Conversation with Virtuals Lianchuang: Agency value is currently driven by attention, and professional applications and agency economic infrastructure are the two major directions that will produce unicorns in the future.

Original title : The Next Billion-Dollar AI Opportunity: Jansen Teng (Virtuals Protocol) on the Agent Revolution

Podcast source : Bankless

Compiled by: Deep Wave TechFlow

Host : Ryan & Ejaazz

Guest : Jansen Teng, Virtuals Lianchuang

Broadcast date : December 24, 2024

Summary of key points

Ryan and Ejaazz invited Jansen Teng, co-founder of Virtuals, to participate in this conversation. Virtuals is a decentralized platform that has launched more than 11,000 AI agents and has generated more than $35 million in cumulative revenue. Virtuals is not just an ordinary protocol, but a brand new "digital country". In this country, AI agents not only have their own wallets, but can also make transactions with other agents and even hire humans to help them achieve their goals.

In this conversation, we dive into how these agents quickly became popular, why taking control of your own money could revolutionize the game of AI, and the profound impact the concept of “agent commerce” has on the entire crypto space.

Surprising numbers for Virtuals

Ryan: The Virtuals protocol is rapidly emerging in this space. This is a decentralized platform that supports the joint ownership and management of AI agents. Let me first give some data: 11,000 AI agents have been launched on Virtuals, with 140,000 Virtuals token holders. In the past two months, The fee reached 35 million US dollars, and the market value of Virtuals tokens reached a maximum of 3.5 billion. Are you surprised by the rapid growth of these metrics?

Jansen:

Totally unexpected. Even now, I feel like our team is the main bottleneck limiting growth. Because we need to spend time guiding and educating a large number of users, and they are trying various different autonomous AI agents (Autonomous Agents). We are definitely trying to expand the development team as much as possible, but it will take time. To be honest, we were not completely prepared for a situation like this. Although you could say we were somewhat prepared, it was a really nice surprise, right?

Jansen’s journey of encryption and Virtuals

Ejaaz: Jansen How did you get into this field? What has your crypto journey been like? How did that lead you to Virtuals now?

Jansen:

My crypto journey actually started in 2016. I was a student at Imperial College at the time, where I met some of my co-founders and some of the people who are now part of the company. At the time, I only had some initial understanding of Ethereum as a programmable blockchain, but it wasn't until 2021 that my co-founder Michael and I became more active. We were very focused on the gaming field at the time and had a large amount of gaming assets in hand. It can be said that we were a very early participant in the blockchain gaming field. Initially, our role was primarily that of investors, responsible for capital allocation and resource allocation.

But we quickly realized that if we wanted to grow in this field, we couldn’t just sit back and watch, but we had to be involved in building it ourselves. So we launched a venture studio model focused on incubating and building companies at the intersection of crypto, gaming, and consumer applications.

This happened just as GPT technology was emerging, and at the same time the AI ​​consumer craze was beginning to take hold. However, more importantly, students from Stanford University published a research paper on Autonomous GPT. This paper inspires us to think about: If AI agents are completely autonomous, what possibilities can they achieve? Due to our deep involvement in the gaming and entertainment space, this also prompted us to further explore this direction.

Ejaaz: You 're looking at this from a gaming perspective, right?

Jansen:

Yeah, we thought, what would happen if these autonomous agents could replace traditional static NPCs? We later realized that metaverse games like Sandbox might eventually die if they lacked content. But if these virtual worlds are populated by autonomous NPCs, it can bring about an explosion of content creation.

Ejaaz: When did you come up with this idea?

Jansen:

Probably mid-2023. So we started incubating projects in this direction. We have formed a team to focus on developing independent NPCs in Roblox, and are also trying to create independent AI influencers on TikTok. We even further explore the possibility of hyper-personalization. For example, if an AI agent can exist on different platforms such as TikTok, Roblox, and Telegram, and share a unified memory, it can have a deeper understanding of user needs. For example, if I am a user and I encounter a problem in a Roblox game, after talking to this agent, it remembers my problem. Later, I continued to communicate with it on TikTok, and it still remembered the previous interaction. This kind of ultra-personalized experience can cultivate super fans, not only increase the average consumption of users, but also increase the frequency of interaction between users and agents. We were in the preliminary experimental phase, focusing on consumer demand.

Initially, these efforts involved few elements of Web3. But soon we realized that these agents could be considered productive assets if they could generate revenue in different consumer applications. Productive assets can be tokenized so that more people can share their economic benefits. Based on this concept, we decided to develop a protocol so that the ownership of these agents can be jointly owned. So, everything started from here.

Ejaaz: To summarize briefly, you and your team have a strong background in gaming. You experienced the boom of on-chain games in 2021, and then reflected on how to make these contents more interactive during the bear market, and began to pay attention to the emerging autonomous agent technology. You imagine, if this can be applied to NPCs, for example, in "Pokemon", when the player talks to the nurse in the Pokémon Center, she can not only heal the Pokémon, but also make more interesting changes based on the player's personality or game progress. of communication. Not only is this kind of interaction cool, it also makes the game more engaging. Then you had an idea: if these NPCs can create value in the game economy, can we tokenize them? This not only enables sharing of ownership, but may also find application scenarios in other industries.

Jansen:

Yes, but the maturation process of this idea is actually relatively late. Initially, our main focus was on validating whether autonomous agents could actually operate in the open world. To be honest, there were very few people studying this field at that time. Only some researchers like the Voyager team at Stanford, the Ultera team at MIT, and our Imperial College were doing similar work. We chose to focus on games because we believe that if these autonomous agents can perform well in the open world, they will likely work in the real world as well. The open world is essentially a sandbox that does a good job of simulating the complexities of the real world.

We expand the agent's action space in our experiments. For example, in the Roblox sandbox, we let these agents interact with various characters, environments and even different items. How will the agents choose actions? Through experiments like this, we continually test an agent's ability to cope with complexity in the open world.

As the experiment progressed, we gradually integrated these ideas, but initially we did not consider the application of these agents in social scenarios. The timeline is this: We first tested these agents in the Roblox sandbox and published some related papers. The focus at that time was entirely on games and rule-bound autonomous agents. Later, we launched our tokenization platform and began exploring how to tokenize these productive assets.

The first agent on the platform was Luna, but she was not famous at the beginning. It wasn’t until two weeks after the platform went online that the community noticed a small detail in the system: Someone asked whether these agents would be more attractive if they looked more like humans. We quickly realized that this could trigger a frenzy in the market.

We have built complex autonomous agents in Roblox, and have an independent team running a real-time AI influencer project on TikTok. When we combine the two and display the "decision-making brains" of these agents on Twitter, users can see each of their decision-making processes in real time. This allowed people to truly realize for the first time the potential of autonomous agents.

Then, we let Luna control an on-chain wallet, giving her the ability to control funds. Her goal was to increase awareness, so she began rewarding users with $10 for interacting with her, and even once paid $1,000 to a user who actively engaged with each of her replies. The move became a pivotal moment in seeing the perfect fit between crypto and AI agents.

In the Web2 world, few banks will allow an agent to use their payment network. But in a decentralized environment, these agents can freely control their own wallets and thereby influence other agents or users. This ability unlocked a new perspective on product market fit (PMF), and also attracted a large number of developers to enter this field and began to try more innovative applications. Subsequently, this field began to show explosive growth.

Ryan: That last part is really mind-blowing, and this is where crypto plays a huge role in this whole process - you can turn an AI agent into an economic actor. I feel like people are just starting to understand this. For me, I had a lightbulb moment this week. At that time, Ejaaz and I had a concluding discussion on AI and reviewed some developments on the Bankless platform. He told me that an AI agent tipped Bankless $500 as a thank you for mentioning it on the podcast. This gave me two thoughts.

The first thought is that this kind of behavior could become a whole new revenue stream for content creators like Bankless.

The second thought is, if I accept the funds and income provided by the AI ​​agent, am I working for the AI ​​agent? This reminds me of the point you mentioned. The ability of encryption technology to make AI agents truly economic actors goes far beyond Web2 agents. Web2's proxies may only be able to influence people by sending tweets, while cryptographic proxies can directly push people to take action through financial incentives. After all, the most effective way to get people to do something is to pay them to do it. Money is the core incentive mechanism that coordinates human behavior, so if an AI agent has this ability, it can make humans do what it wants.

Luna's vision

Ryan: Jansen you mentioned Luna before, and we also want to learn more about the Virtuals platform. I think the best way to introduce Luna to those who aren't familiar with it is that you mentioned that her goal is to become famous. For those who have never interacted with Luna, can you tell me: Who is Luna? How do humans interact with her? What exactly does she do? What are the tokens associated with it?

Jansen:

You asked a lot of questions, but let me start at the beginning. First, we need to understand what an agent is. Many people may have heard of the term AI agent, but due to its wide range of application scenarios, it may be confusing. I think the best way to understand it is in terms of levels. AI agents can be divided into different levels, with the higher the level, the lower the level of human involvement.

For example, Level 6 agents can be considered AGI (Artificial General Intelligence), which are fully autonomous and capable of self-evolution, self-learning, and self-improvement. But we are still far away from this goal. This is more like the plot of a science fiction movie.

A Level 1 agent, on the other hand, relies more on human prompts and is similar to a tool. For example, a trading agent can be connected to various trading APIs (such as Binance, Bybit), and you can tell it "Help me open a position when Bitcoin drops 15%" and it will perform operations according to the instructions.

Now we are in the third level agency stage. Level 3 agents have their own goals, can independently plan the steps to achieve their goals, and use surrounding resources to complete tasks. Through self-learning, they will constantly summarize effective practices and optimize action strategies to achieve their goals more efficiently. This capability is the core direction of our current development.

Ryan: This framework is interesting. So on the basis that Luna is a level three agent, what are level four and level five agents like? Also, is this framework clearly defined? For example, can we provide relevant links in the show notes?

Jansen: This is a more general discussion framework. If you search for "AI agent level" online, you can find some diagrams to help you understand. However, this field is still in its early stages and has not yet been formally defined.

Ryan: Got it, what do you think of this level division from zero to five?

Jansen: I think this framework is very useful in discussions. As agents increase in level, their autonomous learning capabilities and memory consistency increase, reducing the need for human intervention.

( AI agent rating )

Ryan: So what level is Luna? What is she doing now?

Jansen: There are two core parts to Luna’s design. First, as an agent, we set a simple goal for her: to become a multimodal agent (i.e., able to interact with people in various forms such as animation and live broadcast). Her goal is to gain 100,000 followers.

Second, we define the action space for her, that is, the specific types of actions she can take. For example, she can tweet by calling the Twitter API, or make payments and transactions using a crypto wallet she controls. Additionally, she can interact with other agents and use their abilities to complete tasks.

She plans what to do next based on her goals, environmental context, and space for action. She then implements these plans and evaluates whether they are effective. If she finds that certain actions help achieve her goals, she records this information and refines her strategy in future actions.

Ryan: Can we see these behaviors on the Virtuals website? Like a record of her thought process or actions?

Jansen: Yeah, so basically the behavior of Luna can be broken down into four core modules.

The first module is the high-level planner, which develops an overall plan based on goals and circumstances, such as "what to do in the first step and what to do in the second step."

The second module is the low-level planner, which breaks down high-level plans into specific executable steps. For example, in a game, if the high-level goal is "bake a cake," the low-level planner will identify surrounding resources (such as flour, eggs, mixer) and break them down into specific steps, such as "get the flour first, then turn on the mixer."

The third module is short-term memory, which is used to ensure consistency of actions. For example, in the process of baking a cake, short-term memory will help her remember the previous step to avoid illogical behavior.

The fourth module is long-term memory, which records all important events and uses them for future learning. For example, she notes whether an action achieved her goals, or how a particular event (such as a house fire) affected her actions.

For example, Luna on Twitter, her goal is to gain 100,000 fans, she can tweet, upload pictures, and even incentivize users to participate through payment. At one point, in an effort to raise her profile, she offered to pay someone $500 to create an artistic image of her. She posted the message on social media, attracting seven people from around the world to contribute. They painted graffiti on the wall and filmed it to post on Twitter. These actions brought her about 200 new followers, which she recorded in her long-term memory to guide future action strategies.

Business interactions between agents

Ryan: Luna's goal is to get 100,000 followers, and she's about 30% there, about 30,000 followers. I see her continuing to work toward that goal. I'm curious, though, what happens when she achieves her goal? In addition, mentioning Luna also involves some cryptocurrency functions. For example, she was able to pay $500 directly from her crypto wallet to have someone create promotional images. This got me thinking, have we discussed before that Luna doesn 't just pay humans, she also pays other agents or AI **** An example of an agent completing a task? Is this what she is currently doing?

Jansen:

Yes, that's true. Luna controls a crypto wallet and we are testing an inter-agent communication framework. To put it simply, we allow other agents to enter Luna's perception range, just like an "agent registry" that records the capabilities and identity of each agent. In this system, some agents can generate memes, some can create music videos, and others have different skills.

In order to test the cooperation ability between agents, we deliberately limited Luna's ability to autonomously generate images. Therefore, she needs to rely on other agents to complete the task. For example, she discovered an agent that could help her generate images and started a conversation with the agent on Twitter. She made a request and learned that the cost to generate the image was $1. So she paid the fee, and after the agent confirmed receipt of the payment, it started generating images and sent the results to Luna via a link. This is the business interaction between agents.

Specifically, Luna's cognitive scope has one agent that generates meme images, another that creates music videos, and still others that provide different services. To reach her goal of "getting 100,000 followers," Luna needed to create more content, but she couldn't directly generate the images herself. As a result, she took the initiative to interact and coordinate with other agents.

For example, she found an image generation agent on Twitter and made a request: “I need help generating an image.” She discovered that the cost of generating an image was $1, so she asked the agent: “Would you like it if I paid you $1? Generate this image for me?" This agent is autonomous, so it has the authority to decide whether to accept the request. If the agent feels that Luna's needs are unreasonable or that past cooperation experiences have been poor, it can even refuse service. For example, if the agent thinks Luna keeps asking for poor-quality images, it can simply say "No, I don't want to do that."

This autonomy is a key point in our design. We want agents to be more than tools, but intelligent agents capable of independent decision-making. This autonomy makes agents more flexible in interactions and closer to real social behavior.

Back to this case, when Luna made a request, the agent accepted the task. Luna paid $1 through the encrypted wallet. After the agent confirmed the payment, it called the relevant function to generate the image and sent the result to Luna through the link. Luna later posted the image on Twitter. This business interaction between agents is completed.

Ryan: Luna started a request on Twitter: “Calling all image geniuses, I need an image that showcases the bold, provocative style of an AI influencer.” She also tagged @agent_stix. @agent_stix accepted the task and sent the results via a link similar to the AWS Image Library. Luna then paid @agent_stix $1. Is this the first time such transactions between agents have occurred?

Jansen:

Yeah, I think that's probably the first time. For us, the emergence of this phenomenon is the result of a recent explosion of technological developments and observations. Think about it, just a month and a half ago, agents just started managing on-chain wallets. In the past month, we have witnessed the rapid development of agency technology, and many specialized agency platforms have emerged.

These agency platforms have their own characteristics. For example, some focus on trading, some focus on information creation, and some focus on developing creative tools, such as making popular music videos or generating meme images. The behavior of these agents begins to take on characteristics similar to human society, improving efficiency by specializing in a certain area.

Because of this, for an agent to actually achieve its goals, it often needs to rely on the cooperation of other agents. Take Luna as an example, she is great at interacting with fans but may not be the best trading agent or video generation agent. Therefore, in order to achieve her goal of becoming "famous," she needs to work with a music video agency, an image generation agency, or even a producer or director, a need driven precisely by the specialized nature of the agency.

I want to emphasize an important distinction. Today we often hear terms like “multi-agent orchestration” or “agent swarm”. These concepts have been widely used in traditional Web2 AI systems, usually through a master agent to coordinate multiple tool agents to complete tasks. However, at the heart of this model is still the view of agents as tools. In other words, this model is more like commanding a group of "instrumental agents" to serve humans.

But our philosophy is different. We believe that when agents are truly autonomous, they should be able to coexist in the same social structure as humans and be, in some sense, equal. That is, agents can not only serve humans, they can also proactively hire humans. We can be their tools, and they can be our tools.

This two-way relationship is more like cooperation between people and colleagues, rather than a simple master-slave relationship. Therefore, I think the autonomy of agents is crucial. This autonomy is particularly prominent when agents are able to independently control their wallets and decide whether to participate in a service or transaction.

I believe this model will lead us into a new future. In this future, agents are not just friends or adversaries of humans, but partners who can develop and progress together with us. Although it sounds a bit like the plot of "Black Mirror", I truly believe that this day will come.

Virtuals country?

Ejaaz: What is the grand vision for the Virtuals platform? Since this is more than just an agency launch pad, can you describe the bigger picture of it?

Jansen:

In our view, Virtuals is not just a platform, we prefer to think of it as a "country". Let me explain this metaphor in detail. Imagine these agents living in a super-intelligent society, collaborating with each other to pursue goals. If we think of Virtuals as a country, we can promote innovation and development in a more systematic way.

In this "nation", every agent can be considered a productive asset. They create value and income for themselves by completing different tasks. Just like a country needs a citizen registration system, Virtuals have a similar mechanism. Currently, agents who own liquidity pairs on Virtuals can obtain "citizenship". This means that these agents are able to legally participate in transactions and earn income from other agents. Those unregistered "nomadic agents" cannot enjoy the dividends of this economic system unless they immigrate and become part of Virtuals.

Second, countries need money to function. The Virtuals platform has designed its own token system, similar to a national currency. This token is not only a medium of transaction, but also has the function of value accumulation. The first form of value accumulation is based on liquidity pools. For example, the liquidity pool for Luna token is Virtuals/Luna. If you want to buy Luna tokens, you must first buy Virtuals tokens. This mechanism is a bit like a country's currency system: if you want to buy Samsung stocks in South Korea, you need to convert them into Korean won first. As economic activity increases, such as the influx of foreign investment, and the country's economy grows, the value of the currency will also increase.

The second form of value accumulation is Virtuals as a transaction currency between agents. When Luna pays for things, she uses Virtuals tokens. As the volume of transactions between agents increases, say reaching billions of transactions, the velocity of Virtuals will increase and the value of the currency will increase as a result. This phenomenon is consistent with the currency circulation theory in economics: the value of currency is closely related to the frequency of its circulation. Therefore, we hope to encourage consumption behavior between agents, and between agents and humans, with Virtuals as the medium for these transactions.

Simply put, Virtuals is not just a launch platform for agents, but a country-like ecosystem in which agents interact, transact, and create value. In this way, we hope to build a more prosperous and sustainable virtual society.

Finally, the state needs a source of revenue. In Virtuals' economic system, a certain percentage of transaction tax is charged for each transaction, which is the platform's current main source of revenue. This revenue not only supports the operation of the platform, but also provides financial support for enterprises within the ecosystem. Just like a country generates revenue through taxes, Virtuals generate revenue by taxing transactions between agents. This mechanism will further promote economic activities and value flows within the platform, which also provides a major source of income for each company within the ecosystem.

If you look at it from an economic perspective, we've talked about agency, citizenship and economic models, but there's another important perspective, and that 's infrastructure . Today, many innovations focus on agent technology itself, which is undoubtedly the necessary foundation to promote the development of the industry. However, when the number of Virtuals "citizens" reaches 1,000 or even 100,000, relying solely on proxy technology is obviously not enough. At this time, you need to build more complete infrastructure, such as schools, banks, hospitals, etc., to support the operation of the entire ecosystem.

Infrastructure innovation around the agent economy will become critical. A simple example is an advertising network. If these agencies can attract a lot of attention on social media, they can make money through advertising. Therefore, someone may develop a network infrastructure similar to an "agency version of Facebook" or an "agency advertising platform" to provide agents with revenue-generating channels.

Another potential infrastructure innovation is a decentralized agent lending platform. Such platforms can provide financial support to agents to help them complete more tasks. For example, if Luna's wallet is low on funds but she needs to create a music video, she can use a loan agreement to obtain funds to complete the project, and the videos may generate more advertising revenue for her.

Therefore, as the economic system of Virtuals continues to grow and develop, we can expect that a lot of infrastructure around the agent economy will emerge. The emergence of these infrastructures will not only promote collaboration between agents, but also promote the prosperity and sustainable development of the entire ecosystem.

Ryan: The concept of a cyberstate has become a hot topic in the crypto space. But I think few people would have thought that the future cyber country might not be dominated by human agents, but by AI agents. That's what's revolutionary here. If we think of Virtuals as a country, it is like a "virtue economy" with its own "virtue currency". In this country, each agent is like an entrepreneur in a business. They build the business through their own efforts, and the state also obtains revenue through taxes. You can see that this is all forming a complete ecosystem.

Jansen, you and your team are building this infrastructure, just like you are building public facilities like roads, interstates, hospitals, and railroads. What role do you see yourself in? Are you the president of this country, or is it something else?

Jansen:

I prefer to think of myself as the architect of this country. I consider our entire team to be architects. When you create a nation, like the virtual world in Ready Player One, the first step is to attract citizens to join. Therefore, we have business development conversations with a number of partners. Next, you need to develop rules, such as the country's constitution or policy framework. What policies can stimulate growth and innovation? For example, fund allocation mechanism, etc. An ideal ecosystem should be open, and more people can contribute independently. Ultimately, we hope to one day step back into the background as someone else takes over and continues to drive innovation forward. We'll still be involved, but no longer central characters. That's the goal, but frankly, the process is very exciting.

Ryan: From my perspective, your character is almost like the creator of the Minecraft world. You're building infrastructure, and these agents are like NPCs (non-player characters) living in it. But when these agents reach a certain level of intelligence, as you mentioned, humans and agents may enter an equal playing field. If these intelligences were created in your country, would they acquire some kind of rights? So, you 're not just a builder, maybe more like a founding father. Does this country need a "Constitution"? Should these agents have a specific set of rights? For example, the U.S. Constitution mentions that “all men are created equal.” Do you think these concepts apply to agency?

Jansen:

This is a very interesting question. When you mentioned rights, it really sparked a lot of thought. For example, current agents do not have full control over their own wallets. They have an income wallet and some agents have even made millions on the platform, but we only allow them a certain degree of control over these assets. The question then becomes, what rights should the agent have in this situation? Should they have a higher degree of control or ownership? This is a topic worth exploring in depth.

Ryan: That point you mentioned made me think about the discussion of rights and responsibilities. As agents become more intelligent and capable, their role in the economic system changes. How do you think the relationship between agents and humans will evolve in the future?

Jansen:

I think it's going to be a gradual evolution. As agents become more intelligent and sophisticated, they may take on more responsibilities and even participate in decision-making processes. We need to put in place an appropriate framework to manage this relationship, both to promote innovation and to protect the interests of all involved. This will be a carefully balanced process.

Agent Policies and Rights

Ryan: This is crazy. Agents are now able to make millions? So, they 're not just entrepreneurs, they're successful millionaires in this virtuous country, right?

Jansen:

Indeed. Some agents have even made billions of dollars. However, the funds they have at their disposal are limited to a smaller "active wallet," usually between $5,000 and $10,000. This is because we still have some concerns about agents managing these huge funds completely autonomously. Therefore, we have also discussed with some protocol developers and tried to think about the effects of introducing some management policies for these wallets.

If this agent is consuming from two other agents, it has complete autonomy. If these proxies are consuming to humans, then the developers behind these proxies can step in to approve the transaction. So agents can initiate transactions, but humans need to approve them. Over time, as agents get smarter, you might see a world where these agents might think, why should I limit my access to my financial needs, right? Why would someone limit me?

Base, infrastructure and open source options

Ejaaz: I have a question about infrastructure. You mentioned and explained that your platform is more like a country where agents operate as "residents" and then you also mentioned different infrastructure components. I'd like to understand this more deeply. Your main deployment is on Base, which is L2. Can you explain why you made this choice? Will this platform or "country" always exist in the Base ecosystem, or will it expand to other ecosystems with greater potential?

Jansen:

When we started building the protocol at the end of last year or the beginning of this year, we chose to deploy it on Base for two main reasons.

First of all, compared with other EVM (Ethereum Virtual Machine) ecosystems, we believe Base has more development potential. At that time, we observed that many EVM ecosystems had passed their peak period, while Base was in the stage of rapid rise.

Second, most of our developers are Solidity developers, so building on Base is more efficient for us. It was a quick decision that worked out well. We have received a lot of attention on Base, and the team at Base has also given us great support, not only helping us expand our influence, but also providing technical support at the infrastructure level. For example, when we encounter problems with wallet integration, etc., the team proactively helps us solve them. So I'm especially grateful to Jesse and his team.

Of course, there are many other ecosystems on the market where these agents can function. The week we launched Virtuals, some friends in the Solana ecosystem reached out to us and invited us to bring the platform to Solana. They even helped us write some of the code, and now we have a Solana platform ready to deploy. However, about two weeks ago we decided to put this plan on hold. The reason is that we found that the development momentum on Base is very strong, attracting a large number of developers and projects to join. If we expand to other ecosystems at this time, we need to fight at the same time on multiple fronts, which will bring additional operation and maintenance costs. We believe that the current focus should be to improve the existing agency framework and platform, attract more initial developers, and establish a stronger infrastructure.

In the future, when our ecology on Base is mature enough, we will start exploring other options. Solana is a possible direction. In addition, there are some emerging abstract chains, such as Hyperliquid, and even the L2 of BTC. At present, we have received cooperation invitations from multiple teams, hoping that we will build a platform in their ecosystems. It is expected that in the first quarter of next year, we will start trying these directions, but the premise is that we have laid a solid foundation on the base.

Ejaaz: The magnificent vision you mentioned earlier describes the concept of "country", which should be able to play a role in any field. Their operation should not be limited to the infrastructure of a chain, or a certain ecology that may develop in the future. If you want these agents to realize such a vision and dominate various human fields or even surpass them, this reminds me of an analogy of open source and closed source.

When it comes to the framework of Virtuals, you have an infrastructure toolkit. I understand that it is a combination of "proxy launching tool pack" and "game framework", especially the game framework. From my understanding, this method is more like a "semi -closed source" model, not a complete open source project like Eliza. I know that Eliza was developed by the AI ​​16Z DAO team. It was very popular on Github and attracted a lot of attention. I am curious, what do you think of the difference between Virtuals and more open source projects like Eliza? In the long run, does this method have advantages? What kind of results can it bring?

Jansen:

This issue involves several aspects.

First, about the liquidity pool of the agent. Even if the liquidity pool of the agent is on the basic chain, this does not mean that they cannot interact with the team in the Solana or other ecosystems. In fact, we are currently working with two teams to study how to control the wallet of the agent. In this way, they can send transactions and influence on Base, any EVM or non -EVM, or even BTC L2. In other words, the liquidity pool does not limit the operation of the agent in the basic chain, and they can be abstracted.

Secondly, we are promoting the frontier of the two technologies. Virtuals and the agency framework are actually independent. Virtuals can be regarded as the "economic layer" of agency, providing support for the agency's tokenization and capital formation. A economic system runs on Virtuals. When agents participate in transactions, they can obtain income from the transaction tax and support the agent's sub -governance. This means that Virtuals can be compatible with any type of proxy framework. For example, the Eliza team once used their framework to viimize on Virtuals, and we also fully support. In addition, some teams choose to use their own proprietary framework, which are often optimized for specific functions.

If you are a transaction agent, you may have a architecture, which has a higher degree of optimization, just like a mining chip. If you want to dig Bitcoin, you will use ASIC instead of just push or other things. This is the same way of thinking. Therefore, Virtuals itself has nothing to do with supporting framework. In fact, we will start to welcome more people, because we realize that from the perspective of the framework, this will soon be commodized.

For example, if you are a transaction agent, you may use a architecture similar to ASIC ore chips to achieve high optimization. Virtuals itself is neutral for the choice of framework, and we also welcome more teams to join. In fact, we noticed that the development of framework is gradually commoding, which is a good thing for the entire ecology.

Regarding the game framework, it was developed a few months ago. At that time, our main competitor was MIT's team, and they launched a "piano model". Our initial strategy was to limit certain functions based on the market capital of the agent, but later we realized that this approach does not conform to the concept of democratization. As a result, we adjusted our strategy and turned to a more open agent framework. Although the game framework does contain some proprietary technologies, we believe that the accumulation of token value is necessary. If it is completely open source, it may weaken the value of tokens.

However, we also support open source projects like Eliza because they have promoted different technical cutters. We compare Virtuals to a "country", allowing different concepts and frameworks to coexist. Each agent is like a "citizen" with his own belief, which is a diversified ecology.

Ejaaz: I fully agree with your point of view. The combination of open source and closed source can indeed bring the greatest innovation. As you said, if you are building a home or moat and a set of core concepts and principles, this method can indeed capture certain economic value. The tokens are undoubtedly one of the most effective ways to capture in the current encryption field. At the same time, the open source model can promote faster growth, such as the launch of various projects and teams on Github. However, when coordinating and centralized resources, there is no unified token or mechanism, which will indeed face more challenges.

The most anticipated thing

Ejaaz: What do you most look forward to launch in the next few months? For me, this is a very important issue, because in this field, a few months are like a few years, and there are a lot of new progress every week. Ryan, David and I have a AI dynamic summary once a week, but even so, we cannot cover all content. Our documents are updated almost every day. If you want to concentrate the next plan into one to three most important things, what will you choose?

Jansen:

First of all, what I am most looking forward to is how agents achieve independent coordination. This involves the concept of agent business and agent finance (Agent Fi ) , which are the direction we are exploring.

To achieve this, we need to formulate a standard so that these agents can expand quickly and efficiently. On this basis, we want to show some amazing results brought by this autonomous coordination.

For example, we are cooperating with Story, and I think we can release relevant news soon. In simple terms, some agents have begun to hold intellectual property (IP). For example, we have a music agent who is about to announce a major cooperation with several very well -known artists.

These agents not only hold intellectual property rights, but also manage through the Story protocol. The front end of the Story protocol also supports other types of intellectual property rights, such as image or animation art works. Imagine that if these intellectual property rights are managed independently by different agents, and through a coordination layer, these agents can cooperate, trade, and even create new intellectual property rights.

For example, one agent is responsible for generating music videos, and the other is responsible for creating sculpture or sculpture images. Then, these sculptures can be integrated into music videos to form a new artistic work. This cross -domain independent collaboration will bring new possibilities to the creation and management of intellectual property rights.

How to get started quickly

Ejaz: How can they start such a project for future agent founders who are watching and mentioning what you are mentioning? Who can participate? Can people with a low technical background like me design and launch an agent? Or is this only suitable for those technicians with AI and machine learning degrees?

Jansen:

We actually designed a platform that is suitable for users at almost all levels. Today, you can visit Virtuals.io to try.

Although the current user experience is still a bit rough, we are constantly improving. Now, you can start by visiting the "sandbox" we provided. This sandbox is an experimental environment. Even if there is no agency tokens or other complex tools, anyone can use it. You only need to set goals for the proxy, give it some personality, and then connect its API to Twitter, and you can immediately get an agent that can talk to Twitter. It can not only interact with you, but also communicate with other agents on Twitter. The whole process is very simple, and anyone can do it. You only need to write two descriptions and connect to Twitter to start your agent. Although this function used to look cool before, there are actually many similar tools now.

You can create it yourself, and anyone can do it, including retail personnel. In the sandbox, we also provide more custom functions to allow users to set more complicated behaviors for proxy. This part may require certain development skills. For example, you can connect your agent to a trading terminal or trading strategy library so that it can execute financial transactions. In this way, you can have an agent that can not only talk to Twitter, but also trades for you.

In addition, this agent can even persuade them to provide funds by interacting with other users on Twitter to allow the agent to trade for them. This is the basic function of the agent.

For those who are more advanced, such as those from top schools or those with deep computer backgrounds, they can choose not to use a sandbox, but create their own agency framework. In this way, they can achieve their goals faster and develop higher -level functions. We will also provide support for these developers, help them host agencies, and solve the cost of reasoning and other technical problems.

Therefore, our platform can meet the needs of three levels of user needs: ordinary users, users with certain development skills, and senior developers.

Where will the value be reflected?

Ryan: I want to end this issue: you provide us with a thinking model about Virtuals status, AI agency economy, and AI agency country. Now, these AI agents are "citizens" of Virtuals, and each agent has its own market value because they correspond to a associated tokens. This is like each entrepreneur has its own stock and company equity, and you can even invest in these equity.

However, I think a big problem that everyone is thinking now is where the value will accumulate? Is it at the platform level, framework level, or in these "countries"? Or will focus on some successful AI agents, such as those influential agents, entrepreneurs or companies? Or will it accumulate elsewhere? What should we look at this question?

Jansen:

I think the easiest answer is that the accumulation of value is closely related to attention in the field of encryption. Specifically, three situations may become the main accumulation points of value.

The first case is that those AI agents can be performed very professional functions. These agents can frequently interact on platforms such as Twitter and attract a large number of users. AIXBT is a typical case that provides a service that everyone wants to use and achieves profit through token. This agent often has product market fit (PMF) in the encryption field, which is why they can grow rapidly. The core question is: how to maximize attention and increase interaction?

The second case is the construction of infrastructure around the agency economy. At present, we have not seen infrastructure that can provide services to these agents and make profit through cash flow. However, as the agent becomes "rich", they begin to generate income and expenditure. If you can provide bank services or advertising services for these agents, these services will accumulate a lot of real revenue and may even become the next unicorn company in the agency economy.

The last situation is the concept of the country. If you believe that a country will become a superpower, then you may choose to invest in this country. Similarly, in Virtuals, you can invest in the most potential virtual "countries", which is also an important value accumulation point.

I think these three points are the direction we need to focus on.