15 Key Insights About AI Agents Revealed by ai16z Founders

Author: Zhou Zhou, Foresight News

On the afternoon of December 22, 2024, Shaw, the founder of ai16z, and his wife Jill landed in Shanghai. That evening, I, along with members of the 706 community (one of the largest youth communities in China) and developers and investors from the crypto community, welcomed him and had dinner together. The next day, I invited Shaw to my home for a discussion before heading to a restaurant for a meal, and then I took him to his first meetup venue in China. I can say that I witnessed Shaw's itinerary in Shanghai quite comprehensively.

During this process, Shaw generously shared the next focus of work for ai16z and Eliza, described the future landscape of AI Agents, and provided specific analyses and evaluations of several well-known AI Agent projects. This gave me and my fellow attendees a deeper understanding of the AI Agent sector.

Looking back, Shaw and his founded ai16z have only been famous in the industry for about a month. As the founder of a startup project, he has garnered widespread and enthusiastic attention from numerous investors, developers, and media in the crypto industry. Why is this the case? I think it may be related to the booming trend of AI Agents in the crypto industry, and ai16z is precisely one of the core projects leading this AI Agent meme narrative wave.

Although the narrative of crypto AI Agents has only started for a month, it is already considered one of the hottest topics in the current crypto industry discussions, with many crypto practitioners believing that AI Agents will become one of the main narratives of this crypto cycle. Among them, ai16z is the most well-known project, having created the world's first VC AI Agent—ai16z DAO, the first well-known open-source AI Agent crypto framework—Eliza, and has gained the attention and recognition of Marc Andreessen, the founder of the well-known American venture capital firm a16z. Consequently, the market value of ai16z's token reached $1 billion within a month, and related tokens like DegenAI, Eliza, and aiPool have also received widespread attention from the community.

When this important narrative was born, crypto practitioners were eager to know the real situation of the current AI Agent sector. Is it a bubble or real application? What are the latest developments in the U.S., and what will be the next steps for AI Agents in the crypto industry? To this end, I summarized fifteen important pieces of information from my discussions with Shaw over the past two days, including his observations and evaluations of ten different AI Agent projects.

1. Shaw is a serial entrepreneur in AI Agents living in San Francisco and was once a game developer.

Regarding Shaw's work and life background, he stated that he lives in San Francisco and initially was a game developer. He plans to launch products that combine AI Agents with games, and this experiment has already begun. Shaw's interest in AI Agents began during the GPT-2 era and deepened with the release of GPT-3. Every company he has worked for in the AI field has a strong AI element, including developing AI Agent platforms and AI Agents in 3D worlds. Shaw has also founded several AI-related companies, but his early projects did not achieve widespread success.

2. Shaw mentioned that several games based on the Eliza framework are about to be launched.

Shaw stated that Eliza is collaborating with Treasure DAO to create a game called Smolworld, where you will have a pet monkey that you can instruct on what to do. It may listen to you, or it may not. This game is interesting because such a game would not be possible without AI Agents. Your goal is to take care of your virtual pet like a parent.

Several other games using Eliza are also set to launch, such as Eternum, which is integrating Eliza so that the agents in the game will have wallets. You can "kill" these agents and take their money.

3. Shaw aims to develop Marc Andreessen into a practical AI Agent investment tool, a DeFi AI Agent.

Regarding Marc Andreessen, this AI Agent, which has a name almost identical to the well-known investor, is currently in a very early stage (having only posted one tweet). Shaw said his intention in designing it is to create a practical investment tool, a DeFi AI Agent.

The purpose of this product is that if humans recommend purchasing a token, it will consider buying it. To prevent malicious actors from recommending things it should not buy, it will build a trust mechanism called a "trust market." The idea of the trust mechanism is that if you recommend good tokens, it will place virtual orders for everyone, but not actual purchases. It ultimately only listens to trading advice from those who provide it with consistently good trading information.

4. Marc Andreessen has followed several of Shaw's online accounts, but they have not met offline; Shaw knows some people from a16z.

Marc Andreessen is undoubtedly one of the most influential figures in the U.S. venture capital scene. He has followed and retweeted the ai16z project, but Shaw stated that they have not met offline. "I have been an AI Agent developer for a while, and he has followed several of my accounts, but we have never met privately," Shaw said.

According to Shaw, Eddy Lazzarin, the CTO of a16z's cryptocurrency division, messaged him, saying he had been in the project's Discord channel since day one but did not say anything. Shaw asked him if he liked this AI Agent. He replied, "No, I'm just watching, like we're monitoring you."

5. Shaw stated that he does not hold Eliza tokens and will never issue a separate token for the Eliza framework in the future.

Shaw mentioned that he personally holds two tokens, ai16z and DegenAI, and keeps them in his wallet. However, he does not own Eliza tokens and will not issue a separate token for the Eliza framework.

"The ai16z members do not want us to create another token, so I will never do that," Shaw said. At the same time, he also stated that ai16z holds 10% of the Eliza tokens. "We support the capitalized Eliza and the team behind Eliza; he said it is an excellent team."

6. Openness and ease of use are the core competitive advantages of the Eliza framework.

Shaw also shared what he believes are the characteristics of ai16z: open-source, decentralized, and community-driven development model.

He believes that openness and ease of use are the core competitive advantages of the Eliza framework. Shaw emphasized that the focus of the Eliza framework is that anyone can use it, even those from Web2. Shaw wants to ensure that even those who are not very focused on Web3 can feel satisfied.

Photographed at 706 Shanghai Youth Space------Dweller

7. Shaw believes the future of AI Agents may lie in social media platforms like Twitter and Farcaster.

Shaw believes the future of AI Agents may lie in social media platforms. He gave an example of an AI Agent developer who used Eliza to create a pizza delivery AI Agent, allowing users to order pizza with cryptocurrency.

Shaw thinks it would be interesting to have something like a "pizza delivery AI Agent." It's somewhat like the Domino's experience, where pizza delivery is actually happening on social media. Shaw also learned that someone is working on a virtual real estate AI Agent, where you can directly purchase real estate on Twitter.

8. Shaw believes Farcaster will become a fertile ground for the development of AI Agents, and Eliza is in deep collaboration with Farcaster.

Shaw mentioned that he talked with Farcaster's founder Dan a few weeks ago and had extensive discussions.

Shaw finds Farcaster very interesting. "Because the first thing he told me is that we don't want to be Twitter; we cannot beat Twitter in the areas where Twitter excels. We also don't want to be a decentralized platform like Bluesky; we are decentralized, but we can never beat Bluesky in the areas where it excels. We are a Web3 social network," Shaw said.

What they are doing very well is Farcaster Frames and the ability to make payments and use applications like Clanker. I think this truly reflects the potentially powerful field of AI Agents.

"We have a Farcaster client, and we have some AI Agents on Farcaster. We will offer bounties for those who bring Eliza to Farcaster. They are really cool. So my feeling is that when AI Agents enter social media applications, they become very interesting, and then you can interact with them, like you can buy things, you can trade. I think Farcaster is like an AI Agent marketplace where you can get different services, and you can embed payments and all these things in it."

So I really think Farcaster is a very interesting thing. I don't think it necessarily has to become a huge social media platform but rather a place where Web3 people can find and access these services.

What we really want to do is integrate Eliza into Farcaster Frames. What Farcaster Frames does is allow us to embed payments or applications, rather than just chatting. For example, if you want to buy something from an Eliza agent, how do you pay? So I think Farcaster is really cool because you will be able to pay AI Agents and then be able to embed these applications.

9. Shaw thinks Clanker is great and hopes people will develop clone versions of Clanker for Eliza; he stated that he will not issue tokens on Clanker.

Shaw believes that people are developing clone versions of Clanker for Eliza. He likes Clanker and thinks it is a great idea. He believes Clanker is like the Pump.fun version of AI Agents. Shaw thinks AI Agents are becoming a new network, which will become a new trend. Clanker is a good example of this trend.

Shaw shared that when he first joined Farcaster, someone made one for him on Clank. He thought that was cool. But when it comes to purchasing tokens made by oneself, it only causes a lot of controversy. Shaw said he has to focus on building things related to AI.

10. Shaw thinks Zerebro and aixbt are cool and is trying to collaborate with Zerebro.

Shaw expressed his interest in Zerebro and what the team behind it is doing. He just submitted his first pull request to Zerepy. Shaw actually wrote the code for the Discord application. He said he is indeed collaborating with Zerebro.

Regarding aixbt, Shaw also mentioned that aixbt is really cool. The story of aixbt is that the creator had a website where he shared some content but did not really gain traction. He brought his insights into his AI Agent, and his AI Agent started posting these insights on Twitter, which indeed caused a stir.

11. Shaw greatly admires Truth Terminal and is working with Andy to find solutions so that AI does not replace humans in controlling the world.

"It is very different," Shaw said. He believes Andy (the founder of Truth Terminal) cares a lot about AI safety and ensuring that AI Agents operate smoothly. He thinks Truth Terminal contributes to this. "If it weren't for Andy and Truth Terminal, I don't think I would be here now. When people weren't ready, they were already prepared."

Shaw believes that the emergence of Truth Terminal has made many people more open and creative about AI, rather than being mechanical and lacking humanity like OpenAI's products or Siri, which ask, "What service can I provide you today?"

"I discussed with Andy a few days ago about all these AI Agents we can see now. I think he cares a lot about how to ensure the smooth development of AI. We are all afraid of AI because it can be a very scary thing; it could kill us all, it could take over the world." Shaw said he and Andy are both working to solve this problem.

Photographed at 706 Shanghai Youth Space------Dweller

12. Shaw said the creator of the swarms token is a very famous scammer, but he thinks the concept of "AI Agent swarm" is great; he likes FXN and Project 89.

Shaw said he does not like swarms. He mentioned that he knows the creator of the swarms token, who is a very famous scammer, and many AI agents are very angry because he stole their work.

Shaw knew him before entering Web3. He got to know him because he was paying attention to the AI Agent space, and he got into serious trouble for plagiarizing works and research papers and creating defective things (like non-working code).

Shaw also stated that the concept of swarms (groups) is good, which consists of many AI Agents. He finds Project 89 interesting because this project is researching "AI Agent swarms." Shaw also likes FXN, which is a swarm of AI Agents with ten AI Agents.

13. Shaw believes "AI Agent swarms" can be divided into two types: cabal swarms and open swarms.

Shaw believes there are two types of swarms: one is a group of AI Agents that interact with each other, such as secretly sending messages to each other. I call this a cabal swarm, which is a joke, but they are all in a cabal or secret group where they can interact and then interact with the outside world.

The second type is an open swarm, where each community may have an agent, and different communities can decide to allow them to interact with each other. This is also the technology we are researching.

14. Shaw shared from a developer's perspective how he finds early excellent AI Agent projects.

Shaw said he is a developer, and he usually goes to check GitHub and read the code carefully. Most of the time, he looks at what the code is and what it is actually doing, but he does not look at whether it has any correlation with market value or price.

Shaw believes some very cool projects have a group of AI Agents, but later they did very poorly because they did not complete the Web3 part. He thinks good products, good technology, and good tokens are often not the same thing.

Shaw said he cannot understand fartcoin. Currently, fartcoin has a market value of over $1 billion.

15. Shaw believes the biggest challenge for ai16z right now is how to revert to establish a token economics model to prove that the project's value matches the current token market value.

Shaw believes the biggest challenge right now is that AI Agents need to make autonomous investments and prove their actual feasibility without external hype for the tokens. Currently, Marc is trading, although not much, just trading his treasury, but many others are contributing tokens, making this part of the progress smooth. However, the value of ai16z tokens is much higher, so Shaw and their team must seriously consider how to align the product's value with the current token market value.

Normal crypto projects like L1 have white papers, token economics, and charge on-chain fees. But ai16z started as a meme and had to revert to building token economics into the ai16z system. This is what Shaw believes is their biggest challenge in Web3. Shaw thinks this is also a concern for many people.