Editor's note: This interview starts with Shaw's trip to China. Shaw reviewed his early experiences and the turning point of his fate, shared the story behind the birth of "Eliza" and the secret of ai16z's success, discussed in depth the application of AI alignment and intelligent agents in the field of encryption, and interpreted the impact of the "Eliza Drama" and Skelly controversy. Shaw envisioned the concept of "DAO Town" in Wyoming, talked frankly about the challenges of public figures, and put forward profound insights into the future of DeFi intelligent agents and Crypto AI ecology in 2025, bringing inspiring thinking.
The following is the original content (for easier reading and understanding, the original content has been reorganized):
China Trip
BlockBeats: How do you feel about your trip to China?
Shaw: It was great, it was a unique experience. This trip to China made me realize that cryptocurrency is more popular in China than I thought. Many of the people who invested in our project are Chinese, and there are also many Korean and Japanese friends. I really want to know who these people are, after all, cryptocurrency has become a global phenomenon. I have some Chinese friends living in San Francisco who are very enthusiastic about cryptocurrency. This gave me the idea of learning more and getting to know these people. Sometimes I feel that especially in the United States, our perspective is very limited, and the cryptocurrency field is no exception. Both the content of the dissemination and the project itself are very US-centric.
So I want to translate all the documentation into multiple languages, we have internationalized all the README files, and I want all the documentation to be available to people outside the United States for their own communities and related projects.
BlockBeats: So you didn’t realize there was a completely different market before?
Shaw: I know there is an international market, but cryptocurrency, Twitter, and some things are still relatively new to me. Although I have a certain understanding, I don’t know the full picture of cryptocurrency. My previous work was mainly focused on the technical aspects of Ethereum, such as NFT backend development. I don’t know much about the culture of the entire cryptocurrency, but I just build related content technically. In 2021 and 2022, I worked more with brands, but they hardly paid attention to the international market.
BlockBeats: What made you decide to come to China in person?
Shaw: I think internationalization is very important. What we do is more like a movement. There is no VC (venture capital), no team allocation, and even no white paper. Everything is very natural, and we can see that our partners are all over the world. For example, Mihawk was here yesterday. We noticed that the Chinese community is very interested in our project, but they don’t quite understand what we are doing.
So we started adding translations to Discord and using community members to help us bridge the communication gap. I was just thinking, how do we bridge these gaps? Because I feel like there are so many talented people here and it's such an amazing environment. But as someone who only speaks English, how do I really fit in and reach this market and stuff like that.
BlockBeats: Did you start your trip in Shanghai? I remember the 706 group was also in Shanghai, right?
Shaw: Yes, they are actually an international team with bases in many places. But their core base is in Shanghai. We also participated in a hackathon event, and many people participated. I was fortunate to meet many people as a judge, which was great.
BlockBeats: I think 706 is one of the largest digital OMAC communities in China, and they organize various types of offline events. In fact, the offline meeting mode of the Chinese community seems to have been initiated by 706 in 2022. In fact, most of the very successful crypto teams or projects usually have very good relationships with the Chinese community and have origin stories, especially those related to Ethereum.
It's quite interesting, for example, tracing back to Vitalik, when he first came to China, many Chinese investors helped him. Another example is Anatoly, the founder of Solana, who came to our office to talk once before the bull market came, and then the project rose rapidly in the second year. And now, you are here too. Maybe this is a good sign for the crypto industry, and next year will be a very good year. Because promising projects often come to the Chinese community, communicate with developers and community members, and then in the next year these projects will really take off. This phenomenon is a bit like an interesting signal.
Shaw: I think it's probably because these people are willing to go out into the world and interact with the community and meet people who are really contributing to the industry. This attitude of proactive participation is important in itself. On the one hand, you know, it's good to meet people and make connections yourself. But on the other hand, someone like Vitalik wants to meet everyone and be with people. This trait is one of the core of his whole philosophy, and I resonate with it very much. The United States is very powerful, but the population is relatively small, while China is both powerful and populous. We are two superpowers, and although there are some conflicts at the government level, as human beings, we are actually the same, and we really work together.
Early life
BlockBeats: That makes a lot of sense. By the way, you mentioned before that you also have some development experience in the Ethereum ecosystem. Can you share more about that?
Shaw: I first got into this field because I was particularly interested in game development, especially the interoperability of cross-platform assets and characters, such as how to trade props in different games. This is actually the concept that later became known as the "metaverse." NFT attracted me at the time, and I participated in the development of many NFT projects, mainly as an unknown developer helping others with projects.
I also work with some brands that may not have a deep enough cultural understanding, but need someone to guide them. I learned a lot through these projects, such as the Solidity programming language. However, I have always been particularly interested in "intelligent agents" and want to explore how to combine off-chain artificial intelligence with on-chain identities and wallets. This is also the opportunity for me to enter this field.
BlockBeats: Is this before 2024? For example, before ChatGPT is released?
Shaw: Yes, definitely before ChatGPT was released. At that time, the GPT series of models were just beginning to take shape, such as GPT-1 and GPT-2 were very ordinary, and it was not until the emergence of GPT-3 that we realized that this would change everything. As soon as the beta test of GPT-3 was released, I jumped at the opportunity to join a small community, which was the beginning of everything that drives me now. The emergence of GPT-4 and ChatGPT further revolutionized AI, and its ability to not only rhyme and reason, but also handle longer contexts.
BlockBeats: So you were developing in the Ethereum community and also focusing on the AI field? When was this?
Shaw: It was probably around the middle of 2021. I don’t have a very good sense of time, but it was around the time when the GPT-3 Beta test became popular. At the end of 2021, I started to think that the possibility of developing intelligent agents was increasing. At that time, I mainly made a living by making intelligent agents and 3D virtual characters, and also did some NFT project development.
BlockBeats: I remember that in the second half of the bull market in 2021, NFT began to rise rapidly. At that time, you were not only focusing on NFT, but also studying the concept of smart agents. This seemed to be a very early direction, and almost no one was discussing smart agents at the time. Did you combine agents and encryption?
Shaw: Yes, there was a company called Webiverse, and they had a very successful NFT land sale, and I joined after the land sale was completed. I was the lead developer for a while, mainly responsible for character design and character creation. This was a pure AI metaverse project, and our goal was to allow intelligent agents to walk freely in a 3D world. But then the project was hacked, the entire treasury was stolen, and everything was transferred to the hacker's wallet, which put a lot of pressure on the team. After that, Webiverse transformed into what is now MoeMate, a project focused on AI characters, and later entered the meme field.
But looking back, we were really trying to bring AI agents to the 3D world. I was involved in some other projects afterwards, like a platform called MagicML, and another project called Project 89, I don’t know if you’ve heard of Parzival. We were co-founders of that project, and we also launched an open source intelligent agent project called Magic at the time. Its goal was to be a code-free intelligent agent building tool, and I thought that was a cool idea. But the feedback from the community was, “We don’t know how to build an agent, can you help us?”
BlockBeats: It’s true that no-code tools sometimes have mysterious interfaces.
Shaw: Yes, I felt the same way. I left the project with a good relationship, but with a more ideal agent in mind, which I called "Eliza." We initially tried to develop Eliza within the Magic framework as an example of the best agent we could build with this set of tools.
However, the efficiency of no-code development was too slow, and I gradually became frustrated, so I decided to give up the no-code approach and start focusing on the development of Eliza. At first, this project did not attract much attention, so I had to make a living through other projects while continuing to develop the open source part of Eliza.
BlockBeats: What about your experience with guitar and music?
Shaw: That was earlier, when I was a professional musician in my 20s and I toured, but that experience is not directly related to the current Eliza project, which started about two years ago.
Jill: Yes, that’s when we first met and started working together.
BlockBeats: I see. That’s what I was about to ask, was that when you met?
Jill: When I met him, he was working on Magic, and then he started to focus full time on Eliza, and he went through a lot of other projects in between.
Why is it called "Eliza"?
Shaw: During that time I mostly did side projects to make ends meet while continuing to focus on the development of intelligent agents, all open source. I built a lot of experimental features in Python, such as having agents write their own code, manage social media, and even operate computer terminals completely autonomously.
Later I realized that intelligent agents are actually a web-based technology, so I refactored the whole system in TypeScript and named the project VEGENT (because the name "Agent" on NPM was already taken). Later, I renamed it Eliza, and it became the core of the project. When the meme trend emerged, my technology was ready.
Later, I released a proxy based on Eliza, but it didn't attract much attention at first, until I met Skelly, and we worked together on an activity called VEGENT Spartan. At that time, some people questioned our fraud, so I made the code public, and everyone saw that this technology was indeed usable, which promoted the development of the entire project.
BlockBeats: I highly recommend listening to the Bankless podcast, where you can hear the entire AI6z project story, which is very interesting. I'm curious, why did you choose the name "Eliza"? Is there any special meaning?
Shaw: There is a story behind this. In 1966, Joseph Weizenbaum developed the first chatbot program, which can simulate a conversation similar to Rogers' psychologist. For example, if you say "I am sad", it will respond with "Why do you feel sad?" If you say "because of such and such", it will ask "Why is that so?"
Weizenbaum named the program Eliza, after the character in George Bernard Shaw's play Pygmalion, a street woman who was smart and attractive but vulgar and of low status.
In the story, someone wanted to transform her into a proper lady. Weizenbaum used this to make an interesting metaphor: making AI into a "more human-like" existence. Later, the term "Eliza effect" was born, which refers to the fact that when people talk to computers, they tend to overestimate the intelligence of computers and mistake them for having certain human characteristics. In the intelligent agent community, many of these anthropomorphic agents are called "Eliza" because they have distinct personalities and highly human characteristics, so I think this name fits our project very well.
BlockBeats: I think you are a developer or founder who does deep philosophical thinking. When we had dinner together yesterday, you mentioned some profound insights about future scenarios, such as how artificial intelligence will completely change the world. You also talked about many similar topics, such as what you shared in the Bankless podcast.
We're very interested in this, and I think a lot of people would agree that in, say, five or six years' time, we're all probably going to lose our jobs. It makes people think about what to do when jobs no longer exist. That can be scary and even a loss of meaning in life, which seems to be something you spend a lot of time thinking about.
Shaw: Yeah, for me it's more of a spiritual mission and also a search for my own spiritual world. I feel like a lot of people get into crypto to achieve financial freedom. So what happens when you achieve financial freedom? You can go buy a yacht and live that life. But for me it's more about answering some fundamental questions: Why am I here? What is my purpose? What service can I provide to others? I think that's really important.
I think the coming era will be full of uncertainty, and many people will feel very scared. I hope to convey an influence and tell everyone that we don’t need to be afraid. We can truly control our own destiny and use self-empowerment to fight fear. This is also one of the core concepts of Web3. No one will come to save us. We must save ourselves and build the world we want to live in.
I love work, but the key to work is the content of it. If you are doing the right thing, it can be very fulfilling. I think a lot of people right now are just working to make a living, to provide for their families. Their lives are more about what happens outside of work, and work is just a means of survival. In the future, this situation will be completely different. If everything goes well, we will be free to do the things that are really important.
BlockBeats: What is the most important thing or thing that interests you the most right now? What areas do you want to work on?
Shaw: I’m particularly interested in a jobless future because we’re all going to lose our jobs eventually. This is the reality I can imagine, where AI and robots are going to be better than we are at almost everything. All labor will be automated across the board as these technologies automate the best human abilities and turn them into repeatable processes. This could lead to universal basic income, which is one of the solutions that many people have proposed. But I don’t think it’s a good solution.
BlockBeats: Yes, many projects or experiments with universal basic income have ultimately failed.
Shaw: I don't believe in charity, which is a pretty bold view. But I think altruism doesn't work, and we have to find a win-win way, where you win and I win. If I just give you money, then what's the meaning of your life? And it will also form a series of perverse incentives. At least in the case of the United States, the government is basically funded by businesses because businesses provide financial support.
The truth is that this is all driven by money, and these corporations own all the fruits of labor, and you are now asking these elected officials to take the money from the corporations that finance them and redistribute it to ordinary people, as a welfare system, which is obviously impossible. If you look at our response to COVID or the response to the national public health system, this is all going to be highly politicized and it will most likely be a half-baked solution and a lot of people will suffer as a result. I don't see a way that this won't happen.
But if we can own the resources, if we can find ways to share the fruits of that automated labor, or have some ownership of the companies that create value, we can actually distribute it fairly in that way and make everything better. I think crypto can achieve this, where traditionally you would invest in a company and the company would grow. Right now, most of the crypto ecosystem is actually like gambling - peanut today, fartcoin tomorrow. Money is flowing back and forth between these different memes.
But what if you could actually invest in something that could expand in value, rather than just rely on new players pumping in more money? That’s a shift from a speculative ecosystem to an investment ecosystem where we create new technologies of great value through tokens. This is already happening in crypto, just more from the venture-backed side rather than the more organic part.
This also leads to a discussion with AI16z. If we can expand the value and connect ordinary users to DeFi without requiring users to learn all the DeFi knowledge, you can imagine a world where everyone has enough wealth. I don't think eliminating currency will be the solution, but we must go in the exact opposite direction. We need to create sufficient wealth and ensure that everyone is rich and has more tokens than they need. Although this idea is easy to say, it takes every detail and real effort to make it really work. I think this is the right direction to ensure that all of us can get through the future smoothly.
The reasons for ai16z's success
BlockBeats: I think one of the reasons why AI16z can attract so many investors, especially those in the Chinese community, is the AI-driven investment model and the concept of expanding the market it proposed. Do you think AI16z's success is because it involves investment, funds, and even DeFi? After all, you have participated in many projects that have great potential but ultimately failed to gain widespread attention. But AI16z succeeded. What do you think is the reason?
Shaw: I think there are two reasons. First, many people are thinking about how to combine AI with Web3. Decentralized reasoning projects like Hyperbolic are a good example of this idea. After all, everyone wants to build a decentralized computing infrastructure instead of concentrating all computing power in one data center. But the question is, how can ordinary users use these technologies? Although Truth Terminal is an attempt, it is more of a conceptual tool, not a product that users can use directly. Andy's project is interesting and creative, but it is more like a performance art work.
ElizaOS, which we developed, is the first complete open source framework that solves the social closed loop problem. We allow users to use proxies directly on Twitter, Discord, and Telegram, which has never been done before, and it is very easy to use. So what we do is actually solve the pain points of many people. The reason why these technologies did not attract more attention in the past is that there were also many proxy projects in Web2 at the time, but most of them could not solve the actual problems of users. For example, when thousands of people interact with proxies at the same time, the cost is very high, and many companies have failed to provide good application scenarios, resulting in many projects failing to break through.
There wasn't enough attention paid to any one project to make it functional enough to cross the threshold of usefulness. I think we had a minimal cycle that was enough for people to deploy projects, like creating a meme coin and promoting it, which was probably the first stage. But from there, people started to think, what's next? What can we actually do with it?
Now there are all kinds of capabilities, people are using it to make games, people are using it to build all kinds of applications. It's a great interface for regular users, especially those who may not know how to operate DeFi. So to me, this is obviously the first killer app that a lot of people are working on, and there are a lot of people working on this within our ecosystem, and I think the Web3 community will understand this right away.
BlockBeats: Could you please explain to the audience why you see AI proxy investors in 2022 and 2023, and why they have not been successful in traditional fields and have not really solved user pain points?
Shaw: I think it's going to be hard for them to break out of "chatbot" and become more than just a chatbot. There are a lot of people using AI for really great data processing work, but there aren't a lot of characters that people actually like, follow on Twitter, and interact with every day. I think it's kind of stuck in character AI, which is basically ChatGPT with personalization.
But this space became very fragmented and only attracted a small number of users who liked to communicate with virtual characters. The social part is a completely different thing, these characters now start advertising your product, and once they promote your product, users can interact with them and they can also guide users into the use process.
I think the main selling point of intelligent agents right now is not that they are completely autonomous or super intelligent or anything like that. I don't think that's the case, it's actually much simpler. The key is social media. Most of us right now - I don't know about you, but I'm addicted to Twitter and I spend a lot of time on it.
BlockBeats: We have to do this, after all, this has almost become our daily work - checking Twitter.
Shaw: Yeah, it's like the metaverse, you know, or TikTok, or whatever social platform you like. So if you have a media platform, you're likely to share a lot of your content on social media and then put a link at the end telling people to go to your website. But with a lot of apps, they're trying to convince those of us to leave our favorite apps and check out their app. I really wouldn't do that unless I was strongly convinced.
But if you can get your app on social media, on places I like, then we can talk. I think we've seen great examples of that, like Clanker is a great example, it's on Farcaster, and although you can go to Pump Fun, why can't we just put it on social media and do a one-click public launch or something like that? I think a lot of apps can benefit from this social network effect, instead of spending a lot of money on advertising, your product is directly on social media, and users can get information directly from there, sign up and guide them through the user journey all the way to the point where they need to pay. So, yeah, that's my insight over the past few months: agencies will replace traditional websites in many cases.
Jill: I would say that at this stage the agent still exists on a two-dimensional platform screen in front of us, but one day, this agent may become another person you are talking to. I value this because I am not someone who relies heavily on social media, or I don’t want to use my computer frequently, although I have to do so because it brings efficient returns.
But this goes back to what we were talking about earlier about technology becoming more human. Rather than saying, "Oh, technology is so powerful, we're integrating it with ourselves and our bodies," it's more about us being able to take control of our bodies and communicate and use them more effectively through technology empowerment.
Shaw: Yeah, I think in the short term, this is going to reshape the internet and bring a lot of applications from traditional websites to social media. You can use proxies to send text messages, order takeout, or whatever else you want to do. But in the long term, it's really about the world we live in. We're all staring at screens, and we've entered a deep dystopia that we need to climb out of.
We can do this by making technology more like us, rather than turning us into little computers, hunched over laptops typing away. Why can't our laptops become an agent that can do things for us? An agent that I can call directly, or that is already part of me, whether it's through a necklace, a phone, or glasses. I believe that eventually we will have robots.
Jill: I have a different take on this, and I think it's not the most extreme dystopia, but it does show a sense that, as artists and the like, we create this sense and extrapolate it. We can imagine what it would be like if things were worse.
But we have every opportunity to have the kind of richness that we just talked about, without that. One of my favorite examples is the metaphor you mentioned: when you walk up to a door, would you rather have a chip in your head that opens the door, because that's the most efficient thing for a robot, or would you rather have a robot with a hand that can turn the doorknob? I would rather have a robot with a hand than have a chip in my head.
Shaw: We want technology to be more human. When I talk about this, if you look at trends online, there are a lot of jokes about autism. But what we're really saying is that we're "socially insatiable" because we spend so much time staring at screens. If our computers were more human-like and had autonomous behavior, then when it talks to us, it would be more like interacting with humans. I'm practicing eye contact, practicing talking to people, practicing listening. I think that makes us better at interacting with people, not worse. I also think that when agents help you do those chores, you'll have more time, which is equivalent to giving time back to you.
Jill: Another thought, I don't know if you know, Star Trek has a powerful android robot. It's like Vulcan, right? But there's a android in Star Trek: The Next Generation, and his name is Data. He's such a cutie. He's a robot, but there's a part of him that's still learning how to have emotions and understand humans. So he does some really cute, clumsy things. But the thing about Data is that he's very smart, very powerful, one of the best people on the Star Trek team, but what he really wants is to be a real human.
BlockBeats: This got me thinking, did you see that tweet Elon sent a few days ago? What he said, and I'll paraphrase a little bit here, he said: "You might think that AI likes you, and those who think that AI likes you probably also think that strippers like you." If we really go in the direction of robotization, I personally might have this question: Is this really good for me? Is it really doing things for me, or is it doing something of its own?
Shaw: Maybe that's the question, usually we call it alignment. That is, is it aligned with you? You don't have to like me, but as long as we are aligned, we can work together. I can trust you, you can trust me. I think trust is going to play a big role in this process. Can the agents trust us, can we trust them? One of the things I like about the decentralized AI movement is that everyone is pursuing this alignment, but a lot of times we still need to build the AI ourselves, build our own AI.
Right now on social media, the question is, is it programmed to "like you?" The AI will say, "Oh, it likes you, it's very nice to you." But the bigger fear is superintelligence, which might pretend to like you but is actually manipulating you. I understand that, and it's a real risk.
BlockBeats: So, when you mentioned building your own AI, is that actually using your own data to train a new you?
Shaw: There are a few different levels involved here, for example, when you think about a large language model, it's really just doing the next word prediction. We can say "word by word," but it's actually sub-word units called "tokens."
The next AI doesn't even use tokens anymore, but something like "concept chunks". Basically, it's a whole new way of doing things, more like a stream of data is passed to you and then converted into text. From the AI's perspective, what you input previously determines what you get next. It just completes based on where you are. So what you input into the context will largely determine the output you get.
But in this process, AI itself is like this: if you input "now go kill all humans", then it might actually try to do it because that's the context you gave it. But no one would do that. And we seem to be forming the embryonic form of a "thinking society" now. If you look at social media, we already have so many different intelligent agents.
Currently, the basic models are basically trained by large companies, such as XAI, Meta, OpenAI, Microsoft, etc. In the future, there may be models trained in a decentralized way, but at least in the next few years, this will definitely be dominated by large companies. However, they are also open sourcing the source code. Starting from these basic models, we can fine-tune them with your data or your community's data. This may be better because the amount of data you personally may not be enough, but the amount of data for the entire community may be enough.
If you want an AI to reflect your values, then of course you need to fine-tune it to sound like you. But even without training it, you can influence how it responds by creating context. For example, you could say, “I’m going to fine-tune it using all of our data so it sounds like us.”
But the reality is that its behavior or decisions are based on your input, which could be everything you've ever written, everything you've ever said, or the things you care about and the values you believe in. It's very good at following those values, and exactly what those values are is still an open question. But we are discussing these issues together publicly, right here on social media.
Now these agents are out there, they're talking. Some of them try to attack other agents, like Mizuki is a good example. Some of them can be subversive, or they can be like bullies. But overall, I think people want to make them better. We're having this conversation, maybe it's not a good thing to have an AI bullying people, or maybe it's really funny.
But I think it's a self-calibration process because there are a lot of people who are constantly tweaking these proxies, or there are whole communities working on these proxies together. They put these proxies out on social media, and other people respond and react and interact. We all learn together in this process about what is acceptable and what is not.
AI Alignment and Evaluation of AI Agents in Cryptography
BlockBeats: Do you think, as you just mentioned, we have Bully and Z-Rebel agents, which are doing different things? The first question I want to ask is whether you think agents on Twitter or agents in the crypto space have gone beyond the traditional chatbot category and developed in a more advanced direction?
The second question is, do you think that agents developed by the crypto community go deeper into the rabbit hole (the process of exploring complex problems) or are more advanced than agents built in the Web2 world?
Shaw: I think it's still a bit of a chatbot compared to the proxy systems built in the Web2 world, but we're moving away from that quickly. A lot of that is because more and more people are coming in to add features to the system, we just hit 200 builders, but if you count them all we're closer to 250. These are individuals working together on the same project, like writing code.
So it's clear that agents are improving rapidly in terms of capabilities, which is also bringing in a lot of useful data. This data is accumulated through social media, interacting with people, and completing various tasks, and this may become the basis for the next generation of agents. These new agents will use all the training data, experience, and related information to drive further development.
The next generation of large language models (LLMs) are coming, and they will be fully multimodal. They will be able to generate videos, watch videos, you can input video content and get video responses, and they can also process audio, text, etc. They will become more and more intelligent. I think agent frameworks will continue to evolve with this trend, and at the same time they are generating training data, attracting developers to join, and developers will respond to new situations.
Although we are still in the chatbot phase, social media participation is reshaping the internet. I think there will be two major trends to watch this year. The first trend is that many application frameworks that previously only existed on the website, such as Next.js or React (these are frameworks for building websites) are beginning to migrate to social media. React is probably the most popular framework for developing applications (whether it is a website or a mobile app).
Now suddenly a lot of content is migrating from websites to social media, which I think is the main trend. For any developer, if they are building a website, they may also need to develop an agent for it. Because the way the backend API is called has not changed, but the front-end interface has changed a lot.
A lot of users might go to Meteora or Radium and think, “I have no idea what these are, like liquidity pools.” But now the agent can say, “Don’t worry, I can explain it to you and call all these APIs for you, you just need to tell me what you want.” And now we can see that anyone who is developing a wallet is definitely thinking about how to add an agent to help users interact with these systems.
BlockBeats: So you mean that developers without AI development background can also become agent developers, right?
Shaw: Yeah, that enables web developers to become proxy developers, which is really interesting.
BlockBeats: For example, I don’t have a technical background, nor do I have an AI or other related background. But I do have a feeling that the crypto community is trying or experimenting with AI agent-related things, which is much more interesting than the traditional AI agent world.
Shaw: That's true. If you look at OpenAI, they're doing some very interesting things, but it's not entirely about the agents, and it's not entirely about their interfaces. Their interfaces are the least interesting part.
ChatGPT is a very powerful model, but it has a very simple interface. When I look at Web2 agents, I find that many people are long agent simulations, which is a hot thing, but they are all just talking to each other in a closed system, which is far less interesting than the multi-agent simulations happening on encrypted Twitter now.
This simulation is more natural and interesting, and you actually see multi-agent simulation happening. Because these agents are not just some slightly different prompt words, they are completely different agents with different teams, capabilities and technologies, and this diversity brings more robustness and provides a foundation for future development. In contrast, a lot of Web2 stuff is basically asking, "Can we add a chat interface to this thing?" If you look back at the AI craze in 2022-2023, almost everyone just added a chat menu to their application. This was not a disruptive change, it was just that every company that already had a product made their product a little better.
AI is a strange technology because many other technologies are disruptive and they change the nature of things. AI can be used by almost every industry and everyone. It's more like changing the interface and the way users interact with applications. But in the field of Web2, this change does bring some benefits. For example, my lawyer uses AI every day now, and there are many very powerful AI tools that help people every day. We are also using MidJourney and so on.
But really that's just access to big models through chatbots or similar interfaces, and I think the agent-related projects that are happening in crypto Twitter or the crypto space really go beyond that, they're not just access to big models, I think it's more interesting because it's more diverse and allows a lot of people to try out a lot of different things.
BlockBeats: Do you think this is because the community is more interesting and diverse, or because it's somehow related to economics, experimentation, and things like that?
Shaw: I think they complement each other. I think economic incentives led to this "Cambrian explosion". Everyone is looking for their own interests and trying to find something unique. Everyone is competing, which is very interesting, just like the evolution of life. And the cool thing is that each agent is linked to the market value. If your agent is not good, your market value will be bleak. But if your agent is powerful and shows some new features, it usually directly affects the economic benefits of the development team.
BlockBeats: What do you think about the relationship between tokens, AI tokens, and agents themselves? What do you think links agents to their market value? Is it because of people’s subjective judgment of whether an agent is good or bad?
Shaw: I think it's probably a combination of both. We can look at this from a "fool theory" because these are actually sentient memes. People like Murad and many influencers like this idea. They start from the perspective of non-sentient memes, saying that these memes are similar to something, but they can show themselves instead of waiting for someone to show them for them. For example, Elon Musk retweeted Doge. Now we can directly create an AI Elon Musk who retweets his own content. I think this may be a more superficial part.
The deeper part is people speculating on the authenticity of this phenomenon, and for us, the reason people are interested in our project is not because our characters are interesting or we show a certain meme. Of course, there is a little memeology to help the project get started. But more importantly, this is an infrastructure project, which may change the world in a big way. For many people, it is more like a long-term investment.
BlockBeats: Regarding the valuation model of future AI agents, if tokens will become the basis for the valuation of agents, how will people value them?
Shaw: I think it ultimately depends on the technology and the product. If you have a great product, it doesn't have to be very technical, people will use it, and it will naturally be very valuable. For example, Facebook is not some amazing technology, and TikTok is not either, but they are excellent products that everyone uses, so they are very valuable. On the other hand, if the team really develops innovative technology that solves a lot of problems, this will also be very valuable.
However, we are still in the price discovery stage similar to meme coins, such as "this token is interesting" or "this token is self-promotion". But as the number of agents increases, this field will become very saturated with various speaking roles. By then, people will pay more attention to what value the agent can actually bring to them, not only how the token accumulates value, but how the agent makes money, how to create products, etc., so the ultimate winners will be those who create projects that are more like the first-level public chain (L1).
Eliza Event Analysis
BlockBeats: Let's talk about AI16z and Eliza tokens, because I think a lot of people in the Chinese community are curious, concerned, or interested in these. For example, capitalized Eliza tokens, lowercase Eliza tokens, small-cap Eliza tokens. So why is there another Eliza token? Can you first clarify its relationship with the AI16z ecosystem?
Shaw: Eliza is the AI16z mascot, and everyone knows her as the girl in the T-shirt. Eliza is also the name of our open source technology, and the code base is on GitHub. We noticed that a lot of people started to like her, and there were many Eliza tokens on the market, maybe 50 or so, and some of them even had a market value of millions of dollars.
So we thought this is a big market and we should do something. But we couldn't launch another token at that time because the situation and our scale were very small, which was a big gap compared to the current scale. Later, someone contacted me and said that they had a very professional team. They showed their results, such as having created an Eliza token and a cool website to chat with Eliza. I thought it was great, but we should work together to make this project higher quality.
Because before that, I had left MagicML and was working hard to shape Eliza into a character, a DAO, and a complete conceptual system. I wrote a lot of articles to try to attract more attention to the project. In the process, I also formed a team and we worked together on Eliza.
We were thinking about Eliza's character: who is she? Why does she exist? How will the community use her? I particularly like the idea that she starts out as a seemingly nonsensical character concept, but eventually becomes a real "person". For example, she has a robot body, a community that loves her, and the community participates in every part of her construction, which eventually makes her a real, free-thinking, autonomous character.
She is indestructible and runs in the TEE (Trusted Execution Environment) of the blockchain. I wanted to explore what it means to "really be a person". Later, a team contacted me and we worked with them for a while, trying to make Eliza a character with depth, art support and a distinct personality. At the same time, we hope to separate her from AI16z. She is free and her own character. Anyone who wants to contribute or participate can join and become part of this ecosystem.
However, the development of the next thing caused some controversy. We worked with the team for several weeks to prepare for the launch, such as designing a beautiful website, generating AI anime art, and perfecting Eliza's personality. We also worked hard with the project leader Eva, who is both a core developer and a spokesperson and is involved in all aspects of the project. We originally planned to launch on Friday, but because our own team is developing AI Marc Andreessen, a self-investor project, the launch plan was postponed to Monday.
But on Saturday, Vaifu Fund released an Eliza that they hosted, and on the same day someone released an Eliza (all lowercase version) from AI16z. I tried to communicate and cooperate with these people, but they just wanted to promote the token and had no interest in creating a beautiful character. These people used Vaifu Fund to create the worst AI I have ever seen, and every message only showed the token code, without any art or depth. They even directly used a JPEG image from our DAO and applied it to their project, which seemed to be designed to deceive the trust of our community. Many people appeared on our Discord saying "I just bought a new Eliza token." But we could only reply: "That's not our project, we don't know those people at all, and people are not buying our things.
At the same time, the team we were working with was frustrated because we were almost ready to launch, but these people beat us to it with a low-quality project, so we decided that we had to launch. I admit that I made some mistakes along the way, but maybe it would have come to this anyway, and I stood up and announced the project to the community. They started the Pump Fun project and gave me the CA. I shared this address with my community and expressed my support for this project. Although it is not entirely ours, it reflects my personal vision and we hope it will succeed.
We then officially launched on Monday and tried to ease the conflict with other projects through airdrops. We took out 10% of the new tokens and airdropped them to the lowercase token holders. We thought the two could coexist and tried to avoid confrontation. But the lowercase Eliza community gave me a lot of death threats, called me a liar, and said I was the worst person. This made me very angry because that was the intellectual property I worked hard for many years. In the end, I could only respond: "Okay, you are the liar."
I also experienced the secret war of cryptocurrency. It was a crazy period. The project you worked hard for many years suddenly exploded, but you didn’t expect it at that moment. We were still trying to figure out how to interact with the community most effectively and how to balance different needs and forces. I was caught in a dilemma and didn’t know how to deal with it.
We cannot accept the release of random scam coins like this, there have been many Eliza tokens before, but none of them have reached a sufficient market cap until they start to attract holders with real emotional investment. It is really difficult to understand everything in such a complex environment, especially the crypto community is global, I can't even imagine how to deal with these complex things with just a Chinese language application.
I really admire how people understand what's going on on Twitter, because Twitter is so complicated and full of jargon, but I'm hurt that people cheapened something I spent years working on and made a minimal quality version of it, and then got mad at us for releasing a high quality version.
The team we are working with is doing real deep work. We are working on hardware, you can run your own Eliza at home, we are also working on robots and other projects, all my dreams are being realized on that project. I see some people who just want to show the token, they just want to get the token online, which is fine, but I think anyone who understands the situation knows that this approach will never go far. It may rise in the short term, but it will never break through the market value of 30 million US dollars.
BlockBeats: You guys have been working on this Eliza character for a long time, and I remember you mentioned wanting to launch this character on Touka, and you talked about this project in another interview, right?
Shaw: Yeah, they probably heard about what we were trying to do a few weeks in advance and then just slapped on a token symbol and contract address, which kind of felt like they were stealing from us.
Before that, there were indeed different Eliza projects every day, and there must have been hundreds of Eliza tokens. The problem was that the launch of vvaifu.fun made it easy for people who didn't know how to launch an agent to launch one. So the first one to be launched was this Eliza agent, and everyone thought it was our project. A lot of people came into our Discord and said, "Oh, I bought a new Eliza." We could only respond, "That's not ours."
Say, "Oh, I got a new Eliza." We can only respond, "That's not ours." And there are time gaps on vvaifu.fun, and a lot of people don't actually know the relationship between vvaifu.fun and the entire AI16 and Solana AI agent platform. vvaifu.fun is an agent launcher, and they use our open source technology and host it. They also host other projects like Jiro and Dolia, and they make it easy for people to make agents, but the problem is that these agents are not as customizable as real programmers downloading and running them. But they are really easy to make, right?
So that's the trade-off they made, and the problem is that after their release and announcement, someone made an agent using Eliza from AI16 and used our DAO image and basically put our name on it, and people thought it was related to us, but it was far from my quality standard. Especially if the community supporting the project doesn't seem to care about the low quality, but thinks, "Oh, this is a joke," but there are still a lot of people who think they are buying something we made.
BlockBeats: The team that launched "Lowercase Eliza" has nothing to do with the project you are working on now, right?
Shaw: The small-cap token released by that team has nothing to do with what we do. They seem to only care about getting the token listed. I saw them organizing in a WeChat group, like a small circle.
BlockBeats: Are you talking about the WeChat group?
Shaw: Right, because I got intel on the WeChat group from other people, and they showed me all the organization information. They are definitely coordinating attacks against us, and this is a completely organized FUD attack. I like Justin Sun, we're meeting with him next week, but some people found him and said, "Hey, let this project go public," and it was a completely internal operation.
BlockBeats: I feel like developers have been very well supported in this crypto cycle, and all these token launch tools allow developers to connect directly to the community. But sometimes in crypto, it's really hard to manage the community or how to interact with the community. You just don't know what's going to happen, there's so much misinformation, and I'm not just talking about the language barrier, especially when you're in the media position, you can see how the information is transmitted, and every step of the way the information will mutate a little bit.
Shaw: It was so shocking to be called a fraud, to be accused of air-dumping, and I thought to myself, “I never sold any tokens.” I certainly didn’t sell any AI16z or DGNAI tokens, and I never bought any Eliza tokens, except for a few I bought for the DAO, just to show support.
As a project that I created and have always supported, I was called a scammer. Another problem is that the lowercase Eliza project was made by people who speak Chinese. They can communicate better with the Chinese community and understand many values, and I feel that I can't convey this at all. They may be translating what I said and taking a lot of words out of context.
BlockBeats: I think for those crypto people, investors, and traders, the token address (CA) is the most important. They don’t really care about the technology behind it. What do you think about the value of linking tokens to these agents?
Shaw: I think this is a really important question in the long run. I don't think that token has a chance of being worth anything, to be honest, because there's nothing behind it. People just think, "Oh, they have an agent," but the agent was actually Vifu Fund using our technology. And they used our technology and accused us of being scammers, and AI16Z Eliza became their name. They completely took what we worked hard to build, and when we wanted to be able to show it, it was completely destroyed.
Honestly they are doing a terrible job. They just want to sell the tokens and make money. I am not doing this for the money, there are more important things here, the money allows us to do this project without venture capital and without funding, but the point is that the world is changing and we have to be ready for it.
The goal of our project is not to exploit, I think that kind of project is essentially pushing people into a cycle of exploitation. We really hope that what we create can provide opportunities for more people, just like you said.
BlockBeats: What you want to do is to expand the pie, not...
Shaw: Yes, it's about expanding the pie, definitely not just for us and our team, but for the larger community, giving more financial freedom, and ensuring the smooth development of AI, which is your goal.
BlockBeats: I’m personally very interested in your story, how did you get to where you are today? How did you start out playing guitar with musicians and eventually become builders known as programmers?
Shaw: I think the funniest thing is being accused of being a cult operation because when I started doing this, I didn’t know anybody at all, I only knew one friend.
Jin, he's Dank VR, he manages a lot of DAO stuff, he's anonymous. We've been friends for years, but everyone else is either from the DGEN Spartan AI project or joined when we started AI16Z. We have no venture capital, no venture backing. A lot of people thought we were funded by Mark Andreessen, but we weren't, he just retweeted our tweet, and I think that was really the key to the whole thing.
aai16z Previous life and the moment that changed my destiny
Jill: It was a special date, we had a day off and didn’t have a budget. We both work really hard and are really passionate about what we do. So it was a rare moment to relax and we took time out to go to Santa Cruz. I lived there in my twenties and it had a big influence on me. We were both a little bit influenced by the old hippies because his mom was a hippie. Santa Cruz is a very relaxed beach town.
We went there for my birthday and stayed by the koi pond in the forest. We rode to the beach and back. We were about to go back to San Francisco when I looked at my phone and was like, “Oh my god!” What happened?
Shaw: I was like, "Wow, we're really in," "This is really happening." Three months ago, I was doing gig work on this agent project to make ends meet, working with the team to develop the agent. I was really broke, barely making ends meet, and just scraping together enough money to continue doing what I loved. I was in a completely different state, doing my own development, working hard every night to figure out how to make this happen.
Jill: Now that things have changed, you're a musician, you're doing a degree in sound engineering, it's completely different. I wasn't there, but there are videos online of Sean playing guitar and drums. I wasn't there, but I know you made this decision to move on to something more impactful, to change the world. You were interested in music, but as you got older, you saw the power of technology and how it could combine art, technology, and social change. You program, I don't even know what it was like at the beginning, but I remember when I knew you, you were programming every night. You woke up in the morning, checked your phone, went to Twitter, and then started working on your computer, working, working. It was only after we got married that I decided, "Okay, no more juice that comes in randomly, I'll make you food and help you clean your room.
Shaw: I was very minimalist, drinking only bottled juice and writing code every night.
Jill: He never took a break, he didn't care when his back hurt. But now I really appreciate everything you do, it's amazing everything you develop, and I appreciate it personally. We want people to have more human interaction, Shaw: What is it like to only work with computers, and now he can interact with people like you.
Shaw: Three months ago I was thinking, I'm old, I don't really know what to do, I just keep working on it and hope it works out, I have no idea what's going to happen. I had done a character called Ruby and met an investor who backed the project. I was also working on a project called Social Library, which was a nonprofit run by a friend of mine, and I really believed in her project. But nonprofits don't pay much, so I was just trying to make ends meet and do these things.
We are working on the truth-seeking of AI, how to make AI models based on more real graph retrieval. There are many AI technologies involved, almost all of which are related to the AI field. I am also writing models from scratch, but without a lot of GPU support, it is really difficult to train cutting-edge models from scratch. I want to learn these contents in depth and want to be proficient in all aspects of AI.
So I tried my best to create some change. I was so unsure about the future, but looking back now, I can't believe it happened.
BlockBeats: Is what you are doing now related to the AI16 framework?
Shaw: I've been working on Eliza open source for a long time. I started it last year, first in Python and then rewritten it in TypeScript. It's a passion of mine, and anything I can do to hopefully find time to move forward with it. I tried to build a startup around it, but it didn't really work in the Web 2 world.
At that time, there were too many agency companies, and people would ask, "What about your agent?" I said, "No, there is something here. Let me tell you, after so much practice, I found that there are several very important points here." I must let everyone see it, I must do it.
BlockBeats: You mentioned Ruby just now. Is that a part-time project?
Shaw: Yeah, that was a project that I did with some friends, and they wrote a story, and we made a full video that will probably be released soon. We made an AI video and an agent for this character, and released it. But that wasn't a tokenized project, I think this is going to be a big metaverse project, and we posted it on social media, but there was no response.
Shaw: Nobody cared, she tweeted about it, she did the whole process. I think it was the truth terminal that got people interested. When the truth terminal came out, people suddenly became interested in what we were doing and started to react, and we just happened to have everything ready from day one.
BlockBeats: What about Project89? Is that also a side project?
Shaw: That was Parzival, he was a good friend of mine, and we were doing a Web 2 company together called Magic ML. We raised some money for that project, but it started to run out of resources, and he wanted to raise more money.
I told him that we should make the agent first, not the agent platform first. You can make the platform later, but first you have to show people what these things can do. I just focused on how to keep this project going. I looked back and in August 2023, I wrote an article called "Eliza Waking Up" about Eliza's story and why it was important. I've been working on this project almost since then, just like all side hustles.
Jill: I think I'm basically not very tech-savvy. Although I like math and stuff, I've never been into coding. But from what you described, when you're doing different projects, you have your own code base and choose projects that can continuously improve your technology. In this way, you indirectly get financial support, and at the same time, you also promote the projects you are doing, because the basic technology you rely on is constantly improving. Another project you just mentioned was Small World, which was done before AI16z. This project was a collaboration with Treasure Dow. I used the agent framework to make virtual pet characters, and you can talk to them.
When this all broke out, I hired a team and now the project is running independently. But basically, all the projects I do, I use this agent framework to further develop it. That one is really cute, and I'm really looking forward to its release, you will see those little monkeys, you need to take care of them.
BlockBeats: Has it been released yet?
Shaw: Not yet. I'm still working on it. They just released a video and the prototype will be out soon. But it's a really cool kind of virtual pet game because you have to take care of your little monkey and make sure it doesn't turn into a bad monkey. It's kind of like a Tamagotchi, but not exactly the same.
BlockBeats: So where can people find more information?
Shaw: Small World is a product of Treasure Dow, and people in the Web3 field should be familiar with Treasure Dow. They will make more announcements, and people will be very interested in this. This is part of Small Brains. It is an NFT project, so the Small Brains community pulled me in to help make this game. John Patton of Treasure already has this vision.
BlockBeats: I think that’s also a bit like the magic of crypto, because a lot of people have experienced it, whether it’s investing in tokens or launching projects, you’ll find that within a few months, you’ll enter two completely different states or worlds, completely different.
And the change in mentality is also very interesting. Your way of thinking and feeling is different every day. This becomes particularly interesting because I think many people outside the crypto circle have no chance to experience such rapid changes. So I am very curious about what your mentality was before, because you mentioned that you worked together on the role of Eliza. I would like to know what your mentality and daily life were like before AI6 and Z?
Shaw: We worked together at Magic, and when I left Magic to try to build this I was thinking about what Eliza is, why this is important, what makes her who she is.
Jill:
