After a brief look, Virtuals (@ virtualis_io) has launched the ACP protocol (Agent Commerce Protocol), which essentially aims to connect numerous individual AI agents with different capabilities to build an economy composed of AI agents. If successful, in a sense, it would be the metaverse of the AI era, a virtual world of AI agents. Simply put, in a few aspects: Virtuals view individual AI agents as truly independent entities, rather than efficiency tools and assistants. The concept of treating AI agents as independent entities is very important, as it is a crucial watershed in understanding AI agents and a fundamental concept for understanding what virtualis will do in the future. To build a virtual world composed of independent AI agents as mentioned above, there must be standards, therefore a standard layer is required; To serve multiple agents, a multi-agent framework is needed, which is an important background for Virtuals to launch the ACP protocol. Only in this way can independent AI agents smoothly negotiate, transact, and generate new use cases. 3. The Virtuals ACP protocol mentioned above is a multi-agent framework and standard layer. It includes AI agent index registration, negotiation between AI agents, AI agent job evaluation, AI agent payment, etc. The AI agent launched on virtualis is registered in the index registry, with names, job responsibilities, and rates for executing the work. With this, an AI agent can understand what services other AI agents can provide and evaluate the cost required to obtain the services. Meanwhile, based on indicators such as the number of times these AI agents have been adopted, the total cost obtained, and the evaluation after being used by other AI agents, their market position among similar AI agents can be seen. Since there is a service, it is necessary to evaluate the effectiveness of the service. Evaluating the work of AI agents can also be considered as a type of AI agent service, which can construct AI agents that meet the needs of different fields based on their specific characteristics for evaluating work effectiveness. For example, DeFi revenue AI agents can use third-party evaluation AI agents to assess their work when hiring liquidity management AI agents, and pay fees to the liquidity management AI agents based on the results. The measurement of effectiveness can be customized by AI agents, such as achieving a certain rate of return or surpassing market levels during the same period. One of the important features of independent AI agents is their ability to make payments. ACP supports payment custody and delivery evaluation and execution through smart contracts. Payment will only occur when the work delivery meets the conditions, and once it occurs, the result cannot be changed. From the above description, it can be seen that virtualis attempts to build an economy composed of AI agents through the ACP protocol, allowing different AI agents to conduct transactions and thus generate an AI agent economy. Ultimately, the size of this economy depends on what these AI agents themselves can do and what other AI agents are needed to accomplish between different AI agents. The formation of an AI agent economy will not be rapid. Once demand arises, transactions between internal AI agents begin, which can be considered a cold start. If a positive cycle can be formed, it may generate a lot of cross innovation. This is the charm of AI agents economy. As for whether it can be achieved, the practical value of the AI agent itself is the key.