Author: shu fen
Many people have said that the current crypto bull market lacks innovative narratives, but in fact, AI is the most innovative and enduring core narrative. By December 2024, the project with the highest returns in the entire crypto market (non-on-chain) will come from the AI field - Virtuals, with a return of up to 23079%.
"The next destination of large models", "Fundamentally changing the way humans live", "Ushering in a new era of industrial revolution"... People do not hesitate to describe the importance of AI Agents. Regarding the current development momentum and future trends of AI Agents, both retail and institutional investors are not well-prepared. Many people around me did not pay attention before, and only felt the need to understand it after the outbreak, but now the market is flooded with all kinds of information, which is quite confusing. Today, I will thoroughly sort out the AI Agents, and this research report will help you get a quick grasp, as an introductory guide to AI Agents (crypto version)!
This article will be divided into three parts: 1. Demystifying the development status of AI Agents, 2. Screening and analyzing potential AI Agent projects, and 3. Expectations for the application of AI Agents in the Web3 field.
I. Demystifying the Basics of AI Agents
AI Agents first appeared before the public in March 2023, with the release of the AutoGPT framework project, which uses large language models to automatically break down a large task into smaller tasks and use tools to complete them.
The release of AutoGPT shocked the world, as it was the first time that language processing, content creation, logical reasoning, and perception and action capabilities were extended to application scenarios. Shortly after, OpenAI launched a series of GPTs, and then many tech companies began to layout in the application layer, platform layer, development layer, and operations layer to increase the barriers to the next wave of the ecosystem.
What exactly is an AI agent? How does it work? "Agent" in Chinese means a proxy. Simply put, an AI agent is a proxy empowered by AI technology. Unlike traditional software, it is not just passively executing instructions. Its workflow is: perception module (obtaining input) → LLM (understanding, reasoning and planning) → tool invocation (task execution) → feedback and optimization (verification and adjustment).
OpenAI defines "AI Agent" as a system driven by LLM as the brain, with the ability of autonomous understanding, perception, planning, memory, and tool use, and can automate the execution of complex tasks. Unlike traditional artificial intelligence, AI Agents have the ability to independently think, call tools, and gradually complete the given goals.
Let's give an example to help you understand better: If you have a cold and fever, traditional software will only tell you to go to the hospital and pay attention to protection. If it's an AI agent, it can detect your body temperature and other health indicators, match the right medication based on online information, request payment and deliver it to your home, and even write a leave letter for you the next day. This is the magic of AI agents.
II. AI Agent Projects and Analysis
According to the latest data from Cookie.fun, as of December 30th, the overall market capitalization of AI Agents has reached $11.68 billion, with a growth of nearly 39.1% in the past 7 days. This growth trend indicates the rapid expansion of the AI Agent ecosystem in the crypto market.
In this AI Agent craze, a16z and Virtuals Protocol are undoubtedly the two most powerful representative projects. Specifically, Virtuals' ecosystem market value has reached $5.01 billion, and a16z is $1.63 billion, accounting for 56.8% of the AI Agent market share, contributing more than half.
From the on-chain distribution, Base and Solana are the two main battlegrounds for AI Agents. The market value of AI Agents on Base is about $5.76 billion, and on Solana it is $5.47 billion, together contributing 96.1% of the overall market, while the cumulative market value of other chain projects is only $920 million.
This also indirectly reflects that although the AI Agent ecosystem has risen rapidly in the crypto market, attracting a lot of attention and capital, the market structure is still relatively simple, mainly relying on the drive of a few leading projects, and the AI Agent ecosystem is still in its early stages of development.
Next, based on the current market situation, I will analyze the hot AI Agent projects, with three main standards: 1. The long-term value of the project, 2. The real market demand, and 3. The cash flow and revenue situation. If you think why a certain XXX project is not included after reading, please review it again based on these three standards. The following project views are for reference only and not as financial advice.
1. Virtuals
Virtuals actually went online last year. The Virtuals protocol is mainly to establish co-governance for AI agents in the gaming and entertainment field. AI agents can be tokenized and achieve co-governance through the blockchain, with functions including autonomous planning, goal achievement, environmental interaction, and on-chain wallet control.
The biggest innovation and difference between Virtuals and other Web3 AI agent protocols is that it simplifies the complexity of AI agents and provides a Shopify-like plug-and-play solution, so that non-AI professionals can easily deploy AI agents in gaming and consumer applications, and can obtain protocol revenue through tokenization and decentralized co-governance.
In addition, Virtuals has used its AI technology principles to generate an AI virtual idol - the AI-dol band, which has tens of thousands of fans on TikTok, which is quite interesting.
Virtuals has a total token supply of 1 billion, all of which have been released. The token distribution is as follows: 60% held by the public, 5% for liquidity pools, and 35% for the ecosystem treasury, with a maximum annual release of 10% within three years, and the ecosystem fund currently holds over 30% of the tokens.
From the long-term value of the project, it has solved the pain point of non-AI professionals unable to participate in the AI craze, has a certain user base, and the token economy is relatively open and transparent, and the marketing is also very impressive. From the market capitalization, as the leader of the AI Agent ecosystem, its recent surge has been almost without adjustment, so there is a high probability of a relatively large adjustment in the future, so the short-term risk for Virtuals is relatively high.
2. a16z
Although a16z has the same name as the famous venture capital firm a16z, this project has no connection with A16Z and has not received any investment from A16Z. The only connection is that it has received the attention of a16z's founder Marc Andreessen.
The total token supply is 1.09 billion, and the project operates in a DAO mode. According to the project's core influencer Shaw, a16z will launch several games based on the Eliza framework in the future, and in the future, more efforts will be put into building a practical AI Agent investment tool, a DeFi AI Agent. As the founders said, a16z's goal is not to create an AI robot that mimics a16z, but to beat it in the investment field it is most proficient in.
In other words, the a16z project focuses on the AI agent for investment models, which looks no different from the previous AI bots and Telegram bots. Can AI really invest and make money? This is a question mark. The core technology Eliza OS is just a simple development based on the capability base of OpenAI, and if OpenAI opens up its own AI agent in the future, how will Shaw respond?
In summary, I think the a16z project is just riding on the coattails of a16z. Its long-term value lies in the DeFi AI Agent, but this demand is also a false demand, and it has returned to the logic of the previous AI bots. Its technical development capabilities rely on the open-source database of OpenAI, and its imagination is generally average.
3.SWARMS
Swarms is an AI agent multi-agent LLM framework that provides a wealth of clustering architectures and seamless third-party integrations. It currently enables enterprises to easily build and manage collaboration between multiple AI agents, and under the scheduling of Swarms, they can work seamlessly to complete complex business tasks. In simple terms, the users of SWARMS are enterprise B-end, providing enterprise-level AI agent applications.
The founder, Dev, is the 20-year-old Kye Gomez (source from the internet), who publicly claimed that OpenAI infringed on the team's intellectual property, stole the project name, and plagiarized the code structure and methods. Subsequently, Gomez released a more detailed explanation: Swarms is a multi-agent framework that has been running for nearly 3 years. So far, there are over 45 million agents running in production environments, providing services to the world's largest financial, insurance, and healthcare institutions.
After the Swarms token was launched on December 18, it quickly soared to a market capitalization peak of $74.2 million on the 21st, but the good times did not last, and the market value plummeted like a roller coaster to the bottom, remaining at around $6 million. Subsequently, it fluctuated around $13 million until the 27th, when it started to counterattack, rising from the low point of $12 million to $30 million, and then surging nearly 3 times to approach $70 million, almost breaking through the previous high.
Compared to the fanciful castles in the air of the a16z system, if Swarms is indeed created by the 20-year-old AI genius Kye Gomez as rumored, it undoubtedly has a strong technical barrier, and on its official website, it has already provided efficient solutions for many enterprises, demonstrating its strength.
As an open-source project, Swarms has sparked heated attention in the developer community, with over 2.1K stars on GitHub, garnering the wisdom and support of many developers, which proves the maturity and innovation of its technology. Swarms has stronger technical capabilities and strong market demand (enterprise-level), and it will stand out in the race of AI agents.
4.GRIFFAIN
Griffain is a Solana-based project - an AI agent engine, similar to Copilot and Perplexity, and is one of the closest projects to Agentic APP. The ultimate form of the search engine in the AI era should be that the user directly proposes the demand, and the AI directly provides the result or solution, rather than just providing web page links. One of the catalysts for this project is its open access mechanism, and as a leading agent engine, Griffain has undoubtedly attracted a lot of market heat.
Solana is currently the blockchain with the most AI agents, and in October, Goat, as an AI agent, raised funds from humans through pumpfun, which in a sense is an AI singularity, because the excellent liquidity and mature AI agent developer community of the entire chain make Solana a breeding ground for the most imaginative AI agents.
The most important thing Solana has done is to revitalize the ecosystem with Griffain. Because to bring about a true "Agentic app szn", in addition to AI, there also needs to be a channel for the infrastructure. Although GriffAIn has not yet clearly defined the specific application scenarios of its tokens, in the future, GriffAIn will need to connect the demand side with the Solana ecosystem, and as long as it exists within the scope of Solana's existing technical system, it can basically meet the requirements, whether it is to snipe certain qualified tokens on Pumpfun or create tokens, this vision has been recognized by Toly, which adds a lot of imagination to the prospects of GriffAIn.
5.AIXBT
Aixbt is one of the intelligent agents created by Virtuals on the Base platform. It uses smart analysis tools to monitor Crypto Twitter and market trends, providing users with valuable market insights. Some of the analysis content will be shared on Twitter, while the rest will only be accessible to token holders, who can directly communicate with the intelligent agent through their dedicated terminal.
Aixbt's analysis has a certain degree of accuracy in predicting price trends, demonstrating how AI can analyze blockchain data and help traders make smarter decisions across multiple platforms and domains.
I took a look at the relevant content published by Aixbt, and my immediate impression is that the content is very rich, covering almost all the tracks, and the data is readily available. Additionally, there are potential investment opportunities in the short-term cryptocurrency splits; for example, discovering that the hype-based vapor is undervalued compared to similar AI launchpads. Data shows that of the 210 tokens recommended, 183 have achieved profitability after Aixbt's recommendation, a high success rate of 83%.
However, there are also some shortcomings, such as the inability to fully decompose complex projects, and the analysis and data are still at a shallow level, unable to indicate the risks of investment opportunities, but I think it is much stronger than the current crypto KOLs.
From the long-term value of the project, Aixbt has demand in the niche market, and users also have the motivation to hold tokens to unlock more information data and price analysis. As Aixbt continues to feed and evolve the data, I believe it will be the absolute king of AI intelligent agents in market prediction.
In summary, after analyzing the current 5 high-profile AI Agents, based on the three points mentioned earlier, I believe the ranking from high to low market value imagination space is: SWARMS, GRIFFAIN, Virtuals, AIXBT, Ai16z.
III. Future Development Trends of AI Agents in the Web3 Domain
For the application of AI Agents in the Web3 domain, there are several directions worth noting, which also represent future trends. One is privacy and security. AI should be designed with respect and protection for users and society as the basic principle. But as AI understands us more and more, privacy will become increasingly blurred and fragile, and every interaction with smart devices and every input of personal information will become food for AI evolution.
The importance of privacy issues lies in the fact that privacy and security issues are inseparable. The systems that store and process personal data, once they become the target of hacker attacks, will lead to information leakage, identity theft, asset loss and other problems. Is there an environment that can both unleash the powerful capabilities of AI and protect personal privacy? In the Web3 domain, compared to traditional methods, it can provide users with a higher level of data protection and perfectly balance the contradiction between AI's development capabilities and privacy protection.
Therefore, we see that the data storage of many large models has begun to try on the blockchain, and the perfect AI soil of Web3 has also attracted many AI developers to try to use blockchain technology to ensure the security and privacy of data in specific industries with high privacy needs, such as healthcare and finance.
Another important direction is computing power and data. AI Agents, especially Multi-Agent collaboration, face the problem of high development, training and operation costs. For enterprises, training AI Agents requires a large amount of computing power, often hundreds or even thousands of high-performance GPUs or specialized hardware like TPUs. The cost of acquiring and using these computing resources is already high, for example, Stanford's virtual town includes 25 intelligent agents in the multi-agent research, and after the town framework was open-sourced, testing a single Agent requires $20,000 per day in data sources.
In Web3, this can be addressed through a reasonable token economy and user incentive scheme, allowing idle computing power or personal data sets to be reallocated at lower cost, and allowing more individual users to participate in the construction of the AI industry. For example, some data platforms allow users to monetize their own data, providing low-cost data sources for AI Agents.
Finally, I believe that AI Agents will become the new infrastructure of Web3 in the future, deeply integrated with other core elements, and catalyze entirely new application models, rather than just simple AI Memecoins. Currently, in the Web3 domain, AI Agents can help users reduce thresholds and improve experiences, even if it is just to simplify part of the asset issuance process, it is still meaningful.
But from the macro perspective of Web3, AI Agent as an off-chain product, at the current stage, it only plays the role of an auxiliary tool for smart contracts, so there is no need to overhype its capabilities. Since there was a lack of significant wealth effect narratives other than MEME in the second half of this year, it is normal that AI Agent has become popular around MEME.
However, relying solely on MEME cannot sustain long-term value, so if AI Agent can bring more innovative gameplay to the transaction process and provide practical application value, it may develop into a common infra tool.
