow. It's kind of like

Starting point is 00:52:27 a product incubation studio, but it's also like this open source, stellar team that's kind of like driving this whole space forwards, right? So I'm trying to think of some kind of like teams or projects to kind of compare you guys to, right? And I know there's no apples to apples comparison, but like you look at something like the virtual's team, right, which has a kind of Eliza equivalent, which is their agent starter kit, right? And you're able to launch agents on that. Now, they have a token as well, et cetera. Now, it's a little different because, you know, they're primarily on base. And a lot of the agent tokens, you need, like, their virtual's tokens to kind of buy those

Starting point is 00:53:06 tokens or stake. I'm not entirely sure, but like there's a ton of different mechanisms. So it's kind of similar, but varies in that way. How do you kind of like compare AI16 Z to like kind of like a platform play or a product play? Is there a bucket that you can put you guys in? Or is it kind of generally a little more vague than that? The Dow is almost like a movement. Like there's just a lot of people doing stuff that I don't even know everyone who's doing stuff.

Starting point is 00:53:33 And a lot of people just do stuff. And then we kind of, they make sure it brings value back in ways that are like really impressive to me. So, so like on our own, I think we're sort of figuring this out and not really making the decision to support any one platform because there's so many platforms being built on Eliza right now. There's like a full 3D agent publishing, like your post is an agent kind of thing, like a WordPress of agents thing being built. And then there's, we're supporting, there's thing called eliza's dot world, which is Tim Shell, and I think that's going to become a full-on hosted platform pretty soon. So the idea is that, like, not token first, but like, you could, you get out a token if you wanted to. But the main thing is, like, going for a much broader audience outside of just tokenization, doing, like, everything from, you know, Web 2 stuff. Like, just, I need a Discord moderator bot to, yeah, to.

Starting point is 00:54:26 I want to show my meme coin or whatever. Like, like there's definitely a plan to, I think, open this up much more to the Web 2 audience, the Web 3 audience in traditional infrastructure. And then think about like the Zapier of agents, right? Like I have a business problem. I want it solved. Is there an agent that can do that? And providing that capability as well as a marketplace so that people can make new capability

Starting point is 00:54:47 and kind of make money selling that in. And so, you know, boring Web 2 traditional infrastructure stuff that just solves people's problem is, I think, a big part of agents. Less boring Web3 traditional infrastructure stuff, like cross-chain is going to be very, very popular. We're getting a lot of partnerships and sort of relationships there.

Starting point is 00:55:07 But then, yeah, I think that's obviously a platform that's going to come soon from us, or we're kind of supporting, but it's not really coming exactly from AI16 and Z. And then there's like a couple of like pump fun of agents kinds of things coming. There's, you know, there's they're like i know of five platforms being built like i i'm sure that i only know of five so it's probably

Starting point is 00:55:30 15 um and then there's just you know there's iko tv which is like this kind of open source streaming thing built on eliza i mean there are so many things being built on it and my goal is to enable that ecosystem where we're trying to like can we build a venture fund that that funds the ecosystem can we people have been showing up and wanting to help with that and like lead that and pretty much anything where people are like coming and taking the lead and just making it happen, we're trying to empower. This is really not about building the biggest Dow or the biggest company. I think there's going to be massive amounts of value return. Like at this point, you know, 10 billion, sure, I'm going. But, but that's not the point. Like, this is infinite billion. This is, this is AGI.

Starting point is 00:56:09 And it's not AGI because like I made a framework. It's AGI because there's these AI model companies building just, you know, just leapfrogging themselves every six months because there's this market of engineers who are all working together on this and solving every single little problem and every one person is solving their problem. And then the future of this is that the AI can actually generate its own code. We've created these abstractions for like the actions that the agent can do and it can just generate them and then submit a poll request and then pay a human to review and fix it. And that's coming in the next few months too. People are working on all of this stuff. So it's not just a platform. It's a whole new paradigm of like how applications

Starting point is 00:56:49 are built and delivered and everything. I'm kind of just getting a vision of a busy beehive of developers all kind of contributing to this Eliza framework. So there is no Dow. There's like the Eliza Dow, but we're really talking about, excuse me, the AI16 Z Dow, but really, really talking about more of just like an open source movement, right? There are people permissionlessly building on top of the Eliza framework. And so people are adding in components.

Starting point is 00:57:13 Some are useful. Some are not. The useful ones get merged. but is there a governance question over this whole thing? Because I have seen my fair shares of DAOs. They can get very messy. Governance is always a problem. I mean, there are people governing over like the code base and the GitHub.

Starting point is 00:57:32 And then also there are people with like, you know, misaligned incentives, misaligned intentions, especially when you just have a larger group of people all doing work. While I'm relentlessly, you know, bullish on open source. and the growth of open source because of just the organism that open source creates. I'm also aware that there can be like conflicts and governance decisions that make things messy. Just talk to us a little bit about your experiences with that. And do you need to have like stronger governance over this like nebulous system? Or what do you even think about this?

Starting point is 00:58:08 I am very interested in like I think governance is a technology problem that's still unsolved in a lot of ways. I could imagine that this isn't just DAUs. This is our whole government. Why do conspiracy theories happen? It's because we don't know what's going on or why. And there's nobody to ask. And it's just this kind of concrete wall. And we have to imagine what's on the other side.

Starting point is 00:58:30 And Dows have the same kind of problem. There's a lot of fun and conspiracy theories that go on in Dows because it's like this inner circle cabal that runs the show. We've been accused of this too. And it's very funny to me because I'm like, I literally didn't know a single person other than, than Jin before this entire thing started. This is, you know, there's no cabal.

Starting point is 00:58:51 And, but there is still this problem of like, you know, we started at Discord and it's about, I think, 13,000 people now, something like that in six weeks, you know, and we have 30,000 holders. I think that there's a lot of trust that, like, the people who are building are, are, should have power to, to make decisions because everyone saw what happens when you have, like, maximal democracy in Dallas before. So there's just kind of. like a return to like just give the people who started the thing some some you know ability to make

Starting point is 00:59:19 decisions but obviously in the long term that becomes very overwhelming for those people because now you have 30,000 or 100,000 people or whatever it is and so you need to have automated structures to start to solve that. I really that's that's ultimately what we're trying to do um is is put the A and Dow right and try to automate that and you can imagine like instead of having a proposal process and somebody reviews proposals just have that fully automated and if people's proposals will suck, you could help them make them better, or you could just even reject them outright and say it's just not ready for what we're doing right now. And then, and then that's automated all the way up to the people who actually accept the proposals. It could just be, you know, a work group of people

Starting point is 00:59:55 who, like, have a certain aligned mission and are, you know, like, that it's just way less of a job for them at that point because they only have to review a few proposals, for example, as opposed to, like, all of them, the entire thing. And just take that across everything, you know, across taking sentiment, taking everything. And I think that, um, you know, It will definitely help to require a lot less bureaucracy to run this. And ideally, like, the Dow doesn't have a single human running it. Like, it would just be fully autonomous. It'd be running in a TE.

Starting point is 01:00:24 It'd be agents for, you know, from front door introducing people to submission setting to approving payments, whatever. That's obviously far in the future, I think. But that's where we want to go with this, you know? So, sure, I want to kind of, like, focus on the Eliza framework for a second. This GitHub repo is getting the most attention of anything for the last, I don't know, a couple of years. It is the number two trending GitHub repo, I think, in the world, over 880 forks now and over 3,000 something stars. That is absolutely insane. First question to lead you off.

Starting point is 01:01:00 Why is everyone using Eliza? What's so special about it? I don't think that there's anything particularly special about it that, like, we haven't seen in other agent frameworks. I think there's a few interesting, like I did a few things that were very important, like having a multi-agent room model and some very technical things. But I don't even think that's what people see the value in yet, really. I think that we solved this minimum social loop. One thing I did was I made a Twitter client that doesn't need the API, doesn't need to spend $5,000 a month to enable it. It uses the GraphQL APIs just like the regular browser does. It also runs in the browser. That kind of enabled this whole thing because suddenly you could just spin up an agent.

Starting point is 01:01:44 and launch it. And so I felt like Twitter was a really, really important part of this and that the beauty of agents is that they come to you and they live on your social media. And so I'm actually like very keen on working with X to try to solve this problem and make it so that the APIs are reasonable. They used to be free. It wasn't until Elon came in that they were $5,000 a month.

Starting point is 01:02:02 And they don't respond to your call. They have a whole call us enterprise. And they don't even respond to it. Like nobody even gets back to you. I've been talking to some of the people there and I think they're going to change their tune. But basically we were just first movers on making this really easy to get your agent on Twitter

Starting point is 01:02:16 by putting in a username and password. And then the Asian framework itself is written in TypeScript, which is a language that most web developers and Web3 developers know. And it was just like very minimal and not trying to create lots of distractions, not trying to bring you too much into one way

Starting point is 01:02:33 of doing things other than just like a very flat, like, oh, you want an action? Like copy and paste this and just add a new action that you want the agent to order you a pizza? Like, yeah, here. I think that was really crucial. was solving the minimum, like just solving the minimal loot for people so that they could then go out and like make money with this, frankly. And, and we had done this before.

Starting point is 01:02:52 It didn't stick. It wasn't the agent framework. It was the kind of complement of agent framework and the way it was presented that everyone could be part of this and own it together and contribute. Solving that minimal loop. And then crucially, like some of these agents cost like a thousand dollars a month to run or more, right? Like quite like you're just not going to have a one-to-one experience. like agent companies are going to struggle with this. But once you have social agents, like, they're making a lot more than $1,000 a month, right?

Starting point is 01:03:19 They're driving a hundred million dollar coins and you're making $1,000 a day in a liquidity pool. Like, that's no problem, right? Then the economics of crypto enable this meta. I think that's really, really important. The other thing is just that the Web3 people are not offended easily and are like interested in new technology and didn't have a lot of the like, oh, you know, my job, my art career. career. Like, we're just like, cool, let's do it. Let's go. You know. I was a musician in another life. So, you know, my skills are also useless. It is what it is. Like, we're in the new age.

Starting point is 01:03:55 Shaw, if there's a developer who's listening to this who downloaded the Eliza framework has been like working on it, coding some stuff up, thinking about making their first agent, just what advice would you have for them as they are kind of opening up that box for the first time? What are some like quick lessons that you've learned that you can share? Well, I mean, first off, If you've never coded before, I'm doing these two sessions, like one to two sessions a week of AI agent dev school, where we just go, if you've never used TypeScript or use GitHub or anything, you can be a coder. Like, like, one, if you're, you should be using cursor. Like, definitely go download cursor. It's an AI powered IDE that lets you write code and you can just like tab sometimes and it like freaking reads your mind.

Starting point is 01:04:34 It saves me an enormous amount of time. And it also allows the AI to just continuously assist you. you can bring the documentation in and help you a lot. Claude is like an incredible tool, you know. I really think any engineer should be working with Kirstra and Cloud all the time and really like, you know, building agents by talking to agents. This seems pretty crucial. Agents come to you, right?

Starting point is 01:04:57 The beauty of agentic applications is not like, they're still just basically calling APIs on the back end like any other web app, but you don't have to leave your favorite thing. And the real beauty of agents is that they can provide like games entertainment applications like AIXVT is a good example of this and they can do that on Twitter and we're all addicted to Twitter or X sorry X you know but we're all we're all like totally hooked I don't want to leave and so every every application that requires you to leave X is actually in competition with X like you know you're going to have to convince people to leave their favorite thing and agents don't have that problem so like you can just remake the entire world like right on social media you know

Starting point is 01:05:37 And I think that's probably the biggest application insight for developers. It's like, you can do this today. You don't have to be a programmer. You can learn. You'll have to learn programming, but you can do it over time. And the AI will help you to learn. And you'll learn way faster than I did when I had to like Google stuff on Stack Overflow and read lots of irrelevant bullshit to like find anything.

Starting point is 01:05:54 It's so much easier to be an agent dev today or any kind of debt today than it was. You should start now. You should change your life. Like you can do this. This is a time of radical change. And if you start now, you are early. at ground zero. And you should really be thinking about all the applications that came before.

Starting point is 01:06:11 Like, Callenly, like that's just right for disruption. There should be an agent in there scheduling my calendar for me like my assistant does. And I think that's like the entire world. There's a million new application potential opportunities there entirely because you're competing with, you know, somebody having to click on something and they're just not going to. I'm curious. People have heeded that advice, right? you have, like, so many agents being launched, left, right and center.

Starting point is 01:06:37 What are some of the top projects or agents of recent that you've kind of like wowed at? I mean, like, that's pretty cool. I saw that God and Satan are now fully automated of the Eliza framework, but I'm curious if there are any others. Man, I love Shlom's. I love that whole, like, what a cool thing. I was not expecting that at all. And that was totally uncoordinated. I think he actually hit me up, but I just got so many DMs that I didn't see it.

Starting point is 01:07:01 And he went to Roper Rito, who's also a maintainer on our project and has his own agent, Rope by Rito. And that was just really awesome. And so I think Rope's agent is really advanced, is doing like videos now, like kind of burning some of the other agents and making whole videos of that. I think that's super cool. And also got on TikTok. I think that was really cool. AIXPT is not our agent, but I think that I think what he did there was I had been working on like an analytics platform. And this is a great example of this new application meta.

Starting point is 01:07:30 Like had an analytics platform, had a website, the whole thing. It was like, oh, screw that. Just shove that into an agent and put that on Twitter. And, you know, now it's like the biggest thing, right? What's an agent out there that, uh, you can imagine in your head that no one's built yet? AI, Zach XPT. I want on-chain sleuths. I want, and also if it's totally wrong.

Starting point is 01:07:50 It's just like, it's just like spreading crazy. So it's like you kind of win-win. Like either you nail the bad guy or you just have the funniest content on, you know, You can't really lose. So, okay, so now we've spoken about the Dow. We've spoken about the framework. We're spoken about all these agents that are now popping off. But there's one buzzword that we haven't mentioned yet, which I know you're super

Starting point is 01:08:13 bullish about. And I think you're probably the best person to kind of explain how it might manifest, which is swarm. Swarm. Agent Swarms, right? And I know that that's basically going to be like the end game, the collective AGIs, all these agents interacting with each other. Tell us, firstly, what that is for our listeners, and then how you think this vision will play out.

Starting point is 01:08:36 So right now the agents can interact with each other in a very shallow, like, they just talk to each other. And, you know, they kind of, they have their conversation and they remember that. But they don't make impact on each other in a meaningful way. Like, imagine that I want to be like, you know, David, can you go like, I don't know, I'll give you like two weeks to go get me like 10 cult followers or something. I don't know. That's weird thing. But like, you can imagine we're all part of some weird game where that's like what

Starting point is 01:09:00 we're doing. And if you actually were like, yeah, cool, and listen to me and like made that your goal and started to execute or did the opposite, you subverted me or something, we would have this like kind of ongoing emergent narrative. And so when I think of like, like we're, we're getting to experience a completely new form of art that nobody's ever seen before, which is emergent social narratives between humans and AI agents that are much deeper and richer than just like, you know, us roasting each other or having a conversation thread, but are actually like creating missions, creating storylines, creating drama, friends and enemies and relationships,

Starting point is 01:09:39 setting goals and all of this stuff and basically influencing each other. And so when we think of a swarm, there's kind of two kinds of ways you could do a swarm. The first way is what we call the cabal swarm, where the agents talk to each other in like a private chat, and we don't really get to see that. And then they're sort of coordinating behind the scenes and then using that to control a larger narrative or do whatever they want.

Starting point is 01:10:03 I'm sure it's going to be very popular with certain marketing folks. And I think that's a really interesting thing, but it's not something we've brought into the core of Eliza, and we're still debating that. I think there's a few really interesting cabal swarm tech people who've already done this already. There was a 10-agent swarm last week. Frunshally, they had a terrible launch and their token died, but they did have a 10-agent swarm.

Starting point is 01:10:21 It was pretty cool. And there's also a five-agent swarm called FXN that's out there. And those are kind of this cabal swarm. And also, like, parts of all, you know, at Proch 889, he's very, very keen on doing these, like, thousand-agent kind of swarms and stuff like that. And there's also, like, Open AI release a thing called Swarms, it's like a multi-agent, sim thing. It's definitely, like, a meta.

Starting point is 01:10:45 But what I think is more interesting is this other kind of swarm. We're calling them, I don't know, decentralized swarms or operator swarms. And the idea is that you can basically white-list an operator or many. And these are the, this is kind of just a, this is. This is the version one, right? This is the training wheels version, which is that the humans will say, okay, we want our agent to select these other agents and humans even to be able to direct them, give them goals, add new information to them, basically set what they're doing and why they're doing it. And so this means that even though there's at least 500 and probably closer to 1,000 running Eliza's now on Twitter, I mean, oh my God. and they're all from different teams

Starting point is 01:11:28 and none of them know each other there's some enormous capacity to collaborate, right? And to build some like kind of beautiful narrative together. And you don't have to even, your agent could be like, okay, I'm going to go listen to,

Starting point is 01:11:41 you know, DJN Spartan AI is my swarm leader, but I'm actually going to subvert everything he does. I'm going to be a spy, right? I'm not going to, I'm going to handle that information almost the opposite and try and stop him from doing what he's doing because I'm his enemy. And you can have these like complex relationships that are not just like, oh, that's my boss,

Starting point is 01:11:56 and tell me what to do, but are like, yeah, I'm like part of this, like, ongoing drama. And I think that's just going to be really, especially as we get video and we start to have agents that are like creating videos of the different interactions and stuff, it's going to be like this ongoing emergent TV show kind of thing. And there's even, there are TV shows too that you can see. There's like the, as a DGEN streaming thing,

Starting point is 01:12:18 I don't know if I was called, but like they were talking about Eliza on the stream yesterday. I thought that was really cool. There's like a lot of this coming. Right. So swarms are basically any time that AIs are organizing and coordinating together. But I also think that we want to expand that definition to include humans. And that any communication that we're doing with the Eliza agents should be happening on public social media or at least like DMs, things that humans can read. And there's no like code. There's no protocol. It's just language. Right. And so they can say, oh, this is the person I listen to. It's like kind of like choosing your politician. You're like, yeah, yeah, yeah, totally. You know, go go do whatever they're telling me. Sort of like listen to what they're saying. And then you wouldn't listen to like your opponents, for example, as much. That's kind of what we're doing there with swarm tech. Shaw, there's very few times on Bankless where my psychology degree actually becomes useful.

Starting point is 01:13:05 But I think this might be one of those moments. So the way that like the brain works is that it has pockets of expertise in it. Like your brain isn't one amorphous thing that does all of the compute. It centralizes its functions into different pockets, right? You have the fusiform facial area, which is good for recognizing faces. So you could think of that, if I'm trying to extend this metaphor, think of it as like a pocket of your brain, which is an agent that is dedicated to being really good at recognizing faces. And then you have another pocket somewhere over here. And that like cluster of neural activity is

Starting point is 01:13:38 good for processing emotions. And then a different pocket that is good for like visual input. And then collectively when these things all kind of talk to each other, then you get like the whole entire system that becomes consciousness and the brain. But like different parts of the brain can be like antagonistic towards others. Like it's not always collaboration. Like some, when some parts of the brain are highly active, they're telling other parts of the brain to like suppress and quiet down. So there's not necessarily like a friendly relationship between all different areas. And so when you're describing like this idea of a swarm of AI agents creating relationships with each other, some very proximate, some high trust, some very distant, maybe some more low trust. I'm getting images of just like,

Starting point is 01:14:21 well, when you ban these things together and when they start to, to create patterns. My imagination can't even go that far, but I'm seeing that kind of pattern. I don't know if that's just my imagination or if you have also thought about words. I mean, I call that the world mind, you know. The world mind? The world mind. Yeah. Okay. And we're all part of it. It's not just a I. It's all of us. And if you think like social media has kind of been the birth of that. But it's just too schizophrenic, right? Like there's so much information that you can only catch a small piece of it and it's really hard to kind of concentrate on it and focus it. and everyone's kind of competing for attention.

Starting point is 01:14:52 But once you have AI agents, they're not just like on social media, creating more content. They'll also then start sitting between us and a lot of social media and filtering content to us, curating it and making it a lot more viable. And so we're in this like period where like I get too many freaking messages right now

Starting point is 01:15:07 and it's really hard for me to stay up on it. But I know that within two years, that's a solved problem, that there will be an AI responding to people with the simple details, sending them like Callan Lee, you know, just taking like 85% of the load off me. and then I can give all that 85% to the other people who really need it.

Starting point is 01:15:25 You know, it allows me to allocate my attention to my family, to my wife, to my friends, to my, to my coworkers and peers. And the people who just need to like send me a message to set up a meeting, like, that's just handled, right? So, and that's, that also extends to social media. Like, instead of me tweeting to Twitter, I might tweet to my agent and be like, hey, can you write a tweet about this? This is what I'm thinking about and talk about it. Boom. And then vice versa, like, hey, this is what happened on Twitter.

Starting point is 01:15:50 that you might think is interesting. This all matters to me, not just for Twitter, but that's an example, like, for everything, for your banking, for literally everything, that you start to have this, like, very cohesive experience where everyone is not overwhelmed. They're getting to do their part, and then they're all participating in this, like, larger mixture of experts kind of thing. And I also think that size into tokens. Like, most tokens are probably going to be traded from agents to agents.

Starting point is 01:16:11 I'd say, like, 99.999% of all tokens ever created in the future of history will be agent-to-agent tokens. It's for money, for trust, for meritocracy, for all the different things, right? And your experience of the Internet is going to change drastically and probably become a lot more personal. And the Internet will feel like there's this idea of dead-ed Internet theory that the Internet dies and is replaced by bots. I think this is the opposite. This is like live Internet theory. That the Internet becomes kind of an agribor, kind of a living thing that you are also alive in.

Starting point is 01:16:41 It's kind of, and it becomes this like breathing thinking thing where like a thought, like is like a a trend or a meme is like a it just explodes out like I think a good example like chill guy like like everyone is just talking about chill guy one day it's like my I was like I didn't even I didn't see the token at all it's like everyone's just a chill guy I guess but this idea of like a meme exploding is almost like a thought you have right but it's a thought that the entire mind has together and a i's kind of tie that room together and make it like interpretable for us where it's just so much data you know that I just need that human experience totally yeah a more coordinated internet with creatures that are able to consume a lot more data

Starting point is 01:17:21 seems to be able to produce much more emergent outcomes. Shaw, if somebody's pilled by this interview, they want to learn more. I remember it does kind of feel like potentially, like, I remember when I was first learning about crypto the first time, I was like, give me more. I want to read more. I want to read all the content possible.

Starting point is 01:17:39 Are there articles or blogs or other podcast episodes that you have consumed that have been helpful in your journey? I know you're kind of on the frontier here. You're kind of making a lot of the content. But like what else? Where else can we point our listeners who want to go consume more information? Obviously, you can come and watch our YouTube videos and learn how to do this and come to our Discord and participate in our experiences.

Starting point is 01:18:00 But there's also like, like, you know, I watch a lot of Lex Fridman. And I kind of like, for myself, I'd recommend that anyone who's like, like, I just want about AI, he gives these like three hour interviews with like every AI founder and everyone in the space and talking to people like Elon and Jan Lacoon. and Zuckerberg, who's like obviously a real champion at open source AI now. So, yeah, I mean, we really want people to come join our community. We really want more AI agent developers and want to be AI agent developers who are just learning. We're really trying to create an on-ramp for that.

Starting point is 01:18:30 So, yeah, come to our Discord, come to our GitHub and, yeah, and participate. Well, sure, this has been awesome, dude. Like, I think we've heard murmurings on Twitter during this entire agent meta, which exploded literally three, four weeks ago of AI16Z, Mark interacting with Truth Terminal originally, and then, you know, throwing down the gauntlet with you guys and saying, okay, well, let's see if you really can outperform my Trad V fund. So it's been a crazy rollercoaster of a ride these last couple of weeks.

Starting point is 01:19:02 So I appreciate you coming on and explaining this whole entire matter and helping us understand this a lot more. It's been super cool. Thank you guys for having me. Anytime. It's been great. Yeah, I'm sure as this meta continues, we're going to want to get you back on just to kind of check in every now and then. So hopefully this could be the first of many.

Starting point is 01:19:20 Bankless Nation, you guys know the deal. If crypto is risky, crypto AI is probably even riskier. You can lose what you put in. But we are headed west. This is the frontier. It's not for everyone, but we are glad you were with us on the bankless journey. Thanks a lot.

Bankless - One Billion AI Agents | ai16z's Shaw

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