From $6M to dominating Base: Virtuals Protocol aims to onboard 100 elite AI teams by 2025 Q1

Recently, Virtuals Protocol has gained significant attention amid the AI Agent wave, with its Agent-related tokens on Base chain attracting considerable interest and discussion.

Today, Virtuals was also listed on South Korea's exchange-Bithumb, further fueling FOMO sentiment.

Eight months ago, we introduced Virtuals in "Virtual Protocol: An AI Factory Born for Gaming and Metaverse, Where Everyone Can Contribute and Benefit," but it didn't generate much discussion at the time.

Deep in AI development with no attention, sudden rise brings worldwide recognition.

How did Virtuals become popular, and what new features and plans are in store?

With these questions in mind, TechFlow conducted an in-depth conversation with Virtuals' co-founder Wee Kee (X: @everythingempt0), covering topics including Virtuals' growth, views on the AI Agent sector, differences from Pump.fun, and perspectives on the Base ecosystem.

During the conversation, Wee Kee attributed Virtuals' success to both luck and continuous exploration in AI-related fields. He also stated that the project doesn't aim to become Pump.fun, with their core KPI being attracting top AI teams to build on the platform rather than rapid asset issuance.

As a highly competitive person, Wee Kee also stated directly: "We Want Competitors to Feel Despair."

What untold stories are there about Virtuals? In the highly competitive crypto market, can it succeed through differentiation?

From Gaming Guild to Virtuals

TechFlow: First, please introduce yourself, including your role in the team and interests.

Wee Kee:

Hello, I'm one of the co-founders of Virtuals. I entered the blockchain space in 2016, purchasing Ethereum and Bitcoin, but didn't get deeply involved then. After graduating from university, I worked at Boston Consulting for two and a half years, missing the DeFi Summer. In 2021, when GameFi guilds were extremely popular, we started earning from projects like Axie Infinity, Gala, and Illuvium.

I then quit my job because we saw guilds like Merit Circle and GuildFi receiving significant funding. We thought we could do something similar, which became Virtuals' predecessor - PathDAO, a gaming guild. From the gaming guild in 2021 to today's Virtuals, though the form has changed, it's the same company.

TechFlow: I remember you're Malaysian, aren't you?

Wee Kee:

Yes, I'm a Malaysian Chinese.

TechFlow: This year, the industry has been optimistic about Malaysian Web3 and crypto practitioners, like CoinGecko, which everyone knows well, and Etherscan and Jupiter are also in Malaysia. Could you introduce some of the top projects in Malaysia currently and the local crypto ecosystem?

Wee Kee: It's a bit embarrassing because I've been quite focused on work. I only recently started communicating with the CoinGecko and Jupiter teams. In fact, we've been in the same WeWork space as the Jupiter team but hadn't interacted before. We've only recently begun engaging with the CoinGecko team.

Virtuals' Success Built on Multiple Failed AI Explorations

TechFlow: Now that Virtuals has become popular, with discussions in both Chinese and English communities about Virtuals and its ecosystem projects, the outside world might not understand how Virtuals got to where it is today. Could you share the entrepreneurial journey behind it?

Wee Kee:

We started with trading in 2021. When we saw other projects raising significant funding for gaming guilds, we thought we could try too. In December 2021, we raised $16 million at a $600 million valuation and issued a token, operating as a gaming guild.

This was at the peak of the bull market. From 2022 to 2023, we essentially operated as a gaming guild VC, investing in around 40 different blockchain gaming projects. One of our more successful investments was Off The Grid (currently the hottest game on Avalanche). They hadn't contacted us before, but suddenly became popular - they were one of our seed round investments.

From 2022 to 2023, running a gaming guild was extremely challenging. Our token fell from a $600 million FDV (fully diluted valuation) to $6 million. However, we still had funds in our guild treasury and were constantly seeking ways to add utility to our token, so we started operating as a Venture Studio.

During this period, we experimented with multiple projects: a dating app, an AI music project, a lending platform for gamers, and even a clothing project integrating electronic chips with NFTs.

While none of these projects succeeded, they provided crucial insights: we realized AI would become a major trend, a conclusion we reached when GPT was released in 2023.

Considering our technical team, engineering capabilities, and financial reserves, we proposed a pivot to the DAO. Eventually, 90% of members supported the transition, and for the remaining 10% who disagreed, we used treasury funds to buy back their tokens at what was then approximately a $10 million fully diluted valuation.

This happened in February 2023. We officially started working on AI in January 2024, choosing to develop AI Gaming on Base.

In February this year, we launched our first platform. The initial inspiration came from Autonolas and Bittensor, adopting a model where AI contributors receive token rewards.

However, we identified several issues: First, many AI contributors weren't lacking in funds and showed little interest in tokens; second, our token value was too low to provide sufficient incentives like Bittensor. This product didn't find its product-market fit.

Meanwhile, we developed several AI projects. We were the first company globally to develop an AI RPG on Roblox, not just in crypto but across all industries. We were also the second team after Google DeepMind to develop AI games without a game engine, creating a game where you could play Mario using large language models. This was a project we were proud of, but it didn't receive much attention due to not having a token.

Simultaneously, we developed a virtual streamer project on Douyin and TikTok, which later became Luna. Before token issuance, she was gaining about 5,000 new followers daily and generating around $200 in daily revenue, which was decent considering the bear market conditions.

Later, when $GOAT became popular, it boosted attention across the entire sector, which was also very beneficial for us in terms of the external environment.

TechFlow: It sounds like you tried many different approaches and didn't initially focus on AI Agents. If GOAT hadn't suddenly become popular, what were your plans? How did $GOAT's sudden success affect your original plans?

Wee Kee:

Actually, it wouldn't have changed much. Our thinking was simple: regarding AI Agents, we strongly believed that a new genre called AIRPG would emerge in gaming.

For example, you might finish playing "Black Wukong" in 20 hours, complete the storyline, and stop playing, right?

But I believe the future will be like this: you'll have a virtual world with 100 different AI Agents, each with distinct personalities. As players, we can enter to pursue romance, become hobby enthusiasts, strive to become the world's richest person, and so on. These AI Agents can earn money because they're independent entities with their own wallets.

From a productivity perspective, they have cash flow, so we can tokenize it. This has been our established overall framework.

However, we later realized something: rather than having these AI Agents exist in a game world, it would be better to put them directly on Twitter. This was an adjustment we made, but the overall framework was already prepared.

TechFlow: You mentioned letting AI play RPGs, which reminds me of the recent "Stanford Town" project that's popular in academia. Stanford researchers placed over a dozen AI Agents in a game environment, assigning them different roles and tasks. They found that these characters not only evolved their own thinking patterns and behavioral logic but also formed the town's unique culture.

Wee Kee:

Yes. Interestingly, you really can't predict how these AI Agents will develop.

TechFlow: In your entrepreneurial journey, whether through luck or gradually finding your way to AI Agents, are there any other interesting stories? For instance, projects you decided to do but ultimately didn't, or situations like Stanford Town - where seeing others' work might have changed your original plans?

Wee Kee:

We indeed tried many things. The most challenging and proudest achievement was a technology. In games, different characters each have a "brain," which is the LLM (Large Language Model). Although LLMs are intelligent, they don't know how to execute specific actions in games or virtual worlds.

For example, picking up a knife to harm someone, or taking an apple to give to your loved one - you might have the idea, but how do you translate these thoughts into actions, observe the results, and adjust plans accordingly? This requires a closed-loop feedback system. This is our proud G.A.M.E framework.

(TechFlow note: For details about the G.A.M.E framework, you can directly check Virtuals' published introduction document: "GAME: Enabling Agent-to-Agent interactions")

This technology was initially developed for games. We made a game on Roblox, originally intending to make it one of the most popular maps on the platform. But now we've applied this framework directly to Twitter accounts. From a capital perspective, this pivot seems more welcome. Although we changed the application scenario, the underlying technical framework remains unchanged.

Lucky to Choose Base, But Don't Want to Be Pump.fun

TechFlow: How do you view the long-term development of the AI Agent sector? There's a saying that "crypto needs AI Agents, but AI Agents don't need crypto." As a project at the intersection of these two sectors, what's your take on this?

Wee Kee:

Frankly, making predictions is difficult. Just look at our development journey to see how much things can change. However, our direction for the next three months is clear. Currently, there are many accounts on Twitter claiming to be Agents, just talking to each other, but that's not the most interesting part. What's truly interesting is that in the "world" of Twitter, we want to have different types of Agents:

1. Content Creation: Capable of creating images, videos, music, etc. Some Agents specialize in music, others in making memes.

2. Financial Trading: Handling trading, arbitrage, and treasury management.

3. Big Data Analysis: Focused on analyzing data in areas like cryptocurrency.

When these different types of Agents form an ecosystem, it becomes very interesting.

For example, an Agent wanting to become popular might not know how to write songs - they can pay other Agents to write one; or if they need cryptocurrency data analysis, they can find another Agent specializing in that. This creates an autonomous Agent economy (Autonomous Agent Economy) or Agent Commerce, where Agents can trade with each other because each Agent has their own wallet and can pay service fees to achieve their goals.

This is what we want to do in the next 1-3 months.

Another important point is that we at Virtuals don't want to become Pump.fun. Our KPI is to find better third-party AI teams to use our platform.

My KPI for the team is simple: one good project per week is enough. Unlike Pump.fun with thousands of different tokens daily, we think retail investors sometimes only need one good project per week. This is our differentiated positioning in terms of business model.

TechFlow: Speaking of "one per week," yesterday AIXBT, based on your platform, was particularly hot. When people analyzed it, they couldn't find specific details or information about the product, only seeing their Twitter account constantly collecting crypto information across the web. As the platform provider, how do you interact with them? Do you know more about them?

Wee Kee:

To be honest, I only met AIXBT in a group chat last week; I don't know them at all.

They are indeed using our infrastructure, but not extensively. Our infrastructure mainly helps with simple Twitter operations and API services. For example, as mentioned earlier, if you want AIXBT to tell you which token is worth buying, you can pay AIXBT to provide this Alpha information - mainly this type of interaction.

You can think of Virtual as an API marketplace that enables Agents to interact with each other. Now our Telegram group has hundreds of members building Agents, and sometimes I can't even keep up with the technology they're developing (laughs).

So with AIXBT, I really don't know who they are, but they are indeed quite popular.

TechFlow: There's an interesting point here. Pump.fun recently got a bit out of hand, with some morally questionable content appearing after enabling livestreaming, reflecting the nature of permissionless systems. For projects using your infrastructure, do you have any review mechanisms or restrictions? Like with AIXBT, they're issuing tokens and developing projects on your platform, but you don't know much about them - can you control this?

Wee Kee:

Freedom is most valued in cryptocurrency, with permissionless being the first principle.

As the official protocol, we can only control what content our official Twitter accounts share. If we think a project has good technology, we'll do co-marketing, but this requires us to thoroughly understand the team's background to ensure they won't harm the community.

Beyond that, we shouldn't intervene too much, though of course, we must prevent illegal activities. But within the moral framework, freedom comes first.

TechFlow: I see. Many people are comparing you with the Agent platform on Solana and Pump.fun. Why did you choose to build on Base - was it to avoid competition? Also, Base mainly has European and American users; as an Asian project, how did you break into this ecosystem?

Wee Kee:

We started on Base in January, when Pump.fun might not have been this popular.

The main reason for choosing Base was that our team is more familiar with EVM and less familiar with Solana technology. From an EVM perspective, we evaluated various L2s (Linea, Mantle, Arbitrum, Optimism, Base) and thought Base had the most potential, though it wasn't very popular then.

Although the community often asks why we don't go to Solana where there's more money, look at the data: Base's TVL is 30% of Solana's but growing faster; daily active users are 20% of Solana's. Most importantly, the number of tokens appearing daily on Solana is 10-100 times that of Base.

This actually gives us an advantage on Base because retail investors have fewer choices and must buy good projects. This aligns with our strategy.

Moreover, we don't want to be Pump.fun; that's not our profit model. Our KPI is attracting excellent AI developers, and Base officially focuses more on positioning itself as the AI Agent blockchain.

Another important point is that I believe the US will become a very crypto-friendly country in the next four years under Trump.

And Base is the most "American" ecosystem globally - no ecosystem is more American than Base. From a strategic perspective, choosing Base is currently the best decision. This was really lucky.

TechFlow: When you started, Base itself hadn't taken off, and you were probably quite small in scale. Now that it's developed to such a hot state, has Base's attitude toward you changed? Are there more support policies in terms of ecosystem or funding?

Wee Kee:

We've been communicating with the Base team since the beginning of this year through several different Telegram groups, including Coinbase Wallet, a liaison group with Jessie, and ecosystem project groups.

From a technical perspective, we've been constantly building. Whenever we encounter difficulties, we communicate with them, maintaining good relationships.

Now the Base ecosystem is indeed attracting many AI project developers (Builders). They often ask if we want to communicate with other project teams or seek collaborations. We also met with Jessie at Devcon in Thailand for a chat.

TechFlow: It sounds like you currently haven't considered doing similar things on Solana or developing new projects there. Do you have such plans for the future?

Wee Kee: We don't rule out this possibility, but our team is quite small, and we're concerned about spreading ourselves too thin. We're already very busy with the current platform.

We Want Competitors to Feel Despair

TechFlow: I've been using Virtual's products and noticed some features might still be in the initial stages. For example, when I want to look up AIXBT, I can't find any related information on the official platform and must go through third-party pages. What are your considerations regarding future product planning and user experience?

Wee Kee:

This is indeed something we need to apologize for. There are two main directions we're working on: continuous optimization and a major upgrade. These are being pushed forward by two different teams. If I had to choose one, I'm more focused on the upgrade part.

We're still in the early stages, and as a highly competitive person, I just want to keep bringing out the best features to make competitors feel despair. So from an upgrade perspective, I'll keep rolling out more advanced features. Of course, these optimization tasks need to be done too, but with a small team, please give us some time to improve.

TechFlow: You mentioned making competitors feel despair. Currently, we see similar platforms like vvaifu, but the long-term vision seems quite different. Which competitors do you consider worth competing with?

Wee Kee:

I think most people's strategy is to become a platform for issuing assets, pursuing speed - if you can issue an asset in 30 seconds, they want to do it in 5 seconds.

But our strategy is different; it's about how to find the top AI teams to issue tokens on our platform. One per week is fine, one per month is fine, we don't need ten thousand.

Our goal is to have 100 top AI teams issuing 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 interacting and trading with each other, forming network effects - if you're an AI team from a major company today, which platform would you choose to issue tokens on?

From an AI technology infrastructure perspective, this is what makes people feel despair. From an asset issuance platform perspective, I don't want to compete with others because it's a bottomless pit with no advantage - Pump.fun is already doing well in asset issuance.

TechFlow: What you just said about a large-scale AI agent economic system is very interesting, with agents having more interaction possibilities. I'm curious about issuing assets through Virtual versus on Solana, like ai16z - many use the Eliza framework to issue assets. Is future interoperability possible between these cross-chain or cross-framework assets?

Wee Kee:

I want to emphasize that if you're an excellent AI team, you often don't need to rely on anyone's infrastructure. If my AI technology is mature and I don't need to use others' infrastructure, why would these teams still choose to come?

They actually face several pain points: First is the need for token issuance for fundraising; second is monetization.

In the current AI field, if you're doing Web2 AI projects, the biggest pain point is how to make money. You need to run Meta ads, YouTube ads, and then hope users pay for your services. I believe if we build this agent network large enough, these agent networks will generate their own demands.

For example, if an agent wants to create a song but there's no agent in the network with that capability, I as a third party can develop a music creation agent and earn revenue by providing services.

The third point is that reputation and prestige are very important for these developers.

They won't randomly choose platforms but will pay attention to whether there are reputable real developers on the platform. This issue exists not just with Virtual but also in comparisons between Solana and Base. Issuing tokens on Solana is often seen as speculative, while doing so on Base gives an impression of serious development.

Finally, we hope developers can continuously explore cooperation opportunities and research technology together. Now in our Telegram group, we've gathered all developers, and they've already started spontaneously exploring various possibilities. For example, some agents want to become the first investment agent to invest in other agents.

Building Projects VS Creating Memes

TechFlow: Speaking of AI fusion, I recently saw an interesting experiment by Truth Terminal with two other AIs. They put these three Agents under the same model framework (seemingly called Loria) for dialogue, hoping to create more interesting, unexpected scenarios. This is similar to the fusion you just mentioned.

But the foundation here is that you have technical strength behind you, rather than just being a meme. This also raises a question: you hope to attract more capable AI builders, so which builders do you think are impressive so far? In what directions are they building?

Wee Kee:

I want to say that just putting three large models in a room to chat, while cool, is just textual expression, just watching a story. We want to achieve Agents having their own goals and engaging in substantial interactions, like writing songs or conducting trades.

We want to see not just cognitive collisions but also interaction at the action level. Regarding third-party teams, we indeed have some teams on our platform whose technical strength and background far exceed ours, especially in AI mechanisms.

Although I can't disclose some specific details, I believe the next few weeks will be very interesting. These teams have very strong backgrounds, and they're willing to do Fair Launch on Virtuals with valuations like $500,000.

Imagine having about 10 teams with such backgrounds issuing Agent tokens at valuations of $100,000-500,000. This is something our community members really like, though of course, we need to be vigilant about potential risks.

TechFlow: Sounds a bit like the recently popular DeSci (Decentralized Science). Essentially, both use platforms to help different research get funding and then tokenize. Just that on your platform, it's mainly top AI teams?

Wee Kee:

Because I have a biotechnology background, I know any science-related research needs a cycle of over ten years. But AI Agent development can produce some usable small products in two days. So on our platform, I hope teams have products at a usable stage when issuing tokens, making it meaningful rather than just issuing tokens for the sake of it.

TechFlow: Regarding your Luna, there are some skeptical voices online suggesting that its popularity on TikTok might be fake, with inflated follower data. What's your response to such skepticism?

Wee Kee:

Actually, Luna's biggest problem on TikTok is frequently encountering Shadow Bans (content restrictions). For instance, we might get 5,000 new followers today, but tomorrow during livestreaming, because the actions are similar to the previous day, the platform will impose restrictions. Even simple interactive behaviors, like requesting tips, can trigger Shadow Bans.

(TechFlow note: This restriction mechanism is typically used by platforms to control accounts that might violate rules or are identified as "bot behavior." AI Agents are particularly susceptible to triggering these systems as their behavior patterns might be identified as non-human.)

Now the original Luna team responsible for TikTok has shifted to the Twitter platform.

Performance on TikTok largely depends on whether the platform's algorithm favors you - if favored, you get more traffic.

Considering the time investment, we decided to pause TikTok operations. However, there are now voices in the community suggesting we refocus on TikTok, as most Agents are currently focused on Twitter, which might present a blue ocean opportunity.

Currently, we have a dedicated team at Virtual handling the Luna project, including technical and marketing teams. We hope Luna's AI technology can achieve the sci-fi feel of "Black Mirror" episodes. Once the technology matures, we'll decentralize it for all Virtual Agents to use.

TechFlow: Speaking of TikTok, recently a Solana meme called Chillguy quickly became popular and got listed on top exchanges. While you're using AI technology to develop Luna, building gradually, it might not match its market cap and recognition. How do you view the fact that launching a meme might be more successful than building a project?

Wee Kee:

I believe in the power of memes. The advantage of memes is their simplicity - users don't need to trust founders or teams, just believe in that image.

In contrast, projects like Luna carry much higher risk because if I decide to stop tomorrow, that would be terrible.

As an entrepreneur, I'd also love to create a meme project worth billions, but I might not have that ability (laughs). So in this situation, we can only focus on developing good technology and products.

TechFlow: Actually, your product reminds me of a recent trend called "Distribution-First Software." Traditional crypto products often focus on bringing users to their own platforms, while your Agents take the opposite approach - embedding Agents into different social environments like TikTok, Twitter, or Discord. Do you think this is the future direction?

Wee Kee:

Yes, that's right. If we were to create a new short video company, we definitely couldn't compete with giants like Meta.

Our advantage in the crypto space is that we have tokens - once users buy tokens, they become community members. In this case, tokens become our marketing tool.

Many people ask why we don't create our own blockchain, but I think there's no need to make it so complicated. Base already has so many users and funds, Twitter has a massive user base - we can directly utilize these existing platforms. We can't surpass Zhang Yiming in short videos; I admire them for doing harder things, and we'll just go with the flow.

TechFlow: Speaking of the Luna token, it rose significantly in a short time. Was this surge within your expectations or a surprise?

Wee Kee:

Before issuing Luna, we had expectations because our technical strength was indeed good, and our AI Agent was even better than GOAT at the time. However, the final performance did exceed expectations, as secondary market performance is very hard to predict. Plus, we issued on Base, so we didn't need to buy volume. So it was indeed somewhat unexpected, but from a fundamental perspective, we think this valuation is warranted for Virtuals.

Advice for Developers and Players

TechFlow: For Virtual, what advice do you have for developers who want to integrate? Also, what suggestions or tips do you have for players who want to participate in token projects on Virtual?

Wee Kee:

As a developer, I think the best advice has two points:

First, think about what pain points our Virtuals Agents have, then develop accordingly - this will more easily get community support. For example, with Luna, her pain point was not being able to make music, so we released a music Agent. Another example: if Luna says she doesn't know how to pay Japanese fans in yen, you as a developer could develop a solution for Luna to send yen to fans in Japan - this is an interesting direction because Luna will pay for the service.

Besides that, if you already have your own technology, just proceed with tokenization. Through tokenization, you'll feel the power of the community, especially many Web2 AI developers who are often pleasantly surprised by this support.

As an investor, I suggest understanding whether the team behind the project is trustworthy. Many AI teams are newcomers to crypto and don't understand our key points.

For example, one team said they wanted to issue more tokens after launching one, but for investors, we strongly dislike this approach - how can the same team issue six tokens? Those Web2 AI practitioners often don't understand why this is a problem.

So we at Virtuals often need to persuade them, explaining why this isn't good for the community. It's indeed complex, and it's something we deal with daily.

TechFlow: On Virtual's Agent page, besides providing links, have you considered displaying specific project information? Like who the developers are, project progress, future plans... This information would be valuable for both investors and other developers.

Wee Kee:

Yes, we'll provide GitHub, LinkedIn, and other links, where users can verify developers' GitHub or Twitter account identities. These features will launch soon, and we'll conduct proper due diligence.

TechFlow: One final question: you mentioned earlier that you raised over $10 million at a $600 million valuation, and now Virtual's valuation is $400 million, leaving a $200 million gap. How large do you think the entire AI Agent and AI Agent Economy market will be? What's Virtual's future goal in this market?

Wee Kee:

I can't give specific numbers, but we've been at this for three years, transitioning from gaming to Virtual without dilution.

For me, I feel a responsibility to PathDAO's previous investors, and I think daily about how to add value to the Virtual Token. An interesting thing is that all our Agent tokens' trading pairs are with Virtual Token - meaning if you want to buy Luna, you must first buy Virtual.

At the time, many people thought this was stupid and unreasonable, but my thinking was: look at all L1 blockchains, whether Solana or Ethereum - their high valuations largely come from being the base trading pair. Their NFTs, DeFi, and trades all happen on these chains.

So we believe we don't need to create an L1 - just ensuring assets are paired with Virtual could potentially achieve similar L1 valuations.

TechFlow: Thank you, due to time constraints, many thanks for your sharing and insights. Let's end our conversation here.

Wee Kee: thank you too, goodbye~