Recently, after talking with some entrepreneurs and VCs, there is a common feeling that everyone remains firm in their expectations for the AI + Crypto track, but there is a slight confusion regarding the narrative evolution of web3 AI Agents. What should we do? I have outlined several potential directions for the subsequent AI narrative for reference:

1) The use of MEME-based token issuance by AI Agents is no longer an advantage; in fact, there is a growing aversion to discussing tokens. If a project lacks PMF support and relies solely on a set of Tokenomics that is merely spinning its wheels, it will naturally be labeled as pure MEME speculation, just a wolf in sheep's clothing, with little relation to AI.

2) The original sequence of AI Agent > AI Framework > AI Platform > AI DePIN may be adjusted. When the Agent market bubble bursts, Agents will become the "carriers" after the core technologies like large model fine-tuning and data algorithms take shape. Without the support of core technologies behind them, it will be difficult for an AI Agent to showcase any significant capabilities.

3) Some projects that originally focused on AI data, computing power, algorithms, and other service platforms may surpass AI Agents to become the focal point of attention. In other words, even if new AI Agents are launched, Agents created by these AI platform projects will be more market-convincing. After all, projects capable of operating an AI platform have a more reliable team foundation and technical background than a Dev that is merely based on low-cost deployment of a framework.

4) Web3 AI Agents can no longer compete head-on with web2 teams; they need to seek differentiated directions in web3. Web2 Agents focus on utility, so the logic of low-cost deployment and development platforms works, but web3 Agents emphasize Tokenomics. Overemphasizing low-cost deployment will only trigger more asset issuance bubbles. Undoubtedly, web3 AI Agents should innovate and explore in conjunction with blockchain distributed consensus architecture (I have detailed this in my pinned article).

5) The biggest advantage of AI Agents is "application pre-positioning," which follows the logic of "fat protocols, thin applications." But how should the protocol be "fat"? It is essential to mobilize idle computing resources and leverage distributed architecture to drive the low-cost application advantages of algorithms, activating more verticalized sub-scenarios in finance, healthcare, education, etc. And how should applications be "thin"? Allowing AI Agents to autonomously manage assets, autonomously trade intentions, and autonomously interact in multiple modalities is not something that can be achieved overnight. We cannot attempt to take a big bite all at once; we need to break down the demands into smaller parts and gradually implement them. Otherwise, even the maturity standards of a DeFi scenario may take a year or two to develop.

6) The MCP protocol in the web2 domain and Manus automated execution of multimodal interactions, etc., all provide inspiration for innovation in the web3 field. Directly extending development based on MCP + Manus to suit web3 application scenarios, or using distributed collaboration frameworks to enhance business scenarios on top of MCP, is essential. We should not start by talking about disrupting everything; it is sufficient to optimize appropriately based on existing product protocols and leverage the irreplaceable differentiated advantages of web3. Whether in web2 or web3, both are part of the ongoing revolution brought about by AI LLMs. Ideology does not matter; what is important is the ability to genuinely promote the development of AI technology.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。