After waking up, many friends asked me to take a look at #manus, which claims to be a truly universal AI Agent globally, capable of independent thinking, planning, executing complex tasks, and delivering complete results. It sounds very cool, but aside from the anxiety from many in my social circle about losing their jobs, what will it bring to the explosive growth of web3 DeFi scenarios? Below, I share my thoughts:

1) About a month ago, OpenAI launched a similar product called Operator, where AI can independently complete tasks such as restaurant reservations, shopping, ticket booking, and food delivery in the browser. Users can visualize supervision and take control at any time.

The emergence of this Agent did not spark much discussion, as it is driven by a single model and follows a tool-calling framework. When users realize that key decisions still require intervention, they lose the idea of relying on it to execute tasks.

2) On the surface, manus seems not much different, just with more application scenarios, including resume screening, stock research, property purchasing, etc. However, the difference lies in the underlying framework and execution system. Manus is driven by multimodal large models and innovatively adopts a multi-signature system.

In short, AI needs to mimic human execution of the PDCA (Plan-Do-Check-Act) cycle, which will be collaboratively completed by multiple large models, each focusing on specific links. This can reduce the decision-making risk of a single model executing tasks and improve execution efficiency. The so-called "multi-signature system" is actually a decision verification mechanism for multi-model collaboration, ensuring the reliability of decisions and execution through the joint confirmation of multiple specialized models.

3) In comparison, the advantages of manus are clearly highlighted, and the series of operational experiences showcased in the video demo indeed provide an extraordinary experience. However, objectively speaking, Manus's iterative innovation over Operator is just the beginning and does not reach a revolutionary significance.

The key point lies in the complexity of the tasks it executes and the fault tolerance and success rate of the large model's delivery results after non-uniform standard user input prompts. Otherwise, following this innovation, can the web3 DeFi scenarios be immediately mature applications? Clearly, it is not yet possible:

For example, in the DeFi scenario, if an Agent needs to execute trading decisions, there must be an Oracle layer Agent responsible for collecting and verifying on-chain data, integrating and analyzing the data, and monitoring on-chain prices in real-time to capture trading opportunities. This process poses significant challenges for real-time analysis; a trading opportunity that was valid a second ago may no longer exist by the time the Oracle large model transmits it to the trading execution Agent (arbitrage window).

This actually exposes a major vulnerability in the execution decision-making of such multimodal large models: how to connect to the network, retrieve and analyze Real-Time level data from the blockchain, identify trading opportunities, and then execute trades. The networking environment is relatively manageable; many e-commerce websites do not have real-time price fluctuations, which does not significantly disrupt the dynamic balance of multimodal collaboration. However, on-chain, such challenges are almost always present.

4) Therefore, overall, the emergence of manus will indeed stir up 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 let them worry about their own issues.

We must objectively recognize the role of this in promoting DeFi application scenarios in web3:

It must be acknowledged that the significance is indeed substantial, as the proposed LLM OS and the concept of Less Structure more intelligence, especially the multi-signature system, will provide great inspiration for the integration of DeFi and AI in web3.

This actually corrects the major misconceptions of most DeFi projects, which should not start by relying on a large model to achieve complex goals like autonomous AI Agent thinking and decision-making. This is fundamentally unrealistic in financial scenarios.

The realization of the true DeFi vision requires solving complex issues such as the capability limits of individual AI models, ensuring atomicity in multimodal interactive collaboration, unified resource scheduling and allocation in multimodal systems, and system fault tolerance and failure handling mechanisms.

For example: An Oracle layer Agent is responsible for collecting on-chain data and analysis, and monitoring prices to form effective data sources;

A decision layer Agent analyzes and assesses risks based on the data fed by the Oracle and formulates a set of decision-making and action plans;

An execution layer Agent executes based on the various plans provided by the decision layer, considering actual conditions, including gas fee optimization, cross-chain status, transaction sorting conflicts, etc.

Only when this series of Agents are all simultaneously powerful and a large system framework is established can a true DeFi revolution be ignited.

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。