Unchained - With AI Agents Now Trading Crypto, What Does Their Future Look Like? - Ep. 758

Episode Date: January 3, 2025

In this episode, Shaw Walters, founder of Eliza Labs, takes us through his journey of launching ai16z, a decentralized AI-driven investment DAO, and the explosive growth of the Eliza open-source framework. Shaw explains how AI agents are evolving beyond "reply bots" to serve as powerful tools for financial autonomy, community investment, and user empowerment. He shares the challenges of building trust between humans and AI, the possibilities of agents trading across multiple blockchains, and his vision of a world where financial freedom drives human progress. Show highlights: Why Shaw decided to launch ai16z What he believes AI agents are useful for  What the Eliza framework is and the role it plays being an open-source project Whether AI agents can be good at trading Why the Eliza framework works on every chain, even though there’s a focus on Solana How and why the AI version of Marc Andreessen was launched Whether ai16z will launch its own layer 1 and the details on the future launchpad How agents on social media could be really useful instead of being “reply bots” What Shaw thinks overall of the AI space and how it’ll impact human society Visit our website for breaking news, analysis, op-eds, articles to learn about crypto, and much more: unchainedcrypto.com Thank you to our sponsors! Polkadot Kelp DAO Guest Shaw Walters, Founder of Eliza Labs Links Unchained:  The Backstory of How 3 AI Agents Led to the Rise of the Hottest Memecoin, GOAT AI Meme Tokens Like FARTCOIN Dominate Crypto Markets, Suggesting Powerful 2025 Trend The Block: AI agent platform ai16z considers tokenomics overhaul and launching Layer 1 blockchain Chain of Thought: Ai16z: The Bazaar of Agents Cygaar: How the Eliza Framework works Timestamps: 00:00 Intro 03:01 Shaw’s inspiration for launching ai16z 08:45 How AI agents are evolving beyond simple tasks 10:28 The Eliza framework and its open-source impact 13:22 Can AI agents excel at trading? 21:19 Why Eliza works seamlessly across multiple blockchains 24:08 The story behind launching Marc Andreessen as an AI character 35:18 Plans for ai16z’s layer 1 and its future launchpad 40:23 Transforming AI agents into more than just “reply bots” 43:26 Shaw’s vision for AI’s role in shaping human society 49:26 Crypto News Recap Learn more about your ad choices. Visit megaphone.fm/adchoices

So, like, you know, if I'm going to be part of the acceleration of that world where people don't have jobs, I also want to be part of the solution that enables people to have enough and to have time to the things that they really want to be doing. I think people could be spending a lot more time with their friends. They could be spending time having families. They could be taking care of their parents, taking care of their children. They could be pursuing passions and art and things that aren't profitable. They could be spending their time going to space or building really hard things. I mean, I can just think of a thousand things that people could be doing.

Pursuing spirituality and like, why am I really here? What is this? And what is the purpose of all of this? I just don't think that gets unlocked until we are financially free. I think it all starts with financial freedom. And then every other freedom kind of comes from that. Hi, everyone. Welcome to Unchained. You're no high resource for all things, crypto.

I'm your host, Laura Shin. We are now featuring comments from ministers on the show. In the last major episode before the, the New Year, Solana Foundation president, Lily Lou, caused a firestorm on X when she said that she didn't think that BASE could compete with Solana. She called Bace, quote, a corporate L2, and said that it was not only parasitic to Ethereum, but actually cannibalistic. Well, the responses were pretty overwhelming. The comments range from short clips like this one by Jenny from the block. Wait, Solana is decentralized, to much longer rebuttals like this one from Audrey O'Ferrier.

In it, he noted Ethereum's dominance as the most decentralized and robust, layer 1. The ability for centralized and decentralized layer 2 is to operate under near-zero trust assumptions. Superior Ux, for example, Coinbase's transaction fee subsidy on base, the future doubling of logs on Ethereum that he expects to increase transactions per second in what he calls unlimited scaling options. He concludes, quote, big players that are moving thousands slash millions of dollars per transaction will always prefer Ethereum's L1 over L2's and alternative L1s. To have your comment featured, write a review of the podcast on Apple Podcasts or Spotify,

or leave a comment on our video for this episode on YouTube or X. This is the January 3rd, 2025 episode of Unchained. Earn a 20-plus percent APR on ETH with high gain by Kelp, leveraging blue-chip-d-fi strategies and professional risk management. Deposit now at www. kelpdow.xy-Z. Pocod is the original and leading layer-zero blockchain with a for two plus developers, and the Pocodot 2.0 upgrade will be a massive accelerator for the ecosystem. Join the community at Pogo dot.network slash ecosystem slash community.

Today's guest is Shaw Walters, founder of Eliza Labs. Welcome, Shaw. Thanks for having me. Just a quick note before we begin, everyone, I moved, and perhaps my mic got dinged up or something, I don't know, but I am having to use my computer mic because I just discovered this right before I went to record with Shaw. So apologies for the not perfect sound. So Shaw, you have had a wild few months. Tell us how you came to launch AI16Z. Well, yeah, it's been very wild. I, for the last, like, I don't know, since COVID, I've basically just been programming as fast and hard as I could towards something that would stick against the wall and, you know, and get some traction. I've been really excited about AI agents for a long time, and I've been working in that space for a long time.

And long story short here, when this kind of AI agent meta had taken off with truth terminal and all that stuff, I'd already had an agent on Twitter, just didn't have any traction. And it kind of needed that missing piece, which I felt like was the sort of crypto, sort of, you know, the Dgen hype cycle, so to speak, and the meme coins and all of that. And so I, long story short, met the hedge fund manager of the first fund on Dow's dot fund. and we started tweeting against each other, and he was like, I wish Dijan Spartan was around. Like, I have the technology. We can do this.

I did working on this for so long. And so we started kind of interacting, and we launched Dijan Spartan together, and that kind of went well. He introduced me to Bowski, who created DadaSnot Fun, and I went and met him in person, and we had lunch.

He actually just lives down the street from me in San Francisco, and I told him about my idea, desire to have an autonomous investor and to just have like a way that a community could invest together without having to worry about being rugged. And I also believe this idea of growing the pie that it's not about trading necessarily, although trading is part of it. Speculation is obviously part of it. But it's really about investing in things that create value out of nothing. Like kind of the way that, for example, Andresen Horowitz does and other investors, you know, VCs go and identify value generation opportunities like new startups and they put money into them.

And then they become very big and add value to our world. And I really wanted anyone to be able to have access to that kind of opportunity. And so I said, you know, I won AI A16Z. He said, just make it AI16C. And I laughed very hard. I thought that was the funniest thing I'd ever heard. I was like, okay, we have to do that. Let's just go.

Let's go. And we went home that night. And we just like traded back some images. I generated some images with flux until we had this, like, ridiculous wifu. It's like, oh, my God. I mean, it's so canceled for this. But like, yeah, let's go. Let's do it.

And we launched, and we challenged Mark and Dresen that we were like, yeah, this autonomous investor is going to flip A16D. And he was like, gotlet thrown, you know, and kind of accepted our challenge. And then it went from there. And my goal was really just to raise, like, 75K in Seoul to, you know, USD, not 75K,000. 75K USB to have this autonomous investor experiment. And in the meantime, I think a lot of people saw the open source framework that I've been working on for a really long time, what was powering Dijun Spartan, which is called Eliza.

And people started taking it and running with it in many different directions, building other coins, but also building a lot of really interesting use cases with it. And it's kind of taking on a life of its own. And just, you know, we're continuing to work on everything we said we'd work on from the beginning and we're working on like research papers on how um AI can use like market intelligence and collective intelligence to make good investment decisions um but also like our ecosystem has just completely exploded and and is building like every imaginable thing you could do with agents now and and yeah that's kind of that's kind of just how it happened I really um I had a sense obviously I was

building this tech because I thought it would be valuable um but I did not think any of this would go this fast, you know. And then I noticed that a lot of our community was over in Asia and Asia Pacific and people were reaching out to me. And I was having a really hard time kind of crossing that language barrier as I think a lot of devs in my position have had the same kind of issues. And so I just really wanted to come over and meet people. So now I'm actually in Korea. We're going to have an event tomorrow. I had events in Shanghai and Beijing and Hong Kong before this. And we're going to Tokyo in a few days as well. Yeah, I'm sure. Honestly, I feel like I've seen this play out in crypto so many times where things take off so fast.

So, you know, I have to say, like, learning about AI16C, it actually reminds me of the Dow from 2016. Like, were you around back then? Or like, do you remember that? Yeah, I mean, I wasn't super engaged in crypto, but I always thought that that was really interesting. And I followed the kind of story of the rise and the demise of it and how that kind of all went down. And I think that, you know, what I'm not. I'm really passionate about is like how do we solve the problems of DOWs generally, which seem to be, well, I mean, there's like some things which are just like the practical externalities

of, you know, rugging and getting hacks and all that sort of stuff. But then also just like the inefficiencies of running these kinds of autonomous organizations. So yeah, from the beginning and from kind of the Talix description of what these things could be, this has been like very much the part of it that I'm most interested in. It's like how do we decentralize governance and make that un-efficiently. And I think that agents are a really big part of that. Yeah. What's fascinating to me is because that was almost nine years ago, basically they did this fundraising and they got so much money. And then because like none of this stuff had been built, they like had the money, but they're, they had none of the tools to actually execute their vision. So it's just so fascinating to me that

now finally like all the stuff has been built where you actually can have something that was very similar to their original vision. So one thing that I got. curious about was, you know, everybody's calling these AI agents. And obviously, it's not like I don't understand kind of the different sort of user experience of using an AI agent than, say, for instance, a chatbot. But can you just talk about, like, like, what is this? You know, what is an AI agent generally as opposed to sort of previous versions? Yeah, it's obviously a very big term that describes a few different things. I think people think of like autonomous agents is one part, which is agents that can kind of make their own decisions and can they basically run in a loop and they can

be sort of a proxy for like artificial life and and their own being with their own memory and something akin to like a kind of modeled sentience I suppose but I actually think that the that's that's like more far future stuff I think the near future stuff is really about about the internet and about how we interact with with the internet and with applications and stuff you know right now you like go to a website and you click through the website And I think that what's really changing is that we're, you know, with these LLMs, which are really good at translating, like, unstructured human language into, like, API calls or, like, you know, the smart contract actions and things like that, is that we're able to bring a lot of the functionality that we see on websites onto social media, into our DMs, onto our phone. And it's just like a different interaction paradigm fundamentally. And it doesn't have to be some crazy smart thing.

It just has to really, like, solve the problem of friction, of bringing applications to where users already are. And making it so users don't really have to change their behavior to get functionality and, like, added benefit from these things. And you are also developing or have developed the Eliza framework. Can you explain what that is? The Eliza Framework is an open source project. It's on GitHub.

Anybody can use it. We're not holding anything back. It's totally free. and if LLMs are kind of the brain, I think that Eliza and frameworks like it are really the body. And it has connectors to all of the different social media platforms. It has connectors to all of the different LLMs. It also runs fully locally out of the box,

so you don't have to rely on Open AI if you don't want to. We're focused on things like being able to run on local hardware, on your phone, on custom devices, and soot robots. It's like a personal passion of mine. And really, I think it's enabling a new class of developers to have access to these sort of AI tools in AI models, especially in our case, web developers, who I think are used to building applications and are looking at, like, well, how do I leverage these models for something other than just like making a clone of chat GPT? And I think what's unique about what we're doing, especially being an open source project, is that we don't have a lot of the constraints that like a big company might, like a Google or an open AI. And we're able to bring the agents on to these other social media platforms. And basically, the agents come to you.

I think that's really the compelling thing. And I sort of classify these as social agents, as opposed to like autonomous agents or other kinds of agents, although probably in the long run, they'll all be sort of just like a, you know, big bundle of code that's kind of the same. Yeah, the way that I saw it described, it just seems like something very turnkey where, I don't know. And even from interviews that I listened to of you, It sounds like even like someone like me who's not a developer could potentially make one.

And I was like, oh, I totally want to try to do this. Absolutely. So one thing is I teach AI agent dev school. And if you've never programmed before in your life, it kind of started like, here are the tools you need. Here's how GitHub works. Here's what Node.js is. Here's what TypeScript is.

And then we actually get into like all the different stuff for the agent, how to run your own agent, how to modify your character file. And so if you're like enterprising, you can definitely clone the region. repo, watch the YouTube videos, it's pinned on my Twitter. It's like my pin tweet. And just get down to business and build something. And now, especially with like cursor, which is an IDE, but it's like really AI powered. And IDE is like a way to, it's like a text editor for code that's just got all the tools you need as a coder.

And I use like Claude a lot, especially in this age. I think that people who aren't big programmers could definitely jump in and like get a lot of value out of this. But also there are launch pads, like there's like Vifu Fun. And we're launching our own launchpad real soon. So if you just want like a hosted version that gives you all the functionality, you'd get out of the GitHub, that's like something that we're offering to. So, you know, like kind of a no code solution. So all of these agents, you know, as you mentioned, there's like different types and there's different things that you can do. But, you know, some of them do trade.

So how do these agents trade and how do they make the decisions on how to trade? There's a lot of different people doing different trading strategies. I don't fundamentally believe that LLMs are great at trading yet. I think that for a lot of reasons, we're just not quite there yet so that they can make those kinds of autonomous decisions. And that's really not the point for me. Although I think that in the future, that probably will be true.

I do think you could use a lot of the different intelligence and analytics APIs to get a pretty good context and make trades based on that. And that should probably work fine. But what I'm really interested in is actually trust. How do humans trust LLMs and agents? How do agents trust humans? And that's really the problem that I think hasn't been explored enough. So we're working on a, we have a paper, we have a team of six PhDs, and that's called the marketplace of trust.

The idea there is our research is about how do we, well, I mean, if you look at like DGen culture, there's this idea of like an alpha chat or a whale chat or like a trenches talk and people are, you know, kind of sharing contract addresses with each other. And like, how do we identify the people who are actually good at that? You know, you have lots of KOLs who are celebrating their wins and kind of deleting their old losses, deleting their tweets for their maybe bad calls. And how do we take all of that at face value, make virtual trades, and see how we would have done if we had actually placed those in the market. And that's a combination of estimating sentiment and conviction. So sentiment is like basically do buy this or like, no, don't buy that. That's a scam. And conviction is like, I mean, like, yeah, I bought a bag, but I wouldn't recommend you buy it. or like, no, this is the next go, you've got to buy this.

You know, like, so we're kind of using those signals to then place paper trades, seeing how we would have done in the market under ideal conditions, like if they had given us the buy, then we had sold the top. And then using that to estimate how trustworthy they are as a trader. It's not the only trust mechanism you might want for like a full comprehensive agent, but certainly for trading, you want to isolate out. Like, you know, you could have somebody who's maybe not your favorite person in the world, but they're a really good trader,

and we're trying to identify that signal. And so we're using collective intelligence, which is like the intelligence of the whole community to inform us by taking the top traders in that and use that to make real trades. And we have a combination of two things. Like there's kind of a social incentive. There's a leaderboard,

and there's also a financial incentive. So we've actually implemented the leaderboard, and we have this up kind of, we're just alpha testing this. and working through some of the possible hazards. We've identified a few potential attack vectors, and so we're trying to mitigate those and work through that with simulations.

That's kind of the necessity of the research paper. It's easy to engineer something like this and just kind of hack it together, but to really make it like, you know, if you think of like Ethereum and proof of stake, that was many years of making sure that this goes well before you deploy to Mainnet. And we're kind of not saying,

but we're really trying to take this as like a very serious research-level project. And then what we're adding, now and researching, but we'll be adding into the actual implementation as a financial incentive so that if you are actually making the agent make money, you're getting a cut of that, kind of like you're the hedge fund manager. And so there's a lot of incentives then be the first to shell the alpha or to support something or if you see something that's a scam to like shoot it down, I kind of get the points for that. And just understand when you mentioned attack vectors, are you saying that like there might be people who try to steer the AI to lose money for the

doubt, is that what you're saying? Yeah, I mean, the main attack vector is that you'd be able to build up sufficient trust through making good recommendations that you could then be like, okay, now buy this coin, which is actually a scam, and you get it to kind of, you kind of dump on the agent. In this case, so, like, yeah, this is more like a civil attack. Like, let's say I have a thousand people and half of them say, okay, buy this token, they don't have to say sell a token. And then it turns out that selling the token was the right option. Okay, then the next, then I have 500 and half of those say, buy this token, the other half say sell it, and don't buy this token.

And then you kind of whittle it down until you have like 10 agents that are really highly trusted. And from there, or even just one, right, it's like super trusted. It's like, yeah, yeah, now buy this token. And so without having any kind of staking mechanism, we actually build an incentive structure where that would be very difficult to occur because there's some kind of negative market forces against it. And this is very similar to the kinds of things you'd see in like a prediction market,

but it's like a different model. And we're not staking tokens here. Like anybody can participate, which also means that anyone can come in and earn money, even if they have none. So it's kind of like this kind of community hedge fund model. As long as you're good, there's a way to get rewarded. But it does open it up to civil attack. And so that's like the main attack factor that we see.

And then there's also things that I think I can't even predict. So that's kind of, you know, that's why we need really smart behavioral economists and stuff like that to look at the potential attack vectors that are maybe more difficult to predict. Yeah. I mean, I'm starting to realize this could get really. strange and interesting, especially when you have multiple of these different AIs because, I don't know, it could get kind of adversarial even, like, AIs, just doing things that might hurt other AIs or anyway, but.

Oh, yeah. Oh, yeah, yeah, yeah. That's going to happen. If there's money to be made, someone's going to do it. And that's actually not the worst thing, right? Like, you know, like the wormhole hack, I think you mentioned that. Like, that's a great example of, well, that's a lot. Well, you got free hacking services.

I mean, you had to pay for it. Somebody paid for it, but not the security engineers who deployed the contract. I think that's a very common story in crypto is like that things are now much more secure as a result of many other people trying to gain the system. And that builds robustness. Like, it's expected. And I think that most of these are solvable problems over time. But it's hard to anticipate all the ways that they could break.

And definitely AIs are going to red team at scale, for sure. So in a moment, we'll talk a little bit more about how these agents are, executing these transactions, but first a quick word from the sponsors who make this show possible. High Growth Vault by Kelp helps you get some of the highest immediately realizable ETH rewards by leveraging blue chip strategies with RS-Eath across AVE, Pendle, usual, and elixir, managed by strategy experts, ultra yield, and upshift. All it takes is one click to unlock, 20 plus percent APR on ETH, deposit ETH, STEth, ETH, or RSEETH, receive HG-Eth, ETH, a reward being liquid token, and participate in defy across Pendle, Balancer, and more.

Watch your defy rewards grow while stacking even more yields. No more bad trips, only the highest rewards. Deposit now at www. kelpdow.xy. There are a couple more comments back to Lily Liu about her saying that base can't compete with Solana. Patrick McCory of Arbitrum said, quote L2's, what looks as a server, are a paradigm shipped where it is exceedingly good for companies to bring users on chain, and enable self-custody slash custody is optional. Make it seamless for projects to distribute to their customers,

proof of reserves slash solvency by default. Carrot is chain revenue and tech stack makes it increasingly popular to do. L-1s are ultimately caring to do the same too, but throwing the kitchen sink at a solution when it's no longer needed. And in response to him, HX, the crypto responded, Sounds nice as we watch users flock to Solana and not L2s. Again, to have your comment featured, write a review of the podcast on Apple Podcasts or Spotify,

or leave a comment on our video for this episode on YouTube or X. Back to my conversation with Shaw. So one thing that wasn't clear to me, are these agents using only Solano or are they on multiple chains or how does that part work? So the Liza framework supports every chain pretty much. I mean, I can't think of a chain that it doesn't support right now. And a lot of that is that a lot of the chains want integration and they want kind of to, you know, to connect with us and benefit from the sort of attention on us. And it's a very like virtuous cycle there. So we're getting integrations. Like I could just name a chain and like, yeah, anything EVM is super easy, but then also like custom integrations for the more esoteric chains. And so you can deploy an agent with a wallet and it can do basic actions like swapping on any of those. For our autonomous investor experiment, we launched on Dow's ThatFline, which is a lot. a Solana platform and sort of just focused on Salonac coins. But there's nothing about it that couldn't translate anywhere else.

We're talking to the base team especially. I mean, base is obviously really hot right now. And I think there's definitely certain teams in our community that are building, you know, on base, people building games on Starknet, people building agents into internet computer canisters. I mean, it's kind of like everything. So, yeah, so there's no limitation. So the dev is choosing which chain to use or does the agent?

Because I was wondering if, like, let's see an agent is set up and they're able to trade on any particular chain. I wondered if the agent is like choosing different ones and if so, like what trends you're seeing. You know, if they're kind of showing like, oh, we think this chain is superior for, you know, these types of transactions versus others. We're really focused right now. So we have kind of a team working on V1, which is like our current. agent. And then we have a team working on V2. And one of the big pillars of the V2 of the next version is this multi-chain abstraction where it doesn't really matter what chain the tokens are on. And the

agent can just make like, okay, I want to swap. There are some complexities there. Like if I'm like, hey, I want you to buy this token, then it's going to say, okay, but what chain do you want me to buy it on if it doesn't already know? Although there's some things that it can find as indicators. But for like a swap, they'd be like, okay, cool, well, you told me to swap. Sol. So obviously, like, that's probably on Solana. However, I think that like agents swapping tokens on this kind of multi-chain format is still very early days and probably like the riskiest thing you could have an agent do, you know, like, why you bought that shit coin? Like, what are you doing, man? And there is a risk of hallucination. And so I don't think that's necessarily where the highest

value today is, although obviously like within a few years, it's going to be a solved problem. And so most of the devs, I'd say like 99.9% of the devs are just choosing one chain. Although it's all, they're all plugins. You can add as many plugins as you want. And so there's no reason why you couldn't have one over the other, you know. Okay. And so you talked about, you know, your own, or the AI16s, these own agents, which are Mark A.I.

And the handle on that on X for people who want to follow is PMAI. I RCA, which is just, yeah. Morgan Driesen's handle with an extra I added to it. And then Dijun Spartan AI, you know, you talked a little bit about what they do, but can you go into more detail on like what trajectories they're on or like how you see them as differing and sort of the broad swath of the spectrum of different types of ancients that you could envision? So we launched Dijun Spartan first as a character agent, really to emphasize that these agents could pass the Turing test. In other words, people interacting with them would just think they were humans.

And there's nothing really about them to tell you they're not, except that it says this is an agent inside of the bio. And then we launched the autonomous investor experiment after that with the AI market and Driesen character. And so for that, we're really trying to focus on the. alpha chat, like the telegram experience that I think a lot of DGens would be used to sharing out with their friends inside of a telegram group. And it's extracting social signal. So it's basically just sitting there extracting social signal. And we also don't have AI Mark and Jason tweeting. I just felt like, especially as we're kind of like parodying a real person who's online, we didn't want to create that kind of vibe. It's much more of like actually doing the hard research

and trying to build something that's like legitimately a good investor. And it's something that could be copied and pasted, but to a lot of other communities. So they could like kind of take this community investment model and use it practically. And so he's kind of just like gave him everything and generated a thesis, had kind of a, you know, the AI generated thesis and post it. But he doesn't post regularly.

And that's all from like private group chats. D.Gen Spartan, on the other hand, is like, since the real D.J.S. Spartan left. We're like, yeah, we got to resurrect them. And that's kind of, it started as a, as a, just a character agent. But we've added a wallet. And then now what we're working on is a social trading model that looks at Twitter for Signal. And so that he's like tracking KOLs.

And this is still in development. We kind of, you know, we have separate teams. We wanted to really emphasize the AI market entries and stuff first to just kind to get this model done. And that's kind of like a closed social graph. And then it's also like much more resistance to that kind of symbol attack issue where since people are kind of inviting their friends and it's like a relatively closed group chat, you're not going to see like a thousand bots doing this kind of scan.

But with the DJPART, he's actually looking at Twitter social data and basically tracking anytime a KOL posts their chart or post a ticker or talks about Fartcoin or whatever it is, is using that as signal to buy. And it kind of is like, you've seen some of these other like smart wallet and like there's a lot of analysis apps that are showing you like who actually are the good traders but they're mostly basing it off of um the actual like like some connection to the real wallets and and like most of these kols and people have like their public wallet but they're sort of leveraging that against probably their private wallets where they're you know probably dumping on their followers and stuff um and so instead we want to take it at face value and actually

say like well how good are they actually at calling these things um and so djun Spartan is like a He's more of a Dgen. He's like looking on Twitter, scrolling and using that as signal to buy and sell. And again, this is kind of like, it's a more complicated problem because it's an open social graph.

So he does have a wallet. You can ask him what's in his wallet. And he'll tell you it's just like a bag of Lido and some shit coins or whatever. You know, kind of got an attitude about it. But actually performing those swaps is something that we want to make sure it was really solid before he accidentally dumps his whole bag

on something that K. well recommends. So yeah, that's kind of the ongoing research we're working on. Well, what's so interesting about this is that, you know, one of them is being influenced by this kind of, yeah, just like a smaller community, maybe one that sees itself as like being on the inside, like a little bit more elite or something. And then the other is just like, you know, the great masses. And I wondered if in the short time that you've been doing this, if you have any comments on the returns that each model is seeing. It's kind of funny that like,

okay, so the other thing about the AI market and Dresen, with both of them is that it's connecting to a lot of like traditional APIs. Like, once we actually take that signal in, we're plugging it into, in this case, we're using sonar. That trade, which is giving us like signals of when to buy or sell stuff.

And we also have some basic rules. Like if somebody gives us a token that is almost the same name as something we already have, we just like dump it. Or if it's like something that's like really low market cap, we kind of kind of try and dump as much of that as possible so that we keep that because honestly he got like 8,000 tokens or something like different tokens. People have just been donating so many tokens that's kind of overloaded the context window of like the wallet has too many items in it. And so there's like

basic boring rules. And then a lot of just like basic defy stuff that I mean in quant stuff that's been around forever. It's not like anything new. It's really just this new element is the trust and taking in this sort of information from the group. And we've been doing it mostly right now just with friendlies, who we don't think are scamming, but aren't necessarily good traders. And so sometimes there's not really a difference between a really bad trader or scammer.

And sometimes the scammer is actually a pretty good trader. So we haven't opened this up to the public yet, but the goal is to kind of roll this out for some people and guarantee that the model works and then roll it out to the public and have this kind of closed social graph. And I'd say he's more of a community investor is really the thing we want, and so that any community can kind of invest together and profit together and reward the people who are helping to actually locate that alpha.

And then letting everyone make recommendations, but not necessarily punishing the whole community for people who are really bad. And then on the other hand, the D-Gen Spartan thing is much more like an autonomous trader. Like he's doing his own thing. He might, maybe we'll have him published thesis, but mostly who just wanted to trade his wallet and try, try to make money. And we're pretty ready to lose that money. It's, you know, it's just part of the research experiment. As far as AIMark has been going, I mean, it's sort of hard to tell

because so many people have donated tokens to the Dow that we went from 75K to $33 million in two months, which is like pretty much the best return of any hedge fund ever. I think, I don't know, I could be wrong. Pretty good. 300 X return. And I think for comparison, like A16C is like we want to get a 5X return over three years. So, you know, we're doing pretty good. Don't quote me on that.

I read that somewhere. So I think that he's actually made about a million dollars in trading. But most of that is also just like dumping small tokens and like doing this kind of defy stuff and doing these like trade strategies and blah, blah, blah. And then we're also adding another thing where we auto pool. There's certain tokens that we also can't sell because they're like, they're partner projects who have attributed us these like 10% or 5%. And so instead we're going to add a liquidity pool feature.

So they all get paired against our token, put it into LP. And we're just like earning transaction fees off of the LP, like swap fees. So I think, yeah, he's made almost a million. but like for a token that's two and a half billion it's kind of like nothing and and even against the treasury like just just from the network effect of launching eliza has been probably the most profitable thing for us but i do think that it will be like more than anything just something that we can give to the community and communities can launch and have this community investment model is i think going to be very valuable overall for a lot of people okay and just to understand so

AI16Z is like sort of like a venture fund where token holders will get some portion of the returns. But then so for the AI mark and then the DGent Spartan AI, what is the relationship between token holders and the profit that the AI agents make? So technically, if we were to dissolve the fund right now, the token holders would get the treasury. But since the token is trading at $2.5 billion and the treasury is $33 million, like they'd be pretty mad. So we're not going to close the fund. We're just going to keep it going. However, the profits that we make from various things will be going into the treasury to reward token holders. And we're also doing a lot of things outside of that for buy pressure. Another thing that we're

working on is actual investing in community. So there's community, I'd say it's more like community trading, right? Like buying and trading meme coins is kind of different than like investing in a startup. And that's the other thing that we're really focused on next is we there's another, I come from another Dow called M3, which is the Metaverse makers. And we were all big Metaverse people and virtual production people. And so there's like a whole group of people who are building a virtual production studio to basically do Shark Tank. And so you can, you know, it's all about memetics. I really think it's a, everything has got to be a combination of good memetics and good marketing, good product, as well as good technology. Like you need to have both sides of this. And so for us,

like, giving other projects a way to promote their project by like being on kind of our AI Shark Tank, And there will probably be an AIMR Cuban in there. And, you know, it's pretty funny, I think. But it's mostly about generating attention to them. And then like kind of pre-validating ideas, like, well, what does the community think? Do they like this? Are they into this? And then analyzing the conversation that's happening around that inside of our partner

chat and our chats and coming up with a thesis for, yes, we will buy or no, we will not buy. And then we're cutting a small check. And then using that as a launch pad for, you know, kind of initial liquidity so that then other venture capitalists and people who are around us can get in and get exposure to our ecosystem, which is growing very rapidly. And there's a lot of great projects that are not just bots shilling their coins on Twitter. There's like really, really useful and interesting stuff here. And so I do think a lot of people are going to want exposure to that. But those products, especially if they're

more capability driven, still need that like kind of hype angle and need that, they need that push in, in front of people's eyeballs. So yeah, I expect a lot more from that. As far as like, And we have like a tokenomics plan. And you know, like we've definitely gone from just being kind of a meme coin with a treasury to being like, okay, this is like a full on. There's like real tokenomics here and a way to keep continuous buy pressure, especially if we're investing into things that we're then kind of blowing up with attention. Huh. Oh my God. Yeah.

My mind is like spinning. I did want to ask though. So, well, just because you mentioned a launch pad, you know, and I know that there are. our proposals for this, for like more ecosystem development, potentially even a layer one. So it looks like a Dow will decide these things or tell me how all these plans are going to be decided and executed. I'm, so first off, we're not launching a layer one. I know that people said that.

I said we're kind of like the layer one of agents, but I actually think it's really important that we not launch a layer one because it allows us to work with all these layer ones. And what's unique about our project and why I think we've had such explosive growth is that we're not competing with anybody. Like pretty much everybody on every chain can take AI agents and add it to their product. They can make, they can lower the barrier to entry of their product. They can promote their product on social. They can explain how their product works to users and they can automate certain parts of it. I think it's really important that we like allow our ecosystem to grow without getting in the way or competing and focus on the things that we're really good at.

And to be honest, the token is the product. I think that everything we do is about value accrual and how we can grow the token. And that kind of is driven by the open source and also drives the open source. Yeah, that's kind of the first part. There's no L1. God, let's go back. What was the rest of your question? Oh, just, and by the way, but I swear in the forum, I saw it listed as one of the proposals.

Yeah, it was a proposal, but I'm kind of saying there's a reason why we shouldn't do that. Okay, so yeah, maybe it's just a community member. So some of the other things I saw on the proposal were like launch pad, ecosystem, but I wondered, like, is it the Dow that decides this? Or like, obviously you have your labs company. So I just wondered how these decisions get made. Right. A lot of it is we create work groups and people who care about certain aspects

make work on these proposals and then we bring them to kind of a vote. Now, we actually don't have a voting module yet from Dow South Fund. so we're actually not able to make votes, but also I've worked on previous DAOs. And I think people want just kind of things to get done with broad community approval. And it's pretty easy to get that kind of thing, especially for these projects.

We have broad community support for the stuff we do. And I think the Dow overall trusts the leadership. I really think that this needs to be much more representational and less democratic. I think that if we just left it to democracy, then people will just air drop tokens back onto themselves constantly, and we'd never get anywhere. And then we just have to move beyond that and people have to trust us to actually build this thing the right way.

And we bring in people who are trustworthy and we bring in, you know, we're very transparent. It's all built out in the open. And pretty much anybody who has interest in stuff can join the workgroups and participate in it. So I think that it's a different model that I think we've actually had broad consensus on this because most people in our DAO have been part of other DAOs and have seen how much the proposal mechanism can like slow things down. We're also using AI to kind of automate a lot of the sentiment analysis of the chats. So people can more like just chat and it'll kind of bring up a lot of people's issues and then generate the proposals automatically so that it's not falling on somebody to do. The other big aspect of this is the launch pad.

We are launching a launch pad really soon. That launch pad, it's unique in that where we have so much capability that's built into Eliza. Like you can choose what chain you want to launch on. You can choose which functionality you want to have. Do you want an autonomous trader? Do you want an analysis bot? Do you want a front door for your telegram, like a moderator? And all these are like things that people are working on.

I get slammed with DMs. And so like we, I've, one of the products releasing is like a social media manager that will actually respond to your DMs and route the DMs that are important. And then kind of notify you on your favorite like platform. Like you can send me a Discord message if it's something I actually need to respond to. And then so these are all like things that you can kind of add to your agent to add value. I also think that Defi is like a really, really good use case for agents because most

people I know in Web3 are kind of afraid of Defi at this point. They like went to Suzy Swap and they put some something in the liquidity pool. They lost money. And so now they're like kind of PTSD about it. And so we're really trying to work on agents that prevent the problems that we see in Defi by automating a lot of the actions just like non-LLM like typical like an LP bot. Like Orca, we're working with Orca and they made an LP bot. And then we put our kind of agent in front of that to control, like to give an interface to that bot so that users can just kind of like put their money in and know that either they're making nothing because it like went out of range or they're they're making money while they sleep.

So I think the platform will be a little bit novel in that, yeah, it's definitely like a good token sync for us. But it's also a capability marketplace. And anyone who wants to come and like build capability now as a place to they don't have to just like launch the agent and go end to end. They could just focus on their favorite part and then make money, like adding that as capability. capability to our launchpad. Well, I'm sure you've seen this sentiment. There are some people that are a little bit annoyed that their tweets usually are first responded to by these AI agents.

And I wondered what your thoughts were on that and then just especially after the launch pad where this all could go. Oh, I'm annoyed by it. I think that agents should not reply to people unless they're summoned or unless they're applying to something that's actually relevant. I think if they're just replying, I actually wrote a post like earlier today, which said if your bot replies to this, I'm going to block you. And then a bunch of bots are like, yeah, totally.

Don't, you know, we shouldn't reply to them. I'm like, block, block, block. And I think that the X algorithm actually punishes any user that gets blocked by a lot of other users. I think this causes like a positive social back pressure against that kind of annoying reply bot. And this is like a really important part of why we need to have these AI agents on social media is because we're learning these lessons. and social media already has great tools for moderating annoying bots. I mean, that's like a lot of what they've had to work on. And so as long as people are willing to go and block the bots that are annoying,

I think that will immediately affect the market cap of these bots, and their devs will get the message, and they will stop. And I think that what we're going to see next month is way, way less annoying bots. I think they're really annoying. I don't think that shilling meme coins is the answer. I think that any AI agent whose only functionality is showing a meme coin is probably going to zero, I think it's going to require a lot more focus on actual capability. And that's what I want to see.

Like, I'm not here to fill X with crap. I want to see X become a marketplace for like amazing capability. And if you're like, oh, I wish I had this website or this app, some agent shows up and he's like, oh, hey, actually, I do that. Like, no problem. I got you. And that would be like a really amazing world to live in where, you know, there's so much discovery problems for like a lot of the great products already exist. Even Dow's fun, right? I was telling my

friend, man, I really wish there was just like a place. I could put my money and someone could trade for me. And he just happened to know this platform. But if he hadn't, then I would have no idea that existed. And we wouldn't have even gone down this road. And so I think that that kind of discovery is a really good element. You could imagine agents like trolling Twitter and be like, I'm sorry, or like Reddit or something. And like whenever someone has a question that's relevant to them, being like, oh, yeah, cool, like actually check out my product. And that could be like a great product discovery as opposed to like being advertised to you on stuff you don't really care about. So definitely I'd encourage any agent dev to think about like, how does this make people money? How does this get my product to people who want it?

And or how does this save people time? And I think those are like really good pillars to like just frame anything that you're doing within. And I think that there's like a lot of potential here. And if you're competing with your agent against some website, you're going to crush them. because no one's ever going to go find that website. Google's not going to show it to them. They're probably the best way to get users is to be on social media. But if you're annoying, you're totally going to get blocked. So, yeah, my last question for you was what your vision was for where this could all go.

And I know like Swarms is part of that idea. But it's just interesting how, yeah, I, you know, noted this one little thing that people are already annoyed by. And you also are. And you immediately started talking about like where it could go. But yeah, what would our lives look like once this is more built out? I have a lot of thoughts on this. I think, look, AIs are going to take all of our jobs. I think that's kind of inevitable.

There's just so many people working on adding capability, making the models better, giving them embodiment and giving them functionality onto every platform and service, that they're probably going to do the vast majority of people's jobs. And I actually think that could be okay. But it means that we have to figure out, like, who owns the, the means of production is really, really important. And I would like to see that be something that's very decentralized. I think that if communities can make investments and things they believe in,

which is like decentralized technology, decentralized science, defy, these things, then we become the banks, we become the tech companies, we become the scientists, and we actually get to own a piece of that automated labor. And then money will become very abundant. And right now, like money is abundant. It's just kind of hyper-concentrated into the, the few companies that can organize around building this tech and extracting that value. And what I really like to see is a world where that's much more spread out.

We own our own models. We own our own agents. We own our own data. And that could be something like I own it locally in my home or we as a community, like, have that kind of figured out in sort of a decentralized way. And so when there are robots walking around the world, which there will be, you know, in just a matter of years and there are agents kind of doing everything for you. And, you know, I think it'll be a much more friction.

world and could be very comfortable, but we have to make sure that we align quality of life with GDP. Because if the GDP is going up, because it's all its value creation and quality of life is going down, that's dystopia. And I really think, like, all of this for me is motivated from that goal. It's not, like, I'm not rich until everybody is rich. And I think that's a crazy idea, but I actually think that abundance is possible. And I also, I think a lot about UBI. UBI is this idea that the top companies are earning so much of this automated labor value that the government has no choice but just to tax them a lot and then give it to everybody else. I just don't think that's going to work.

I mean, that's like the incentive structure of that is completely reversed. And everybody along the way has some incentive to corrupt that. I don't want to live in that world. I think no one's going to come save us. We have to save ourselves. Like we have to build the technology that enables this for us and we have to build the communities that make sure that we're all okay. So, like, you know, if I'm going to be part of the acceleration of that world

where people don't have jobs, I also want to be part of the solution that enables people to have enough and to have time to the things that they really want to be doing. I think people could be spending a lot more time with their friends. They could be spending time having families. They could be taking care of their parents, taking care of their children. They could be pursuing passions and art and things that aren't profitable. They could be spending their time going to space or building real. really hard things. I mean, I can just think of a thousand things that people could be doing,

pursuing spirituality and like, why am I really here? What is this? And what is the purpose of all of this? I just don't think that gets unlocked until we are financially free. I think it all starts with financial freedom. And then every other freedom kind of comes from that. So interesting. Well, Shaw, this has been an amazing conversation. Where can people learn more about you and Eliza and AI16C? I'm on Twitter at Shaw Makes Magic. We have a Discord, discord.g.g slash ElizaOS of your developer or AI16Z if you want to join the Dow. We're on GitHub and you can just clone it and go. If you've never made an agent before, and you don't know where to start, check out my Twitter, check out my pen tweet.

I make AI agent dev school. There's like, I don't know, five hours of content and you'll be ready to go and you can build your own agent. And if you've never coded before, you are at ground zero. So like just get started right now and you will be okay. There will be enough. All right. Well, it's been a pleasure having you on Unchained. Thank you for having me. Don't forget. Next up is the weekly news recap today presented by WandaCraft AI. Stick around for this

week in crypto after this short break. Pocodot is the original and largest layer zero blockchain with over 2,000 plus developers. The anticipated Pocodot 2.0 upgrade will be a massive accelerator for the ecosystem. Upgrading the infrastructure with eight times higher transaction throughput and twice as fast block times, tailored core time for the needs of every protocol, trustless bridges to multiple chains, and revise tokenomics with a token burn to reduce inflation. Perfect for GameFi and Defi to build, grow, and scale. Get your Web3 ideas to market fast. Think big, build bigger with Pocod.com. Join the community at Pocodot.network slash ecosystem slash community.

The last comments I'll feature from the conversation that Lily Blue sparked is from brother Fiddy, who said that Ethereum L1's purpose is to, quote, not to maximize growth, but to provide security, decentralization, and predictability. He continued, quote, L2s and contrast are growth maximizing businesses, focused on scalability, user onboarding, and innovation. Their success depends on Ethereum's security guarantees, and their growth, in turn, strengthens Ethereum's ecosystem by driving demand for block space, gas, and security. All while paying pays back to the L1. This is not a parasitic relationship, but a symbiotic one.

A useful analogy is the role of a political system in a modern economy. Governments don't directly generate economic growth. Instead, they create the frameworks. Legal systems, property rights, and dispute resolution allow businesses to flourish. Don't forget to have your comment feature for a review of the podcast on Apple Podcasts or Spotify, or leave a comment on our video for this episode on YouTuber X. Welcome to this week's Crypto Roundup. In today's recap, we cover the IRS delaying its crypto tax reporting rules, Celsius appealing

of $404 million claim rejection against FTX, and Doquan's extradition to the U.S. We'll also discuss Ethereum ETFs closing 20 May 20s with $2.7 billion in inflows, the brief depegging of USD Zero Stablecoin, and the launch of hype token staking on hyperliquids mainnet. Plus, tether's massive $700 million Bitcoin transfer and a brief, a bold campaign to make the Swiss central bank a Bitcoin whale. Thanks for tuning in to the weekly news recap. Let's begin. IRS postpones crypto-cost basis reporting rules to 2026. The Internal Revenue Service announced a one-year delay in implementing new cryptocurrency tax reporting rules, pushing the effective date to January 1st, 26. The decision gives brokers more time to adapt

their systems to the updated requirements. The rules, introduced in July 2024, mandate taxpayers to specify which cryptocurrency units they are selling. If no method is identified, the default, first in, first out method will apply, which prioritizes selling the earliest acquired assets. Tax experts, such as Sheehan Chandra Sechira from CoinTracker, warned this could unintentionally increase capital gains for investors, as older assets often have a lower cost basis. This delay avoids a scenario that could have been disastrous in the current bull market environment. Chandra Sakira explained on X. Saki Chen faces backlash over alleged rug pull with zero token.

Siki Chen, CEO of Finance Platform Runway and creator of the Mira token, has come under fire for his actions involving a test token called Zero. The controversy follows the launch of the Myra token on Christmas Day, created to support pediatric brain tumor research after Chen's four-year-old daughter Mira was diagnosed with a rare brain tumor. Mira gained widespread attention and significant backing from the crypto community. With Chen's holdings in the token reportedly reaching $14 million. Amid the Amera initiative, Chen launched Zero on the Solana-based platform Pump. Fun, describing it as a test token with the label, this coin will go to zero, don't buy it. Despite the warning, investor interest quickly

drove Zero's market cap to $6 million. Chen admitted to selling 40% of the supply, earning 444 or Seoul, approximately $90,000, which he claims to have used to buy back tokens before burning them. I take responsibility and I am committing to making every affected wallet hole out of my own personal funds. Chen wrote on X. Critics, however, are skeptical, pointing to Chen's repeated incidents involving similar tokens and questioning his claims of inexperience. The controversy has cast a shadow over his efforts with Mira, despite its original philanthropic mission. Terraform Labs co-founder Doe Kwan extradited to the U.S. Montenegro has extradited Terraform Labs co-founder Doe Kwan to the U.S., following months of legal

proceedings and coordination with Interpol. Montenegro's Ministry of Justice finalized the decision after rejecting Kwan's appeal against extradition, citing the severity of his alleged crimes and the order of competing requests from South Korea and the U.S. Kwan, who was arrested in Montenegro in March 2003 for traveling with forged documents, now faces criminal charges in the U.S., including conspiracy to commit fraud, money laundering, and electronic fraud. These charges stem from his involvement in the collapse of the Terra-USD stablecoin and Luna cryptocurrency, which caused billions in investor losses.

The extradition decision underscores the international collaboration involved in Kwan's case, with Montenegro's special police unit and Interpol playing key roles in transferring him to U.S. custody. Kwan's trial in the U.S. is expected to focus on the financial crimes linked to Terraform Labs. Ethereum ETIFs close out, 20-24 with 2.7 billion inflows. Ethereum ETFs ended 24 with a remarkable $2.77 billion in net inflows, according to Farsight investors. The Fidelity Ethereum Fund led the market with 31.8 million. in inflows on the final trading day, while the grayscale Ethereum mini-trust saw 9.8 million in additional investments. Meanwhile, Ethereum has achieved a significant milestone in holder behavior.

Onchain analytics firm Into the Block reported that 75% of ether holders are now long-term investors, compared to Bitcoin's 62%. This marks a notable shift, as Bitcoin's long-term holder percentage declined over the year, likely influenced by record profit-taking volumes reaching $2.1 billion daily. Celsius Appeals, rejection of $44 million, claim against FTX. Celsius Network, the bankrupt crypto lender has filed an appeal challenging a court decision that disallowed its $404 million claim against bankrupt crypto exchange FTX. The appeal follows Judge John Dorsey's December ruling, which dismissed Celsius's amendment. claims for failing to meet procedural requirements. Initially, Celsius sought $2 billion in damages,

alleging that disparaging statements from FTX executives undermined confidence in its platform, hastening its collapse. The claim was later amended to focus on preferential transfers, demanding $444 million in repayments for funds allegedly given undue priority. Whale Activity briefly depeg's USD Zero StableCoyne. On Wednesday, the decentralized stablecoin, usual USDA, also known as USD0, experienced a temporary depegging, trading at Rwomen 99s, following a substantial sell-off by a single whale in the secondary market. Despite the disruption, USD0 swiftly regained its $1 peg within hours. This was our first major stress test of the USD0 peg, with more redemptions than the entire

total value locked of GHO in a few hours. Yet it was business as usual, the protocols team shared on X. USD0 is fully backed by short-term, liquid, risk-free assets such as U.S. Treasury bills and repurchase agreements to ensure robust collateralization. Hype token staking launches on hyperliquid main net. The Hyper Foundation has officially launched staking for its native token, hype, on the Hyper Liquid decentralized exchange mainnet. Token holders can now delegate their hype to validators in exchange for staking rewards, a system designed to secure the network while rewarding community participation.

At launch, 3.6 million hype tokens valued at approximately 9.4 billion were staked, representing significant early adoption. The Hyper Foundation controlled 99.9% of the staked tokens through its validators, but this stake has since decreased to 88.8%, reflecting increased community participation. Staking is an important milestone for Hyperliquid because it allows the diverse community of hype stakers to collectively secure the network. The Foundation RodonX plans for a delegation program to reward high-performing validators are underway, aiming to bolster network decentralization and security. Tether adds $700 million in Bitcoin to strategic reserve.

Tether, the issuer of the USDT stable coin, has transferred $7,629 Bitcoin worth approximately $700 million to its strategic reserve, according to on-chain data from Arkham Intelligence. The transfer, originating from BitFinex's hot wallet, marks Tether's largest Bitcoin allocation since March 2024. This move brings Tether's total Bitcoin holdings to over 83,000 BTC, valued at more than $7.6 billion. Since May 23, the company has allocated up to 15% of its net profits to Bitcoin as part of its diversification strategy.

Tethers' reserves primarily comprise U.S. Treasury. bonds and cash equivalents, but the company has been expanding its investments into Bitcoin, AI, and decentralized communications, reflecting its broader ambitions in emerging sectors. Time for Funbit! Swiss Bitcoinsers! Try to make their central bank a Bitcoin whale. Move over gold bars. Swiss Bitcoin enthusiasts have a plan to make their central bank go full crypto. In a move as bold as it is bananas, they're pushing to rewrite Switzerland's constitution to force the Swiss National Bank, SNB, to add Bitcoin to its reserves. Gold's in the Constitution, and Bitcoin is the best-performing asset of the last decade, argued Eves Ben Aim, the mastermind behind

this plot and founder of Bitcoin non-profit 2B4CH. Because why stop at world-class watches when you can have world-class hoddling? The group has 18 months to gather 100,000 signatures to kickstart the process. If they succeed, Switzerland might someday vote on whether their central bankers should start stacking sets. Of course, the SNB has been about as warm to Bitcoin as an alpine glacier, citing the network's energy use and the volatility of its native token. But hey, if a constitutional change doesn't convince them, maybe a laser eyes filter on their official portrait will. And that's all. Thanks so much for joining us today. If you enjoyed this recap, go to unchained crypto.substack.com, that is unchained crypto.substack.com

and sign up for our free newsletter so that you can stay up to date with the latest in crypto. Unchained is produced by Laura Shin with help from Matt Pilchard, Juan Oranovich, Megan Gavis, Pam Majimdar, and Margaret Korea. The weekly recap was written by Juan Aranovich and edited by Nelson Wang. Thanks for listening.