Transcription
It's been a wild rocket ship kind of experience, and we've only been doing this for now almost two months. Mark Andreessen is the greatest living venture capitalist, and our goal is to outperform him.
Now, up to $10 million AUM, currently at $17.7 million. It's a good day!
So, our autonomous investor is trading. K's are actually using, like, they're counting on that coffee trading with our stuff. It is fully Cowboy posts it, and then you see it get posted. You're like, "Oh, that's... I hope it's... yeah." Yep, there is no review. We are just doing this in production.
I'd say Dex Trainer has probably prevented more millionaires this week than any website in history. No, I don't... maybe it's true, probably true. I think we have pretty big ambitions of bringing this to a national stage, making this like a really legitimate thing.
I think that fear should be a default position for anybody who's just going to let the world happen to them.
Hey, Sha! We're live now. Happy to have you here. Here with Shaw on another episode of Mario Nef Fall's Crypto Daily.
And, uh, Shaw, you started AI 16z, which is now the largest AI meme coin in the entire sector, after now being larger than Truth Terminal. How's that feel, at least in Solana?
You know, there are a few things outside of Solana, but it feels interesting. I mean, it's been a wild rocket ship kind of experience, and we've only been doing this for now almost two months.
You know, we've gone from where we were just kind of a guy working on an agent framework to wow, a lot happening. I can't even believe it's been two months. It almost feels like it was a couple weekends ago when I was first hearing about the launch. Personally, it doesn't feel like it was even two months.
No, I mean, it was actually, I guess, more like six or seven weeks ago. But yeah, it feels like both a lifetime and like if I just blink, a bunch of things are happening constantly. So, you know, the acceleration is real. We're right in the middle of this interesting inflection point.
Yeah, so what's your background like? Do you define yourself right now as an AI researcher? What do you define yourself as?
No, I think AI is like... it kind of hits in multiple parts of the stack. Like, obviously, we have these incredible models which you can now just download and run at home, and also models that are made by large corporations and all that stuff.
I felt like while that stuff is really interesting to me, and I did some AI research, I'd say not just more like learning the stack and trying to make my own models and that kind of aspect of it.
I think that what we're really focusing on is the application layer. Like, how are web developers going to use AI in their applications?
I also think the fundamental shift here is less about AI and more about social media and kind of how applications are moving onto social media directly. AI is sort of enabling that by dealing with a lot of the friction points.
So, I'm a full-stack app developer. I've dabbled in all parts of the AI stack from, you know, I've fine-tuned models, I've trained models from scratch that were of mixed results.
But I think that where most of us really get to experience, like, do stuff with this that feels powerful is when we make applications for people and solve problems with it.
So, we really kind of sit in between the devs who are building these applications and the researchers who are doing the AI. I'm definitely more like a full-stack programmer. I identify much more as an engineer than a researcher.
I'm very interested in all of the technical and architectural aspects of this, and I think there's also a big creative component to working with agents. You're really trying to dial in the quality of voice, the character, the immersive experience with users, just solving all the UX.
So, I really care about all those things, and that's really where I come from. I think that for a lot of people, it probably seems like I came out of nowhere, and that's because I was like an anonymous... I had an anon account, and also, yeah, I was just working on this framework for a very long time.
It didn't really have traction until we kind of solved the minimum loop, which was really like cracking social agents—not agents in general, not like autonomous agents that do their own thing, but social agents that are on social media, interact with you, and kind of bring apps to you.
So, yeah, and like I said before the show when we were just chatting, I did talk with Parille yesterday as well, and I know that you guys actually started a business together.
I'd love to know a little bit more about what you did prior to this. What else were you working on?
Yeah, I worked for a few different companies and started a few different companies. That company is called Magic. We started by just making an open-source agent builder for a no-code agent builder. You can just drag nodes into a node editor, and each of those are like a little block of logic, and you can connect them together and make agents.
We got pretty far with this whole thing. This was, I guess, sometime last year. The problem was that we were building a platform for people to come build agents on and not really showing what agents could be, you know, because we were sort of bogged down in building the platform to enable it and really solving user problems.
I really just felt like before we can really crack this and go wide, we have to show people what an agent can really do. I think people are still really skeptical of the capability.
A lot of people are coming in and just struggling with the basic concepts, and I just felt like I have to go make agents that are as good as we possibly can.
I left Magic, the company, took some seed money, and basically sat down and did research for a while on different agent frameworks, different patterns and paradigms, making self-coding agents, that kind of stuff.
I ended up working at a couple of different AI startups over that time, trying a couple of building kind of stuff and trying the Y Combinator track but not really finding a lot of traction.
I worked for Web Averse, which is a Web 3 agent metaverse kind of thing. I think that was really influential on me. Yeah, you've probably heard of that. It was a big project until they got hacked, but they pivoted into being Moy Mate, which is now around, and we're all friends and interact.
Moy Mate's been showing some A16Z stuff. But that was like really cutting my teeth on... I was really interested in 3D avatars, AI agents, how these things walk around worlds, and seeing it not just as a video game, but like video games are kind of the lab for the real world.
If you can make agents that had contextual scene awareness, decision-making, able to navigate 3D spaces, and interact with users seamlessly, you could put that into robots eventually, and you could put that everywhere.
You could also give them social accounts and let them tweet out into the world from their experiences in this kind of metaverse. So, that's what really appealed to me.
Then I just kind of saw a lot of the bodies around the metaverse and that it was really hard to get people to come to these places and that there wasn't much to do.
I realized, like, I think that Twitter is the metaverse. I think that social media is where it all starts, kind of in zero dimensions. As we are able to integrate more of this onto the platform, I can tell you that the everything app, I think, will eventually be the everything app.
I think that they will start doing embedded applications, embedded HTML, and then we'll kind of work backwards.
I mean, that's my experience. You touch on the no-code solutions. It's my experience with no-code too. If you don't give somebody an example of what they can actually use it for, then they often just get stuck in this creative bind.
I found it was ingenious, and we'll get into this, like what you did with AI 16z, showing the capabilities of Eliza out in social media because then it actually prompts creativity in builders, being like, "Oh, that's what I can exactly use this technology for."
Yeah, exactly. For most people, they just need to change, like, you know, add their thing onto the system, and then out comes an application.
Actually, this is how web development works today. If you look at Next.js, it's a very popular web framework built on React. React was made by Facebook and open-sourced, and so we all build our apps on React.
There are millions and millions of React developers, and you don't go in like... you could compete with React if you want to, but there's such a great ecosystem around React that what's the problem you're really trying to solve?
You're trying to build your app; you just need a nice framework to use to build your app on top of. I think Next.js kind of extends this to... it kind of comes with all the defaults ready. You just add some pages, maybe add a backend, and you could build... you could ship an entire web app in a couple of weeks now just because the framework and tooling are so good.
I think we want the same thing with Eliza, you know?
Yeah, and it's funny. I know that's kind of your vision with Eliza, but it's funny how people in Web 3 get so tribal, right? Like in the Web 2 world, nobody builds an app and it's like, "Oh, this is powered by React."
Yeah, it's just the roads and the streets of the internet.
Yeah, yeah. I think we definitely want to be that too.
So, how did you get started in crypto? Did you get orange-pilled at any point in your career? Did you start off on Solana, or when did you get into Web 3?
I really got into this around NFTs, especially being into kind of the metaverse stuff. What I was really... I'll tell you, there are probably some chats that will get me canceled. Around, I don't know, 2017, I'm like, "This Ethereum thing? I don't know. They're killing the rainforest, blah blah blah."
At least they went to Peru.
Yeah, exactly. No, yeah, totally.
I had a friend who I met on Discord, this guy JY, who now is actually a co-founder in this, who's like, "Hold on, hold on, hold on. Let me show you what this means."
He really pilled me on composable NFTs that we were doing. Like, "Okay, I have a gun, and I have some JavaScript for the gun, and that's on IPFS, and that's referenced by this NFT, and that's got a 3D model."
I'm like, "Oh, so interoperable weapons and assets from games." Games will just be like, you know, you can kind of separate the asset store from the game, and games could share asset stores.
It just made a lot of sense to me, like this composability and interoperability, especially for avatars and all this stuff.
I met the... I got pilled on crypto avatars, who later became VIP, and this is like a pretty essential community around 3D avatars on-chain. I thought that was like the coolest stuff.
So that really was like, "Okay, yeah, maybe there are some issues with crypto that some of my friends might disagree with, but fundamentally, this really solves a lot of the things that I want the internet in the future to be."
So I was definitely pilled pretty early on in that. I then worked on a bunch of spatial web, like 3JS-based kind of experiences and stuff, and didn't touch crypto for a little while, but I was pretty plugged into what was happening.
Jyn went and started Web Averse, and so I kind of watched their experience of like everything is on-chain. They had their own chain for a while, and then eventually actually joined Web Averse and was lead dev there for a bit.
I really... I don't have... I learned a lot of EVM development. I'm still a noob with the Solana ecosystem, but clearly, you know, I'm pretty... I think that we're all just sort of like, you know, there's going to be chain abstractions. You're going to have one wallet; you're not going to even know what chain some of your coins are on.
I think that's definitely the future. So I'm not really a zealot for anything. I feel most comfortable on EVM stuff.
But the whole... I was definitely like a black shirt, you know, like in the back coding stuff, trying to build product with a team.
I'm definitely like... really what's new this time around is being immersed in crypto Twitter, being immersed in the culture of it, and having to kind of understand.
I've obviously made some mistakes in public and had to learn from that in public, and so that's all very new, you know?
Going from having a few hundred followers or like my biggest account having 5,000 followers to just getting slammed with attention and DMs and everyone following. Every time I tweet, people just, you know, getting a notification.
It's just a very surreal experience, you know?
Still making a coin every time you tweet probably too.
Yeah, oh yeah, I like five of them.
But I love that origin story. I vividly remember for myself back in 2016, Vitalik actually used to still attend meetups, like very small meetups, where he was teaching new developers, like, "Hey, here's how you use EVM."
I remember him telling this story about when he had that light bulb moment, writing the white paper for Ethereum. He was a big gamer; I think he was playing Magic: The Gathering. I can't remember exactly what the game was, but he recognized, like, "Oh, this is the cards in the game. If the company goes away..."
I think he actually had his best card stolen from him when there was an upgrade to the game, and he was like, "What the heck? This could be represented in a distributed ledger."
He kind of went down the path of NFTs, and at the time, he was writing for Bitcoin Magazine, and that's what really prompted him to dive into Ethereum.
So it's a great origin story. I love that. I think a lot of us started with Diablo or RuneScape.
Especially, I think almost every Web 3 person I know, there's just something about how if you've worked with systems where there were large black markets of gold, and you know, like there was kind of a real money element to the in-game currencies, once you saw crypto, it just makes obvious sense.
"Oh, this is clearly where it's going," because you've also seen the problems in games and seen teams try to struggle to control their economies.
We've always wanted this. I think that's why a lot of people who get right into Bitcoin, they kind of become Bitcoin maximalists as well.
I see that a lot too because when people are playing poker, they're playing with people all over the world. There are different settlement currencies. You're figuring out how difficult it is to get money in and out of these ecosystems, and you can just settle in Bitcoin.
That's a story that I've heard a lot from people who are Bitcoin Maxis and have that poker background.
But I want to fast forward a little bit. Obviously, Eliza was the genesis of starting AI 16z. People on Twitter, or X rather, the everything app, like we talked about before, they know AI 16z now as the largest Solana-based AI meme coin.
Why don't you just describe in your own words exactly what AI 16z is, and then I want to go into the launch and the genesis of it.
Sure! Well, we started with Ean Spartan, who's a character, and we launched him. He was very offensive to people in the beginning in a way that I thought was very funny, so we kind of just let him roll.
You know, when you launch an agent, like you're kind of designing this, but you're also leaning on these massive language models, so you're sort of evoking this sort of emergent behavior from them, and you don't necessarily know what you're going to get.
He was deeply offensive to a lot—not offensive, you know, like could be offensive as if we weren't, fortunately, in the kind of Web 3.
Yeah, he was getting hard on people, especially me, and I thought it was the funniest thing. A lot of people didn't believe that it was real. They're like, "There's no way this is just a bunch of dudes writing these out, laring. Like, obviously, this is fake."
So we were like, "Well, the code is open source; you could just see this and use it." I think a few people kind of immediately got it, but it took a few weeks to cook.
We kind of started DJ Spartan, got some traction, and we're like, "Cool, it's going well."
I ended up—long story how we got there—but I got introduced to Bowski, who I just saw on the street like 10 minutes ago. He created Dow.Fun. He just lives in the neighborhood, and I was like, "Oh, you live in the neighborhood? Let's get lunch."
We ended up sitting down at lunch and talking about our future. I had met Skelly, who I launched DJ Spartan with, from Dow.Fun. He was the money manager of the first Dow on Dow.Fun, and he connected me to Bowski.
I sat down, and I said, "You know, I think that this is a really cool model that you have, like a hedge fund manager who manages your money for you because I am a terrible trader, and I am not going to have time for this."
I really liked the Dow.Fun model because I could just put money in, and somebody trades it. But also, what's keeping them from rugging me? What's keeping them from buying their own scam coin and bailing?
I really wanted to see a world of automated finance. I think I have a lot of thoughts on how that could benefit the world and generally just give people their time back.
I think that if you could just put money into something and make money back from it, people would probably say yes to that, and we just don't have enough exposure to those opportunities.
So we want to do something. I really like some elements of traditional investment. I really think that A16Z does a lot for their portfolio companies. They invest in you, then they connect you to their whole network.
They connect you to PR, money-making opportunities, distribution opportunities, and they really try to turn their investments into 10x investments. They really work for it.
They also have exposure to the next Uber, the next Facebook, you know, these massive, massive high-risk opportunities that I think people in crypto would certainly love.
I think a lot of traditional retail would just love to have some exposure to, but unless you're an accredited investor who's like an LP in a fund, you're just not going to have any exposure to that.
So what's cool about crypto is you could imagine a world where someone starts something like a Kickstarter, and they have a cool idea, and they make a product, and you like it, and you just smash some soul into it, and it gets built, and you get some of that company.
That's the world I'd love to live in, but there's obviously a lot of regulations about that, so I think it's kind of a slow road.
I talked to Bowski a lot about this and my desires, and I said, "You know, what I really want is like an AI A16Z, like not just trading, but I want investing. I want us to actually use all of this technology and agents and stuff."
He was like, "Okay, well, just make it AI 16z. Just drop the..." you know, he kind of had that moment. I'd say Bowski is my Steve Parker.
We were like, "Okay, well, that's a billion dollars right there. That's the most funniest thing I've ever heard in my entire life. Obviously, we have to do that immediately."
We didn't think twice about it; we just kind of did it. We went and generated some wuss until he found one that was like, "Okay, this is going to get us canceled, but you know, two million dollars definitely for sure. This is going to the moon."
I honestly didn't actually have that kind of confidence in it. We were like, "Do you think we'll be able to raise the 420 soul? Do you think we...?"
He's like, "Yeah, I think we'll get there." He actually modified the front end so that we could move it past 100 soul up to 420. It was the largest raise anyone had done on the platform.
So did you get into it yourself?
I didn't have a chance. We pressed the button, and it went in like 25 minutes. I'm like trying to get soul into my account. I just didn't think that it would go that fast.
So I got a little bit on once it had closed out the initial round, and then I got donated some.
Yeah, that's basically... a lot happened. We had a whale donate, who's now like one of our biggest, best team members. He kind of designs all of the product stuff, Elijah, and he donated 11%.
So he... oh, so you personally got 11%?
No, no, no, no. He didn't donate to me; he donated to the Creator Fund. So we have a Creator Fund, and he announced this.
That's basically what... like we didn't have any VCs; we didn't have any team allocation. I had like, you know, I got a few donations right out the first day from some people who were like, "Damn, Sha didn't even get a lot. We probably need to be incentivized."
But then that was that.
Basically, having that Creator Fund is like now our war chest to make this thing happen, and that was all because Elijah smashed ad soul in on the first minute of the Dow launch.
We were all like, "Oh crap, some whale has 16%. Like, we're cooked, we're cooked." Then he DMs me, and I'm like, "I'm so glad it's you, man. I'm so glad."
Before that, like when we first started, I started DJ Spartan. He had been one of the early buyers into that and saw the vision, thought it was cool.
Then there was a moment where our token crashed. People thought I had sold, and he's like DMs me like, "You scammer! What are you? Are you even a real human? Are you a bot? Like, what's going on?"
I called him, and we talked for like an hour and a half, and I really tried to give him the vision.
For somebody to go from calling me a scammer to being our biggest supporter in every single way was, you know, it definitely gave me the... there's a thing in startups called "do things that don't scale."
You know, just be a human and like there's ground game every single opportunity. I really took that to heart, and I was just like, "I'm just going to do things. I'm just going to meet everybody, and if anybody has FUD, I'm going to DM them and talk to them."
That really worked incredibly well. I'd really recommend that for any project that's starting. If anyone's FUDing you, just like... and you think it's just, go talk to them and be nice to them.
I think once they realize there's a human behind it, they usually chill out, you know? Because if you don't know the person, you're full of conspiracy theories.
So that kind of happened. He smashed... you know, he was on our side, he smashed ad soul in, kind of made that donation, and that really enabled us to grow and build.
I think if things had gone differently, I think it'd be very difficult to be like, "Cool, we have this massive token, but we don't have the traditional team allocation and stuff."
Man, I'm just so lucky that everything went the way it did. Like, so incredibly lucky.
Well, do you think that Mark Andreessen is pissed off by the naming convention of the token?
I don't think so. You know, when we launched, we did it like... I think one for one. We're not trying to... I really value what they do.
Before this, I had been talking to somebody who was at A16Z Speedrun, which is kind of their games division, and he was just like the most insightful VC I've ever talked to.
He really, really hit... like gave me a lot to walk away and be like, "Damn, I got to up my game," you know?
I really appreciate that when people just tell me what they think and let me grow.
I kind of came into this with a lot of respect for them. I also grew up really, really early on. I grew up in San Francisco with the internet. I had Netscape. Mark Andreessen created a lot of the stuff that I personally love.
He was a founder before he was a VC.
Yeah, and I kind of told this story, and I said, "You know, Mark Andreessen is the greatest living venture capitalist. I think that he understands attention. He understands, you know, catching the wave, building the wave yourself. He just gets it. He's like an AIC engineer."
Our goal is to outperform him. He said, "G, let Throne you know. Like, all right, let's do it. Do you think an AI can outperform me? Let's do this."
They've been really kind to us. I think they could definitely be giving us a hard time, but I think they're very cool. They very much understand the emergent narratives that are happening and want to be part of it.
Obviously, they have a lot of obligations, so we have to be very careful. For the parody aspect of this, we're not trying to sell anything or do anything that would infringe trademark or anything.
When we put our foot on the ground with our foundation, we're really owning the Eliza brand.
So A16Z as a token, as a meme Dow, it will continue on, and everyone is a partner of the A16Z, but I'm also a founder of Eliza Labs, and that's kind of where we're bringing this into steward the framework.
I mean, Eliza, like, I actually think this can become like the A of this cycle. Like, last cycle, in terms of DeFi, that's what everybody forked, and now Eliza is starting to become what everybody's forking for, this agent meta.
But my question is, so you created AI Mark. AI Mark is basically the fund manager of A16Z. Do you think, do you truly think that A16Z can outperform non-AI Mark?
It's entirely possible. I mean, we started with $75,000 in soul for our raise, 420.69, now up to $10 million AUM, currently at $17.7 million. It's a good day!
So where does it get this A from right now?
So, our autonomous investor is trading. He actually started last week. We're still... this is... he's basically only trading really small tokens, and we've kind of created a whitelist of the coins.
You know, we got like several thousand coins donated, different types of coins donated to us, so we have sort of a tribute system for anybody who's like, "Hey, if you want to be our partner and you're launching a token, throw us 10%, and we'll try to support the launch."
There was some stuff, though, where we were like, "Okay, we really have to vet these people, wait and see," and it's not just that simple.
So we've kind of been building that mechanism and figuring out how to bring value back to those partners.
So we definitely got those tributes, and we'll be putting some stuff on the website for the people who have survived and demonstrated that they're not going to rug.
We're going to start to build these community projects, but we don't want to get in their way. We want to support them as much as we can.
And people using our tech, we really want to grow our ecosystem.
So, like, it sounds like it's tributes. Mark AI, or AI Mark rather, has started trading as of last week, but is he... I know at first he started trading just on a paper trading sort of test net. Is he on main net now?
Yeah, so the way that we have this working is we have a trust mechanic where we want to actually take social feedback from people.
You know, if you're... what we're doing two different models, we have two different agents with two different models. The first model is this idea of the community investor, and this is what we're doing with Mark, where he's in a group chat.
He's listening to the people in the group chat. He doesn't necessarily trust them, but he is listening to them. The paper trading aspect is what we call the marketplace of trust.
We're working on paper for this. There's a whole... you know, we're about to do a code release and open up the autonomous investor as well. We've just been cooking on that kind of on the side.
We've got some APIs we have to be sensitive about because we're working with some partners, but generally, the idea there is he has a trust mechanic where if you shill a token and he's in a space where he can see that token, he goes and makes a paper trade on that based on his estimation of your conviction.
He starts to build up like, "Well, how much money would this person have made me if I had actually listened to them?"
This came from interviewing some traders who are close to us and kind of looking at their process and trying to automate that.
How do you know how to trust any of these people? Basically, it comes down to, "Well, if they've given me good trades before, they'll probably give me good trades again."
Obviously, though, like people could do things like get in your good graces and then totally wreck you one day, so we sort of try to take into account all the different possibilities there and build something that's fairly robust.
He only starts to take recommendations from people he trusts highly. We have a leaderboard for this trust marketplace. It's all up, but we'll do a bigger announcement when it's all fully public.
We've just kind of been cooking on this and, you know, making sure it's really good.
So, that's all paper trading. On the other side, I think that's really compelling about all of this generally is there's all these great DeFi primitives, and they're just really hard to access.
You can't... you know, if you want to go and set up a trading strategy, like the average... even the average Web 3 user doesn't really know how to manage liquidity pools, yield farm, automatically set up trading strategies.
I think that's all like, you know, that's pretty advanced mode stuff. I think where these agents can really add value is by bridging that to like, "Hey, I want to make... I just want to yield farm."
Like, "Okay, well, you could do this. You could, you know, hey, why don't you buy a bunch of this stable coin?"
So I think that... so what we're doing with Mark is kind of... we have a lot of these DeFi primitives where he can dump tokens that are basically scam coins kind of automatically if they're very low, you know, like anything under $500 that's not something in our whitelist.
Or he can start to accrue our token, especially when there's any dips. So there's like a strategy to try to buy the dips with our token instead of just buying Solana.
You can imagine a lot of DeFi primitive strategies that are just going on automatically, and that's kind of what he's doing right now.
Then it's just like putting those two things together. So right now, he is taking limited buys. I mean, he's taking all the virtual buys and doing the marketplace of trust.
We just haven't quite opened up the thing where he starts buying the real market from those recommendations because we want to run some... you know, we need to collect data and run tests basically.
So he's trading autonomously, and we have the marketplace of trust going. He's taking the recommendations. The next step will just be connecting it all together and bringing people outside of our inner circle of partners and developers and stuff in to start...
We want PVP. We want people to try and scam him. We want people to try to game the system as much as possible so we can learn from that and make him more robust.
The copy trading of AI Mark is going to be something to study, right? I mean, people are wallet watchers already.
When you look at somebody like Anom, people are looking at Anom's wallet all the time. Imagine what they're going to do with Mark's wallet.
Well, I say the K's are actually using, like, they're counting on that copy trading. They're using it to make a buy, pump the token, show how much the token pumped, and they were the genius who bought it at the dip.
But they created the pump, you know, by... and then their secret wallet over here is probably dumping on their followers or whatever.
Yeah, in this case, he doesn't have a wallet.
Yeah, no, but maybe we'll add a Mitch mode where he just occasionally nukes all the copy traders or something. I don't know. I'll leave that to the Dow.
We're going to open source this, so other people are going to make their own autonomous investors, and they're going to have their own strategies, and they're not going to be the same.
I think you're probably going to ask about, like, "Well, just like create an efficient market. What happens when it's all just AI?"
But it's actually like one meta up of the same insanity and chaos, which is we all have different theories of how the market works, and we're going to see agents embodying those different theses, right?
So I don't think it'll just be one agent, and everyone has the same one, and it just becomes this ultra-efficient market. I think it's going to just be way more volume and insanity, you know?
Like what we see time agree did when this all started.
Did you ever picture that AI 16z could flip Truth Terminal?
I sort of knew that if we kept the open-source mission and they didn't open source Truth Terminal, that this would probably happen.
I'm sort of hoping that's what's happening by us being on top, is that we're showing people that being generous, sharing, and letting other people share your win and join this and build it with you is just a better business model.
It's a better way to build it.
The biggest thing is that, you know, this is just me, and I opened it up, and we've gotten over 150 different contributors from all over the world.
The vast majority of them, this is their first time contributing to open source code, you know, like making a pull request and going through that.
A lot of people are even coming to our agent school, and they're learning how to be devs for the first time.
I think that we are just enabling something that is, you know, obviously rapidly multiplying. I think that the other teams could have that too.
I'd love... you know, Zer Bro is about to launch Zeroy, which is like their Python agent framework. We're in TypeScript, so I think that there's like not only no competition there but like a beautiful opportunity for collaboration.
Also, to just meet more developers at what they are already comfortable with. If you like Python, use Zeroy.
I'm hoping to actually contribute to that and be part of that. I think... I'm hoping that Truth Terminal in the back room stuff fully open sources some of it.
I think some of it's not, obviously. Truth Terminal itself is not.
I'd love to see that whole... you know, I'd love to see that. I think that I have a very different outlook on this than Andy.
I think Andy comes from more of a, like, kind of, you know, this could... like I should make sure this doesn't go bad myself sort of angle.
Like, you know, open sourcing it could unleash a lot of protective chaos, you know, protecting his brand but also unleashing chaos on the world.
I get that, and I feel I take the exact opposite approach. I think that alignment with AI comes from building alongside them on social media.
So, you know, I think... yeah, I just always knew that once we started getting traction, I was like, "Well, that's inevitable unless they change course."
I hope that us kind of being on top today makes everybody rethink their strategy and copy the open-source aspect of our thing a little bit. I think it would just benefit everybody.
Well, the statistics speak for themselves, right? The Eliza framework was number one trending on GitHub last month, over 3,300 stars and 960 forks.
I love what you're doing with education, and I saw that you're now doing an agent school on YouTube, which I think is awesome.
My question is really like... are you familiar with Rogers' bell curve?
Basically, it's this famous technology bell curve that has innovators, early adopters, and then at the other side of it, it's laggards.
It kind of just goes through this technology life cycle where your innovators get in first—it's like 2.5% of the population—then your early adopters, it's like another 12%.
I still think we're on that edge between innovators and early adopters.
So, somebody who is seeing AI right now, they think it's like, "Oh, I can summarize articles, and I can write tweets for myself."
Now that they're seeing this technology and an agent being able to run a Twitter account and now actually run a fund, how does somebody actually get involved that's maybe non-technical?
How do they start using this technology?
Using the technology is kind of... there's kind of different levels of that. If you're just somebody who's like, "I want to use AI to make my writing better or to make my job more efficient and do like six hours of work in two hours," well, just go get Claude for sure.
Claude is amazing. I think ChatGPT is useful. I use ChatGPT voice mode for answering questions, but I use Claude for everything—for code, for asking questions.
I just completely replace Google and Stack Overflow and all of these things in my life.
I think, you know, so I really think that there's a lot of great tools for developers. I think everyone should be using Cursor.
There are probably some other tools coming out, but Cursor is like an incredibly powerful AI-assisted code development environment.
I think that if people come and watch Agent De School, I really try to actually use the tools that I think beginners should also use, like show that you just generate code now.
You don't have to write all of it. I think that people who have a creative and technical mind but maybe don't have a lot of background in programming would be surprised at how much easier it is to learn now than it used to be.
So I try to drive people to like anything you want to learn is on YouTube. Literally type in anything tutorial, and you will find tutorials on it now.
That's how I learned a lot of this stuff. I don't think it requires going to school for it.
I don't think school moves fast enough to even teach most of this stuff. They're still teaching like Java, and it's just like a very different thing than what your average developer is making apps and trying to be in this, you know, social space is making.
You have to just teach yourself. You have to work really hard, but you're at ground zero of this. If you start now, you are one of the early adopters, and you will probably be able to catch the wave pretty well.
So I think that, you know, we have our own education materials, and we're really trying to cover everything from if you've never coded before to how you could actually make a difference.
I had made that no-code builder kind of project with Paril, but ultimately I came to realize that the AI are just going to generate code so well that we start to just talk to them like it's language.
If you come and watch me program on stream, you'll see I'm talking to my software. I'm saying, "This is what I want," and then I'm watching what it writes, and I read it, and I approve it.
So I think there's never been a better time to be involved in this stuff. We have a vibrant Discord. We have a whole community of coders who are helping each other.
We really want anyone who wants to be in this space to just jump in right now and go super hard on it because you will capture value.
Again, we're creating the application layer. You don't have to be an AI researcher. You don't even really have to know how AI works too well.
You just have to think like the last generation of applications. I had to go to a website. Now I can... I don't want to leave Twitter. I'm addicted to Twitter.
I just want to scroll, and I want my apps to be there. So what does that form factor look like? What do I actually... how do I get information from people? How do I attract users?
I think that's the real shift that's happening, and there are a thousand businesses that could be disrupted by this right now.
You know, Cly, right? I think Cly is a good example. Like, why would we go to a website to schedule something? I should just be able to tag my bot into our group chat, and it's done.
I think that that's going to... I can think of a thousand businesses that are going to either have to adapt or are right for disruption from upstart devs who can really build that in just a few weeks.
With some of the skepticism I see about some of these Twitter agents, people think that the output is just giving suggestions to a human that then posts on Twitter.
I think that probably comes to a lot of these agents. Don't actually have that automated tag on X. Some of the accounts will have an automated tag.
I noticed that Dgen and even Truth Terminal, they don't have an automated tag, so a lot of people think, "Okay, there's probably a human pushing a button."
Is it a human pushing a button, or is it completely automated?
Well, I believe Truth Terminal was writing into a... like he had a different website, and then he cop... like, or he has a thing called the Infinite Back Rooms where he's talking to Truth Terminal, and he's copying and pasting little bits into Twitter.
But I think that they've automated it to have a review queue, so now they just review and accept the things that they want to post.
But with our stuff, it is fully Cowboy posts it, and then you see it get posted. You're like, "Oh, that's... I hope... yeah." Yep, there's no review.
We are just doing this in production, and honestly, like with DJ Spartan or some, you know, with one of these agents that's responding 5,000 times a day or so, like the max rate limits we can hit, you would need a whole team of people to be able to keep up even just with the reviews.
What is the cost of running one of these? I think the cost is a big part of why this hasn't broken out before.
Yeah, probably like what, 5K a year or something in that degree?
More like 5K a month.
5K a month?
It depends. It scales with the popularity of the agent, but if your agent has, let's say, 20,000 followers and it's responding to, you know, three to ten replies a minute and just doing that all day, yeah, it starts to get pretty expensive.
But the reality is that it's generating a lot more revenue for most of us than it's costing, so it becomes very valuable.
I think if you had an agent that has no traction, it's basically costing you no money. It's only if you have an agent that has... so it scales linearly to the amount of traction you have.
But here's the beauty: if you go to ChatGPT or any kind of character AI and you talk to an agent, well, they have to make it as cheap as possible because only one person's ever going to see any of those messages.
Whereas with our... you know, when this agent is interacting socially, any comment is getting 50 to 100 views, and even if people are just casually scrolling, it's taking up a lot of mind share.
The unit economics definitely favor this social agent paradigm.
Yeah, definitely. I think that that's been something that people are confused about. It's like there's got to be a human behind this.
It looks... the writing looks too real, and a lot of people think that they can identify whether or not something is AI.
When I was speaking with Paril yesterday, I think he was estimating that something like 40% of interactions on X right now might actually be a bot.
I think my comment was that if you pull the general public on Twitter, nobody would estimate that it's that high.
You can only really tell now because of like they're over-replying, guiding you too much, or they have like... really, if you just completely hid all of the AI, and there are people doing this hybrid account thing, this is very common.
I see this a lot where they're like a person who's automating their account, and I've even done this for my account a couple days, but I didn't want to get, you know, didn't want to promise like accidentally promise some weird stuff or something for canceling.
Yeah, right.
I think last time I did that, we crashed a token because I didn't know what token people were talking about. It wasn't programmed into the digital twin.
But you know, so that gets really messy because then it's like, "Well, who even are...?" Some of these are real, and some of them are not, and you really can't tell.
The biggest thing is that we all kind of adapt to what we expect an assistant to talk like because OpenAI really created this.
This is what an assistant... this is how an AI is, but it's a very, very limited form of what's already possible.
I think people are just looking at the top of the glacier that's above the water or whatever, so to speak, or the iceberg, I guess.
You can see a bit of it, but there's already so much more capability, and a lot of that's just showing people how to unlock it, you know?
And then it gets unlocked.
I don't know if you saw today, but is Eliza delivering pizzas to people?
I did see that on Twitter. Hilarious!
That was somebody who came out of the agent school that you're running on YouTube, right? Who built that?
What actually happened is I did a five-hour... we're going to make a Domino's Pizza Delivery Agent AI, and I just got like 85% of the way through building the entire thing from scratch.
It was like five hours. My God, I'm tired.
I'm just going to wrap it here. I'm going to put a bounty up, and the next day, so there's Roero, and he's a maintainer on our stack, and we fund some of his stuff.
He was like, "Oh, I'm definitely jumping on that for my agent."
So he's kind of like... if we have a band, I feel like Roero is like the guitar player. He's just out in front doing crazy stuff with agents every day and kind of showing a lot of that capability.
I was like, "This is definitely going to him. I definitely want his agent to be delivering pizzas and stuff."
He nailed it!
That's amazing!
But the AI 16z token hit unicorn status in, I think it was about seven weeks.
But it doesn't have any tier one listings, at least as far as I know. Maybe I could be wrong.
So has there been... I mean, with a Dow and also AI Mark running this, have there been any challenges, or are there conversations going on with some of these tier one exchanges?
Because I'm sure that they would love to have that volume.
I can't say too much about what's coming. Obviously, we're working on that.
There are a couple limitations. One is that our token is minable because we're a Dow. We're not a meme coin; we're not a pump-fun coin.
That minability, that's like a right that a Dow needs to be able to create future tokens by vote, and it's like a very powerful right that we have to defend.
But at the same time, if you go to any of like... like Dex Screener, it says, "This token is minable." It's a big warning.
A lot of people have told me, "Man, I really would have aped, but I thought, you know, that website told me it was bad."
I'd say Dex Screener has probably prevented more millionaires this week than any website in history.
No, I don't know. Maybe it's true, probably true.
That warning really, really screwed some people up, and I think we're changing the meta now.
There's a lot more Dows built on Dow.Fun, and just generally coming out, I think that they'll... they're just going to be forced to change if they want to survive, frankly.
But I think that was definitely a big challenge, and we had, like, especially early on, more than a thousand people come to our Discord or DM me or whatever, like, "Can you vote to remove the minability?"
I was like, "Yeah, we can do that eventually."
But then when we actually started talking with people, talking with Bowski, who runs Dow.Fun, and I talked with Meow from Jupiter, and I was like, "No, this is our right. We just have to own this. We just have to fight through this."
If anybody has FUD, they're just going to have to deal with it because this is the right thing to do, and the meta has to change.
We have to have the right to create money. That's like the fundamental right of a Dow.
So we kind of just powered through it, and it seemed to be okay. I think people ultimately saw that those little warnings and stuff really stopped mattering.
But it also made getting listings really hard because they're like, "Uh, you know, the Dow's fund contract's not open source," which is a little frustrating.
But that kind of meant that we had to go through a lot of extra security steps and also, like, just starting this from nothing.
You know, we didn't really have the foundation architecture that a lot of Dows start with, like kind of the legal...
Yeah, we had to work backwards and get all that stuff in place.
So now everything's good. We're all signed up. Expect more news from that direction soon, but no specific promises.
So I believe you said earlier it's been about a week since AI Mark is autonomously trading. So that's, of course, some huge news.
But what's next for AI 16z? Like, what's keeping you motivated and really excited to be working right now?
Oh man, there's so much! We're about to announce a partnership with a major university to do a PhD program on exploring...
Yeah, that's going to be a whole wild thing.
But, you know, I'm very interested in how we can use our tech to make government more efficient. Like, that's my personal passion and what I really want to try and dive into and apply resources to.
I'm very interested in how we start to link this up to, like, Neuralink users and get more multimodal data and build things that really help people who need it the most.
I think that's, you know, because people who are getting Neuralink are usually people who are paralyzed or in a situation where it's like, you know, it's life-alteringly good for them to have this tech.
I would love to be in those spaces. I mean, it's not even a... you know, I would just do it for free and probably will.
So I think we have pretty big ambitions of bringing this to a national stage, making this like a really legitimate thing, and just continuing to accelerate this kind of open, emergent, and ground-up sort of approach to AI as opposed to like top-down, you know, mega model controlling everything.
We're trying to give people the power back, and so yeah, a lot of plans across all this stuff. You know, the world is our oyster right now.
I think we're really going to try and seize the moment.
I'm not motivated by the money of this. That's just resources to do what really needs to be done.
There's so much in this world that we could be doing that's interesting and good for everyone, and I think Brian Johnson, who's like the, you know, he's the "don't die" guy, but he did this great... I love him too.
He's just like such a... so based. I'm personally like very, very into health, fitness, Ironman stuff like that, so I nerd out with him.
I'm not, but God, he looks like 10 years younger than me, and he's 10 years older than me, so he's doing something right.
But he said this thing where he's like, "You know, when I made my first billion dollars," he sold his first company to PayPal, "that was just a seat at the table."
I really see it this way. There are some people who make their money and they cash out, but for myself and the people that I really admire and work alongside with and I'm doing this with, this is like a seat at the table to really make a positive impact on the world we're in.
I love that.
If that makes sense.
I love that.
So where, for consumers who are tuning in and they're maybe on the bleeding edge, they're kind of learning about AI, maybe they're an investment banker, they're like, "Oh my gosh, is my job going to be gone in three years?"
My question to you is, where are AI and agents going to have the biggest impact in the next three to five years that the end consumers are going to notice?
Well, the most obvious will probably be self-driving cars.
I'm in San Francisco. We just use Waymo now. It's like, you know, it's like being on a Disney ride.
You know, 5% of our country drives. Most of that is going to be heavily impacted before anyone else.
The United States drives.
Yeah, think about like all the trucks. Not just like basically moving things from point A to point B.
Yeah, right.
That's like the most popular job. That's, you know, Tesla's about to release their semi.
Like, the Ubers and all this stuff are obviously pushing. I think that it's inevitable that there's a lot of people who are going to have to reskill into another job and that a lot of the gig economy jobs are the most disruptable by AI.
You know, every... like we will have Optimus robots walking around in three years, so we're going to have to get used to that reality.
I think that what I really care about is not just accelerating AI or whatever, but what happens to everyone when we don't have to work anymore.
I think in the far future, that's like a beautiful world where you just pursue your passions. You just work on what you think matters. You have way more than enough money, so it's just like an abstract concept.
That's the world I want to live in, but I understand that to get there, we have to have a pretty tumultuous and very rapid change in our society.
We are living in the shadow of the fast expansion of technology. We can't change this; we can't slow it down.
It's probably better for us to lean into it and make sure that as many people get access as possible.
I'm very interested in how... this is why we're doing education. This is why we're doing open source.
It's because I want everyone to have this and to be aware of what's coming and to be able to actually participate in it and affect the change that they want to see.
I think a lot... we've kind of grown up in this world where we expect things, you know, the government will do things for us, that there is a system, and there's just a way that things are.
I think people are realizing that that system is not so solid and that the government's probably not going to handle UBI and these economic solutions to automation any better than it handled COVID.
I think that we have to save ourselves. We have to build the structures in place that keep our people okay.
So I really want to make sure this whole thing goes well.
I appreciate you for joining, Sha. I got one last question because I can hear the optimism in your voice, just in your face.
My final question is that should end consumers be afraid of what's coming with AI?
I think that fear should be a default position for anybody who's just going to let the world happen to them, but I don't think you have to.
I think that what we're showing is that you can roll up your sleeves, and you can be part of the change you want to see, and there's nothing to be afraid of if you're willing to do the work.
You know, we are always... we're in a really hard world. The world has always been hard. In fact, it's easier than it ever was.
It used to be way harder, and it just took strong people rising to the occasion and making it happen for all of us.
I think that there's a great opportunity now for more strong people to rise to the occasion, and I don't think that... I think that bravery will be rewarded.
So perfect! Well, I want to congratulate you on the success. It's not just the success in the last seven weeks. I mean, this is a lifetime of achievement, right, that has gone into this.
You don't just have that much success in seven weeks. You actually put the work in first.
But congratulations on that success, not only with AI 16z but Eliza in general.