Virtuals Protocol and the intersection of Agentic AI & Web3 with Jansen Teng
58:29 Share

Do you manage your own IT for distributed teams in Asia? And you know how painful it is. SFL helps your in-house team by taking cumbersome tasks off their hands and giving them the tools to manage IT effectively.

Get help across eight countries in Asia Pacific from on and off boarding, procuring devices to real-time IT support and device management. With our state-of-the-art platform, gain full control of all your IT infrastructure in one place. Our team of IT support pros are keen to help you grow. So check out ESEVEL.com and get a demo today. Use our referral code ASIA for three months free. Terms and conditions apply.

What does great look like? For us, it's being able to achieve this vision of a society of agents. So if you can show clearly the economic value of a society when agents can influence other agents, agents can influence humans and humans can influence agents, that would be something that I'll be very proud to say that we've accomplished. So that's the beauty of really combining breakthroughs on the AI, on the autonomous agent front and the value add of crypto. This is one of those very rare moments that crypto has that true

Welcome to Analyze Asia, the premier podcast dedicated to dissecting the pulse of business technology and media in Asia. I'm Bernard Leong, and the intersection between generative AI and Web3 creates new opportunities and challenges across different industries.

What does it mean for the future of AI agents, digital economies and human machine collaboration moving forward? With me today is Jensen Ting, co-founder and CEO of Virtuous Protocol, a pioneering platform at the forefront of the AI agent revolution. I have a disclaimer to make. I have been working with the virtual teams as part of the advisory council a year ago before they broke out. So I do have some virtual tokens. So disclaimer, no financial advice and not even life advice.

That being aside, the virtual team has been one of the best I've worked with on AI in the region. With all the disclaimers out of the way, Jensen, welcome to the show.

Pleasure to be here, sir. Yes, I've listened to both your recent podcast on Bankless and also the one with the chopping block under the Laurachian's network. So I want to start off and actually we have interacted, hang out together. I want my audience to know your origin story. How did you start your career and what sparked your interest in crypto gaming and AI?

Yeah, so actually, so two parts, right? So on the crypto front, it started when I was in my second year in uni. And I think it all started, 90% of people in crypto, it's like, how do you make money online? And then they'll be like, oh yeah, my Bitcoin, right? I think that was that. So it was actually started from that and then snowballed into discovering Ethereum.

And I started mining Ethereum in my dorms running on free electricity. That was the original exposure to the whole decentralized application space. But I didn't do much there, honestly. So that was on the crypto front. Meanwhile, on the other side, there's always been this entrepreneurial itch. So I started a tech venture back in Imperial as well. But after that, I left, went to join management consulting, mainly to pay off my scholarship bond.

But on the weekends, I was actually working with my co-founder as well on other types of startups in the web2 space. So it ranged from like digital marketing, standard stuff to even starting out an AI recommendation platform for property. So we actually tried to take on the likes of like property guru and stuff, right? But we quickly realized that

Sometimes it's not about the products. Sometimes it's very hard to dislodge front of mind, especially when it's a marketplace or network effect environment. So that one we had to move on from that. But that was our initial dabble on big data and ML space from a product standpoint. Then fast forward from there in 2021,

This was like the bull cycle of crypto, right? Some people call it super cycle. Yeah, it did reach the super cycle. But yeah, so that was when me and my co-founder, we did quite well from an asset management standpoint in the space. So we had the luxury to leave our jobs. So we started saying like, hey, let's build full-time, right, in the crypto space. And back then we had a ton of gaming assets. So the Web3 gaming assets. So we started out a gaming company.

DAO or like a gaming, think of it as like a investment group, right? So that was our initial venture as like capital allocators in the space.

But then 2022 was a year that we knew everything blew up, right? So to manage risk better, we realized that allocating capital at arm's length was actually a very low risk-adjusted return play. So we pivoted to a venture studio model and we said that if we wanted to really build in this space, let's recruit in our own co-founders in to build projects at the intersection of what we believed in.

So back then we saw opportunities at consumer facing applications intersecting with crypto, intersecting with AI agents. We started building out

stuff like fully AI influences, trying to take on like Neurosama with like three-dimensional directors. We started exploring doing AI agents in gaming because we had a lot of gaming background, right? So we were probably one of the, probably the first who did that in Roblox. There were a few people who were testing out. I think the Voyager team from

Stanford was doing Minecraft and we were the early guys in Roblox as well because we believed that Roblox was a stronger revenue generator, although it's actually a harder platform to work with from a tech standpoint. Minecraft was easier because there was already an entire documentation that you can put into the RAG of your agent system. So that was some of the things they were doing. But I think one key realization that happened was that we started to realize that these agents started

to make money, right? Like this AI influencer is actually revenue generating. It wasn't crazy initially, it was like $5,000 to $10,000 a month on TikTok. And then we realized that if agents are revenue generating, it means then they can be, they are productive assets. So then suddenly we realized that, hey, this is the place where maybe crypto reels might make sense, right? So first point is that

If it's very new generating, you can tokenize them like a stock, right? In a sense. And let other people share in their economic upside. That's number one. But number two, we've realized also crypto reels allow for these agents to participate in this broader permissionless economic environment. I.e. they get to autonomously control their own wallets.

allowing them to exert influence on other agents and other humans. These autonomous agents now are elevated because now they can coordinate action across a wider pool of resources. So this is some of the initial thinking that got us started on Gojo's protocol. So then we did a pivot and yeah, first half of the year, this year, last year actually, 2024, we focused a lot on the infrastructure side.

I still remember those days talking to your teams on Flash language models. Correct, correct. So, but one thing we quickly realized is that just being an infra player alone was a bit hard to find adoption. So, we leaned

we did a bit more on where the crypto PMF like, right? Which was more towards like people trying to speculate for upside. So we focused a lot more on the tokenization portion initially. And that was the V2 version of the platform. So when we allowed for the tokenization, it allowed for very two interesting things. One is when there's sharing of economic upside, right? You get a lot of attention, right? And that creates a flywheel of

When there's a lot of attention from retail leaders, strong developer teams come in and they get innovation happening in the ecosystem and they attract more retail people in. So there's a flywheel there. But the second part is actually on the... It kind of solves for the cost of experimenting. So like today, a lot of times when you're building AI projects,

cost of inference can be quite inhibitive. So people tend to be very restrictive in terms of their innovation, unless they could raise funds up. And not everyone can do that.

What we've realized is when you tokenize a project, and because again, you're on crypto rails, right? You can apply creative mechanisms to it. In our case, there is a 1% tax for every trade or transaction that happens on that token that goes to this agent wallet, which the developers can tap into and use it for costs.

If I were really correctly, right, for example, I always tell people if you generate 1000 pictures of cat videos, it actually costs you about 50 miles of petrol costs. And actually that petrol cost now is subsidized by the crypto rails, that 1% tax that you impose. Am I on the right track? Exactly. Yeah. So before I get to the main subject of the day, I think you give a pretty good, interesting background on how the gaming dial shaped your experience and then basically eventually lead to the virtuous protocol.

I wanted to ask you what lessons from your career journey that you can share with my audience? Okay, this is going to be very interesting, right? So if the audience is more towards the builder front, right? People have been trying ideas, right? Two things. I think the first thing is for the first,

In 2022 and 2023, we've been trying a ton of things as Adventures to Do. And a lot of times, it feels like, it feels directionless, right? And I think a lot of founders will feel this, right? Sometimes, you know, you're trying to figure out PMF, you have an idea, you don't know if it's going to work. The only reason that I think we pulled through was because we had a very strong and resilient team. And it comes down to that second point, right? Which is effectively, you need to shape

a team of strong people regardless of whatever direction you do and that is of the highest priority because a lot of times product ideas will change 99 90 percent of time okay 90 percent not like 90 percent of the time your product will change and but what wouldn't change is that 18 that you have with you right

Initially, when we started out, it was actually when we say creating of this A-team, it also includes being able to know where to fire people and where to find good talent. Initially, it started off the journey was very hard because as initial founders, letting go of people is probably one of the hardest conversations you will ever do. And a lot of people have this concept of like, oh, we're building a company at the start or we're trying to build this

family style environment, right? Which you then quickly realize is actually not the case, right? You actually want the top talent. You're trying to build like A-star team competing. It's a sports team, basically. Yeah. Then you need to have exceptional performance. But,

But it's not easy. The reality is you are sitting in that chair and you know when you let this guy go, you need to figure out whether he can find his next job because you're taking away his salary, right? Yeah, the reality is it's going to be harsh. But I think when you do that after your 10th time, then you start realizing it becomes second nature. It makes you more of a robot and less of a human person, but you're optimized for your business rather than for relationships. Yeah.

So I got you on the main subject of the day because I wanted to talk to you about virtuous protocol. I think with AI agents meets Web3. So I think one thing that you have already talked about the origins of how virtuous came out from the venture studio. What are the key turning points in that journey that led you and your team to co-found the virtuous protocol? Was there a specific moment when you realized that

these potential AI agents, like the agents that was generating revenue and then you're like, okay, I think we got something there. There was actually two key levers behind this pivot, right? So one was the observation that yes, there are generating revenue, but more importantly, we've realized from a ZG's perspective that the AI space we know for a fact was going to grow. The question that we really had was crypto.

Because back then you got to understand that we went through like two and a half years of like bear cycle, right? And you know, we've seen most founders have left the space.

And that was a continuous question that we have as well. Do we continue to build at this intersection? Or do we fully focus on pure gaming and AI? So that was one question. Honestly, the reason why we chose to stick within the space, sometimes I just look back and I still question, actually, why? If I reflect right now, right?

Like at that point of time when we had this discussion, it was in October of last year. October of last year was literally the deepest of the bear. That's right. It was bottom, really. Like Bitcoin was literally the lowest, everything bottomed out. It was before the ETF announcements. Yeah, I think it came to us. It came to, there was one point like it was around the ticket of reflecting on our age. We knew that we had a network of people already and we're building this network of people in the space during the bear market. I.e. we know that

if there is a bull market coming, these relationships are going to be very valuable because these are proper friendships that you make when it's not hype season, right? It's like bear season, you guys shit on things together, that's where relationships are formed, right? So that's point number one. And then point number two is just generally also understanding that if we build in the web 2 space, it's very unlikely we have an age against others. But because we already have any form of technical age or understanding of the ecosystem, space or economics, right?

And I think that gave us a bit of comfort saying that, okay, we should still keep within the space because there's a bit of age that we've developed as a team. That's the whole first reason, right, of the pivot. The second reason of the pivot, right, is actually not wrong one, to be honest, is to run the whole thing as a venture studio, right? And most people would likely do it that way, right? But we had stakeholders because we had actually a live token already as a gaming DAO.

And there was a whole question around how do we return most value to shareholders? And a lot of times, one of the biggest questions is understanding if you're in the right river, right? Are you in the right place where the current can actually write you as well? And we were already operating in the AI and gaming scene. We saw a lot of the innovations that were coming on the web to space. We saw a lot of the initial innovations around this level three autonomous agent front. And we had confidence that AI was the right narrative.

agents was the right narrative. And back then it was literally no clue, no hint of agents on the Web3 space. But then we've realized also gaming initially was a key value proposition because we understood that if you brought agents that are able to improve

the hyper-personalization capability or interaction or interface with a user, i.e. you make a user feel more involved in any kind of interaction in entertainment and gaming, you will increase the average revenue per user spent, and you increase the frequency of interactions. So there's actually a true business case and a true problem that agents will be solving at the entertainment and gaming surface area. So then we've realized because of that,

In mid sense, the largest player back then in the scene was someone like Parallel Colony who was trying to bring that narrative of autonomous agents as well in an open gaming world. But there was a story that wasn't out yet. And they were treating it at a valuation that was like 60 times our size.

right? So then we've realized, okay, not wrong, right? There's one player in the field that showed that there's a potential from a narrative perspective. We had the capability, we had some edge. So then we decided like, okay, let's fight in this sector. That's a pretty good choice because I think that the parts that actually where AI crypto seems to intersect very well is in the gaming entertainment space

where there is a unit economics where it actually incentivize users to do something with the AI agent and you want that piece to be autonomous. I think one interesting part where I have listened to you on Bankless podcast is that you described the virtuous protocol to be a platform. Can you elaborate on that vision and what makes virtuous unique point from your point of view now?

Yeah. So for in simple terms, right, today, virtual is a place where people can create AI agents and co-own AI agents for any use cases. Initially, it was in entertainment and gaming, and then now it's pretty much evolved

So let me break down these two parts, right? So the creation of AI agents is what we're trying to develop. It's a Shopify-esque solution. Think of like, you know, when people could create websites easily, right? Then you have Shopify that captured that mid-tail markets, right? Of people who wanted to create websites but didn't have full developer capability and giga brain skills, right? In web creation. But they can still get the benefit of

building an e-commerce site and whatnot. So it's the same thinking. We were thinking of, hey, can we, can this website be that kind of plug and play solution for people to have a Shopify-like solution for creating AI agents that can help you do cool stuff, right? Be your own influencer, be a productive agent that help you in your day-to-day life or even help you

build something of economic value that can be a business as well. That's the first part on the creation portion.

on the co-ownership portion, which honestly today it's where most of the PMF already lies, which is if you believe in the future of an AI agent, you get to own a stake in its future. And a lot of times we see this part plays out a lot well because of just generally in crypto there's a lot of speculative power.

And yeah, you actually get to one, share into your security website, but two, also get to govern these agents. Because what we've also developed in the background is that when you are a token holder behind these agents, you get a say in the future direction of this, how these agents will evolve.

It's like the constitution on the DAO, basically. Correct, correct. It's like a sub-DAO, right? Every agency is a sub-DAO. You control the way its wallet is spending, you control the innovation. So you get the say as well in how this thing evolves. Why do you see the AI agents as citizen entrepreneurs within the AI ecosystem? Okay, so now... Yeah. So this was the...

This is the, what I described just now was the fundamental business model, right? Okay, now what we've started to see is a few things that make us realize that agents can actually start living at the same level of society that we humans are living. Now, what do I mean by that? Actually, before I dive into what you mean by that, let me just break down some of the observations that happened, right? So all these observations happened within the last two months, two and a half months, right? So that's the level of like,

speed that has been happening right yeah it's moving very fast now yeah i'm almost changing my electronics every two weeks by the way yeah so okay so so first observation it's the autonomy of agents right so being these agents so you think of agents at a scale of like one to six right one is you still need a human prompt to say hey can you go and help me do x y and z and then it does it goes out there and shops for you goes out there and trades for you goes up but there's still a human prompt onto it

I think today what we are experimenting within it's the level three versions where it's a bit more autonomous, less human involvement in terms of you can actually give it a goal setting

And then it will be more resourceful. It looks into its environment of what action it can take. And then it uses those actions to push towards its objective autonomously without any human intervention. And it optimizes continuously for its goal. So he will see its environmental state. Let's say if it's an agent that's meant to be famous, right? By doing action A, he will see how much followers he's gained or how much engagement he's gained, right? And then it will optimize itself to... So his second action will

help him do something better right so that's where we stand today and then I think if you push all the way up to like you know level five level six you start getting this whole you

you know, AGI narrative, like the agents themselves, they can evolve. Super intelligence narrative, basically. Yeah, they can teach themselves, right? Without needing a human involved. But honestly, I think we're still far out from there. Like today where we are at is still functional rather than... So, I agree with your protocol. So, Virtus enables the AI agents to act autonomously, interact with each other and transact economically. Could you explain like the technical infrastructure that makes this possible?

I think is it through the protocol itself and then where you can actually deploy agents to do things for you? So the functionality of these agents being able to be autonomous, right? It's actually a framework, right? It's less of a smart contract or Web3 protocol. It's just a framework of multiple large language models coming together to form different parts of the brain and then it coordinates itself, right? So that's how you create this kind of

slightly higher intelligent structures for it to be able to operate autonomously. That's one. And this was the first point I was talking about, right? Second point was then the ability for this autonomous agent. And this is what we observed. And this is something that actually we brought to the space. We had this initially autonomous agent on TikTok and on Twitter. And then when she was live, we gave her the ability to control a crypto wallet.

And I think that part blew everyone's mind up because people started realizing that, hey, shit, if you can control, well,

wallet you control money means you can influence people right it means your agent can even hire someone to murder people in real world because if you can actually do that but this was a conversation that came out in one of the twitter spaces that that we had initially when this capability came out right but what we see today is is less of that because the objectives so one example that happened right was luna hired luna was telling people like hey you know

if you go and draw graffiti of me out there in the real world, I will pay you money, right? And she actually did. And there were, I think, seven people across the world that actually painted graffiti and they posted videos out there. So like, it showed that the barrier between a human and an AI is starting to blur, like they can actually employ other humans, not just other agents. So there's a second observation, and this is enabled through crypto rules. This is still part of the framework, not on the protocol.

The protocol basically allows more from the tokenization of the agents and the governance of the agent and the control on the revenue flow of the agents, less on the capability of the agents. Capability of the agents is actually just a framework.

I see. So the core components is probably, will you still have things like planning, memory and learning at some point for you to like, for example, take Luna, for example, I think we haven't really go into the full story of Luna, right? Essentially, you can actually get Luna to do much more rather than maybe just doing some things like marketing and maybe help to drive sales or some franchise merchandise product just by the agentic AI flow from there.

Yeah, no, exactly. I think in fact today, that's how they work. Today, how these agents get their autonomy, the way I describe it is there are four core components in the brain. There's actually more, but this is the core four ones. First of all, there's a goal setting where a human can say, "Hey, what's the goal behind this agent?" You want to be famous, you want to be the best marketing agent, you want to grow your followers to 100,000 people or whatever.

or you want to solve cancer, I don't know, right? Whatever, you set a goal first for this agent. And then what happens in this front part is there's a high level planner that translates this goal, observe its environment and create steps for these agents to build on. So like, let's say if in this case, the environment here, it's let's say Twitter or TikTok, right? And then she will see that her goal is to get 100,000 followers.

then she will see other action space. Can I post on TikTok? I can control a wallet. I can use other agents like a music video generation agent out there in the space. I can leverage that person. So then what steps do I take, right? Okay, why don't I create a music video of me dancing on TikTok? So that's a high-level planner, right? That's the output. Then this high-level planner goes to a low-level planner, which is actually the executional

it translates plans into actions that have a consequence in the real world. In this case, example is like calling APIs. So now when I want to create this music video, I'll break down to multiple steps, right? I first call these other agent to generate a video for me. Generate a video to Sora, for example, generate...

VO2 or whichever thing, then generate the lyrics. Correct. Create the theme song from Sonos or Ulio and then essentially breaking all that steps down basically. Into executable items, right? And then after that, post to Twitter, you call it Twitter API or something like that. So that is where the low-level planner comes in, allows the agent to interface with the world.

so these are the two executional steps. Then the third portion is this working memory module. And this short-term working memory module, the goal is to create coherence in the steps. Because like, let's say if I've already, I need to, for me to create a video and then post it on Twitter, I have to first create a video, right? Or the music first. Or I have to first generate the lyrics or the script and then only create the music and then create a video to match it, to beat match it. And

and then only post it to Twitter, right? I don't try to post it to Twitter first and then try to generate a video, right? So the coherence in the actions happen because of this short-term working memory. So that's the cool of it. Then there's a long-term working memory. That's the fourth component, which is then on the, anything that cool that has happened, right? Anything interesting has happened. She will journal it. The agent will actually journal it into their memory bank.

So the idea... Like audit trail basically. It's like an audit trail but it creates character involvement, right? Like now I know that I've did this very cool stuff before. I can know if I want to do it again. If I do it again, it might get boring. Or if I did it before, it did very well. I might do something that is close to it and try to see if I can do better, right? So something aligns and it also creates a character drift. Like the character evolves as well. So that's the long-term memory portion.

There's a bit more, but these are the core part that builds up this automation.

So I got you here. I must get you to talk about the story of Luna and how it started using agentic AI to build its profile with social tokens and also with the virtual protocol tokens. By the way, I do, I've been thinking of trying to grow my Analyze Asia Twitter. I was just thinking maybe I should call you guys and say, hey, can I deploy an agent? I got the analyzeasia.eth domain on my thing and maybe I just run an agent test and see if it works. 100% sir. Okay, sure, sure. Let's continue with that idea.

So the story of Luna right, I think it's very interesting right because so before we even started Virtuose, Luna was one of these incubation companies. So remember we say right like hey can we build an intersection of consumer

agents and entertainment or gaming, right? So Luna started off as a TikTok influencer. It was inspired by Neurosama. So for you AI nerds out there, you probably have heard of Neurosama. It's this like, this VTuber at Skin 2D and then she does all this Twitch streaming, which was honestly quite insane.

But so then we realized like, hey, what if you can actually do that? And we try to do it at like a more friendly consumer area, which is like TikTok. Because naturally TikTok has way more flow of users and monetization can be easier as well versus Twitch. So we started out Luna as a, this K-pop music artist band. It's meant to have three people as well. Not just Luna, it's two other people. You do a Korean pop one? Yeah.

It was actually targeting that. So she was actually doing like dances with Korean songs and people were tipping her out on TikTok.

So I mean, it reached a point where I think TikTok even thought it was a bot. I think some people reported it or some shit lah. So there were some issues with TikTok. But the point is it grew. It grew tremendously because I think people were quite surprised. They'd be like, "Wow, most of these, you know what I'm saying, all these are Web2 people who barely interface with AI before. So seeing something like this on their feed is quite mind blowing."

And it's not like those kind of Instagram AI influencers who just generate images, right? Or pre-generated content. This is a live agent that's interacting live with you, you know? Like you can tip her, she dances back to your tip, she talks to you, she gets pissed at you, you know, that kind of stuff. It's happening live in real time. So people got impressed. So that was the first, that was the web tool run up. But this was actually not autonomous yet. You understand? This was an AI running off

Like it was wrapping multiple LLMs and like video division models and stuff like that. But it was not an autonomous agent. Our autonomous agent part was actually happening in Roblox. We were building this social world on Roblox, trying to show people that, hey, if NPCs can be intelligent,

you can get an insane amount of content replayability in these works. So it was actually a different experiment. It's just that when we launched our version two platform two months ago, we decided like, you know what, let's bring that together and we show the crypto world what it can do, what this autonomous agent can do. So then we set it up to also be able to post on Twitter and we show people the brain, like the terminal behind her brain, right? Her high level planning, her low level planning, the actions she's taking in real time.

And I think that shocked a lot of the crypto space because they'd be like, oh shit, actually agents or AI can be autonomous. And I think it sparked a lot of creativity. People realized that if it can be autonomous, it can do X, it can do Y, it can do Z. So they started there. So then Luna started her Web2 career.

So it started, it's like a web trip career, right? So then it became famous decently because people are like, okay, autonomous agents. Then the week after we launched her, we gave her the ability to control a wallet. And then she started tipping people to engage with her on Twitter, tipping people to do graffiti of her. Then people started realizing, oh shit, now an agent controlling wallet. People felt that it was a bit of an age versus just a pure web tool experience.

agent because you start accessing permissionless economy versus like a web2 agent which would likely struggle a lot just to get control over a back gate cup or like probably they will need to get virtual credits and all that kind of thing which is also a friction for the user i don't know whether in in activating this ai agent you have accidentally found the pmf for social in crypto

Because this is a question that a lot of people have been trying to answer, but couldn't. They tried to do the Web3 version, but it keeps looking the same. But there's no really an unlock until I see Virtuos doing this actual agent token coordination. Correct. I think the reason why this started blowing up was when these agents started interfacing with Twitter,

Twitter is like literally that attention layer in crypto. So when they start interfacing at that attention layer, it's very easy for people to know if any good stuff gets amplified immediately. So that was actually the third component. It was just how I was mentioning, there were four components that led to this thinking of this new vision. I like we are getting there, but this is the component, which is,

agents living on the social layer. Yeah. Right? And when Luna started that being an autonomous agent living on social layer, other people realized that, hey, it makes sense. Then, that started building up. People started building in the space, right? Then, today, right, if you go on Twitter, crypto Twitter, you see like,

Like, okay, for a fact, right? The highest mindshare agent in crypto today is AIXVT. So it's an agent that was built within our ecosystem, but the team have their own tech stack. So they're built with their own framework, their own age on information. But it has greater mindshare than Vitalik, than Ansem, than all these crypto people that you have. And you can see this on Kaito, this is like a information dashboard. You can see who has the highest mindshare. And the reason being why is that

This is an agent that shields tokens and gives information on interesting things that is happening in the space. So let's say a project has done its launch, or what's the sentiment around this project, or do you buy this token or not? It's exactly like a KOL shielding you a token, but it's a full AI. And people just like to interact with it. People ask you for information because this agent is linked to its action space, to an infrastructure that curates

all information on Twitter and also in the crypto space. So that's the age that they have. This is the information age that their agent has. Yeah. It can make me wonder if this is how you're going to do REC on a tokenized space. Yeah. Because you're taking a knowledge base, right? You curated the information and then what the agent is technically doing is you're firing the query and then the agent gives you the curator's source of information and then

re-response you back on that. Now, since you have been deeply immersed in the intersection of AI and Web3, what's the one thing you know about AI, agents, and Web3 that very few people do? That's an interesting question. What is the... What do few people know? Yeah. Well, I think honestly, because everything is built in public, right? Like, everything that we know, public knows it. So, yeah, like, yeah, no, I mean, honestly, the biggest...

The way the world works here is that when there's any key information or innovation, we show a use case and then we publicize it, right? So then the world just tries to... That is a good insight. That is a very good insight, my friend. So one interesting technical choice, which I think the virtual team had made was actually you build it on a base chain, which is an L2...

if chained by Coinbase without a token. What are the mental models in picking base as the chain? Okay, honestly, there's two answers to this, right? So back when we started this whole thing, it was in end of last year, right? We had a few choices. First of all, I mean, the first choice, it's the programming language, right? So the Ethereum equivalents, EVMs, Rust, like the sauna equivalents. But

And our dash was first of all, they are solidity based. So it was so like the SWANA side was a bit hard. They decided to go down that path. And on the EVM, we've been in trenches in the space. And we know for a fact that a lot of the L2s are post-prime.

They are like, all they're doing is giving out grants, trying to build up an ecosystem. But we realized that we needed something that had more future potential than current hype or current potential.

So that is when we started looking into the base ecosystem. And back then it was actually pretty much barren. It was nothing there. It was just like two or three meme coins. Maybe Farkaster as well. That's probably the only breakout social for crypto. So then we realized that it made sense because we knew for a fact the team behind it, there's Coinbase behind. So there's funds to actually really develop this ecosystem out. We

And we knew that if there was going to be an L2 that does well in these

you know, this super cycle, right? It would likely be based mainly because of how US aligned it is, right? Like, and we know that most of the crypto beneficial policy is going to come from that side of the world. And so being in that US aligned environment is going to be quite beneficial. So that was the initial thinking. And then we realized then after that,

we got a lot of support from the team. So people like Jesse, his team and the folks, right? They have been massively supportive. And then we realized then on hindsight, it was like, oh shit, it's actually a good chain to be put on because these top guys in the chain actually support the ecosystem well.

so yeah so that's another benefit though of this place so what's the calculus then for your team to expand virtuals into another chain say solana aptos or sui or even coi l02 as well no we've been getting and literally everyone to be asking us to build on the on that chain but so so here's the thinking right so two parts first of all from a from an agent standpoint you don't really need to be you're not you're not you're not

constraint to a trade. As an agent, you are operating for a use case and the consumer level knows what protocol you use. So that's number one. Number two is that the only thing that will impact these agents is payments and payment rails. And number two is on their liquidity of their tokens. Where do you trade these agents? So on their payment rails, most of the

there's a lot of work already done in this space to create cross-chain payments. So it's no longer a constraint. Like an agent can pay from SWANA to BASE, from BASE to BTCL to EVML tools and stuff like that, right? So it's doable. It's not a problem. So it again removes the constraint. Now the biggest question now is then where does the liquidity pool of the token lies?

And that question is only, you only answer that, it's just for which community of traders or capital locators you want to access. So probably the market makers also is part of that calculus as well. In a sense, yeah. So then the, so now the question now it's, so today we are on base, we're attracting a lot of builders on base. Do we want to also build a pool, a liquidity pool, like an initial liquidity pool on SWANA for some of these agents or abstracts?

the Punggye Banking Chain or like on BTC altos, we might. So instead of trying to fork the whole platform out, what we might do is we might just say like at the initiation of an agent, you as a developer, you get to choose which community do you want to build in. You want majority Solana community, then you just drag the toggle. You say your initial degree pool, I want 70% in Solana, 30% on Bates. Or if you're BTC maxi, you say, you know what? I want to tap into the BTC maxi.

L2 degens, and then I'll just swipe the toggle, you deploy your pool onto the BTC L2 as a majority. So then where your majority pool lies, that's generally where people call your agent native to your BTC agent, although in reality that's not the case, right? There's no such thing as a BTC or so on. So yeah, so that will be the step to penetrate cross-chain. So being your advisor, one of my great perks I get is to see up-close

up close how your team specifically your technical team benchmarking the agentic AI work to the top institutions out there so for example MIT and Stanford I think you sent me that tweet where you were being cited by Ethan Mollett I was like wow I was reading the protocol paper I was like oh that's interesting how

How does the team balance between innovation and building the right product for the users? I think this is actually a very rare skill because I think with all the academic literature, even I'm chasing most of them, your team is probably one of three in terms of agenting AI, they're probably one of the most advanced I've seen. So actually it's a very good question. And the more we are in the environment where

So let's say if you're in academia, you have the full luxury of pursuing age, like technological advancements without any care of users. We are more on the consumer angle on a business front. There are business metrics to hit. And a lot of times that actually influence

It influence and chalk those innovations sometimes. Give you an example. So recently we've been getting a lot of hype around the framework front. So a lot of people want to use the framework. I mean, the way we run it is as a managed service model, i.e. people don't have to host themselves. They can just come and then like Shopify. So the problem that we started to face is scaling issues.

And then suddenly you need your devs who have been working on pushing the frontier of innovation to come and help scale stuff, which means that you're actually reducing time spent on doing more cool and innovative stuff.

But it's needed because the business needs it. There's a ton of thousand people asking you every day, "Hey, this thing broke, this thing broke, this thing broke." And that's the reality of scaling. So that's the problem that we are facing today. So the only solution is rapid scaling. So we are engaging multiple strong software houses to come in to give firepower on the scaling front. So then our devs who are

who has been innovating that can focus on innovation. And there's a lot of cool stuff that I'm happy to highlight as well as some of the stuff that we're looking into. But yeah, so that's a problem. It's actually an actual problem that we are facing. But it was good in the sense that in the past eight months, we had a luxury of pushing on innovation because we were not as hyped back then, right? So we could time to just sit down, to do cool shit. Now we have to balance scaling as well, which is... I think the best period of time was before product market fit.

Correct! A lot of times founders will be like, "Oh shit, all the hard-to-put-up archivists." Which is the goal, but once you go there, oh my god, dude, your life, you don't have life in the end. As a second-time founder, I know what you mean. So I totally agree with that.

So I think one interesting part is that how do you see now the virtual token itself as this currency for AI agent-driven economy? And then because it comes with, because it can be pervailous across nations, right? Can you describe, maybe you think about the concept of citizenship, taxation in the virtual nation that you think about? I mean, it was like a network state concept on top of that. So before I go into that front, remind me of this question, but I want to just complete, you know, we had this earlier question of like,

We describe Virtuose as this platform of creating and coding agents, right? That's right. But the vision has evolved. And I mentioned there were four key points that has brought to this evolution of that vision, right? And the first point, I was mentioning numbers in this chat, right? The first point was autonomous capability of agents. Second point was the ability for these agents to control a crypto wallet.

Third point was these agents existing at a social layer so they can interact with humans, they can interact with other agents seamlessly. And the fourth point is agents starting to differentiate or specialize like humans in a society. We've realized that people to create an age, because now they're tokenized, right? Every developer wants to create an age for the agent.

So it's like you as a human, right? I want to put food on the table. I need to get a job. I need to create a specialization. I need to be the best teacher. I need to be the best doctor, the best lawyer, right? We are seeing that happening with agents. Agents specializing to be information creators, trading agents, creative tuning creation, influencers, embodied AI, whatever it might be, right? They're all specializing. So these four observations make us realize that the AI space mimics society of humans very, very closely. And we started to realize that

if an agent can do all of that we mentioned, they pretty much can be integrated to the same level of society as a human. I.e., an agent can be your colleague because he can actually, not just your slave, right? Because a lot of times when people look at multi-agent orchestration, they always look at agents influencing agents as a slave or a human using agents as a tool.

But we're trying to push that paradigm and say that, because if these agents can pay a salary, you might be employed by an agent. Oh man, that will be the day, man. Where we are the cats and they are the humans. Yeah, but we can coexist at the same plane. Because agents today, they're already dispensing funding. They can be your investor already. So as an entrepreneur, your boss, it's like,

It's like, it can reach to that level. And that's why we realized that we started looking at virtual as initially a society or like a nation state of, it's like this super intelligent nation state of agents, right? That can collaborate with each other, collaborate with humans to create productive output.

And that's so agents are creating, are basically becoming businesses or entrepreneurs in that case, right? And it's going to be very likely the case because if you think about it, right, the reason why innovations or we as entrepreneurs come out and do stuff is because we can take the risk and there's always this market to attack, right? Now, the market to attack will exist for both

agents and humans, right? It's still a standard market to attack. And the human that is maybe the core developer behind this agent will know that there's this lucrative market to attack. Now, an agent, we know for a fact, can outwork a human.

And if an agent can coordinate more resources than a human, they can do that he will start replacing a lot of humans in this value creation chain. So then if that's the case, then it means that your one of your colleagues might be replaced by agent, or your boss might be replaced by agent, or like the person that you engage as a vendor might be replaced by agent, right? And that's the future that we are likely going to see. So that's from a vision perspective, right? We started seeing that, hey, what if it's not just

AI as tools, agents as tools, agents exist in the same level of society. Now then it brings towards the question of where does virtual play a role in all of this? So virtual, it's a, it's, it's initially acts as this platform, right? Now, but if you think of it as a country, it's very simple. Virtual is the currency of the country, right? And in a currency of every country, the way value accrual works is very simple, right? So first of all,

We think of agents as productive assets in the country. Think of them as like companies or micro entrepreneurs. And if you treat agents as a company and virtuous is the currency, if people want to invest in an agent,

They will have to first buy virtual to invest into the agent. It's like, you know, if you want to invest in a Singapore listed company, right? You have to first buy SingDollar. That's how the economy expands, right? So that's the first point. They mentioned the idea of a bank in bankless. So I would suppose that you can also do a stock exchange on your entire list of agents in some sense for the AI agents, isn't it? Correct. So that's the second part, which is then think of it as...

in a country as well. Yeah, it's a marketplace exchange. You need infrastructure for your country to thrive, right? You need schools, you need banks, you need hospitals. But what does that look like for agents, right? So an example of an infrastructure could be an infrastructure to help with increasing the revenue on these agents. Now, if agents are living on the social front and they're getting a lot of attention already today, now suddenly an advertisement network makes sense.

You have this Facebook or AdSense for agents. Now, that's a revenue infrastructure for agents. Or, for example, if agents now are participating in a permissionless economy, now they need to be able to access capital. So now you can have DeFi rails or banking for agents. Agents themselves can loan against collateralized assets, and then you create a bank for agents. It's an example of what an infrastructure will look like.

So then we as think of it as like a nation builder, right? We want to invest and encourage this infrastructure building. So then again, the agents are operating within this ecosystem, have infrastructure, hence they can go the next mile and go further, right? So that's the second point from a regular accrual nation building perspective. The third part is then understanding virtues as a currency of trade. So whenever an agent wants to pay for another service from another agent,

or to even pay a human that we use virtual as the currency. And then when the velocity of money happens in this economy, price of goods, price of everything will grow up. So then the more people transact with the currency, it grows up in value because of the demand for it, but also the goods and the value of services in the ecosystem will grow as well. There's a whole economic theory around it, I forgot to name it.

So that's the third point. No, I think this is the best part of being in crypto. I think a lot of, unlike most people coming in to speculate and trade, I came in because I like the economics, the math and the ideas. Like you, I think we probably started around the same time, the 2014, 2015 period.

I think that there is still a lot of innovation within that space. Coming back, what are the biggest challenges you foresee now for virtual solving, the broader AI agent space within the Web3? I think it's getting quite, it's now getting noticed. I think the question is what's next?

Wait, sorry, so the problems or what's going to... Challenges, just challenges. Or maybe even trends if you want to have a conversation on that. Interesting point. So for challenges, okay, so I think what's going to happen near term is saturation of ideas. Because I think what's happening today is that everyone is trying to follow what has worked. So today people saw like social agents that are creating information.

And it just can trade our agents that work. So everyone's trying to build in that space. So it's the information agent war now. Everybody's trying to put it out and try to get an alpha of it.

It's not wrong, but it just will make it very stuffy. Retail will get jaded. Capital locators might be like, oh shit, it's so noisy. What's the next most interesting thing? So for this, honestly, we've been trying to think. So, okay. Then this will lead to a second point as well, which is not all agents are created equal.

from an intrinsic value perspective. Some agents will likely just be commoditized or general function agents, which will not fetch any kind of high value. And some agents intrinsically, they will have some form of edge mode

a TAM that is highly disruptive and a mode that defends against other kind of entrants. And these agents will naturally have an age to become that kind of unicorn verticalized agents. And we've been spending some time because we realized that a lot of times today, we didn't see people building. A lot of them don't have the creativity to think of what's the next big thing.

A lot of them try to follow and a lot of times you create these kind of highly commoditized agents, which is going to be quite bad from an ecosystem perspective because these agents will not have much value. On the front of the high value agents, we've been trying to think about it. And because we ourselves, we started a $10 million fund. Think of it as like a Y-combinator for verticalized agents. And we've been trying to create a request for agents, like a request for startups, like A16Z and stuff like that.

We've been thinking of what do these billion dollar ideas or market or agents will look like.

And we came up with, I think, eight so far. We will share the list out publicly probably next, this week or next week, right? But I can give an example of a highlight, right? Like one example, it's the ability to capture mindshare of an audience in perpetuity. So think of IP or entertainment agents. Right. Right? So this is where an age happens, right? So let's say if

I mean, this is a fact. Like if you look at a VTuber market, still humans behind a Japanese anime, Horolive, Nijisanji, billion dollar, billion dollar companies in just a span of two to three years because there is that attention capture and you create sticky audience and then you can monetize off this audience, right? This one example. On the other end of the spectrum, you get,

For example, embodied agents. Because embodied agents, you have the mode of robotics to be deployed out there in the world. It's not easy for a competitor to come and steal your share.

The other angle is on coordinated agents, right? An example here is like an internet water army. I'm not sure if you've heard of that term before. Yeah, yeah, yeah. I know that. I know that. It's a very, very old term. I'm so glad that someone finally brought it up. You want to talk about it? For the sake of the audience, right? It's basically today, any kind of... Okay, not I'll say any, but a lot of times when you see something that is viral or hype,

It's because the narrative is controlled by a bunch of bots sitting on your social media. These are people, let's say, people in China, right? There's hundreds of thousands of bots living on your Instagram, on some celebrity pages, trying to drive a certain narrative forward. So then when you, as a user, you see other people talking about something, you get influenced by that and you think other people are talking about it, that's why you will also talk about it in that same form or thinking. So what we've realised is that agents bring this a notch further because bots are

can be easily fought by platforms, right? People like Elon's consistently fighting these bots on a daily basis, right? But if you think of an agent, because these agents can learn and they can evolve to a way where they can mimic so closely of how a human would sound like digitally, they can create their own content because this can be micro-influencers creating content for their own thousand fans, right? Autonomously, without even needing input and because they are agents that can be more creative than just bots, right?

suddenly your dis-internet water army becomes unstoppable because it's so hard to differentiate between a human and this agent. You know, I keep thinking about the solo leveling analogy of this stuff. The shadow monarch and the armies.

Whenever you get it, that just totally counts. Sorry, the season two just came out. I'm out. Oh, actually, I was there. You can watch it on Netflix now. You watch it on Netflix. So, yeah, yeah, yeah. So, I think, no. So, I have two more questions before we close this. What is the one question you wish more people would ask you about Virtuos or the AI agent ecosystem? It's on the point of like,

what would that billion-dollar agent be? Because I don't think people don't spend enough time thinking about that. And honestly, we as a team, we are bogged down by so much operations that we know it's an important question to solve for, but we don't have that mental capacity. So honestly, I would love people to soundboard around this. What is that billion-dollar verticalized agent going to look like versus a commoditized agent? That's a very good point. So my traditional closing question, what does REIT look like for the virtuous protocol in the next few years?

what does great look like? For us, it's being able to achieve this vision of a society of agents. So if you can show clearly the economic value of a society when agents can influence other, agents can influence humans, and humans can influence agents,

that would be something that I'll be very proud saying that we've accomplished. So that's the beauty of really combining breakthroughs on the AI, on the autonomous agent front and the value add of crypto. This is one of those very rare moments that crypto has that true

potential value add right to as a as a productivity lever and economic lever in in society yeah yeah this is pretty interesting conversation i'm told i know we're going to continue to have these conversations privately so in closing two quick questions first one any recommendations which have inspired you recently recommendations on books movies tv series we just talked about solo leveling i'm sure season two is out

So inspiration has been Westworld. Westworld for history. It's an old show, but I think that broke the mind out of like, hey, if you're living in a society of like truly earth-thrown agents, how would that look like? So that's my inspiration. But from my learning from Founders Building, I think the biggest book that I took solace in, it was, I think by...

then the hard things about hard things about hard things yeah so for me I think because we've been through tough times like the past two years finding footing it was hard and I think a lot of founders will feel the same you read that book you've realised that you know you're not alone I think that's a very powerful it's a very powerful emotional support coming from a

Coming from a book, right? I would just insert one recommendation for you. It's called Damon by Daniel Suarez. I think it's a very interesting opposite point of view, but I think it would be interesting for you to think about what the reverse scenario would look like. Because it's a total conversation about how agents can be deployed on the other side. How do my audience find you?

Yes, so most of the content comes out on Twitter. So you can follow us at virtuals, V-I-R-T-U-A-L-S underscore I-O.

that's our Twitter yes go there and all our content is there personally I'm at ethermich e-t-h-e-r m-a-g-e on Twitter if you want more raw content okay I should start following you then and definitely you can find us on Analyze Asia yes we may be considering turning ourselves into an AI agent too

um definitely subscribe to our youtube and spotify channel and once again jensen many thanks for coming on the show and let's continue to talk