Author: Haotian
Some market misunderstandings about the existence of AI Agents, a simple discussion of my views:
1) AI Agents are already mature enough in the web2 domain, what comparability do web3 AI Agents have?
For friends who raise such doubts, I always admit that they are right, because at present, it is indeed a fact from any perspective. However, the problem behind this is not that we do not understand web2, but that some people's understanding of web3 is always alienated from prejudice, and they should delve deeper into web3.
2) AI Agents cannot escape the MEME attribute, and AI MEME is the ultimate destination?
In the short term, from the perspective of asset issuance narrative, AI Agents intuitively feel like MEME, but originating from MEME, and becoming popular because of MEME, its ultimate settlement must be the application landing value. The culture of financial nihilism cannot support the grand narrative of AI. The application value of AI Agents themselves, as well as the disruptive value of AI Agents landing in the existing blockchain big infra framework, are the vast oceans for AI Agents.
3) AI Agents in the web3 direction should mainly develop monolithic AI, because Agents with Utility have the meaning of construction?
The imagination space of monolithic AI in the web3 field is far less than that of web2. Since we want to plan for the future in the web3 field, the goal of AI Agents should be how to run within the blockchain system, and the imagination space of "chaining" is greater than anything else. Focusing only on Utility application landing that is no different from web2 will not give full play to the ultimate potential of AI Agents, and will also miss many opportunities for framework-level innovation in the blockchain circle.
4) The various protocol standards driven by AI Agent economic models and technical frameworks are confused?
Although Virtual, zerepy, ELIZA, ARC, and Swarms are all classified as AI Agent frameworks, there are also differences between specific scenarios and universal scenarios. Virtual's closed-loop Tokenomics growth flywheel, Eliza's open-source without Tokenomics constraints, zerepy's expertise in monolithic AI practical applications, ARC's performance in LLM large model optimization, and Swarms' challenges in enterprise-level multi-model interaction, etc., each has its own shortcomings and strengths, and it is wrong to confuse and compare them, and there is no result.
5) AI Agent docking with blockchain is a pseudo-proposition, and there is no need to combine with traditional infra to achieve "chaining"?
AI Agents must be combined with blockchain to find a decentralized asset custody solution, a decentralized data Memory processing solution, and a decentralized Solver intention transaction automated execution Marketplace, etc. The existing public chain infra's DA, interoperability, chain abstraction, and DeFi infrastructure will become the prerequisites for AI Agents to dock with the blockchain. It is necessary, and we must take this step.
