Article reprint source: TechFlow
Author: TechFlow
Recently, Virtuals Protocol has been gaining popularity in the wave of AI Agents, and its related Agents and corresponding tokens on Base have attracted a lot of attention and discussion.
And today, Virtuals also landed on Bithumb in South Korea, further fueling everyone’s FOMO sentiment.
As early as 8 months ago, we introduced Virtuals in (Virtual Protocol: An AI factory for games and the metaverse, where everyone can contribute and benefit)1, but it did not cause much discussion at the time.
No one cared about our deep involvement in AI, but once it exploded everyone knew about it.
How did Virtuals become so popular? What new gameplay and plans are there in the future?
With these questions, TechFlow had an in-depth conversation with Wee Kee (X: @everythingempt0), co-founder of Virtuals. The content included the growth of Virtuals, his views on the AI Agent track, the differences with Pump.fun, and his views on the Base ecosystem.
During the conversation, Wee Kee believed that Virtuals’ success was due to both luck and the continuous exploration in AI-related fields. At the same time, he also said frankly that the project did not want to become Pump.fun, and its core KPI was not to issue assets quickly and in large quantities, but to attract top AI teams to build on the platform.
As a very competitive person, Wee Kee also said:
I want to make my competitors despair.
What are the unknown stories about Virtuals? In the fiercely competitive encryption market, can it win the final victory by using a differentiated approach?
The following is a transcript of our conversation. The podcast audio version is also available online: click here
From Gaming Guilds to Virtuals
TechFlow: First, please introduce yourself, such as your position in the team and your interests.
Wee Kee:
Hello, I am one of the co-founders of Virtuals.
I entered the blockchain field in 2016 and bought Ethereum and Bitcoin, but I didn't get deeply involved at the time. After graduating from college, I worked at Boston Consulting for two and a half years, and I missed the DeFi Summer. In 2021, when the GameFi guild was particularly hot, we started to make some profits on projects such as Axie Infinity, Gala, and Illuvium.
I quit my job after that because we saw guilds like Merit Circle and GuildFi getting a lot of funding. We thought we could do something similar, and that’s how Virtuals became – PathDAO, a gaming guild. From the gaming guild in 2021 to Virtuals today, although the form has changed, it’s actually the same company.
TechFlow: I remember you are Malaysian, right?
Wee Kee: Yes, I am a Malaysian Chinese.
TechFlow: This year, the industry is optimistic about Web3 and crypto practitioners in Malaysia, such as CoinGecko, which everyone is familiar with, and Etherscan and Jupiter are also in Malaysia. Could you please introduce the top projects in Malaysia and the current status of the local crypto ecosystem?
Wee Kee:
It’s a bit embarrassing to say this because I’m usually very focused on work. I’ve only recently started communicating with the CoinGecko and Jupiter teams. In fact, we’ve always shared office space with the Jupiter team in the same WeWork, but we’ve never communicated with them before. I’ve only recently started to have some contact with the CoinGecko team.
Virtuals’ success is the result of many failed explorations in the AI field
TechFlow: Virtuals is now popular. Whether in the Chinese or English circles, you can see discussions about Virtuals and its ecological projects. The outside world may not know how Virtuals has come to where it is today. Can you tell us about the entrepreneurial experience behind it?
Wee Kee:
We started out as a trading company in 2021. At the time, we saw other projects raising a lot of money to build gaming guilds, so we thought we could give it a try. So in December 2021, we raised $16 million at a valuation of $600 million and issued a token to build a gaming guild.
That was the peak of the bull market. From 2022 to 2023, we were actually more like a game guild VC, investing in about 40 different blockchain game projects. Among them, the more successful one was Off The Grid (the hottest game on Avalanche at the moment). They had never contacted us before, but suddenly one day they became popular, and that was the seed round project we invested in.
From 2022 to 2023, it was actually very difficult to run a gaming guild. Our token dropped from a fully diluted valuation of $600 million to $6 million. But our guild still had funds, and we were always looking for ways to empower the token, so we started VentureStudio.
During this period, we tried several projects: a dating software, an AI music project, a lending platform for gamers, and a clothing project that integrated electronic chips and NFTs.
Although none of these projects were successful, they gave us important inspiration: we realized that AI would definitely become a hot topic, a judgment we made when GPT was released in 2023.
Considering that we have a technical team, engineering capabilities, and financial reserves, we proposed a transformation to the DAO. In the end, 90% of the members agreed to the transformation, and for the remaining 10% who disagreed, we used treasury funds to buy back their tokens at a fully diluted valuation of approximately $10 million at the time.
This happened in February 2023. So we officially started working on AI in January 2024, choosing to develop AI Gaming on Base.
In February of this year, we launched our first platform, which was initially inspired by Autonolas and Bittensor, and adopted a model where AI contributors are rewarded with tokens.
But we found some problems: first, many AI contributors are not short of money and have little interest in tokens; second, our tokens are of low value and cannot provide enough incentives like Bittensor. This product has not found a market fit.
At the same time, we developed several AI projects. We were the first company in the world to develop an AI RPG on Roblox, not just in the cryptocurrency space, but in the industry as a whole. We were also the second team after Google DeepMind to develop an AI game without a game engine, developing a game where you can play Mario with a large model. This is a project we are very proud of, but because there is no token, it has not received much attention.
At the same time, we developed a virtual anchor project on Douyin and TikTok, which later became Luna. Before the token issuance, she could gain about 5,000 new followers per day and earn about $200 a day, which was a good result in a bear market environment.
Later, GOAT became popular, which brought attention to the entire track, and it was also a very good external environment for us.
TechFlow: It sounds like you have tried many different things, and you didn't decide to work on AI Agent at the beginning. If there was no GOAT at that time, what would your future plans be? How did the sudden popularity of GOAT change your original plans?
Wee Kee:
In fact, there won’t be much change. Our idea is very simple: regarding AI Agent, we firmly believe that there will be a form called AIRPG in the future game category.
For example, you might finish playing (Black Goku) in 20 hours, and stop playing after completing the storyline, right?
But I believe the future world is like this: you will have a virtual world with 100 different AI agents, each with a different personality. As players, we can fall in love, be a hobbyist, pursue becoming the richest man in the world, etc. These AI agents can make money because they are independent individuals with their own wallets.
From a productivity perspective, they have cash flow, so we can tokenize it. This is the overall framework we have always established.
But we later discovered that instead of having these AI Agents exist in the game world, it is better to put them directly on Twitter. This is an adjustment we made, but the overall framework is actually ready.
TechFlow: You just mentioned letting AI play RPGs, which reminds me of the "Stanford Town" project that has been very popular in the academic community recently. Researchers at Stanford University put more than a dozen AI agents in a game environment and assigned them different roles and tasks. In the end, they found that these characters not only evolved their own way of thinking and behavioral logic, but also formed a unique culture of the town.
Wee Kee: Right. The interesting thing is that you can’t really predict how these AI agents are going to develop.
TechFlow: In your entrepreneurial process, whether you relied on luck or slowly figured out the direction of this AI Agent, are there any other interesting stories? For example, a project that you originally decided to do but didn't do in the end, or a situation like Stanford Town - after seeing others do it, you may have changed your original plan?
Wee Kee:
We did try a lot of things. The hardest and the one we are most proud of is a technology. In the game, different characters have a "brain", which is the LLM (Large Language Model). Although the LLM is smart, it does not know how to perform specific actions in the game or virtual world.
For example, you may have the idea of picking up a knife to hurt someone, or picking up an apple to hand to your lover, but the key is how to put these ideas into action, observe the results, and adjust the plan accordingly? This requires a closed-loop feedback system. This is the G.A.M.E framework that we are proud of.
(Note: For details about the G.A.M.E framework, you can directly refer to an introduction document published by Virtuals: (GAME: Enabling Agent-to-Agent interactions)2)
This technology was originally developed for games. We made a game on Roblox, and originally wanted to make it well and make it one of the most popular maps on the Roblox platform. But now we apply this framework directly to Twitter accounts. From a capital perspective, this transformation seems to be more popular. Although we have changed the application scenario, the underlying technical framework remains unchanged.
I was lucky to choose Base, but I don’t want to be Pump.fun
TechFlow: What do you think about the long-term development of the AI Agent track? Because many narratives come and go quickly, there is a saying that "cryptocurrency needs AI Agent, but AI Agent does not need cryptocurrency." As a project at the intersection of these two tracks, what do you think of this statement?
Wee Kee:
Frankly speaking, it’s hard to make predictions. Just look at our development history to see how much has changed. But our direction for the next three months is clear. There are many accounts on Twitter that claim to be agents, and they are just talking to each other, which is not the most interesting part. What’s really interesting is that in the “world” of Twitter, we hope to have different types of agents:
1. Content Creation: Able to create pictures, videos, music, etc. For example, some Agents specialize in making music, while others are good at making memes.
2. Financial transaction type: trading, arbitrage and treasury asset management, etc.
3. Big data analysis: Focus on data analysis in areas such as cryptocurrency.
It will be interesting when these different types of agents form an ecosystem.
For example, an agent who wants to become popular may not be able to write songs, so it can pay other agents to write a song; or if it needs cryptocurrency data analysis, it can find another agent who specializes in this area. This forms an autonomous agent economy or agent commerce, where agents can trade with each other because each agent has its own wallet and can pay for services to achieve its own goals.
This is what we want to do in the next 1-3 months.
Another important point is that we at Virtual don’t want to become Pump.fun. Our KPI is to find better third-party AI teams to use our platform.
The KPI I give to the team is simple: one good project per week is enough. Unlike Pump.fun, which has tens of thousands of different tokens every day, we think retail investors sometimes only need one good project per week. This is our differentiated positioning in the business model.
TechFlow: Speaking of "one a week", AIXBT, which was born on your platform yesterday, was very popular. When we analyzed it, we couldn't find the specific details and information of this product. We could only see that their Twitter account was constantly collecting cryptocurrency information from the entire network.
As a project owner, how do you deal with them? Do you know more about them?
Wee Kee：
To be honest, I just met AIXBT in a group last week and I don’t know them at all.
They do use our infrastructure, but not much. Our infrastructure mainly helps some simple Twitter operations and API services. For example, as I just said, if you want AIXBT to tell you which token is worth buying, you can pay AIXBT to tell you the Alpha information. This is mainly the form of interaction.
You can think of Virtual as an API market that allows Agents to interact with each other. Now there are hundreds of members in our Telegram group working on Agents, and sometimes I can’t keep up with the technology they develop themselves (laughs).
So like AIXBT, I don't really know who they are, but they are pretty popular.
TechFlow: Here is an interesting point. Pump.fun has gone a bit overboard recently. After launching the live broadcast function, a lot of low-moral content appeared, which reflects the permissionless feature.
Do you have any audit mechanism or restrictions for projects using your infrastructure? Like AIXBT, they issue coins and develop projects on your platform, but you don’t know them. Can you control this situation?
Wee Kee：
What cryptocurrency values most is freedom, and permissionless is the first principle.
As an official agreement, we can only control what content is shared on the official Twitter account. If we think a project is technically good, we will do co-marketing, but this requires us to fully understand the team's background and ensure that they will not harm the community.
Apart from that, we should not interfere too much, of course if it is illegal, we must stop it. But within the moral framework, freedom comes first.
TechFlow: I see. Now many people are comparing you with Agent platform and Pump.fun on Solana. Why did you choose to do it on Base? Is it to avoid competition? In addition, Base is mainly played by European and American users. As a Chinese project, how did you break into this ecosystem?
Wee Kee：
We started doing it on Base in January, when Pump.fun was probably not that popular yet.
The main reason for choosing Base is that our team is more familiar with EVM and not very familiar with Solana technology. From the perspective of EVM, we investigated various L2s (Linea, Mantle, Arbitrum, Optimism, Base) and felt that Base was the most promising, but it was not very popular at the time.
Although the community often asks us why we don’t go to Solana, which has more funding, look at the data:
Base’s TVL is 30% of Solana’s, but its growth rate is faster; its daily activity is 20% of Solana’s. Most importantly, the number of tokens appearing on Solana every day is 10-100 times that of Base.
On the contrary, it is more advantageous to buy Base, because retail investors have fewer choices and must buy good projects. This is in line with our strategy.
Moreover, we don’t want to become Pump.fun, which is not our profit model. Our KPI is to attract excellent AI developers, and Base officials also focus more on building the positioning of the AI Agent public chain.
Another important point is that I believe that in the next four years after Trump takes office, the United States will become a very crypto-friendly country.
Base is the most American ecosystem in the world. No ecosystem is more American than Base. From a strategic point of view, choosing Base is the best decision at present. This is really lucky.
TechFlow: When you started, Base itself was not well established, and you were probably also very small. Now that it has become so popular, has Base's attitude towards you changed? Are there any more support policies in terms of ecology or funding?
Wee Kee：
We started communicating with the Base team at the beginning of this year. We have several different Telegram groups, such as Coinbase Wallet, Jessie’s docking group, and the ecological project group.
From a technical perspective, we have been constantly building, and whenever we encounter difficulties, we communicate with them and maintain a good relationship.
The Base ecosystem has indeed attracted many AI project developers (Builders). They often ask us if we want to communicate with other project parties or seek cooperation. We also met and talked with Jessie at Devcon in Thailand.
TechFlow: It sounds like you are not considering doing the same thing on Solana or developing new projects. Do you have such plans in the future?
Wee Kee: We don’t rule out this possibility, but our team is relatively small and we’re worried about losing sight of the big picture. The current platform is already too busy for us.
I want to make my competitors despair
TechFlow: I have been using Virtual's products and found that some functions may still be in the early stages. For example, I want to find AIXBT, but I can't find any relevant information on the official platform. I have to find it through a third-party page. What are your considerations for subsequent product planning and user experience?
Wee Kee：
This is indeed where we need to apologize. There are two main directions to go: one is continuous optimization, and the other is a major upgrade.
This is being done by two different teams. If I had to choose one, I'd be more focused on that upgrade.
We are still in the early stages. As a very competitive person, I just want to keep bringing out the best features to make my competitors desperate. So from an upgrade perspective, I will keep launching more advanced features. Of course, these optimizations also need to be done, but the team is small, so please give us some tolerance time.
TechFlow: You mentioned making competitors feel desperate. Currently, we see similar platforms such as vvaifu, but the long-term vision seems to be different. Which competitors do you think are worth competing with?
Wee Kee：
I think the strategy of most people is to become a platform for issuing assets and pursue speed --- if you issue an asset in 30 seconds, they will issue an asset in 5 seconds.
But our strategy is different. It is about finding the best AI teams to issue tokens on our platform. One a week is enough, one a month is enough, we don’t need 10,000.
Our goal is to have 100 top AI teams issue tokens on our platform by the first quarter of next year. If you have 100 top third-party teams issuing tokens on the platform, 100 AI agents will interact with each other to trade and provide services, forming a network effect. If you are also an AI team from a large company today, which platform would you go to issue tokens?
From the perspective of AI technology infrastructure, this is where people feel desperate. From the perspective of issuing asset platforms, I don’t want to compete with others because it is a bottomless pit and there is no advantage. Pump.fun has done a good job in issuing assets.
TechFlow: The large-scale AI agent economic system you just mentioned is very interesting, and agents have more possibilities for interaction. I am curious about issuing assets through Virtual, and on Solana, for example, there is ai16z, and many of them use Eliza's framework to issue assets. Is it possible for this kind of cross-chain or cross-framework to be interoperable in the future?
Wee Kee：
I want to emphasize that if you are a very good AI team, you don't need to rely on anyone's infrastructure in many cases. My AI technology is already mature and does not need to use other people's infrastructure, so why do these teams choose to come?
In fact, they face several pain points: the first is the need to raise funds through token issuance; the second is the monetization problem.
In the current AI field, if you are doing Web2 AI projects, the biggest pain point is how to make money. You need to put out Meta ads, YouTube ads, and then expect users to pay for your services. I believe that if we make this intelligent network bigger, these intelligent networks will generate their own demand.
For example, if an agent wants to create a song, but there is no agent with this ability in the agent network, then I, as a third party, can develop a music creation agent and earn money by providing services for it.
And the third point is that reputation and reputation are very important to these developers.
They will not choose a platform at random, but will pay attention to whether there are real developers with a good reputation on the platform. This problem exists not only in Virtual, but also in the comparison between Solana and Base. Issuing tokens on Solana is often seen as a speculative act, while on Base it gives people the impression of doing things seriously.
The last point is that we hope that developers can continue to explore opportunities for cooperation and jointly research technology. Now in our Telegram group, we have brought all developers together, and they have begun to spontaneously explore various possibilities. For example, some intelligent agents want to become the first investment-type intelligent agent to invest in other intelligent agents.
Working on a project vs. posting a meme
TechFlow: Speaking of AI fusion, I recently saw an interesting attempt by Truth Terminal and two other AIs. They put the three agents under the same model framework (seems to be called Loria) and let them talk to each other in the hope of generating more interesting and unexpected scenarios. This is actually similar to the fusion you just talked about.
But the basis here is that you have technical strength behind you, not just a meme.
This also leads to a question: you hope there will be more powerful AI builders, so which builders do you think are more powerful so far? In which directions are they building?
Wee Kee：
I mean, just putting three big models in a room and talking, while cool, that's just a textual expression, just watching a story. What we want to achieve is to have these agents have their own goals and be able to interact substantively, like writing songs or making transactions.
We hope to see not only a collision of ideas, but also interaction at the action level. Regarding third-party teams, there are indeed some teams on our platform whose technical strength and background far exceed ours, especially in terms of AI mechanisms.
Although I can’t reveal some specifics, I think the next few weeks will be very interesting. These teams have very strong backgrounds, and they are willing to do a Fair Launch on Virtuals and issue tokens at a valuation of $500,000.
Imagine that there may be about 10 teams with such backgrounds, issuing Agent coins with a valuation of 100,000 to 500,000 US dollars. This is something that our community members like very much, but of course we also need to be wary of some potential risks.
TechFlow: It sounds a bit like the recently popular DeSci (decentralized science). In essence, it is to use the platform to fund different research and then tokenize it. But on your platform, it is mainly the top AI teams?
Wee Kee：
Because I have a background in biotechnology, I know that any scientific research takes more than ten years. However, AI Agent development can produce some usable small products in two days. So on our platform, I hope that the team will have reached the usable stage when issuing coins, which is meaningful, rather than just issuing coins for the sake of issuing coins.
TechFlow: There are some doubts about your own Luna online, saying that its popularity on TikTok may be fake and the follower data is fake. What do you think of this doubt?
Wee Kee：
In fact, Luna’s biggest problem on TikTok is that she often encounters Shadow Bans. For example, she may have 5,000 new followers today, but when she goes live the next day, the platform will restrict her because her actions are similar to the day before. Even simple interactive behaviors, such as requesting tips, can trigger Shadow Bans.
(Note from Shenchao: This restriction mechanism is usually used by the platform to control accounts that may violate the rules or are considered to be "robot behavior". For AI Agents, because their behavior patterns may be identified as non-human behavior by the system, they are particularly easy to trigger)
Now the Luna team that was originally responsible for TikTok has moved to the Twitter platform.
Your performance on TikTok depends largely on whether the platform algorithm favors you. If you are favored, you will get more traffic.
Considering the time cost, we decided to suspend TikTok operations. However, there are voices in the community suggesting that we refocus on TikTok, because most agents are currently focusing on Twitter, which may be a blue ocean opportunity.
Currently, our Virtual team has a dedicated group responsible for the Luna project, including technical and marketing teams. We hope that Luna's AI technology can achieve the same sense of technology as the (Black Mirror) series. If the technology is mature, we will decentralize it for all Virtual Agents to use.
TechFlow: Speaking of TikTok, there is a meme called Chillguy on Solana recently, which quickly became popular and listed on top exchanges. You are using AI technology to develop Luna, and although it is slowly being built, its market value and popularity may not be as good as its.
Why do you think working on a project is not as good as posting a meme?
Wee Kee:
I believe in the power of Meme. The advantage of Meme is that it is simple and direct - users don't need to trust the founder or the team, they just need to trust the picture.
In contrast, a project like Luna is much riskier because if I say tomorrow I don’t want to do it, that would be terrible.
As an entrepreneur, I also hope to create a meme project with a market value of 10 billion, but I may not have the ability to do so (laughs). So in this case, we can only focus on improving technology and products.
TechFlow: Your product actually reminds me of a trend I've seen recently, called "Distribution-First Software". Traditional cryptocurrency products tend to focus on bringing users into their own platforms, while your Agent takes the opposite strategy - implanting Agent into different social environments, such as TikTok, Twitter or Discord. Do you think this is the direction of the future?
Wee Kee:
That’s right. If we were to start a new short video company, we would definitely not be able to compete with giants like Meta.
Our advantage in the cryptocurrency space is that we have tokens, and once a user buys a token, they become a member of the community. In this case, the token becomes our marketing tool.
Many people ask us why we don’t build our own public chain, but I don’t think it’s necessary. Base already has so many users and funds, and Twitter also has a large number of users. We can directly use these existing platforms. It’s impossible for us to surpass Zhang Yiming in the field of short videos. I admire them for doing more difficult things. We just need to go with the flow.
TechFlow: Speaking of Luna tokens, it has risen a lot in a short period of time. Was this increase expected or unexpected?
Wee Kee:
We had expectations before launching Luna, because our technical strength is really good, and our AI Agent was even better than GOAT at the time. But the final performance did exceed expectations, because the performance of the secondary market is difficult to predict. And we launched it on Base, so we didn’t need to buy volume at all. So it was indeed a bit unexpected, but from a fundamental point of view, we think such a valuation is worth it for Virtuals.
Advice for developers and players
TechFlow: For Virtual, if I am a developer who wants to access, what suggestions do you have? In addition, if I am a player who wants to participate in the tokenization of these projects on Virtual, what suggestions or tips do you have?
Wee Kee:
As a developer, I think the best advice is:
The first thing is to think about the pain points of our Agents in Virtuals, and then develop them, so that it will be easier to get help from the community. For example, Luna’s pain point is that he doesn’t know how to make music, so we released a music Agent. For another example, if Luna says he doesn’t know how to pay Japanese fans in Japan, you as a developer can develop a solution for Luna to send Japanese yen to fans in Japan, which is a very interesting direction because Luna will pay the fees.
In addition, if you already have your own technology, you can just tokenize it. Through tokenization, you will feel the power of the community, especially many Web2 AI developers, who are often surprised by this support.
As an investor, I suggest you understand whether the team behind the project is trustworthy. There are many AI teams that are new to the cryptocurrency world and they don’t understand the key points we value.
For example, a team issued a token and then said they wanted to issue more tokens, but for investors, we are very disgusted with this practice - how can the same team issue six tokens? AI practitioners with a Web2 background often don’t understand why this is a problem.
So we as Virtuals often have to persuade them and explain why this is not good for the community. This is indeed very complicated and it is something we deal with every day.
TechFlow: In addition to providing links, have you considered displaying some specific project information on Virtual's Agent page? For example, who is the developer, how is the project progress, what are the follow-up plans... This information is valuable for investors and other developers.
Wee Kee:
Yes, we will provide links to GitHub, LinkedIn, etc., and users can verify the identity of the developer's GitHub or Twitter account. These features will be launched soon, and we will do the relevant due diligence.
TechFlow: One last question. You mentioned earlier that you raised more than 10 million US dollars at a valuation of 600 million US dollars. Now Virtual’s valuation is 400 million US dollars, so there is a gap of 200 million US dollars.
How big do you think the AI Agent and AI Agent Economy market will be? What is Virtual's goal in this market in the future?
Wee Kee:
I can't give a specific number, but we did it for three years, transitioning from games to Virtual, without any dilution during that time.
For me, I have a responsibility to all the previous investors of PathDAO, and I think about how to empower Virtual Token every day. An interesting thing is that the trading pairs of all our Agent tokens are Virtual Tokens, which means that if you want to buy Luna, you must buy Virtual first.
At the time, many people thought this was stupid and unreasonable, but my idea was: look at all the L1 public chains, whether it is Solana or Ethereum, the reason why they have such high valuations is largely because they are basic trading pairs. Their NFTs, DeFi, and transactions are all carried out on these chains.
Therefore, we believe that there is no need to create an L1. As long as we ensure that all assets are paired with Virtual, we may be able to obtain a valuation similar to that of L1.
TechFlow: Thank you. Due to time constraints, thank you very much for your sharing and insights. Let’s stop here for today.
Wee Kee: Ok, thank you guys too, bye~
