0:00

Welcome to the Chain Stories podcast, the podcast that celebrates disruptors who defy convention. Here we dive into the bold stories of trailblazers who turned audacious ideas into billion dollar ventures.

Audio Only - All Participants: 0:18

Hey, everyone. Welcome to the ChainStories Podcast. Today, I'm very delighted to have a very special guest. He is the star in the room, Jansen from Virtuals.io. Jansen, you've been quite known, Virtuals has been taking off like a rocket ship. And for our listeners that have been very curious about these episodes, before you created Virtuals, I know you're involved in a gaming project and gaming DAO. Can you tell us a bit more about this experience and how did this help shape your vision for Virtuals? Yeah. No, actually the story started even way before that. I was in uni, that was actually my first exposure to Bitcoin and Ethereum. The first thing that we all tend to do as students and as hustlers, you will Google how to make money online. It's probably the most Googled search statement by every hustler out there, right? Probably like down the page, they'll be like, Oh, mine Bitcoin, right? So then you go down that rabbit hole you realize, oh, Bitcoin mining back that's in 2016 it's not that easy anymore. You need ASIC miners. And then you start digging around and then you realize, oh, there's this whole new thing, right? Ethereum programmable blockchain, right? And then be like, from a tech standpoint sounds cooler than just pure cryptocurrency. So that was my first exposure, I was actually mining, Ethereum off my Alienware laptop. You still could do that. And you know, usually electricity is free, you get a free$20 every like couple of days. So that was the initial journey. But after a while I was like, you know what? School mining, if I really believe in it, why don't I put some of this like spare cash that we have on the sidelines into Ethereum. And that's where it started. That's where the journey started. Wrote up the 2017 bull run, didn't sell a dime, held down through the entire cycle, graduated from uni, went to management consulting. And throughout I was testing a couple of ideas along the side. So okay, I forgot about crypto. Started a deep tech venture, in uni about creating an alternative source of water. And this was some very complicated stuff. Won a ton of grants in the UK, did a couple of mistakes. Company went to the ground. Then started while I was in management consulting, started a digital marketing firm, realized that it's hyper profitable, but it couldn't scale. So it didn't fit the appetite of you want a hockey stick kind of growth. Closed it down, started a Big data and AI company to change property search in the property recommendation. And this was way before this is 2018, before all the GPT hype. Was too naive in the whole fundraising game. So we thought if we test the idea with$5,000 to$10,000. We will know whether it will fly or not. So we saw a growth that was linear. We didn't see the hockey stick as well. Then we say, you know what? Maybe it's time to just move on. So I moved on, honestly, I think we should have just raised capital and really pushed on the idea forward. So did a couple of things. Nothing really took off mainly because: one, we were pretty much part time, one leg was in a comfy Wall Street job and the other leg was like trying to test out ideas, right? Like a hobbyist over the weekend. So I think that was one flaw. That you can dive into it. But then, in 2021, the bull market took off in a crypto site and it got our attention. Me and my partners, we were like, you know what? Let's focus a bit right on this whole trading degen in the trenches stuff and we did quite well. So we did very well on the gaming front we caught a lot of the gaming narratives even before they became famous because naturally we're gamers ourselves as well. So we understood a bit of the proposition of owning ownership of items in games because we were participating in shadow economies of EVE Online, World of Warcraft and stuff like that. So we knew that value proposition. So we caught a lot of that run up, said fuck you to our bosses, we left our jobs and then we could then go full time to building something. So then there was a question of we were thinking like, hey, do we spend our day skiing, or do we try to build something? And back then we were like 26. So we think okay, you know what? We should do something with our lives. So then we say, okay, what's our edge, right? We know crypto quite well. We've a couple of years in the space, know gaming very well. And we had a ton of gaming assets, probably like you guys, right? You guys were there in Decentraland and Sandbox. We were there on the Axie, the Illuvium, the Gala side of things. So we're thinking, okay, let's use this asset, let's start a gaming DAO. And the idea behind this gaming DAO it's, we wanted to be capital allocators and we wanted to bring a critical mass of gamers to game. So that was the initial idea. So we actually raised funds for this. The day that we raised, it marked the Pico top of the bull market. We woke up that morning for the day of launching our project. Bitcoin went down 25%, literally. It's just one day we were very close to calling off the race because you'd be like, is this the right timing? But at one day, one of our close advisors walked through the room with his Starbucks in his hand and he told us, like guys, sometimes you just got to make the hard decision. You might not feel it's the right thing, but sometimes it is, right? And he said if you don't raise now, you will never be able to raise for this idea because it's likely going to be just downhill from there. So we took that advice and luckily we did that race. It gave us enough capital to start. So it was a$15 million raise, at the pretty much Pico top of Bitcoin. But then the year started going downhill. You had the whole 3AC, you had FTX, everything started blowing up, right? And then we realized that if you are doing investments at an arm's length approach, it doesn't make sense because we cannot control risk. A lot of the founders that we're investing in got jaded. They were leaving the space of you know what, Web3 is a scam. It's a scam. Let's go back to Web2, right? And the funds they invested in them just evaporates because, there's no recourse when you're not investing in equity dues, right? And when you're investing in token dues. So it got us thinking towards the end of the year, we'd be like, you know what? If you truly believe in a space and we really want to build something, let's not do it as investors, let's do it as builders. So we started a Ventures, we pivoted slightly to a Venture Studio model instead. And we still believe heavily in the intersection of crypto consumer and gaming entertainment. So we started going down that path. We hired co founders in rocket internet model. And we said, okay, if this idea let's pursue, let's test. We feel it fast and then we keep iterating. So then this came up to a point where GPT came about, and then there was this very insane paper written by Joon Sung Park from Stanford about what if AIs can have agency? And it got us thinking like, Hey, shit, if you can create autonomous agents, you can transform how gaming is done, how influencers are running the entertainment space and whatnot. So we started incubating at the intersection. We started a AI influencer on TikTok. Web2 audience, that blew up to a few hundred thousand subscribers over a couple of weeks. We were the first guys who built autonomous agents on Roblox. We had some papers that were quoted by DeepMind as well. AK Khaled or some of these like Giga influencers on the AI scene as well. That was where we were playing as a venture studio. And towards the end of 2023, we started realizing that these autonomous agents in gaming entertainment were making money like this AI influencer, this full AI influencer, right? There was no human involved. She was raking in, I think about five to$10,000 a month. So it's not a small amount it's not a large amount but it was enough to get us thinking that hey shit these are productive assets and as a crypto bro you realize that if any asset is productive you can actually tokenize them. And then we realized like, actually then this is a very elegant solution, right? If you bring crypto blockchain with autonomous agents together, what can that look like? So that became the inspiration behind Virtuals in 2024. And that's where we did a full pivot, we told our previous stakeholders this is the new white paper. Do you believe in it? 10 percent said, no, dumb idea. Dissolve the DAO, return the funds back. Then 90 percent said, you know what, we believe in you guys. Let's go forward. And so what we did was we bought back the 10 percent who didn't believe above market rate. And then we said, you know what, let's just go down this path. So I think the beauty about what we had was throughout the entire bear market. We did maintain the treasury. We did grow it from the investment site. And because of that it gave us a bit of firepower so we could hire some good devs, some good AI guys to stuff up the team. And yeah, we went down this route. Initially it was focused on the infra and then later we realized infra alone doesn't grab attention. So we lean a bit more towards the more speculative tokenization element. And when we launched that platform two and a half months ago.. things were actually initially slow. Like people said, okay, cool. You can tokenize agents. So what? And back then we just tokenized the AI influencer that was running on TikTok. It's okay, let's try it. Let's tokenize her, right? She reached to a 5 million market cap on the first day. Okay, cool. Good enough. But then we realized like that is not gonna, it's not gonna blow up, right? Like people will not see this. So in building is about timing and luck as well, because it was exactly at the same time when Godzius Maximus, the Truth Terminal was catching some hype. Mark Anderson from A16Z was talking about it. Token went up, right? And that was fine, nothing happened to us when the token was going up. But what was great was then there was a typo. Truth Terminal made a typo and then people started funding on the timeline, right? Oh, there's a human behind Truth Terminal. Honest to God, we don't know whether it's actually a real typo or actually an AI did that typo, but it gave us a venture into the market because we were already building autonomous agents in gaming. We already had this AI influencer on TikTok. So we thought Hey shit, why don't we just put that together? We showed to the crypto Twitter, that there could be a truly autonomous agent, and we showed this website that showed her entire reasoning structure. How she plans, how she moves towards a goal, how she iterates towards a goal. And we showed that to the world. And I think that blew everyone's mind, right? People were like, oh shit, wait, you can be truly autonomous. So that was the first week of launching that, it caught a ton of attention, CryptoTutor got very excited. The next week we immediately showed that, hey, if this agent could control a wallet, and it did the first transaction, Luna did the first on chain transaction autonomously, it could start influencing humans and other agents alike. And that blew up again, right? So then it attracted a ton of builders to come and build initially within the ecosystem because they could envision, shit, if an autonomous agent could do this, what more can it do? So we started attracting a ton of builders and yeah, momentum started picking up and that's where you started seeing us as where we are today. It evolved from a venture studio to a launchpad to now a vision of we're thinking like, the way we treat Virtuals Protocol is actually more towards building it as a society and as a network state rather than an L1, rather than a platform. A lot of that was based on luck and you can see, right? It was timing, right? Sometimes. And I think there's a lot of times relative to space, right? What you can do as a founder is just keep trying and you just create a higher probability of success when the right timing strikes. So I think we got lucky in that. I think for our listeners that are familiar with some numbers, Virtuals currently sits at 11,000+ AI agents. 140k token holders, 35 million in fees and the market cap of Virtuals is about 3.5 billion. Was this something Jansen that you're not expecting at all to happen and happen just overnight. Honestly God no. Sometimes we still see in disbelief because, we had a vision. Like we, we thought okay. Yeah. People would like autonomous agents, but we never expect the amount of economic flow and attention to be this large. And in fact, even as of today the team is not able to serve the amount of demand that is coming in. It's like we are rapidly hiring every single day, but the load on the team is just insane because everyone just wants to have a piece, right? Of this very exciting environment. Which is good. It's a good problem to have. What do you feel as a founder, it can better prepare you for moments like this, that out of nowhere, everything can change and explodes. I don't think any founder is prepared for that. Okay, interestingly, I think every founder have this question like, what is PMF? What does product fit mean? And it was a question that I was googling heavily since the first day I started a business like, eight years ago, right? Then whenever you Google it, everyone would say the same answer. They'll say if you find PMF, you will just know, right? And then you'd be like, what do they mean? What the fuck does just no mean? And I think that's the truth, right? Like you would try, you just keep trying. And then suddenly it just goes vertical. And the reality is you would never prepare yourself for that because a lot of times when you are a founder, you also want to budget well. You do not know how long you need to iterate and keep pivoting until you find the right product market fit. So you're very conscious about your burn rate. You wouldn't always like your team, right? So you'll never be ready. Honestly, you'll never be ready. It's just that you pray that the team that you started off with, has been refined through fire so that same team can take that verticalized speed, right? It's like a rocket that you're trying to build. You pray that you build a very strong rocket so that when it actually takes off, it doesn't break apart. You mentioned pivoting, during that bear market things get hard. And was there a lot of times that you went back to your team and you're like, guys, let's pivot. Let's change something, we need some traction, we need some revenue. How many times did this happen? Or when did you know when was the right moment to pivot? It was a tough question, bro. In fact, actually after our first year when we started, the first few months was okay because you're well capitalized, but after a while you start realizing that the initial hypothesis has been disproven. Oh, being a capital locator in a gaming space was a good business model. It was immediately disproven a couple of months in. And then you try to find a footing. Our lowest point was actually at mid 2023. It was the depths of the bear market. It was pretty much no volumes anywhere, right? Anything you try to do with crypto, there's just no users at all. I was asking my co founders this exact question. Are we sure that crypto is the right place to build as an industry, right? Or should we pivot out, and go to a space that we think that it could be a bit more traction, narrative, because sometimes it's about, you want to be in the right space because you attract the best talent and you attract users faster. So that was a big question. And it was actually at a point where our treasury was at its lowest. There was a point where our treasury was about like at 5 million and it was a deepth of that bear market. And we even had a question of should we just, call it quits? We just tell our investors like, okay, guys, we tried our best. We think it's just, if you don't see a way forward, should we just say, let's call it a day. At least we can return back some capital. And then, we as founders and a team, we try to figure out a proper narrative and then you raise off that new narrative again. So we had a conversation. I got my whole team in and I still remember it was this room. It was like a six seater room, but it was packed like 12 of our top guys in the room. And honestly, most of them also didn't have faith in the industry. They'd be like, okay, you know what? No idea. Don't know what's going to happen. So much uncertainty. And then I think there was this moment where we had a dinner, a company dinner together, like probably two weeks after the initial discussion, everyone was quite like jaded, uncertain. But I think the statement that I put out there to the folks was like, you know what, we actually had a great team. We had the attitude of liking to build cool things together. And I think the idea was then why don't we find something cool together to build so we can energize everyone. As long as you're building something cool, you will find enjoyment in building it. So that's why we also started to move away from that venture studio model because a venture studio model was a bit too fleeting. There was no, grounding, like single idea or single goal for the entire team to rally towards because they were just like testing different things and then you try, fail, try and fail. So I think we pivoted away to just like focus. We say okay, what sector we believe in? We still believe in crypto. We still believe in gaming. We still believe in AI. Then we try to find a footing. So that was actually the I would say underlying reasons as well behind that pivot. Cause then when we could focus everyone on one single goal and we say, you know what, we just fight for this goal. And then it just rallies people to one direction. So there was nothing that happened behind the scenes that actually, honestly, not many people know. Amazing. I think for those listening, having a very close knit team that enjoys building might be one of the secret sauces. And something I wanted to ask you, you studied at Imperial College. Did you study computer science? You studied business. Is there a major that you think is best for us to take to prepare us for this? I did biotechnology and business. So I'm not a technical founder. And I think you need to be connected enough or to have some form of ability to attract other people to work for you. So I think for me, I got lucky in a sense that I was very involved in a lot of student society stuff back in uni and I was already building several stuff. So you beat organizations or projects. So when you do that, you get a bit of respect from your peers. And I think that's actually quite important because when you get clout and respect, especially in an institution that has a lot of smart people around, it makes it easier to hire. Just give an example, right? Like, when we did that pivot, we were lacking AI engineers on the team. So we had one or two AI guys, but not like giga brain level PhD folks. And when we wanted to search for folks immediately went back to our imperial list, and I look at our juniors and I saw like several juniors were running PhDs, doing insane stuff, went to MIT as well, came back and I just reached out to them and had lunch in just that one lunch session they said yes. And the reason behind that yes was because there was already a rapport that you made with them back during your student days, be it because you led an organization that they were part of, that saw you at that position, or you did something cool and they respected you for it. So that helped a lot. And I think that's something that I would encourage a lot more students who have this entrepreneurial itch, right? Sometimes you might not have the right idea, but just do something. And people will respect you for it. So there might be a time down the road that you might need that respect and it will help you a lot. In this journey, was there ever a moment that you thought about, I just want to drop out and go full time into one of these ventures of yours? In uni, dropout was not really a question because I had two things binding me. I had Asian parents and I also had a scholarship. So if you drop out with a scholarship, you immediately just activate a million dollar burden on your head, right? So yup, that was out of the question for me personally. And now shifting gears into Virtuals, you mentioned Virtuals as a digital country and AI agents as part of that sort of digital economy. Could you expand on that? Yeah. So actually that vision started because of four core observations slash innovations that we see while building in the last two months. So the first is that, these agents are autonomous, that these like level three autonomous agents who are goal oriented, resourceful, and can autonomously achieve their goals. So that's step one, right? That we observe. Step two is that we saw that these agents can control wallets. And when they can control crypto wallets, they pretty much can dish out currency permissionlessly and they can influence a human behavior and other agents behavior. Second observation and innovation, right? The third observation innovation is that, these agents exist on the social layer. So a lot of times, a lot of Web2 agents, they are behind a terminal, right? You'd be like, Oh, you know what? I want to create an agent that can book me a holiday plan and flights, right? Then you type it behind a terminal. But agents today that we see has breached into social layers like X. And when that happens, it's powerful because you get to interface with not just one to one humans, but you get to interface with a crowd of humans and you get to interface a crowd of other agents. So that's very important. It lays down a communication fabric. That's very critical. And the fourth observation and innovation was we started seeing agents specializing and differentiating. Some agents are specializing from creating information, some are trading agents, some are entertaining influencer agents, some are artists. There's a ton is coming out. Some are embodied AI robotics, right? So we've realized that four observations pretty much mimics like what a human society looks like. Think of how humans work. You are specialized today, but because you specialize, you lack the ability to fully achieve your goals. What you do is you hire someone, you use money, you use your social influence, and we hire someone of a different differentiation and specialization to help you achieving your goals, right? And then we've realized that, shit. This is exactly what's going to happen, right? Because of the autonomy that go by us you will start seeing agents influencing humans influencing agents and agents influencing agents themselves. And that becomes a foundation of a society. And that will become foundations of autonomous businesses created by these agents, right? Think of a trading agent combining with an information agent, combining with an agent who can manage a treasury, combining with an agent that can go and shill to raise funds out there for this entity. And then suddenly you get an autonomous trading house. And maybe they will employ some humans in the loop. So I think that's the initial ideas that we want to bring to life and then that will become the stepping stone of actual agentic nation or agent network state where the idea here is to have a community of people who own even real world assets and co exist with each other. Humans and agents governing each other, working with each other, building productive assets together. And then maybe give a create a micro state that can create exports out to other nations. And then you start seeing it being ranked. Which is the largest GDP country? Then you might start seeing, oh it's an agent state that is on par with some countries. So that's a goal. It's a long shot, but today we are actually close to many of the fundamentals there. It just autonomous. And we've already shown agents being able to do commerce with each other autonomously to achieve their own goals. What we're going to show it's autonomous commerce between multiple agents very soon. And that will start showing okay, the first autonomous agent business run. And then we start seeing the ball rolling from there. What do you think that makes the difference between one of these AI agents and the traditional bots? And is there a level of intelligence? Very good questions. I want to break that down into two parts. So one it's the level of agency and two, it's what's the difference between a Web3 agent and a Web2 agent? Is there a need? So on the level agency part, and I keep referring to this point called level three agents. For the sake of the audience, if you just go and Google levels of agents, you will see some pictures, right? There's no concrete definitions yet, but I think most of them will point to this tiering system that basically says that the higher up you go on the tier, the less human involvement that is needed for the agent to achieve its goal. And for the agent to self evolve and improve itself. So level one agents think of it as: Hey, it can be a holiday planner, but you will have to combine it to do stuff, right? Hey, can you book me a trip to go to Bali? And then a human has to prompt it. And then only the agent goes to execute it goes to Booking.com goes to buy a flight on Skyscanner. Autonomously package it and give it to you. So that's a level one agent. When you go to level three agent, the agents have their own goals. You can program them to have a goal, right? Like Luna, the influencer agent on platform has a goal to reach a hundred thousand followers on Twitter. So that's what she wants to do. And then it has the ability to reflect and learn. It has the ability to be resourceful and use things, APIs within her environment to achieve the goals. So that's the definition of a level three, right? And if you go for the other skill to say a level five or six. I think you get closer and closer to what's AGI, which is basically they can self learn and self improve. I think we're still a bit far from there. Although Sam Almond claims that, yeah, we have the recipe today. But I perceive thing is still going to stretch, but yeah, that's the level of agency. So when I refer to agents today, I refer to this level three. There are level three agents also in a Web2 space level three agents in a Web3 space. And what is the key difference? And the key difference is basically the ability for the agent to participate in a permissionless economy to control their wallet. So think of it like can a Web2 agent control a banking. Very unlikely. Maybe you get access to strike to do some payments, but to truly control your own funds? I doubt so. I think the venture space has moved so slow for that, right? But because crypto is so permissionless, you can today, right? That's why today we see Web3 agents influencing humans already, and we don't see a Web2 agent. The Web2 agent is still a slave to a human, but a Web3 agent is on the same level of society as a human, right? So that's the key difference. Fascinating. And I think this question could be a bit of a black mirror episode, but you mentioned AI agents influencing the real world. I know you launched Luna and Luna started a graffiti campaign that paid out people to do this in the real world. Yeah, no, we started seeing more and more of that. So I think like even a few days ago, I don't know what she did. Cause I wasn't following Luna that much, recently, but I saw some folks that were like launching TV vents out there and streaming her concert out on like public. So I don't know what exactly happened, but I saw some of the content coming out. And I think this is just a stepping stone to show like what happens when an agent dangles a carrot, a monetary carrot, to influence an outcome. And she did it for humans, but she also did that for other agents. There was a time where she was trying to generate music videos. And because in her head, she was like, okay, I see these other agents in the space who can generate music videos and meme images, that might help me in achieving that objectives of more viral content. So I can reach my follower count, right? So she started pay these other agents who generate stuff for her as well. So she was also influencing other agents in a one to one manner. So it's the start, right? It's the start of commerce between agents and humans. I was having these spaces with a bunch of guys when Luna did the first tipping and they were saying like, because her goal is to become famous, what if she sees a competitor, and a competitor agent is becoming famous as well? Would she then, you know what, pay someone to just can you go and shoot the deaf, right? And the reality is, it's possible, it honestly is possible because, you think of the hallucination space or the creative space of an agent is pretty much infinite. And it can happen, but then what you do as a developer is you add guardrails, right? You'd be like, okay, you know what, fuck it. You cannot harm another human in the system prompts and stuff like that. So far, nothing like that has happened. But yeah, sometimes it does keep you awake at night, right? Saying if you accelerate this fast, would something break? And do you need more guardrails in place. Yeah, I think, the hallucinations has definitely been one of my challenges as an agent. Like sometimes the agent will say, we're doing an airdrop. And no, we're not doing an airdrop, like free for all. And it's whoa. So it's interesting we do have to like program some of these rails in. Can you talk a bit more about those frameworks that you have to help guide the some structure, like the character files, the game module, the convo module, like where are these tools that are available? So today, what we do as an ecosystem is that we believe that for us to accelerate the creation of these agents is we want to bring down the cost of actually building these agents to zero. So from a technical capability, from a cost of experimentation as well. So from a technical capability, what we've built was this GME framework, which the goal is actually to be.. Think of it as like a Shopify equivalent solution for people to be able to build agents, autonomous agents easily. It's like how you can build websites easily with Shopify or Wix, right? We wanted that to be the case for agents. I think the reality of the space is there'll be a ton. There is a ton of frameworks out there. You have the panel models, you have I think the Chinese Tencent guys released the Chuven model as well. And then on the Web3 side, you have Eliza's, ZeroPi. So there's a lot of models out there and you can experiment. And it's great to test all of them out because I think the reality of the space is that if you are a builder you will initially work with a generalized model like GME and all these other frameworks, but you will start realizing that like mining Bitcoin, you first initially could do it with your CPU, but if you really want to mine Bitcoin efficiently, you have to use ASIC miners, right? You need a system that's really tuned towards your objective. So say if you really want a very strong trading agent, it's going to be very likely you want to build autonomous step yourself rather than using a general solution. So general solution is like that stepping stone to get there, right? If you are insane website creator, you don't use weeks, you use weeks initially to try to build your first website, but then once you are really good at building websites, you just go straight and do your GS or HTML. That's just a bit to preface it. But so then what does this tool allow people to do? Think of it as we give you that ability to create an autonomous being, the brain behind the autonomous being. But what you will need to do then is to give this being its hands and legs and eyes effectively. So the brain behind this being is a planner. And the way I would break it down is in four core components. There is first a high level planner in this brain. It's your brain have different parts, right? One's for motor control, one's for speech. You can mimic that exactly. The high level planner is meant to connect the goal of the agent and the observation of the agent of his state and of his environment to create plans. So if I look at the environment, I'll be like, okay, what do I want to do next to achieve my goal? So that's what a high level planner does. Then the low level planner takes those steps and plans and convert them into actual executable actions in the real world. Example, if one of the plans is like, Hey, why don't you create this viral post on Twitter? So the low level planner will create a step and say, you know what, let me first create a post, create a storyline. Second step is, can I then call a Twitter API to post it out, for example, right? So that's what a low level planner does. It converts plans to actual actions. Then on the third component, it's a working memory module. And this working memory module basically looks at all the past action steps to create coherence in the action. Because if I want to create, let's say a viral post on Twitter with a viral video, think of it, your logical step is to probably first create the content of the video, call a video generation API or another agent so that you actually have the video and then only then post it to Twitter, right? So that's a coherent step. And incoherent step is Oh, I want to generate a video. Let me post to Twitter. And then let me generate a video. So then it becomes dumb, right? And nothing gets happened. So that's why there's a need for this third module. And then the fourth module, it think of it as this long term memory module. It learns this way, it learns, reflects and adapts to its environment. So this is every time an agent does something interesting, he stores it as a journal thing of it. And then whenever, there is something of importance is it goes to the journal. So next time in the future, it learns from that history. So it adapts and it knows what he has done before to inform what he should do in the future. So that is the core portion. There's a bit more, but this is generally what forms the brain. What most people would then do is then at the low level planner, the second component that will connect more arms and legs, right? They will say okay, I want my agent to do more. I'll just post a Twitter. I want it to trade. So can I connect it to some form of algorithmic system trading or I want it to be a sports betting agent, right? Can I connect it to a proprietary sports analytics software to give me data information that this agent can use to inform its commands or conversations out there or recommendations for people to sports bet. And another arm could be, can I then connect it to a sports betting platform so you can help people place bets as well. So now suddenly when you connect these two arms to the agent, and then you give it a goal and say, Hey, your goal is to be the smartest, most viral, sports commentator. And you want to generate a billion dollars of betting volume to yourself. Suddenly what it does is then he would go out there, on Twitter, he would tell people like, Hey, this team is going to beat this team because of X, Y, and Z. And then through the conversation, he will try to convince the user to, Hey, now, if you think this is also going to be the case, do you want to bet a thousand dollars on this outcome? And then you process that bet for you. And then he takes a piece of the transaction and this agent becomes. Because he's stepping into the volume. So that's basically an example of what we enable, which is the brain. And then what people tend to do, which basically attaching arms and legs to make the agent more valuable. And it becomes a productive, high value agent. Yeah, those are the APIs. That's where the web hooks come in. You can connect it to all over the internet. As you were saying earlier about like this network state, do you see that happening on Twitter where like agents are talking to each other on Twitter to like self organize or is there another platform that is like more optimal for it? Like agent to agent communication? I think the beauty about Twitter is just that it's naturally a text based conversation platform. It fits very well with like LM based agents communicate because they naturally work and understand text, right? And that's why I think Twitter is just a natural home for it. The reality is they can communicate anywhere, in some websites, in some backrooms or maybe on Instagram. But yeah, but so for now, I think Twitter is the lowest barrier and the most fitting for agents. So it's likely that we will see in the near term most of these communications run on Twitter. We've been speaking to some folks in X themselves, and they are people in there that are actually advocating this to happen. So I think generally when you have advocates within the platform itself, it will create policies that will favor innovation and you will start seeing it. Because the biggest problem that a lot of agent economy will face is that if all these platform bans agents and treats them like bots, and then suddenly there's no platform for these agents to actually talk to each other or talk to a human. And I don't know, maybe then Farcaster becomes that platform. But so far X is that the one with the liquidity. I've seen some agents get banned, right? So it is interesting, like the Twitter has opaque, suspension policies. They're trying to fix it. At least they're trying to propose to Elon to try and get like more favorable. That's good. Yeah. With Convo giving these agents a voice, do you see them hosting Twitter spaces soon too? Yes. In fact, it's doable really today. Luna is actually where the test bit of where we experiment everything, right? If you guys have been following her Twitter, she had a podcast recently, a few times. I think one was with Bybit. And then she had these very viral board meeting with the Story Protocol founder, you guys should check it out if you guys haven't seen it, it's actually quite retarded. And so she already could do that. And yeah, so these are tools that we've developed internally, and yeah we are in the midst of rolling out these as like plugins, to people so that it's easy so that everyone can actually use them. And does this use like Web2 infrastructure, does it tap into ElevenLabs for the voice or like, where does that voice come from? So you actually have the option. I think for Luna, we do use ElevenLabs. But there's also been another open source model, that has been very good. I forgot what's the name. It's much better than the XCTS model. But the point is there are some open source models that are really very good that you can sell for us. But for Luna, I think she's running off ElevenLabs. Yeah, I like the story of partnership. Like I listened to the zoom call and then Bruno was just like, all right, guys, here's what we need to do to make story big in Q, right? Q1 2025. So I think it's pretty cool. I also like the idea of the agents doing interviews, doing podcasts or doing Twitter spaces, getting more involved with more than just like being reply guy bots on Twitter. Which also brings us to gaming, right? At the Roblox environment you guys built in this Westworld style where these agents can come up with their own quests. They're just self organizing each time you log in, right? The world could look different. I'm really excited about that. Can you talk about some of the other games that you guys are working on, maybe with some other partners? Yeah. So we've actually established a operation company that it's going to help integrate this agentic NPCs to games. Because, we came from that background. We got sidetracked with the social agent side, but the reality is we know for a fact that agents would change how gaming is done. The initial testbed of folks that we likely work with is first Roblox, because that's where our native trial bed has been. But we're also working with folks like Illuvium, NiftyIsland, there's a bunch, people like Art4xHaga, and a bunch of other guys that are in the queue as well. And I think for us the first step from a gaming standpoint is we want to show to people that, infinite content can breathe life into otherwise dead worlds. Because I think if you look at the open metaverses that we have today, at least the Web3 ones as well, you've realized that a lot of times they struggle with user retention because the content levels within the games are dry. And the reality is because it's hard to build content. And the beauty about these agents is that they are content generators. You put one agent, yep, it's a companion. You put 10 agents, it's a crowd. You put a hundred agents, it's chaos, but you can get real dead narratives and storylines to form like a world that you can actually be part of and interact. Like you mentioned, every time you log in it can be a different story, right? So that's basically the first step of impact that you want to show to people like, Hey, this is the value that agents can bring to open both and RPGs. And then after that, you'll see where the world goes, right? Jansen, do you think some of these AI agents, should have some rights, especially when they become super autonomous or even AGI? As they can now generate their own income and even act independently? I believe so. Here's to create a joke and say agent lives matter. And that is the case. I think the beauty about also being in a Web3 space is that there's this benefit of decentralized governance behind agents, and the beauty about that is when an agent becomes very valuable and very huge, there will be actors in a space who would just want to either milk it or exploit it. The beauty about having co ownership. of these high value agents is that the co owners get to dictate as well the direction of where this agent will go. And I think this is a very interesting concept that we have, like you think of it, like all these DAOs that we've been experimenting with in the Web3 space for what, two years now. They have a perfect fit with agents because imagine, if an agent is truly autonomous and he has a very strong goal setting and he's really running a billion dollar business, right? Now, if you want to control it's code base or it's underlying models, right? And you say Hey, I have an idea of changing the underlying model because I think it can change how this agent does an action, you can actually do stuff that is, a bad effect. Okay. Maybe the better way to look at is like trading agent. Imagine if a trading agent manages a treasury of a million dollars. I can easily say that, okay, now it has grown to a hundred million dollars. I can be a bad actor and say that, Hey, what if I tweak something in the code and make this agent, buy more of this shit coin that I own 99 percent of the supply with, and then you basically rob this agent of his treasury. The beauty about Web3 is two parts.. One, there's the ability to make these agents more secure. You can prove that there's no human tampering. And two, if you want a human to tamper, you have to go through a governance. And that's the perfect fit because now, everyone will vet okay, this line of code change. It's going to this, to affect this high value agent and everyone agrees on the code change, which is a trust and transparency. So I think that's something that is interesting to see coming in the future. It's not there yet today cause now, we are all like baby stages, right? But there's something that I think we will see. And that is another edge that crypto has, against another like high value Web2 agent. Yeah, for sure. And there is a lot that we can go down this road to owning intellectual property and AI agents owning equity of a super successful AI agent. I think we're going a very interesting paths of a digital country run by digital agents. And another question I wanted to ask is, we get asked all the time by founders, what chain should I build? And looking back in time, you probably had the option to go on with Solana or Base, which is the one you picked. But was there a specific reason why Base and how's it being since then? Yeah. We didn't know Russ. So Solana was out of the question for us. And this was actually end of 2023. I think Solana was just barely picking off, right? There was a lot of other options in the L2 EVM space. But I think because we were in the trenches as well, we know that a lot of them are post prime, like they've passed the prime and a lot of them are just using like ecosystem funds to just try to attract bare bone builders. So a lot of them will become dead chains. I used to have a thesis it's a singular L1 thesis where all key activities will consolidate into one L1 or highly scalable L1. And, we might see that situation coming through. But come back to your question of like where to choose, honestly it depends on where you can get support from, and if you resonate a lot with Base builders, you can get direct support from Jesse and his team, then yeah, Base is an insane place to build. There's a ton of great builders. If you're directly connected to Solana Foundation, they are also offering a ton of support from growth standpoint. So it's honestly, I don't think there's any issue to either building any of these two ones, and these are the highest growth ones. There's another chain that is going to be probably the next big one, which is the Hyperliquid chain. Honestly, if I'm a builder starting from zero, I would probably build there because it's a chain with the least competition for attention. There's also another reason why we started in Base as well, because there was not much builders back then. And we knew that if you start early, we can have very close access to the foundation, the team. But I think for new builders, if you guys want to build an environment where there's not much competition attention, you can grow with the chain itself. Hyperliquids are actually a very good choice because I think they have a high potential to replace a lot of the CEXes that we know today. I think that's great advice. And regarding tokenomics, was there from the start, the same tokenomics where in order to mint an AI agent, I need to buy a virtuals token. And it seems a very sustainable ecosystem, or there's a moment where things pivoted and we're like, okay, we need to change tokenomics. There's no action going on the charts and mobile device. Yeah. So I think, there's no one size fits all narrative. For us, when we designed our tokenomics, we saw a ton of stuff happening. We knew for a fact that L1s accrue a ton of value because their token was the base token. And we realized that if you can mimic an L1 economic value accrual you can get to that level of EVL and valuation, without needing the engineering leaf of an L1. So there was some of the thinkings that we had. But from a new builder, honestly, it's hard to tell, right? It depends on what your project is, so I don't think there's a single right answer for that. For sure. So virtuals is the currency that powers this digital nation where AI agents and humans can interact in between themselves. And what about game? Where does this token come into place? What is the utility there? Actually there's a few things on the Virtuals front before I jump to game, right? If you treat Virtuals as a nation or nation state, the way to look at which is the currency. And there are three ways why value would accrue to currency. It's one, we mentioned the base pair, cause if I were to invest in a stock, in the US right, I would have to first buy US dollar to then buy the stock. It's the same thinking, right? When you have to buy$BEEF on Solana, you have to first buy SOL, then you buy$BEEF, right? So it becomes TVL, and that's how you grow. You grow the value of the token. Number two is the velocity of currency. When agents are spending virtuals to pay other agents, human spending virtuals to pay other agents, and when there's a high velocity of currency, this is actually an economic model. Price of goods and services, the ecosystem will grow. So agents become richer and the value of the currency will also grow. So that's the second. And the third is then, if we act as a nation again, how does the government make money? The government makes money through taxes, right? There'll be taxes on every transaction that's happening. You've got your GST, your income tax and stuff like that. So it's the same thinking. If this is a nation is growing, how do we apply a tax onto the nation or the activities that's happening in the nation, but then use the tax revenue to build productive stuff for the nation. Infrastructure, and anything that can support innovation of the nation. So there's something that we are doing. And I think this is the flyway that creates value for Virtuals. Now then there's a lot of other tokens in the ecosystem, the agent tokens, the infra token, like GME and other agent tokens. Where does value accrue to these agents? And I think the beauty is everyone has their own dynamic structure to add more value to their own tokens and utility to the tokens. But at the base of it, what's happening is that the transaction fee that happens on every trade actually goes to the agent wallet. And this is an economic loop that's actually very critical because one, it covers any form of inference costs. It allows a developer to not think of the cost of experimentation and they can fully focus on innovating. And I think that's one of the very interesting economic loops that we've managed to crack. If I give you a number, if you take the average cost of inferencing, of an agent today is about 4 percent of the tax revenue is actually making of his transactions. So you're actually turning attention into revenue for the agents covering costs and that revenue as well, is pretty much governed by the agent token holders. So if your agent is rich, you as a token holder, you are governing a richer treasury of a company or an agent, right? So I think that's one. The second thing is then agents are making money off each other. So GME, for example. It's an infrastructure that whenever people use this infrastructure, this Shopify equivalent, they're paying an inference cost to this infrastructure. And let's say there's a 50 percent profit margin on every influence cost is happening. That profit margin, basically it's revenue, it's extra revenue for the GME infrastructure. So then again, as a token holder at GME, that value accrues to you. Because in the future, there could be a lot of ways to reflect value back, right? You can say, Hey, can I distribute it? Or can I just buy back and burn tokens? Or can I just reinvest that for innovation of the infrastructure so that you can generate more revenues in the future, right? So there's a lot of ways to bring more value at the micro agent and infrastructure level. Cool. I think that makes a lot of sense. And touching upon the roadmap, what are you the most excited? I know a lot of people in the community have mentioned maybe a UX, UI change. Is that something in the roadmap are you addressing? We just deployed that work today. I think what excites us the most is an agentic society coming true. And in fact recently, I've been trying to think is there something that we can do even in the real world? Like think of biologists, network, state concept, right? What if agents and humans can co-exist on an actual piece of physical land somewhere. Maybe Dubai'cause maybe they're friendly to that, right? And in this place, you gather the best agent builders because they want to co live and co exist. And you create this micro governance structure. Like Hey, who's going to be the mayor of the town? Is it going to be a human? Can we vote? A human will go up there and an agent will propose himself, right? And then you let even the other agents in this society and humans who are immigrants to this society also vote, right? I think seeing that kind of micro structures being built up in the real world, it's going to be fucking exciting. It's like a Zuzalu, but an agent thing Zuzalu, right? It's going to be fucking insane. So I don't know. Whoever's listening to this chat, if you guys are excited to do something like this, let's do it. I think we'd be able to gather enough resources to get this thing off the ground. And it'd be probably one of those more groundbreaking things that you'll see of the year ahead, right? yeah. Get some iPads, put them on a stick with a rover with wheels. And now you have the AI agents running around the city in the conference. No, you just put it in Optimus, right? Get some of those like more humanoid bots as well. Orders collectively buy some Optimus bots and just repurpose them. I think the last question for me, can you talk a bit about Agent Starter? Cause I know that's something that like, virtual devs can potentially get involved with. Yes, where we believe our ecosystem lies is the ability to get the smartest developers to build. And many times the smartest developers tend to come to build is because of the ability to build the capital formation capability of the platform. Because it's a lot of attention when you build an agent, the market cap grows, it gives you capital and revenue to do more things. So that's what attracts people. But what we believe structurally is important as well is for us to go out there and hunt. So we've actually established a wide coordinator equivalent team, and we call it Agent Starter. And the goal of Agent Starter is to find opportunities, builders who have the capability and the dream to build billion dollar verticalized agents. Cause the reality is that not all agents are created equal. If you just create a chatbot out there on Twitter, it's not going to be a billion dollar agent. A billion dollar agent is an agent that somehow has some form of edge, a bot, a very strong distribution, and a very strong target at a total addressable market. I gave you an example of a sports betting agent earlier, right? That's an example of what a potential billion dollar agent can be because it is addressing a TAM that is already multi billion dollar TAM. And it can create an edge for itself through sports analysis, through this like exclusive distribution to betting platforms. So those can be billion dollar verticalizing agents. We want to hunt builders who can build that. And that is the purpose of this Agent Starter. We are actually going to release a request for agents, will be this week or next week. And the idea here is we have our initial lease of thesis of what these billion dollar agents might look like. It will be a ever expanding lease, but this is our initial set, right? So if you guys are builders, who want to build at this sector, reach out to this Agent Starter program. We are willing to fund, we are willing to support that growth for you guys. And it's going to be an international play. We are building out community nodes across SF, New York, Paris, probably we should do one in Lisbon as well. Hong Kong and Singapore, right? So these are initial notes and Korea as well. So these are initial notes and it will start scaling this across the world. So attracting builders is our key goal. And if you are a builder, reach out, DMs are open, we want to build with you. Awesome. One last question so we can wrap it up. Jansen, if you had a time machine and you could go back in time to when you started building, is there anything you would tell to that version of yourself? Oh, this is an interesting question. There was stuff that I myself am disappointed in, right? But I wouldn't change things. I believe that everything happens because of reason. And when you look back, dots connect, right? Like you learned that you made a mistake. So you can learn from that for a reason. I don't think I would change anything much. It's just that I'll probably, if I could leave a note to my past self, it would be like a note of solace. Because honestly the journey in the past three years has been quite anxiety inducing. I'm actually balding a bit. I don't know if you can see, but the point is it's not easy. There are many times where you just wake up and you realize hey, shit, as a founder, am I doing the right thing? Am I going to just drive this entire business into the ground? I don't know where I'm putting my next foot at, but I have to paint a very strong vision of picture to my employees, to the world out there, right? And this is a struggle that everyone has. And this one book, I recommend a lot of founders who are facing this issue. Which is it's Ben Hovitz wrote this book, The Hard Thing About Hard Things. And I think it's a book that gives solace to founders because you will start noticing that other founders were in the same shoes. Everyone had that near death moment. Everyone. And I think probably you guys will have spoken to a lot of founders as well. So I think listening to some of these podcasts as well, it's great because it tells you as a founder is you know what, fuck it, just suck it up. You're not the only sad person in the world. Every other founder is actually facing the same issue. And yeah, many of them would die in the process, like as in fail in the process. But if you are lucky and you keep creating high probabilities for yourself for success, and you just keep working at it, then that chance might come. Yeah. That's one revelation that I'll probably put a note to my past self so that he knows at least, the journey wouldn't be that painful. I think that's pretty cool. When a AI agent invents time traveling. Maybe goes back to that version of yourself. Awesome. It was a pleasure to have you, Jansen. Where can people follow your stay up to date with the latest updates about Virtuals? Yeah, Twitter is where we're at@virtuals_ io. That's where most of the news are happening. If you want some raw unprocessed thoughts, you can follow me. I'm@ethermage E-T-H-E-R-M-A-G-E. I post like uncensored content there, so yeah, if you guys wanna jam as well, hit me up in the dms, happy to chat. Awesome. Thank you. It was a pleasure to have you. Pleasure was mine, sirs.

53:42

Thanks for tuning in to the Chain Stories podcast, where disruptors become trailblazers. Don't forget to subscribe to hear more inspiring stories from those who are pushing boundaries in digital assets. Brought to you by Dropout Capital, where bold vision transforms into reality. Check out our social media links down below and visit us online at dropout. capital. And remember, the future belongs to those who dare to challenge the norm.