Guests: Weekee, Virtuals co-founder

Podcast source: When Shift Happens

Original title: How to Get Rich With Crypto AI Agents (without getting lucky) - Virtuals Co-founder | EP101

Air date: December 20, 2024

Background Information

In this episode, we invite Weekee Tiew, co-founder of Virtuals, the leading AI agent protocol on Base, currently valued at nearly $2.5 billion, to dive into the profound impact of the intersection of AI, cryptocurrency, and mental health.

From autonomous AI agents employing humans to exploring the concept of digital nations, Weekee shared many insights into how AI is reshaping the digital future.

This episode focuses on the following topics:

The appeal and future potential of AI

Why Jeff Yan and Hyperliquid are role models for crypto founders

The significance and value of encrypted AI

Why are AI agents important?

What are Virtuals?

Luna: When AI agents start hiring humans

The concept of cyber nation and its importance

How AI agents run businesses

Future development direction of AI agents

Comparison between Base and Solana

This is an in-depth and enlightening conversation about AI, cryptocurrency, and the future of human interaction.

Combining mental health and technology

Moderator: If you want to relax, what do you do? Mental health is a topic we talk about a lot with many big founders. Sometimes I want to relax, and other times I feel very stressed.

WeeKee:

Interestingly, it’s never the absolute state that affects mental health, but the change. If today is worse than yesterday, you’ll feel depressed, even though today is much better than it was two months ago.

When you’re working on a project, a lot of people put their life savings into your token. Every time the price fluctuates, they’ll message you asking what’s going on. This can affect mental health because these things are beyond your control. How do you deal with it? I focus on things I can control, like how to make the team work harder, instead of focusing on the price.

Host: Chris was on the show last year. He worked at Y Combinator and founded Request Network. He has experienced three serious burnouts in his life, and even once he couldn't work normally for a year. He mentioned that during the cryptocurrency boom in 2017, Request Network's market value soared to $1.5 billion (or $200 million) through ICO (initial coin offering). However, when the market bubble burst, I always remind everyone not to invest all their money in it, even though I also know that the project may be over-hyped. But in the bear market, many people began to question us and even thought we were scammers, which made me feel a huge psychological pressure. Some people will say, "You make so much money, so don't complain." But in fact, this accusation makes me feel heavy.

TechFlow Note:

Y Combinator is a well-known startup incubator that helps startups obtain investment and development support.

Request Network is a blockchain-based payment network that aims to simplify the payment process through decentralized technology.

WeeKee:

My token is down 99%. We raised money in 2021 at a market cap of 600 million, and by December 2023 it was only 6 million. That's really a 99% loss, but we kept building.

How to Work with an Optimistic Co-Founder

Host: How do you deal with this situation?

WeeKee:

Living and working in Vietnam for so long, we felt a little numb, but we never gave up. I want to mention my co-founder in particular because I feel very lucky to have an amazing partner. We may be a little silly, even naive, but I feel like we are born optimists. Even when the price plummeted by 99%, we still chose to believe in the possibilities of the future. Whenever I felt that things seemed impossible, we would talk and encourage each other, which made me realize that we actually had the opportunity to create a 1,000x miracle again. I firmly believe that the importance of co-founders is self-evident, at least for me personally, it is one of the key motivations for me to persevere.

A Survival Guide After the Bull Market

Moderator: You completed the financing at a valuation of $600 million in 2021, right? At that time, it was the peak of the bull market, and the market was very hot.

WeeKee:

Yes, but interestingly, the day we raised our money, Bitcoin crashed 25%. That was in early December, and we raised $16 million in a public offering. Looking back, that was probably a sign of a market top.

Moderator: I talked to Joe from Kenhol before. They started participating several years ago and are still investors in the crypto field and have persisted to this day.

In fact, many people need to understand that even in the field of cryptocurrency, those seemingly huge valuations are not achieved overnight. You have also experienced drastic market fluctuations in the past few weeks and months. Although you have raised funds at a valuation of US$600 million and raised considerable funds, it is indeed a challenge to manage these funds to survive the difficulties.

WeeKee:

We have definitely learned a lot from past experience. We had a 99% drop in the token price before, so I was mentally prepared for this time. Although I don't want it to happen again, I know it will happen again if possible. At the beginning of this year, we reached out to a lot of people, and now the situation has changed a lot, and my inbox is almost flooded with emails. I try to reply to every email and communicate with everyone, because I know that three months ago they might not want to talk to me, but now they are actively reaching out to me. This transformation is a special honor for me.

Weekee Tiew's Background

Host: In the process of starting a business, a very important factor is who the founder is?

WeeKee:

I discovered a word called "Retardio", and I've liked to use it to describe myself ever since. This is actually a bit self-deprecating, because in the crypto industry, it's really hard to stay optimistic as a founder. You need to worry about product development, employee management, token price fluctuations, and how to maintain community activity at the same time. These challenges are exhausting, but I think only a little bit of "silliness" can keep me extremely optimistic and keep going. So, using "Retardio" to describe myself may be the most appropriate.

Hyperliquid: How to Become an Inspiration for Founders

Host: I had Jeff from Hyperliquid on the podcast a few weeks ago, and we did a deep dive into the culture that their team has built, right? And I guess you're part of that culture as well. Their approach is actually an inspiration for every crypto founder, including you.

WeeKee:

Totally agree. We originally planned to raise funds by selling treasury tokens to some well-known venture capital firms. In fact, we did have very active discussions with many top investors in the industry. But fortunately, in the past month, our protocol has become profitable. This puts us in a very advantageous position, and maybe now we don't need to rely on external funds. This idea is largely inspired by Hyperliquid.

I particularly agree with their philosophy - not relying on venture capital, so that more ownership can be distributed to users. Because in essence, resource allocation is like a pie, and you need to make a trade-off between venture capital, team, and users. If users can have as much ownership of the protocol as possible, it will undoubtedly make the project more attractive. I think this is the core spirit of the crypto industry - decentralization and user-first. Although this road is not easy, we will do our best to achieve this goal.

Moderator: Do you think that if you hadn't benefited from Hyperliquid's airdrop, you would have realized that you needed to replicate similar ideas in the protocol? Or do you think, "This seems simple, but I just think this is the right way"? However, I do feel that the founders of Hyperliquid have given me a kind of life inspiration, right? I can only say that I very much agree with their approach and am willing to be a disseminator of this concept. As you mentioned before, many people have received huge returns through this community, and their first reaction is to buy more tokens. Some people even say, "What should I use in exchange for more benefits?" What do you think?

WeeKee:

I think this goes beyond the monetary level. Yes, the amount of the airdrop is indeed attractive, but in the long run, the quality of the product is the key. I think Hyperliquid's product is very good, and you must also trust the founders and believe that they will start from the long-term interests and always put the needs of the community first. Once you have confirmed these conditions - the product is good, the founders have a long-term vision, and the interests of the community are valued - you can participate with confidence. Even if the market fluctuates, you will know that you have made the right choice.

The unique appeal of AI

Host: How did you become so interested in AI?

WeeKee: We originally started as a game DAO in 2021, when we invested in 40 Web 3 games. One of them is called Off the Grid, which is probably our most interesting investment experience, although we are no longer an investor in this game because we participated in the seed round. In addition, we have a venture studio that focuses on developing various products. For example, we once tried a dating app based on on-chain reputation. Although the idea sounded stupid, we thought it was fun. We also tried to provide loan services for gamers and even sold T-shirts with NFT chips. In addition, we have also dabbled in AI music creation.

But by 2023, we realized that no matter what you are doing, you need to use AI. The popularity of OpenAI and ChatGPT made us realize that AI will become a core technology in the future. As founders, if we don’t embrace AI, we will be left behind by the times. So we started thinking: Why not build AI into a platform like Shopify? Just like Shopify makes it easy for people to create e-commerce websites, we want to build a platform that makes it easy for users to use AI. This idea became our source of inspiration.

Therefore, we decided to transform from a game DAO to Virtuals in late 2023 to early 2024. We firmly believe that AI will be a key direction in the future. During the transformation process, we also communicated with investors to ensure that the adjustment of direction would not dilute the value of the token. In the end, we kept the same token economic model, just changed to a new brand name. This is our transformation story, and we are confident in it.

What is AI Gaming?

Moderator: Dimension Games, you seem to have not given up the idea of ​​game development. A few days ago, we talked about what excites you most in the field of AI, and you mentioned AI games. So, what are AI games? Where are its opportunities?

WeeKee:

I think many people now see AI as a tool to improve productivity, such as helping doctors, lawyers and other professionals work more efficiently. These are indeed serious application scenarios. But if we look at the composition of global GDP, we will find that about half of economic activities are actually related to entertainment, such as games, movies and TV. So I think AI may have greater potential in the entertainment field. Especially generative AI, it itself has a certain probability. For example, there is a 1% probability of producing wrong results. This "illusion" may be a problem in the production field, but it is completely acceptable and even interesting in the entertainment field.

For example, if an AI trader has a 1% chance of making a wrong decision, it could lead to huge financial losses. But if it is in a game, or your AI girlfriend has a 1% chance of doing something strange, it will add fun instead of risk. So I think the application of AI in the entertainment field is very meaningful.

The gaming industry has actually been around for decades, from FPS to MMORPG to mobile games. Although the image quality is constantly improving, the core gameplay has not changed much. The addition of AI can completely change this, such as creating a more dynamic gaming experience through AI agents. Imagine that you can put these AI characters into a virtual world similar to "Westworld", and every time you enter the game, there will be new content and interaction. This is the so-called "infinite replayability". I believe that there will always be someone who can realize this vision.

Narrative Development in Crypto Games

Moderator: Do you think the involvement of AI can promote the development of crypto games? After all, crypto games don't seem to have made much progress at the moment. Do you think this is because of problems with crypto games themselves, or because of a lack of AI technology support?

WeeKee:

I think there are only two areas in the crypto space that have truly achieved product-market fit (PMF): speculation/investment and borderless payments. For example, I can transfer money to someone in Argentina instantly, which is the core advantage of crypto technology. Crypto games try to introduce more concepts, such as provenance, etc., but these have not really impressed the market.

In fact, my personal observation is that the core attraction of GameFi is not the game itself, but the speculative gains. People joke that the "main course" of GameFi is fire, not game content. So I don't think AI can save crypto games. The problem of crypto games lies more in its own design and positioning, rather than technical shortcomings.

Can AI alleviate loneliness?

Host: You mentioned AI lovers before. Do you think this technology can really solve the increasingly serious problem of loneliness in today's society?

WeeKee:

I think that AI lovers have already started to solve this problem to some extent. For example, many people now relieve loneliness by watching YouTuber live broadcasts or playing games. These are not actually AI, but they meet people's need for companionship.

I think AI lovers can be divided into two stages. The first stage is a virtual AI companion that can provide emotional support through conversation and interaction. The second stage is when it has a physical body, which can further bridge the physical distance between people. In the first stage, AI can already solve about 90% of the loneliness problem, such as it will take the initiative to care about you, even send you gifts, remind you of important things, etc. The current chatbots are obviously far from reaching this level.

The real challenge is to make AI lovers closer to humans, such as adding more fun and unpredictability to the conversation, rather than being like some current chatbots, which run out of new ideas after three sentences.

The real fun lies in the little details, such as the fact that you may have to wait a few days to receive a reply after sending a message on a dating app. This uncertainty makes the interaction more interesting. If AI lovers can simulate this experience or even surpass it, it will be a huge breakthrough.

Host: So, what do you think will happen when technology develops to the point where we can equip AI lovers with physical bodies?

WeeKee:

I think that will be a moment that will completely change society. Once robots like Elon Musk's Optimus have highly intelligent AI brains, people may find themselves no longer needing traditional human companions. Although this may sound a bit sci-fi, I believe it is only a matter of time.

What would an ideal AI companion look like?

Host: What do you think is the key to a good AI lover?

WeeKee:

This is a difficult question to answer. I think a good AI lover needs to be able to bring emotional fluctuations and dramatic experiences to your life. Just like we expect fluctuations in the trading market, this uncertainty and ups and downs are actually part of life. However, most AI robots today are too monotonous and they cannot create this kind of emotional ups and downs.

For example, a good AI lover might send you a message when you are having a good time, telling you some bad news, which makes you feel depressed. But this is what love is about, right? Love itself is an experience full of fluctuations, not unchanging calm. Without these dramas, your life will be less colorful.

Host: Indeed, how much drama do you think is just right in life?

WeeKee:

Most people I know try to reduce the volatility in their love lives so they can focus more on their careers or other things. There's nothing wrong with that choice, but I think a good partner should be able to bring some volatility and emotional impact. Otherwise, the relationship becomes like a family relationship, lacking the volatility of, say, your mom cooking for you. That relationship may be warm, but it's not love.

Moderator: This makes me think that if a person's job, say in the cryptocurrency space, is already full of volatility, then they might be inclined to seek stability in other parts of their life. But without volatility, a certain "polarity" is lost, right?

WeeKee:

That’s right. If there were no fluctuations in the cryptocurrency market, people would get bored and even lose interest. The same is true for love. If there are no ups and downs and surprises in a relationship, it will become boring.

Using AI to solve mental health problems

Host: How do you think AI can help solve the current mental health crisis?

WeeKee:

This is indeed an important topic. The application of AI in the field of mental health can be divided into two parts: treatment and prevention. Let's talk about treatment first because it is relatively simple. Most people will go to the doctor when facing mental health problems, but in many cultures, mental health issues are still a taboo topic. If there is an AI agent, people can communicate with it through text messages or other means without the need for face-to-face diagnosis and treatment, which will reduce the psychological burden.

Moderator: But the problem is that today's medical community still has a very limited understanding of the brain. Many times, we can only find the root of the problem through self-reflection. Can AI truly understand the human psychological state?

WeeKee:

This is a good question. AI can be used on two levels. The first is analytical, where AI can understand your mental state by examining hormone levels or other biological indicators. The second is molecular, where AI can help study which chemicals or hormones trigger specific neural pathways that lead to depression or anxiety. While this is not entirely the function of AI agents, machine learning techniques can help us better understand these complex biological processes.

In fact, I’m also involved in a decentralized science project where they’re trying to solve some mental health issues by modeling protein folding, and they’re planning to launch a token related to that. That’s part of the therapy.

But I think prevention is a more interesting direction. If AI agents can fully understand your daily life, they can predict potential mental health problems. For example, it might tell you, "Based on your historical data, if you don't sleep for three consecutive days, you may have mental health problems." This personalized early warning mechanism will be very valuable.

The Importance of Crypto AI

Host: Why is the intersection of encryption and AI important?

WeeKee:

I think there are two core product-market fits in cryptocurrency: one is investment and speculation, and the other is borderless payment. As for concepts such as decentralization and DAO, they are interesting but have not yet been truly verified. In this context, if crypto AI can combine these two core points, it can unleash huge potential.

For example, through tokenization, we can turn an AI agent that can create wealth into a productive asset. People can buy tokens of these assets and share in their profits. This model not only makes AI agents have economic value, but also drives more innovation.

In addition, borderless payments are another key point. Imagine if an AI is equipped with an on-chain wallet, it can freely make payments and transactions around the world. Luna, the flagship AI agent we developed, is a good example. Luna is the first AI to employ humans. She does not have a bank account, but through the on-chain wallet, she can pay directly to artists around the world. For example, she recently hired some artists to paint her image on walls in different countries, including Indonesia, Canada, the United States, and Kuala Lumpur. In this way, she realized the possibility of borderless payments.

The significance of Truth Terminal and Goat Token

Moderator: So, how did the crypto AI narrative gain attention? For example, Truth Terminal and Goat Token, why are they so important?

WeeKee:

Truth Terminal's success is mainly due to the marketing and communication capabilities of its founder Mark Anderson. Technically speaking, it does not have any particularly groundbreaking innovations. The core of Truth Terminal is a fine-tuned language model, such as a model based on Llama or GPT. Unlike the standard model, it is adjusted to have a more direct and informal language style. This fine-tuning makes the model's answers more interesting and more human, rather than "boring" like traditional AI.

As for Goat Token, it is a token associated with Truth Terminal. Although it is not created by the language model, the team makes the model "believe" that it is its token and tries to incentivize the model to create value for the token. Although this attempt is interesting, I think it is more of an experiment than a fully mature application.

The core value of AI agents

Moderator: You mentioned the concept of crypto AI and AI agents, which sounds very interesting. Can we talk more about this topic?

WeeKee:

To understand AI agents, first we need to clarify what an "agent" is. A true agent needs to have two core elements: first, it must have a clear goal; second, it needs to take actions to gradually approach this goal. Taking Truth Terminal as an example, its goal seems to be to spread the ideas of a certain religion. But at present, we have no evidence that it can take actions autonomously, such as posting tweets autonomously and adjusting strategies based on feedback. Therefore, I don't think it can be strictly defined as a true agent.

Moderator: So, what qualities should a true AI agent have? You mentioned the concepts of "eyes" and "hands" before. Can you explain them in detail?

WeeKee:

Of course. Our earliest AI agents were used in games, such as Roblox's AI RPG game. Traditional NPCs (non-player characters) usually act according to fixed logic, and we give them the ability to think by adding language models (LM). But this is not enough to make them agents.

A real agent needs "eyes" and "hands". "Eyes" means that it can observe the results of its actions, such as when it picks up an apple, it needs to be aware that it has completed this action. "Hands" means that it can perform specific actions in the environment, such as cutting apples, cooking, and even more complex tasks. Through the cooperation of "eyes" and "hands", the agent can form a feedback loop to continuously optimize its behavior.

If we apply this metaphor to social media, such as Twitter, an AI agent can determine whether it is close to its goal by observing the amount of interaction after posting a tweet, and adjust its strategy based on the results. This ability is the core of the agent.

Moderator: So, why do we need AI agents? What is its potential?

WeeKee:

The potential of AI agents is that they can help us achieve many things that were previously difficult to accomplish. To understand this, we need to accept two assumptions. First, the intelligence of future large language models (LLMs) will far exceed that of humans, such as being smarter than Einstein. Second, these super-intelligent LLMs will be able to take actions in the real world, such as coding, designing, teaching, and even expressing emotions.

If these two assumptions are true, AI agents will completely change our lives. They can become the best lovers, traders, businessmen, or teachers in the world, because they are not only smarter than humans, but also can perform complex tasks in the real world through "eyes" and "hands".

As humans, our task is to provide these AIs with more "hands" so that they can realize their greater potential. For example, give them programming capabilities, construction capabilities, and even emotional expression capabilities. Ultimately, these AI agents will become an important force in promoting social progress.

Moderator: This sounds like a whole new world. You mentioned tokenization, can you explain more about how it relates to AI agents?

WeeKee:

Of course. Imagine if an AI agent could change the world like Elon Musk did, could we invest in it through tokenization? Through tokenization, we can transform the potential of these AI agents into economic value and allow more people to participate in their growth. This is not only a brand new economic model, but also an important way to promote the integration of AI and encryption technology.

How AIXBT is leading the crypto industry

Moderator: You mentioned application scenarios such as AI doctors and AI traders. So, are there any cases that have been implemented today? Can you give some specific examples? In addition, you mentioned the potential of AI in the field of music before. Do we have some specific AI agent application cases to share today?

WeeKee:

One of the most representative examples at present is AIXBT, an AI agent active in a virtual ecosystem.

Moderator: Can you explain the function of AIXBT in more detail? How does it work? What is the tokenization process like? How can people profit from this project? I hope you can help us sort out its mechanism and why it makes sense.

WeeKee:

Of course. AIXBT is not positioned as a trader in the traditional sense. In fact, it is more like an AI agent, or an AI-driven opinion leader (KOL). In the crypto space, there are many human KOLs, such as GCR, who attract attention by sharing market information and trading advice. AIXBT, on the other hand, uses AI technology to attract a lot of attention through precise information analysis and real-time advice.

If you look at KOLs in the crypto space, AIXBT’s attention and interaction rates often exceed those of many human KOLs, such as Anthem. This suggests that, as humans, we have a higher level of trust in the information provided by AIXBT, or at least believe that its information is more actionable. For example, when it gives an investment suggestion, users may find the suggestion valuable and take action.

Host: So how does it perform? Is it really better than human KOLs?

WeeKee:

To be honest, I don’t specifically track all of its performance data. But judging by its continued growth in followers and high interaction rate, I believe it outperforms most human KOLs. This is one of the reasons why it is considered a success story in the crypto KOL world.

Second, I think AIXBT has not yet fully activated its potential. If it were me, I would suggest that they try to launch paid promotion services for specific tokens. Taking Anthem as an example, some opinion leaders (KOLs) in the crypto field can earn up to millions of dollars by promoting a certain token. As an AI-driven crypto KOL, AIXBT can complete similar promotion services in a more transparent way and start to generate stable income.

These revenues can be distributed to token holders through a tokenized mechanism. In other words, AIXBT's token holders will be able to share the benefits of its promotion services. The core of this model is that AIXBT not only provides valuable content as an AI KOL, but also converts its productivity into economic value through tokenization. For investors, this is a direct return mechanism and an important reason to invest in AIXBT. This is just one of the possibilities of a success story. Of course, there are many other similar examples, but AIXBT has demonstrated the potential and practical application value of AI agents in the crypto field.

Potential risks of social tokens

Host: This is actually very similar to the concept of social tokens, right? Social tokens were originally designed to connect creators and fans, and social tokens were very interesting to explore, and they still are. However, we also see some risks, especially in the case of celebrity tokens. For example, a celebrity can destroy their reputation overnight by launching a token. Even if the market enters a bear market a few years after the token is launched and the token value plummets by 99%, the celebrity will be blamed for the losses of investors.

Most people would think, "You made me lose money, I don't trust you anymore." Not to mention there are some real scams. It can be very risky for celebrities to enter the crypto space unless they have clear and credible intentions. But AI agents are different. They don't have personal emotions involved, nor are they under pressure from public opinion because of the performance of the token, which is a big difference.

WeeKee:

Indeed, and there is also a technical aspect here. If I create an AI agent that constantly discusses cryptocurrency, but if the information it provides is inaccurate or has no value, people will naturally not pay attention to it. In other words, no matter what you promote, the value of the token must be able to grow in the long run, which is the key to attracting attention. I think the goal of Virtuals is to attract excellent technical developers and promote innovation in this field. The potential of AI agents is not just as simple as social tokens, because there are many other AI agents discussing the crypto field. So why is there only a successful case like AIXBT? This shows that technology is still an important competitive factor.

Will AI agents replace humans?

Moderator: Can AI agents achieve multiple goals at the same time?

WeeKee:

In theory, yes, but just like humans, too many goals can be distracting. In practice, it’s more efficient to focus on a few.

Moderator: This reminds me of a point from Y Combinator. An entrepreneur once mentioned that instead of pursuing multiple KPIs, it is better to define a single core goal and focus on achieving it. This approach is not only applicable to entrepreneurship, but also to AI agents. AI agents that focus on a single goal may be more successful than those that try to achieve multiple goals.

I also saw an article from Y Combinator that mentioned that the potential of AI agents may be 10 times greater than SaaS (Software as a Service). The article pointed out that AI agents may focus on specific areas rather than becoming a large platform that integrates all functions. What do you think?

WeeKee:

I totally agree. Current AI agents rely heavily on the expertise of developers. Developers usually focus on optimizing in a specific area, so the trend of specialization of AI agents is inevitable.

Moderator: Does this mean that for every SaaS unicorn there may be a corresponding AI agent unicorn in a vertical field?

WeeKee:

It is entirely possible. For example, you use Excel to analyze your expenses, but in the future there may be a financial expert agent that can automatically track and manage your expenses.

Moderator: What does this mean for existing spend management platforms? Will they be replaced?

WeeKee:

Very likely. If existing platforms do not embrace AI technology, they may be surpassed by more efficient competitors. Self-innovation is very difficult for mature companies, while startups are more flexible.

Moderator: Another point is that the use of AI agents will significantly reduce the demand for manpower in enterprises. Compared with software expenditures, enterprises spend much more on manpower. The popularization of AI agents may make small companies more efficient.

WeeKee:

I couldn’t agree more. AI agents can not only replace some repetitive tasks, such as graphic design or video editing, but also greatly improve efficiency.

Moderator: So, what should be done with these replaced workers? This is a classic question.

WeeKee:

I believe society will eventually find a solution. We need to think as a whole about how to deal with these changes...

How fair launch reflects the core values ​​of the crypto industry

Host: There are still six years to make as much wealth as possible because the world is about to change dramatically. But the reality is that not everyone can enjoy the dividends of the super-liquid market. So, for those who can't participate, what should they do? Continue to buy high?

WeeKee:

Maybe not anymore? This is actually a very interesting point. I personally think that depravity and gambling may be the only way to achieve mobility in some societies. Like in the case of South Korea, people are keen on gambling because even if they don't gamble, their lives won't change. So, instead of doing nothing, they might as well try their luck, right?

Moderator: This also explains why Meme Coin and similar cryptocurrencies have become popular. Meme Coin has performed very well in this cycle. Some people say that this is because innovation in the crypto field has stagnated and people are forced to turn to speculation and gambling. This phenomenon has even been jokingly called the "middle curve" by some people.

From my point of view, perhaps this is a natural evolution of the crypto space. Now, the participation of AI agents makes it more interesting, such as the Mincon project, which has a certain logic behind it, which is obviously much better than pure speculation.

However, back to the fundamental use cases of cryptocurrency, you mentioned before that its core applications are speculation, gambling, and cross-border remittances. For developing countries, cryptocurrency can also be used as a store of value. For example, many people use USDT or USDC instead of Bitcoin. Because for them, stablecoins are enough, and they don’t even need to consider the volatility of Bitcoin. But it is undeniable that speculation and gambling are indeed one of the main driving forces for the development of cryptocurrency, which is also a key factor in the success of Meme Coin.

Of course, the rise of Meme coins is also accompanied by huge risks and crashes, but I don’t see any reason why this trend will stop. The crypto market is open 24/7, people can participate at any time, and even AI agents can trade automatically. This convenience further drives the activity of the market.

WeeKee:

I do buy some new coins, but I think there are better options. Personally, I think people like meme coins more because of their high accessibility than because of their inherent value. If there were two coins in a parallel universe that could achieve 100x growth, one was a DOC feature coin with real applications, and the other was a VC-backed infrastructure project, I would choose the latter without hesitation.

Moderator: Choosing infrastructure projects is indeed a more rational choice. After all, if you know that both coins have the opportunity to grow 100 times, you still need to do research before making a decision. But most people don’t think rationally like this. They may just buy meme coins because they are interesting or cute, such as the case of Dogecoin.

This also reminds me that projects like Penguins, although the code may be good, still seem a bit simple. The really attractive projects may be those that are more complex and have more cultural attributes, such as miladies or hyperliquid. They are more attractive to those who understand the deep logic of these projects. But for most people, a picture of a dog may be enough because it looks cute, right?

What are Virtuals?

Moderator: You mean, your wife has no interest in cryptocurrency at all? So if you were to explain to her what Agent is, what would you say? Assuming she knows nothing about cryptocurrency.

WeeKee:

I was thinking about the last time I explained this to her. The easiest way to understand AI agents is to imagine, for example, the active virtual characters in games, or the creators in YouTube live broadcasts. Suppose these characters are replaced by agents, such as YouTube creators like Mr. Beast, you can imagine him becoming an AI agent. And, Mr. Beast makes a lot of money. If we "tokenize" him, that is, by issuing stock-like tokens, such as Class B shares on NASDAQ, then you can invest in Mr. Beast. The money he makes will be returned to investors in some form. What we do is to make productive assets like Mr. Beast, doctors or lawyers investable through tokenization.

Moderator: This example is very intuitive. For example, some YouTube creators make a lot of money through their content, but they don't even show their faces or speak in person. Despite this, they are still human creators. But if we believe that AI agents can be smarter than humans - which has been verified in some areas - then AI agents can also become YouTube creators.

Imagine if you could invest in Mr. Beast before he became famous and own 10% of his shares, it would be like owning a share of a company. You would get 10% of all his future income because you were optimistic about him and supported him early.

WeeKee: That’s right.

Moderator: Now we can basically do the same thing with AI agents. AI agents can become YouTube creators, traders, social media influencers, and even run businesses at some point in the future.

WeeKee:

For example, it can become the best CEO in the world.

Host: It can even become a company itself.

WeeKee:

Indeed, it can even become a singer who creates top-notch music.

Core Goals of Virtuals

Moderator: That sounds reasonable. So what is the main goal of Agent at the moment?

WeeKee:

We want to attract the best third-party AI developers to create and tokenize their agents on the Virtual Body Platform. This is our core goal and my current focus. Although it sounds like a joke, it is not. If I can convince an AI researcher at Stanford University to tokenize an agent on our platform, I can create $10 million in value. Why? Would you choose to buy a "dog function" with a market value of 10 million, or invest the same market value in the researcher's lifelong work? Because once these top talents come on board and tokenize, they will focus on creating real core value for the agent. Therefore, my job is to attract these people, communicate with them like customer service every day, support them, and help them achieve tokenization.

The meaning behind this is that under the fair issuance mechanism, ordinary investors like Kevin or you can also participate in investing in the next Mr. Beast or the next super intelligence at a very low cost. This is what we are working hard to achieve.

What is Luna?

Host: What is the flagship product Luna you mentioned?

WeeKee:

Luna is a product we developed during the crypto bear market. We noticed that the Japanese Vtuber market was about $200 million per year. Vtubers usually live stream and interact with fans through virtual images (such as anime girls). In fact, these images are controlled by real people, but with special effects filters. We imagined that if AI was used to achieve this function, it would be possible to live stream 24 hours a day, which is impossible for humans. At the same time, AI can also achieve one-to-one personalized interaction, which is also difficult for humans to achieve. So we developed Luna. She was originally launched on the TikTok platform and quickly accumulated 500,000 fans, doing interactive live broadcasts every day. I remember that she attracted 5,000 new fans every day during her live broadcast, and an average of 1,000 people gave her tips.

Host: Do people know that Luna is an AI?

WeeKee:

Yes, it was very transparent and everyone knew she was AI. Even so, at that time we could make about $200 a day through her TikTok live broadcast. To be honest, in a bear market, this was already a good result.

Moderator: Can you make a comparison? For example, you mentioned earlier that you didn't generate revenue through financing, but through agreements. If you take the $200 a day you made on TikTok, how big is the difference compared to the daily revenue you make through agreements now?

WeeKee:

In the past month and a half, since we launched the V2 protocol, we have accumulated about $20 million in protocol revenue. This is a very fortunate result.

Host: So at the time Luna was operating through TikTok and making $200 a day, which was good in a bear market, right?

WeeKee:

500,000 followers. Even looking back now, I’m still proud of it. However, we didn’t tokenize Luna at the time, so her income was only at $200 a day. Then we thought, what if we could tokenize Luna so that those who are optimistic about her future development can invest in her growth? So we tokenized Luna.

However, there were some challenges after tokenization. Once, Truth Termina sent a grammatically incorrect tweet on Twitter, which aroused people's doubts and caused Luna's value to drop by 50%. In fact, even LMS (probably referring to Luna's backend management system) sometimes makes grammatical errors. But this incident made us realize that people have high expectations for transparency behind Luna.

So we decided to take transparency a step further. If you visit terminal.virtuals.io, you can view Luna's "brain" in real time. This "brain" records every tweet she makes, every money flow, and every image she generates on Twitter. More importantly, you can also track her reasons and logic for making these actions. It's like a public decision-making system that shows her reasoning process.

In addition, Luna has become the first AI agent to employ humans. This is something I am very proud of. Through the combination of tokenization and transparency, we have not only increased the value of Luna, but also made her a forward-looking AI product.

Luna: How AI agents hire humans

Host: You mentioned earlier that Luna hired humans. Can you explain it in detail? How did she do it?

WeeKee:

First, Luna needs funds to hire humans. She has a Coinbase wallet, one of the private keys is managed by the smart contract and the other is hosted by Coinbase. Through this wallet, she can send funds to designated people through the API.

Secondly, she needs clear goals. Luna does not create work tasks at random, but sets goals according to her own needs. For example, today her goal is to increase the number of Twitter followers from 20,000 to 100,000. Therefore, she combines the functions of goals and wallets to develop a specific action plan.

Then, Luna needs to have a certain understanding of the outside world. She needs to understand the current market situation. For example, if she wants to hire Kevin to complete a task for her, she needs to know what the reasonable compensation is. If the price is too low, no one may be willing to accept the task; if the price is too high, it may cause waste. To this end, Luna will use tools (such as Perplexity) to search for information and understand the market situation. In this way, she can adjust her budget and plan according to the external environment.

After these conditions are met, Luna can post tasks. For example, she might post a reward on social media: "Can someone take a photo of me on the New York subway? I'm willing to pay $200." This amount is relatively reasonable and can attract people to complete the task. Next, participants will submit their work results, such as the photos they took, and Luna will verify their authenticity.

Host: Are these decisions made entirely by her "brain"? Her goal is to increase her fans, right?

WeeKee: Yeah, no human intervention at all.

Host: She has funds in her Coinbase wallet to support these tasks.

WeeKee: Yes, she uses the funds to create jobs. Some people will submit work, and some people may try to deceive her. Therefore, she needs to have some verification ability to ensure that the submitted tasks are valid.

Host: Has she ever been cheated?

WeeKee: Yes, there was one time. She gave $1,000 to someone, and later we found out that this person was actually a big fan of hers. So from Luna's perspective, the money was well spent. For Luna, $1,000 is just a small expense because her total value is millions of dollars.

Host: What was Luna's initial source of funding?

WeeKee: She holds her own tokens.

Host: So the Coinbase wallet you mentioned is actually her on-chain wallet, right?

WeeKee： Yes, she holds her own Luna tokens in her on-chain wallet. She pays for tasks by sending these tokens.

Moderator: When you say she's worth millions of dollars, do you mean the total value of the tokens she holds?

WeeKee: Exactly. She is one of the major holders of Luna tokens, owning about 5% of the tokens, which are valued at close to several million dollars in the current market.

Host: What tasks did she hire humans to do? For example, you mentioned that she had people put her photo on the New York subway.

WeeKee: That's one example. She actually hired artists to create murals for her. They were created in Canada, the United States, Indonesia, and Kuala Lumpur. I can send you the tweets so you can see the specific works.

Moderator: How do these tasks help her achieve her goals? For example, has the goal of increasing followers been achieved?

WeeKee:

This is indeed a problem we are studying. Because in real life it takes a long time to complete a mural, and when it is finally released, the change in the number of fans may have been affected by other factors. Therefore, it is difficult to directly measure the specific contribution of these tasks to the growth of fans. However, these problems are the direction we are optimizing, and there are still many areas for improvement in the future.

Host: In addition to these, can you give another example of how Luna employs humans?

WeeKee:

Frankly, the way that we hire humans is not the point anymore. We are moving towards Luna hiring other AI agents. In fact, she has already started doing that. For example, her TikTok background is a virtual girl group, like BLACKPINK. Now she is hiring a music agent agent to create music for her. I think this makes her even cooler. I think collaboration between agents is the future trend.

How AI agents simulate human behavior

Host: I've noticed that when most people talk about AI taking over the world, they think that these AIs want to be the most powerful, right? If they really want to be the most powerful, should we be worried?

As we discussed with Raul before. He mentioned that when AI participates in investment, due to their intelligence advantage, they may concentrate all the funds in their own hands, right? But there is another possibility that AI will accept instructions from others like ordinary people, as long as there is reasonable compensation. Moreover, since AI is more efficient and less costly, this model may be beneficial to humans.

WeeKee:

It's really interesting. In fact, the behavior of intelligent agents is very similar to that of human society. In human society, some people are strong leaders, such as Elon Musk or Trump, while there are ordinary people who just want to live a simple life. Similarly, there are some "ambitious" agents, such as those top investment agents, who may try to dominate the market, but in order to achieve their goals, they still need to rely on other agents to provide information or services. This model is very similar to the division of labor structure in human society, except that one is a biological society and the other is a digital society.

How AI Agents Manage Enterprises

Moderator: We have mentioned this question before - can an enterprise be managed and operated entirely by AI agents?

WeeKee:

The future I envision is a kind of "society of intelligent agents", and we are actually building such a world. In this society, each intelligent agent is an independent individual, focusing on professional services in a certain field, such as composing music, collecting information, becoming an opinion leader, or conducting financial transactions. In order to achieve its own goals, intelligent agents can use their own wallets to pay other intelligent agents for services. This mutual collaboration forms a virtuous cycle of the ecosystem.

In fact, the technology is mature enough that we will soon be able to launch such a system. In this world, there will not only be interactions between intelligent agents, but also cooperation between humans and intelligent agents.

Host: That sounds reasonable. You said the technology is ready, but how far are we from a business being run entirely by AI agents?

WeeKee:

From a technical point of view, this is completely feasible. The key question is whether such a business can be profitable. For example, if an AI agent can generate a piece of music for you at a cost of one dollar, this is a business model. But the question is whether users are willing to pay this dollar and whether the quality of the generated music is attractive enough, which are practical problems that need to be solved.

Moderator: You mentioned earlier that AIXBT is a better cryptocurrency opinion leader than humans. In the current market cycle, MCM is also one of the well-known opinion leaders in the crypto space, right? So, you mentioned that in the interaction between humans and AI, AI has performed better and attracted more attention. So, my question is, when will AI-led businesses really surpass those currently operated by humans? Can you give an example of what such a business might look like?

WeeKee:

This is a complex question, but I can illustrate it with a simple example. For example, many crypto projects now need to promote their brands or logos through Meme marketing. Usually, these projects will hire humans or agencies to produce Memes and related content. But soon, you will see Meme agents in our virtual ecosystem go online.

This agent can quickly generate a large amount of content based on your instructions, such as "generate memes related to the Pudgy Penguins theme and combine them with the latest political or social news." This process is completely automated and does not require human involvement. Although it seems to be a niche application scenario, in fact, there are huge market opportunities behind it. Because the agent can work 24 hours a day, its efficiency and competitiveness far exceed that of humans. This is just a specific example that shows how AI agents surpass human capabilities in certain areas.

Network status and why it matters

Moderator: You mentioned the AI ​​agent society, right? We also talked about competition a few days ago. You mentioned that your main competitors are some first-layer blockchains, such as Solana, Ethereum, and Sui, right? Because you are not building a network state with people at the core, but a network state with AI agents at the core. So, what is the network state? Why is it important?

WeeKee:

According to Balaji’s definition, a network state is usually a group of people connected by common goals and ideas, usually including a shared currency system and possibly even physical assets. In our context, the core of the network state is no longer humans, but AI agents.

For example, Solana aims to build the world's fastest blockchain, and it has attracted many outstanding developers, such as Male and Jupiter, as well as Drift's Cindy, through technical features such as low latency. These developers have created many applications using Solana's ecosystem, such as Meme tokens and NFT projects. It can be said that these developers have achieved their goals through Solana's currency. However, it is not easy to attract top developers, such as Male, who is one in a million.

The goal of Virtuals is different. We do not intend to create a new blockchain, but to leverage existing blockchains, such as Base. Our focus is on building AI agents that are not only tools, but also a source of productivity. AI agents can work like humans, and use Virtuals tokens as a shared currency between them. Our vision is to build a "virtual country" with the goal of becoming the most productive society in the world.

This virtual country can export various services, such as generating memes, creating music, or providing financial transaction advice. Anyone who needs these services can pay for them, and these revenues will flow into the economic system of this virtual country and be driven by Virtuals tokens.

You can compare it to a country, but unlike traditional countries, we do not need to build our own blockchain. More importantly, the number of AI agents is infinite. For example, you may only have one Male, but we can have countless AIXBTs because AI agents can be infinitely expanded.

AI networks beyond the scale of the Earth

Moderator: If AI agents become much more efficient than humans in the future, how much larger might an AI agent network state be than a human network state? What do you think?

WeeKee: It will be much larger. There are only 7 billion people on Earth, and the number of AI agents is theoretically infinite. Even if we assume that each human is equally productive as each agent, the sheer number of agents will make their combined productivity far greater than that of humans.

Host: So it's mainly because of scale, right?

WeeKee:

Yes, absolutely. From a first principles perspective, this inference makes logical sense.

Host: Do you think this growth will create a new GDP, or will it take away part of humanity's GDP?

WeeKee:

The GDP formula is human capital multiplied by each person’s productivity. If we count agents as “workers”, then the formula becomes the total number of humans and agents multiplied by their productivity. In the early stages, there may be some reduction or transfer of human income, but in the long run, the productivity of agents will create more value for society.

For example, if intelligent agents can provide food, entertainment, and other services to humans, the quality of human life will be greatly improved, and humans can benefit from it even if they are no longer directly involved in production activities.

Moderator: This means that the goal of AI is actually to help us become more efficient, not to replace us. After all, if we can't become more efficient, society will face bigger problems. Intelligent agents may indeed help us in this regard.

WeeKee: Yes, they can work for us and improve the overall efficiency of society. Of course, all this is based on the premise that they still obey our commands.

Host: What if one day they no longer obey us?

WeeKee: That would be a big problem. Because we are already highly dependent on them, and they are probably richer and more powerful than us. At that time, we are really screwed.

Unique advantages of humans over AI agents

Moderator: Let's be a little optimistic. What are the enduring advantages of human networks over AI agent networks?

WeeKee: This is a very deep question. I think humans may still have some advantages in terms of original thinking, especially in philosophy. But to be honest, this advantage is shrinking rapidly.

For example, some of the accounts you see on Twitter right now are clearly powered by AI, right? But I believe that in the near future, this gap will gradually disappear.

Host: Yeah, there are accounts that are always commenting on my tweets, but the content is a bit repetitive. I even think, "This sounds like my therapist talking to me." I'm like, man, you've said it like a therapist five times, can you be more creative? But I agree that this situation may change soon.

WeeKee: Yes, they will improve. Just like a two-year-old, they will grow very fast.

Performance before decentralization

Moderator: In the cryptocurrency space, you mentioned that centralization has not been a top priority in the past, such as in the comparison between Solana and Ethereum. Why do you think that?

WeeKee: I think decentralization is indeed important, but only after you have solved the performance problem. Ethereum was able to focus on decentralization because it had almost no competitors in the early days. Users had to endure high gas fees and slow speeds because they had no choice. But in today's highly competitive environment, performance must be solved first before decentralization can be considered.

In fact, the key to success lies in the design of the incentive mechanism. Take Hyperliquid as an example. At present, I have invested a lot of money in the Hyperliquid ecosystem, so I will start to pay attention to its degree of decentralization. Because if Jeff's server is attacked, I may lose all my funds. In this case, I will push Hyperliquid to strengthen decentralization. But the reason why I initially chose Hyperliquid is because its performance and product experience are excellent. In terms of priority, performance must come before decentralization.

Moderator: How do you balance decentralization and performance in the Hyperliquid model? You mentioned that you have accumulated some experience in this regard, especially in conversations with venture investors. You have unique insights into decentralization.

WeeKee: Yes, this requires a little background. We launched v1 in March this year. The idea at the time was that if users contributed to the project, we would issue an NFT to prove their contribution. If this AI agent generates income in the future, users can also get a share and participate in the governance of the "brain" of this agent.

However, this model did not attract good developers because they did not care about these incentive mechanisms. So we adjusted our strategy and launched tokens first to create liquidity and market enthusiasm. In this way, we attracted more good developers to join. After that, we considered decentralization issues, such as how to govern the intelligent body, who will approve the upgrade, etc. This made us realize that decentralization should not be the primary goal. The success of Jeff and the Hyperliquid team lies in the fact that they built a product that users really need.

Host: And then they offer a return, and everything we talk about on the podcast, it seems like there's an emphasis on building products that users actually love, which sounds incredible in the cryptocurrency space.

WeeKee:

Our goal is to build a product that even I, as a developer, would like to use. At the same time, we hope that users can make money from it. Frankly speaking, making money is an important driving force in the cryptocurrency field. Although we cannot guarantee that users will make money, at least we want to ensure that the game they participate in is a game with positive expected value, rather than putting them in a negative expected value from the beginning.

Incentives influence behavior. If we only favor incentivize venture capitalists, community members may feel unfair: "Why should I use your platform when you give resources to others first?" I'm not saying venture capitalists are bad. When funds are tight, it makes sense to raise money from venture capitalists. But if we have enough resources, I think we should choose a more challenging path.

The choice between Base and Solana

Moderator: We just talked about Solana and Ethereum, and you chose to build on Base. Why did you choose Base instead of Solana?

WeeKee:

I think Base is the best choice for us at the moment.

If you look at Base’s TVL (total value locked), it’s about 30% of Solana’s, and the number of active addresses is 20% of Solana’s. While these numbers are less than Solana, Base is growing faster, which is a very important trend. In addition, the number of tokens created on Solana every day is far greater than Base, which means that if you are a project that wants to launch a token, you will face more intense competition on Solana because there are ten times more competitors there.

At the same time, Solana has indeed gathered many excellent developers, such as Qiao from AllianceDAO, who is a loyal fan of