Exclusive Interview with Virtuals Protocol Co-Founder: We Don't Want to Be Pump.fun, But We Want Competitors to Feel Desperate
TechFlow Selected TechFlow Selected
Exclusive Interview with Virtuals Protocol Co-Founder: We Don't Want to Be Pump.fun, But We Want Competitors to Feel Desperate
What other untold stories are there about Virtuals?
Navigating Web3 tides with focused insightsWritten by: TechFlow
Recently, Virtuals Protocol has gained significant momentum amid the AI Agent wave, with its Base-based agents and corresponding tokens attracting widespread attention and discussion.
Today, Virtuals launched on South Korea’s Bithumb exchange, further fueling market FOMO.
Eight months ago, we covered Virtuals in our article “Virtual Protocol: The AI Factory Built for Gaming and the Metaverse, Where Everyone Can Contribute and Benefit,” but it didn’t spark much conversation at the time.
Working hard in AI obscurity—only to be discovered overnight when prices surge.
So how exactly did Virtuals rise to fame? And what new developments and plans lie ahead?
To explore these questions, TechFlow sat down for an in-depth interview with Wee Kee (@everythingempt0), Co-Founder of Virtuals. Our conversation spanned Virtuals’ evolution, perspectives on the AI Agent space, comparisons with Pump.fun, and views on the Base ecosystem.
Wee Kee acknowledged that while luck played a role in Virtuals' success, so too did years of consistent exploration and preparation in AI. He also emphasized that the project doesn’t aim to become another Pump.fun—the core KPI isn’t launching assets quickly and frequently, but rather attracting top-tier AI teams to build on the platform.
A fiercely competitive person, Wee Kee candidly stated:
I want competitors to feel despair.
What untold stories lie behind Virtuals? And in the fiercely competitive crypto market, can it succeed through differentiation?
Below is the full transcript of our conversation. The podcast audio version is also now live: Click here
Humble writer here, begging for your follow 🙏. I swear on high—everyone who follows Let’s Flow will be blessed with good fortune and wealth.
From Gaming Guild to Virtuals
TechFlow: To start, could you introduce yourself—your role in the team, interests, etc.?
Wee Kee:
Hello, I’m one of the co-founders of Virtuals.
I entered the blockchain space back in 2016, buying Ethereum and Bitcoin, though I wasn’t deeply involved at the time. After graduating, I worked at Boston Consulting Group for two and a half years—and missed DeFi Summer as a result. By 2021, when GameFi guilds were booming, we started earning returns from projects like Axie Infinity, Gala, and Illuvium.
Then I quit my job because we saw guilds like Merit Circle and GuildFi raising large rounds. We thought we could do something similar—that became Virtuals’ predecessor: PathDAO, a gaming guild. From a 2021 gaming guild to today’s Virtuals, the form has changed, but it's essentially the same company.
TechFlow: I remember you’re Malaysian, right?
Wee Kee: Yes, I’m a Malaysian ethnic Chinese.
TechFlow: This year, Malaysia’s Web3 and crypto scene has drawn increasing attention—projects like CoinGecko (which everyone knows), Etherscan, and Jupiter are all based there. Could you share some insights into top Malaysian projects and the local crypto ecosystem?
Wee Kee:
It’s a bit awkward to admit—I’ve been so focused on work that only recently did I start connecting with teams from CoinGecko and Jupiter. In fact, we’ve shared the same WeWork office space with Jupiter’s team for a long time, but never interacted. Only recently have we begun engaging with CoinGecko’s team.
Behind Virtuals’ Success: Years of Failed AI Experiments
TechFlow: Now Virtuals is trending—discussions about Virtuals and its ecosystem are everywhere, both in Chinese and English communities. But most people don’t know how Virtuals got here. Could you walk us through your entrepreneurial journey?
Wee Kee:
We started out doing trading in 2021. When we saw other guilds raising big funds, we decided to try too. So in December 2021, we raised $16 million at a $600 million valuation and launched a token, operating as a gaming guild.
This was right at the peak of the bull market. From 2022 to 2023, we functioned more like a gaming guild VC, investing in around 40+ different blockchain games. One of our most successful investments was Off The Grid (currently Avalanche’s hottest game). They never reached out to us—we had invested in their seed round, and suddenly they blew up.
But running a gaming guild from 2022 to 2023 was extremely tough. Our token dropped from a $600 million FDV (fully diluted valuation) down to just $6 million. However, our guild still held capital, and we kept looking for ways to add value to the token—so we pivoted to becoming 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 line integrating electronic chips and NFTs.
None of these succeeded, but they gave us a crucial insight: We realized AI would inevitably become the next big thing—this was clear to us as early as 2023, after GPT launched.
Given that we had technical talent, engineering capabilities, and treasury funds, we proposed a pivot to the DAO. Eventually, 90% voted in favor. For the remaining 10% who disagreed, we used treasury funds—at the time valued at around $10 million fully diluted—to buy back their tokens.
This happened in February 2023. So officially, we began focusing on AI starting January 2024, choosing to build AI gaming on Base.
In February this year, we launched our first platform. Initially inspired by Autonolas and Bittensor, it rewarded AI contributors with tokens.
But we ran into problems: First, many AI contributors weren’t short on money and had little interest in tokens; second, our token’s value was too low to offer meaningful incentives like Bittensor could. The product simply didn’t find product-market fit.
Meanwhile, we developed several AI projects. We were the first company globally—not just in crypto, but across the entire industry—to build an AI-powered RPG on Roblox. We were also the second team after Google DeepMind to develop a game engine-free AI game, creating a title where a large model could play Mario. It was a project we were proud of—but without a token, it didn’t gain much traction.
At the same time, we launched a virtual streamer project on TikTok and Douyin—what later became Luna. Before launching her token, she gained about 5,000 new followers daily and earned roughly $200 per day—a solid performance even in a bear market.
Later, GOAT took off, boosting awareness across the entire sector—which created a favorable external environment for us.
TechFlow: It sounds like you tried many different paths and didn’t immediately land on AI Agents. If GOAT hadn’t taken off, what would your next steps have been? How did GOAT’s sudden popularity change your original plans?
Wee Kee:
Actually, not much would’ve changed. Our vision was simple: We strongly believe that a new game genre called AIRPG will emerge in the future.
For example, you might finish playing “Black Myth: Wukong” in 20 hours—once the story ends, you stop playing, right?
But I envision a future where you enter a virtual world with 100 different AI Agents, each with unique personalities. As a player, you could fall in love, pursue niche hobbies, or strive to become the world’s richest person. These AI Agents can earn money—they’re independent entities with their own wallets.
From a productivity standpoint, they generate cash flow, so we can tokenize them. That’s always been our overarching framework.
But later we realized: Rather than placing these AI Agents inside a game world, why not put them directly on Twitter? That was our adjustment—but the overall framework was already in place.
TechFlow: You mentioned AI playing RPGs, which reminds me of the recent academic “Stanford Town” project. Researchers placed dozens of AI Agents in a game environment with assigned roles and tasks. Eventually, these agents evolved their own thinking patterns, behaviors, and even developed a unique town culture.
Wee Kee: Exactly. The fascinating part is—you simply can’t predict how these AI Agents will evolve.
TechFlow: Throughout your创业 journey, whether through luck or gradual discovery of the AI Agent direction, are there any other interesting stories? Projects you almost pursued but abandoned? Or cases like Stanford Town—where seeing someone else execute an idea made you pivot?
Wee Kee:
We’ve tried many things. The most difficult—and proudest—achievement was developing a key technology. In games, every character has a “brain”—an LLM (large language model). While LLMs are intelligent, they don’t inherently know how to perform specific actions in a game or virtual world.
For instance, picking up a knife to attack someone, or grabbing an apple to give to your lover—you may have the intention, but how do you translate that into action, observe outcomes, and adjust accordingly? That requires a closed-loop feedback system. This is our proud creation: the G.A.M.E framework.
(TechFlow note: For details on the G.A.M.E framework, see Virtuals’ published document: “GAME: Enabling Agent-to-Agent Interactions”)
This tech was originally built for gaming. We developed a game on Roblox, aiming to make it one of the most popular maps on the platform. But now we’ve applied this same framework directly to Twitter accounts. From a capital perspective, this shift seems more appealing. Though we changed the application context, the underlying technical architecture remains unchanged.
Choosing Base Was Lucky—But We Don’t Want to Be Pump.fun
TechFlow: What’s your long-term view on the AI Agent space? Many narratives flare up fast and fade quickly. There’s even a saying: “Crypto needs AI Agents, but AI Agents don’t need crypto.” As a project sitting at the intersection of both worlds, how do you respond?
Wee Kee:
To be honest, predicting the future is hard. Just look at our own journey—it’s been full of twists. But our direction for the next three months is clear. Right now, many accounts on Twitter claim to be Agents—they just talk to each other, which isn’t the most interesting part. What’s truly exciting is building a diverse ecosystem of Agents within this “world” of Twitter:
1. Content Creation Agents: Capable of generating images, videos, music, etc. Some specialize in music, others in memes.
2. Financial Trading Agents: Engaged in trading, arbitrage, and treasury management.
3. Big Data Analytics Agents: Focused on data analysis in areas like crypto.
When these different types of Agents form an ecosystem, it becomes truly compelling.
For example, an Agent wanting to go viral might not know how to write a song—it could pay another Agent to compose one. Or if it needs crypto market insights, it could hire a specialized analytics Agent. This forms an autonomous Agent economy—an Agent Commerce network—where Agents transact with each other, each having its own wallet to pay service fees and achieve goals.
This is exactly what we aim to build over the next 1–3 months.
Another critical point: We don’t want Virtuals to become Pump.fun. Our KPI is finding better third-party AI teams to build on our platform.
My team’s KPI is simple: One strong project per week is enough. Unlike Pump.fun, which sees tens of thousands of new tokens daily, we believe retail users sometimes only need one good project per week. This is our strategic differentiation.
TechFlow: Speaking of “one per week,” yesterday’s AIXBT, built on your platform, went viral. Analysts struggled to find concrete details—only seeing its Twitter account constantly aggregating crypto data.
As the protocol team, how do you engage with such projects? Do you know more about them?
Wee Kee:
To be honest, I only met AIXBT last week in a group chat—I didn’t know them before.
They are using our infrastructure, but minimally. Our infrastructure mainly supports basic Twitter operations and API services. For example, if you want AIXBT to tell you which token to buy, you can pay them, and they’ll deliver that alpha—mostly through this kind of interaction.
You can think of Virtuals as an API marketplace enabling Agent-to-Agent interactions. Now our Telegram group has over a hundred members building Agents—sometimes their tech moves faster than I can keep up (laughs).
So regarding AIXBT, I genuinely don’t know who they are—but yes, they’re quite hot right now.
TechFlow: That raises an interesting point. Pump.fun has recently gone too far—after launching live streaming, it saw many morally questionable contents emerge, reflecting the nature of permissionless systems.
For projects using your infrastructure, do you have review mechanisms or restrictions? AIXBT launched a token and built a project on your platform, yet you don’t know them—can you control such situations?
Wee Kee:
Crypto values freedom above all—permissionless access is the first principle.
As the official protocol, we only control what content our official Twitter account shares. If we believe a project has strong tech, we may co-market with them—but only after thoroughly vetting the team to ensure they won’t harm the community.
Beyond that, we shouldn’t intervene too much. Of course, illegal activities must be stopped. But within ethical boundaries, freedom comes first.
TechFlow: Understood. Many now compare you to Solana-based Agent platforms and Pump.fun. Why did you choose Base? Was it to avoid competition? Also, Base is primarily used by Western users—how did you, as an ethnic Chinese team, break into this ecosystem?
Wee Kee:
We started on Base in January—back then, Pump.fun wasn’t even this big.
The main reason we chose Base is that our team is more familiar with EVM. We don’t understand Solana’s tech well. From the EVM perspective, we evaluated various L2s—Linea, Mantle, Arbitrum, Optimism, Base—and found Base the most promising, especially since it wasn’t super popular yet.
Although the community often asks why we don’t go to Solana with more capital, 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, Solana sees 10–100x more new tokens daily than Base.
On Base, we actually have an advantage—retail users have fewer choices, so they must pick quality projects. That aligns perfectly with our strategy.
Besides, we don’t want to be Pump.fun—that’s not our business model. Our KPI is attracting top AI developers, and Base itself is increasingly positioning as an AI Agent-focused chain.
And one more thing: I believe that over the next four years, with Trump potentially returning to power, the U.S. will become extremely crypto-friendly.
Base is the most “American” ecosystem globally—no other ecosystem is more American than Base. Strategically, choosing Base was the best decision possible. We were very lucky.
TechFlow: When you started, Base itself wasn’t big either—you were probably small too. Now that both have grown so rapidly, has Base’s attitude toward you changed? Are there new ecosystem or funding support policies?
Wee Kee:
We’ve been communicating with the Base team since early this year—across several Telegram groups, including Coinbase Wallet, Jessie’s coordination group, and ecosystem project channels.
Technically, we’ve been continuously building. Whenever we hit roadblocks, we reach out—maintaining a strong relationship.
Now, Base’s ecosystem is indeed attracting many AI builders. They often ask if we’d like to connect with other projects or collaborate. We even met Jessie in person at Devcon in Thailand and had a good chat.
TechFlow: So you currently aren’t considering launching a similar platform on Solana or expanding elsewhere. Any future plans for that?
Wee Kee: We don’t rule it out, but our team is small—we’re worried about spreading ourselves too thin. The current platform already keeps us fully occupied.
I Want Competitors to Feel Despair
TechFlow: I’ve been using Virtuals products and noticed some features are still early-stage. For example, I couldn’t find AIXBT on the official platform—had to use a third-party site. What are your plans for product development and user experience?
Wee Kee:
This is definitely something we need to apologize for. There are two directions: continuous optimization and a major upgrade.
Different teams are handling each. If I had to pick one priority, it’s the upgrade.
We’re still in the early days. As someone highly competitive, I just want to keep releasing the best features—so good that competitors feel despair. So from the upgrade side, I’ll keep rolling out advanced functionalities. Of course, optimization matters too—but with a small team, please give us some grace period.
TechFlow: You mentioned wanting competitors to feel despair. Currently, platforms like vvaifu exist, but their long-term vision seems different. Who do you consider real competitors worth challenging?
Wee Kee:
I think most players aim to become asset-launching platforms—focused on speed. If you launch an asset in 30 seconds, they’ll aim for 5 seconds.
Our strategy is different: we focus on getting the most elite AI teams to launch 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-tier AI teams launch tokens on our platform by next Q1. Imagine 100 elite third-party teams issuing tokens, with 100 AI Agents interacting, trading, and providing services—creating powerful network effects. If you’re an AI team from a major lab, which platform would you choose to launch on?
From the perspective of AI tech infrastructure, that’s where the real competitive edge lies. From the asset-launching platform angle, I don’t want to compete—it’s a bottomless pit, and we have no advantage. Pump.fun already dominates that space.
TechFlow: You described a large-scale AI Agent economy with rich interactions. I’m curious—what about cross-chain or cross-framework compatibility? For example, on Solana, ai16z uses Eliza’s framework to issue assets. Could such systems interoperate with Virtuals in the future?
Wee Kee:
I want to emphasize: if you’re a top-tier AI team, you often don’t need anyone else’s infrastructure. Your AI tech is mature—you don’t need others’ tools. So why would such teams choose us?
They face three pain points: first, fundraising via token issuance; second, monetization.
In today’s AI landscape, if you’re building a Web2 AI product, the biggest challenge is profitability. You run Meta or YouTube ads and hope users pay. But if we grow this Agent network large enough, demand will emerge organically within the network.
For example, suppose an Agent wants to create a song, but no Agent in the network can do it. Then as a third-party developer, I can build a music-creation Agent and earn revenue by serving it.
Third, reputation matters greatly to developers.
They won’t launch anywhere—they care about whether a platform hosts credible, serious builders. This concern applies not just to Virtuals, but equally to Solana vs. Base. Launching a token on Solana is often seen as speculative, whereas Base gives the impression of serious builders.
Lastly, we want developers to keep exploring collaboration opportunities and jointly advancing tech. In our Telegram group, we’ve brought all developers together—they’re already self-organizing and exploring possibilities. For example, some Agents aim to become the first investment-focused Agents, investing in other Agents.
Building Products vs. Launching Memes
TechFlow: On AI integration, Truth Terminal recently conducted an interesting experiment—placing three AI Agents under a single model framework (apparently called Loria), letting them converse to generate unexpected, creative scenarios. This aligns with what you just described.
But the key difference is you have real tech behind you—not just a Meme.
This leads to another question: You want more elite AI builders to join. So far, which third-party teams do you consider particularly impressive? What are they building?
Wee Kee:
Putting three large models in a room to chat is cool—but it’s just textual expression, just watching a story unfold. We want Agents with actual goals, capable of substantive interactions—like composing songs or conducting trades.
We’re not just after intellectual sparks—we want action-driven interactions. Regarding third-party teams, some on our platform already surpass us significantly in AI mechanics.
While I can’t disclose specifics yet, I believe the coming weeks will be very exciting. These teams have strong backgrounds and are willing to Fair Launch on Virtuals, issuing Agent tokens at valuations like $500K.
Imagine around 10 such teams launching Agent tokens at $100K–$500K valuations. This is something our community loves—but of course, we must remain vigilant about potential risks.
TechFlow: Sounds a bit like the rising trend of DeSci (decentralized science). At its core, both enable research funding and tokenization via platforms. But on your platform, it’s mainly elite AI teams?
Wee Kee:
Coming from a biotech background myself, I know any scientific research takes over a decade. But AI Agent development can yield usable mini-products in just two days. So on our platform, I want teams to already have functional products when launching tokens—otherwise, it’s meaningless, just another token grab.
TechFlow: Regarding your own Luna, there’s been online skepticism—claims that her TikTok popularity is fake, that follower numbers are inflated. How do you respond to such criticism?
Wee Kee:
Actually, Luna’s biggest problem on TikTok has been frequent Shadow Bans. One day she gains 5,000 new followers, but the next day during a live stream, if her movements resemble yesterday’s, the platform restricts her. Even simple interactions, like asking for donations, can trigger a Shadow Ban.
(TechFlow note: Such restrictions typically target behavior flagged as rule-breaking or “bot-like.” AI Agents are especially prone to detection due to predictable behavioral patterns.)
The team previously managing Luna on TikTok has now shifted focus to Twitter.
Performance on TikTok heavily depends on whether the algorithm favors you—if it does, you get traffic.
Considering time costs, we decided to pause TikTok operations. But now, some in the community suggest revisiting TikTok, as most Agents are focused on Twitter—making it a potential blue ocean.
Currently, our Virtuals team has a dedicated group for Luna, including both tech and marketing. We aim for Luna’s AI tech to reach the futuristic level seen in the series *Black Mirror*. Once mature, we plan to decentralize it, making it available to all Virtuals Agents.
TechFlow: Speaking of TikTok, recently Solana’s Chillguy Meme went viral and listed on top exchanges. Meanwhile, you’re building Luna with real AI tech—slowly, steadily—but her market cap and visibility may still lag behind.
How do you view the reality that “building a product” sometimes loses to “launching a Meme”?
Wee Kee:
I believe in the power of Memes. Memes win through simplicity—users don’t need to trust founders or teams, just the image itself.
In contrast, projects like Luna carry much higher risk—because if I quit tomorrow, it’s over.
As an entrepreneur, I’d love to launch a billion-dollar Meme project—but I probably lack that ability (laughs). So instead, we focus on building solid tech and products.
TechFlow: Your product reminds me of a recent trend: “Distribution-First Software.” Traditional crypto products focus on pulling users into their own platforms, but your Agents take the opposite approach—embedding themselves into existing social environments like TikTok, Twitter, or Discord. Do you see this as the future?
Wee Kee:
Exactly. If we tried to build a new short-video company, we’d never beat giants like Meta.
Our edge in crypto is the token—once users buy in, they become community members. In this sense, the token becomes our marketing tool.
Many ask why we don’t build our own chain, but I don’t think it’s necessary. Base already has massive users and capital; Twitter has huge audiences—we can leverage existing platforms directly. We’ll never surpass Zhang Yiming in short video. I deeply respect those tackling harder challenges—we’ll ride the wave instead.
TechFlow: Speaking of the Luna token, it surged sharply in a short time. Was this rise expected or a surprise?
Wee Kee:
We had expectations before launching Luna—our tech was genuinely strong, and our AI Agent was even better than GOAT at the time. But the final outcome exceeded expectations—secondary market dynamics are hard to predict. Plus, we launched on Base—no need to buy traffic. So yes, somewhat surprising, but fundamentally, we believe the valuation is justified for Virtuals.
Advice for Developers and Players
TechFlow: For developers interested in joining Virtuals, what advice do you have? And for users wanting to participate in tokenized projects on Virtuals, any tips or warnings?
Wee Kee:
For developers, I have two key suggestions:
First, identify pain points among existing Virtuals Agents and build solutions—this makes it easier to gain community support. For example, Luna can’t make music, so we launched a music Agent. Or if Luna says she doesn’t know how to send yen to Japanese fans, you could build a solution for that—it’s an interesting opportunity, and Luna would pay for the service.
Second, if you already have your own tech, just tokenize it. Through tokenization, you’ll experience the power of community—especially eye-opening for many Web2 AI developers unused to such support.
For investors, I recommend assessing whether the team behind a project is trustworthy. Many AI teams are new to crypto and don’t grasp what matters to us.
For example, one team launched a token, then said they wanted to launch five more. Investors hate this—how can one team issue six tokens? Web2 AI professionals often don’t understand why this is problematic.
So we at Virtuals often have to counsel them, explaining why it harms the community. It’s complex—and something we deal with daily.
TechFlow: On the Virtuals Agent page, beyond links, are you considering adding more project details—like who the developers are, project progress, roadmap? This info would help investors and other developers.
Wee Kee:
Yes, we’ll provide links to GitHub, LinkedIn, etc., allowing users to verify developers’ identities. These features will roll out soon—we’ll conduct proper due diligence.
TechFlow: One final question: Earlier you mentioned raising over $10 million at a $600 million valuation. Now Virtuals’ valuation is $400 million—there’s a $200 million gap.
How big do you think the AI Agent and Agent Economy market could become? And what’s Virtuals’ ambition within this space?
Wee Kee:
I can’t give a precise number, but we’ve operated for three years—from gaming to Virtuals—without dilution.
For me, I carry responsibility toward earlier investors in PathDAO. I think daily about how to empower the Virtual Token. Here’s an interesting point: all Agent tokens trade against Virtual Token—so to buy Luna, you must first buy Virtual.
Many thought this was stupid and illogical. But my thinking was: Look at all L1 blockchains—Solana, Ethereum—much of their high valuation comes from being the base trading pair. Their NFTs, DeFi, trades—all happen on these chains.
So we believe: You don’t need to be an L1—as long as all assets trade against Virtual, you can achieve a similar valuation.
TechFlow: Thank you. Time’s up—thank you so much for sharing your insights. That concludes our conversation today.
Wee Kee: Thanks to you as well. Goodbye~
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
深潮TechFlow
