What will Manus bring to the big outbreak of web3 DeFai scenarios?

The realization of the true DeFai vision requires solving complex problems such as the capacity limit of single AI models, atomicity guarantee of multimodal interaction and collaboration, unified resource scheduling and allocation of multimodal systems, system fault tolerance, and fault handling mechanisms, etc.

Authored by: Haotian

After waking up, many friends asked me to take a look at #manus, which claims to be a globally universal AI Agent that can independently think, plan and execute complex tasks, and deliver complete results. It sounds very cool, but apart from the anxious voices in many friend circles about losing their jobs, what will it bring to the web3 DeFai scene? Below, I will talk about my thoughts:

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 takeout orders in the browser. Users can supervise visually and take control at any time.

There hasn’t been much discussion about the appearance of this set of Agent, because it is driven by a single model, or the framework of tool invocation. When users think that critical decisions still need to be intervened in, they lose the idea of relying on its execution tasks.

2. On the surface, manus seems to be not much different, just adding a lot of application scenarios, including resume screening, stock research, real estate purchase, and so on, but in fact, the difference lies in the framework and execution system behind it. Manus is driven by multimodal large models and innovatively adopts a multi-signature system.

In short, AI is to mimic the PDCA cycle of human operation (plan - do - check - act), which will be completed by multiple large models working together. Each model focuses on a specific stage, which can reduce the decision-making risk of a single model and improve operational efficiency. The so-called ‘multi-signature system’ is actually a decision verification mechanism of multi-model collaboration, which ensures the reliability of decision-making and execution by requiring the joint confirmation of multiple professional models.

3. In comparison, Manus’s advantages obviously stand out, coupled with a series of operation experiences demonstrated in the video demo, giving people a truly extraordinary experience. However, objectively speaking, Manus’s iterative innovation for Operator is just a beginning and has not yet reached the significance of a revolutionary revolution.

The key lies in the complexity of its execution tasks, as well as the definition of the fault tolerance and successful delivery results of the large model after the non-uniform standard user input prompt. Otherwise, following this set of innovations, can the DeFai scenario of web3 be maturely applied immediately? Obviously, it’s not achieved yet:

For example, in the DeFai scenario, the Agent needs to make transaction decisions, and there needs to be an Oracle-level Agent responsible for on-chain data collection and verification, as well as data integration and analysis. It also needs to monitor on-chain prices in real time to capture trading opportunities. This process poses a great challenge to real-time analysis. It is possible that a trading opportunity that was useful a second ago no longer exists after the Oracle’s large model is transmitted to the transaction execution Agent (arbitrage window).

This actually exposes the biggest weakness of such multimodal large models in making execution decisions, how to connect to the network, trigger chain calls to access real-time level data analysis, identify trading opportunities from them, and then capture trades. The networking environment is actually fine, as the order prices of many e-commerce websites do not change in real-time, which does not easily cause significant dynamic balance problems for the entire multimodal collaboration. However, if on the chain, such challenges exist almost all the time.

4）So, overall, the emergence of manus will indeed cause a wave of anxiety in the web2 field, after all, many high-repetition clerical and information processing jobs may face the risk of being replaced by AI. But that’s their anxiety.

We must objectively recognize the role of this matter in driving the application scenarios of DeFai in web3.

It must be acknowledged that the significance is definitely significant, after all, the LLM OS and Less Structure more intelligence concept it proposes, especially the multi-signature system, will provide great inspiration for the combination of DeFi and AI to expand web3.

This actually corrects a major misconception of most DeFai projects, don’t rush to rely on a large model to achieve AI Agent autonomous thinking + decision-making and other complex goals, which is simply not practical in the financial scenario.

The realization of the true DeFai vision requires solving complex problems such as the single AI model’s capacity limit, atomicity guarantee of multi-modal interaction collaboration, unified resource scheduling and allocation of multi-modal systems, system fault tolerance, and fault handling mechanisms, and so on.

For example: the Oracle layer Agent, responsible for collecting on-chain data and analysis, monitoring prices, and forming an effective data source;

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

The execution layer Agent executes multiple solutions provided by the decision-making layer, taking into account the actual situation, including gas cost optimization, cross-chain state, transaction ordering conflicts, etc.

Only when this series of Agents are synchronized and a huge system framework is in place, can a true DeFai revolution be sparked.