After waking up, many friends showed me manus, which claims to be a globally universal AI agent that can think independently, plan and execute complex tasks, and deliver complete results. It sounds very cool, but apart from many anxious voices about losing their jobs on social media, what will it bring to the explosion of the web3 DeFai scene? Below, let me share my thoughts: 1) About a month ago, OpenAI launched a similar product called Operator, which allows AI to independently complete tasks such as restaurant reservations, shopping, ticket booking, and food delivery in a browser. Users can visually supervise and take over control at any time. The emergence of this set of agent has not been discussed by many people, because it is a single model driven framework or a tool invocation framework. Users lose the idea of relying on it to execute tasks when they think that key decisions need to be intervened. 2) On the surface, Manus may not seem much different, but it has added many application scenarios, including screening resumes, researching stocks, purchasing real estate, and so on. However, in reality, there are differences in the framework and execution system behind it. Manus is driven by a multimodal large model and innovatively adopts a multi signature system. In short, AI needs to mimic the PDCA cycle of human execution (plan execute check act), which will be completed by multiple large models working together. Each model focuses on a specific link, which can reduce the decision-making risk of individual model execution tasks and improve execution efficiency. The so-called 'multi signature system' is actually a decision verification mechanism for multi model collaboration, which ensures the reliability of decision-making and execution by requiring joint confirmation from multiple professional models. 3) By such a comparison, the advantages of manus are clearly highlighted, and coupled with the series of operational experiences presented in the video demo, it truly gives people an extraordinary sense of experience. But objectively speaking, Manus' iterative innovation of Operators is just the beginning and has not yet reached the revolutionary significance of subversion. The key lies in the complexity of its task execution, as well as the definition of the fault tolerance and delivery success rate of the large model after the non-uniform standard user input prompt is entered. Otherwise, following this innovation, can the DeFai scene of web3 be maturely applied immediately? Obviously, it is not yet possible to achieve: For example, in the DeFai scenario, if an Agent needs to execute trading decisions, it needs an Oracle layer Agent responsible for on chain data collection and verification, data integration and analysis, and real-time monitoring of on chain prices to capture trading opportunities. This process poses a great challenge for real-time analysis, as trading opportunities that were useful a second ago may disappear after the Oracle large model is transmitted to the trading execution Agent (arbitrage window); This actually exposes the biggest weakness of such multimodal large models in making execution decisions, which is how to network and touch the chain to retrieve and analyze Real Time level data, analyze trading opportunities from it, and then capture transactions. The online environment is actually quite good. Many e-commerce websites do not have real-time changes in order prices, which is not easy to cause huge dynamic balance problems for the entire multimodal collaboration. If it is on the chain, such challenges are almost always present. 4) So, overall, the emergence of manus will indeed trigger a wave of anxiety in the web2 field, as many repetitive clerical and information processing jobs may face the risk of being replaced by AI. But make them anxious about themselves. We need to objectively understand the role of web3 in promoting DeFai application scenarios: It must be acknowledged that the significance is certainly significant, as its proposed LLM OS and Less Structure more intelligence concept, especially the multi signature system, will provide great inspiration for web3 to expand the combination of DeFi and AI. This actually corrects a major misconception in most DeFai projects, which is not to rely on a large model to achieve complex goals such as autonomous thinking and decision-making for AI agents. This is simply not practical in financial scenarios. The realization of the true DeFai vision requires addressing complex issues such as the upper limit of individual AI model capabilities, ensuring atomicity of multimodal interaction and collaboration, unified resource scheduling and governance of multimodal systems, system fault tolerance and fault handling mechanisms, and so on. For example, the Oracle layer Agent is responsible for collecting on chain data and analyzing it, monitoring prices, and forming effective data sources; The decision-making agent analyzes and assesses risks based on data fed by Oracle, and develops a set of decisions and action plans; The execution layer agent executes various solutions provided by the decision-making layer, taking into account the actual situation, including gas cost optimization, cross chain status, transaction sorting conflicts, and so on. Only when this series of agents are synchronized and powerful, and a massive system framework is established, can a true DeFai revolution be ignited. Note： The attached manus video demo can be viewed in conjunction with the above considerations 👀。 I think the content is useful. Could you please help me with one click three-way support? Thank you!