GuShiio.com Be closest to the user and make the best AI Agent.
Authors: shiyun, Zhang Yongyi
The saying that 2025 is the first year of AI Agents was fulfilled in the early morning of March 6, Beijing time.
“After DeepSeek, another sleepless night in the technology circle.”
Many users commented this on social media.
Everyone stayed up all night, using the invitation code only for one of the products, which was the world’s first AI Agent product “Manus” developed by Monica.im
According to the team,”Manus” is a truly autonomous AI agent that can solve various complex and changeable tasks. Unlike traditional AI assistants, Manus can not only provide suggestions or answers, but also directly deliver complete task results.
Manus’s introduction video is only 4 minutes long, but it is amazing
Photo source: Monica.im
As the name “Manus” implies, it symbolizes “hand” in Latin. In other words, knowledge must not only be in the mind, but also be able to be executed with hands. This is the essential advancement of Agent and AI Bot products.
Where is the Manus cow? The most intuitive thing is to look at the use cases displayed on the official website and the spontaneous display of users. The Geek Park section is organized as follows:
Travel planning: Not only does it integrate travel information, it also creates customized travel manuals for users. For example, users plan their April trip to Japan and provide personalized travel advice and detailed manuals.
Stock analysis: Conduct in-depth stock analysis and design visually attractive dashboards to showcase comprehensive stock insights. For example, conduct in-depth analysis of Tesla stock and create visual dashboards.
Educational content creation: Create video presentation materials for middle school teachers to explain complex concepts such as the Momentum Theorem and help teachers teach more effectively.
Insurance policy comparison: Create a clear insurance policy comparison table, provide best decision-making suggestions, and help users choose the most suitable insurance product.
Supplier procurement: Conduct in-depth research throughout the network to find the supplier that best suits user needs and serves users as a truly fair agent.
Financial report analysis: Capture changes in market sentiment towards specific companies (such as Amazon) through research and data analysis, and provide market sentiment analysis over the past four quarters.
Organize startup company list: Visit relevant websites to identify qualified companies and organize them into a table. For example, compile a list of all B2B companies for the YC W25 batch.
Online store operations analysis: Analyze Amazon store sales data to provide actionable insights, detailed visualizations and customized strategies to help improve sales performance.
When the Agent finally outputs an extremely complete and professional result through a long series of thought chains and tool calls, users began to sigh that “it can really help humans do things.”
According to official website information, Manus achieved new state-of-the-art (SOTA) performance at all three difficulty levels in the GAIA benchmark test, which evaluates the ability of general AI assistants to solve real-world problems.
To sum up, what Manus wants to do more is to be your literally “agent” in the digital world. And it did.
Just as you might think, Manus was launched in the early morning and woke up all people in the AI circle!
01 Manus, your digital agent
First of all, Manus ‘experience is the biggest difference from previous LLM:
It emphasizes the ability to deliver the final result directly, rather than just giving a simple “answer.”
Manus currently uses the Multiple Agent architecture, which runs in a similar manner to the previously released Computer Use by Anthropic, and runs completely in a separate virtual machine. At the same time, various tools can be called to write and execute code, browse web pages, operate applications, etc. in the virtual environment to directly deliver complete results.——
In the officially released video, three work cases completed by Manus in actual use scenarios are introduced:
The first task is screening resumes.
From 15 resumes, recommend suitable candidates for the intensive learning algorithm engineer position and rank candidates based on their intensive learning expertise.
In this demonstration, you don’t even need to decompress the zip file and manually upload the resume file one by one. By this time, Manus had already shown his human “intern” side, manually extracting files, browsing each resume page by page, and recording important information therein.
Like an intern, Manus automatically understood the hidden command of “unzip the packaged files thrown by the boss.”
Photo source: Geek Park
Among the results given by Manus, there are not only automatically generated ranking suggestions, but it also ranks candidates into different levels based on important dimensions such as work experience. After accepting that users prefer to present them as Excel tables, Manus can also automatically write Python scripts on site to generate corresponding tables.
Manus can even use his memory ability to record information such as “users prefer to accept results through tables” during this practice. The next time he processes similar task results, he will give priority to using tables for presentation.
Manus remembers user preferences during the content generation process
Photo source: Geek Park
The second case, which is more tailored to the Chinese people, is the selection of real estate.
In the case, the user wants to buy a property in New York. The requirements entered are that they want to have a safe community environment, low crime rate, and high-quality primary and secondary education resources. Of course, the most important budget is enough to afford it with a fixed monthly income.
In this requirement, Manus AI breaks down complex tasks into a to-do list, including researching safe communities, identifying high-quality schools, calculating budgets, searching real estate, and more. We also searched through the Internet to carefully read articles about the safest communities in New York and collect relevant information.
Second, Manus writes a Python program to calculate an affordable property budget based on user income. Combined with relevant house price information on real estate websites, filter the list of properties based on budget range.
Manus can automatically search and filter out listings whose conditions do not meet user requirements
Photo source: Geek Park
Finally, Manus integrates all collected information and writes detailed reports, including community safety analysis, school quality assessment, budget analysis, recommended property lists and links to relevant resources, just like a professional real estate agent. And because Manus comes with the attribute of being “completely based on user interests”, its use and even experience are better.
In the last case, Manus demonstrated analytical capabilities for stock prices.
The task given by the case is to analyze the correlation between the stock prices of Nvidia, Maiwire Technology and TSMC over the past three years: it is well known that there is a close correlation between these three stocks, but it is difficult for novice users to quickly sort out the causal relationships.
Manus’s operation is very similar to that of a real stock broker. It first visits information websites such as Yahoo Finance through the API to obtain historical stock data. At the same time, it also cross-verifies the accuracy of the data to avoid being misled by a single information source and has a significant impact on the final results.
In this case, Manus also used the ability to write Python code, perform data analysis and visualization, and also introduced professional financial tools for analysis. Finally, through data visualization charts, combined with detailed comprehensive analysis reports, the user feedback on the causal relationship is really like daily work done by an “intern” in the financial field.
Not only that, Manus’s official website also displays more than a dozen scenarios that Manus can use: directly use Manus to help you organize your itinerary, customize travel routes, and allow it to learn and use various complex tools to complete daily tasks in a streamlined manner.
In this process, what really makes Manus different from usual tools is its autonomous planning to ensure the ability to perform tasks.
The ability to learn independently also makes Manus’s working ability improvement logic more like a real human being. Even at this stage, it may not be able to achieve expert-level mastery in a specific field, but it can already see huge potential.
With the addition of autonomous learning capabilities, the versatility of AI Agents has been greatly improved. In the actual test of Manus by users, you can even directly describe relevant content in a video screen to it. Manus can finally directly According to the corresponding information, it can overcome the limitations of platform content on search engines, and accurately find links to certain Douyin Short Video.
Since the current version of Manus runs completely asynchronously based on the cloud, Manus’s capabilities are not limited by factors such as the end-side platform form or computing power you use. Users can even temporarily shut down the computer after giving instructions to Manus. When Manus completes the activity results, you will be automatically notified of the results.
This operating logic is also very familiar, like a person calling an intern on WeChat after getting off work,”Organize the files and send them to me”. However, now, this intern can really respond to you 7 x 24 hours a day, and you don’t have to worry about him “rectifying the workplace.”
02 Multi-agent + self-inspection, running through the AI Agent flow
From the above cases, it is not difficult to see that Manus’s real trump card is not the concept of “AI Agent” that has appeared in Computer Use, but its ability to “simulate the human way of working.”
Rather than “running calculations,” Manus’s working logic is more like “thinking and executing orders.” It doesn’t do what humans really can’t do currently; that’s why some users who have experienced the current version of Manus describe it as “an intern.”
The Manus official website displays many tasks that Manus can complete, including a case that shows how to use Manus in a B2B business. Quickly and accurately match your order needs with global suppliers.
In conventional products with similar needs, it is a common logic in the industry to integrate global supply chain enterprise information within the platform to help users complete supplier/demand-side matching. But in Manus’s case, you can see a completely different way to achieve this.
Manus AI uses an architecture called “Multiple Agent” that runs in a separate virtual machine. Through the division of labor and cooperation mechanism of planning agent, execution agent, and verification agent,. To greatly improve the processing efficiency of complex tasks and shorten response time through parallel computing.
In this architecture, each agent may be based on a separate language model or reinforcement learning model and communicate with each other through APIs or message queues. At the same time, each task also runs in a sandbox to avoid interfering with other tasks and support cloud expansion. Each independent model can mimic the human process of processing tasks, such as thinking and planning, understanding complex instructions and breaking them down into executable steps, and then invoking the appropriate tools.
In other words, through Manus’s multi-agent architecture, it is more like having multiple assistants assist in completing tasks such as retrieving resources, docking, and verifying whether information is valid to help you complete the entire work process. In fact, it is not only like you have hired an “intern”, but more like you have directly become a miniature version of a “department director.”
In the case of B2B business, Manus uses web crawler and code writing and execution capabilities. Manus will automatically search in the vast ocean of the Internet, and based on your own needs, identify potential suppliers in terms of product quality, price, delivery capabilities, etc., to match you with the most suitable source of goods. Not only can the conclusion be visually presented in front of you in the form of a chart. Further detailed operational suggestions can be given on these data.
Manus fulfills the requirements in B2B scenarios and may be better than built-in tools on a single platform.
Photo source: Geek Park
As for how and what technology the Monica team uses to achieve the video effects, according to the news, the team may announce it for you on March 6, Beijing time.
03 The ultimate of “stitching” is bursting
What company is Monica.im behind Manus?
Monica is an All-in-One AI assistant. Its product form has slowly expanded from browser plug-ins to apps and webpages. The mainstream usage scenario is that when users click on its small icon in the browser, they can directly use the major mainstream models they access. Through an accurate understanding of user needs in segmented scenarios, Monica has picked the “low-hanging fruit” of the large model.
Its founder, Xiao Hong (nicknamed Xiaohong, Red in English), is a young serial entrepreneur born in 1992 and graduated from Huazhong University of Science and Technology. In 2015, he started a business after graduation. His early entrepreneurship was not smooth (such as campus social networking and second-hand markets). In 2016, he founded the operator of Weixin Official Accounts to provide editing and data analysis tools, gained millions of users and achieved profits. The final product was sold to a unicorn company in 2020.
After the big model wave in 2022, he officially founded Monica, focusing on overseas markets. Through the independent developer product ChatGPT for Google, the product quickly completed a cold start.
In 2024, Monica will allow users to obtain the latest SOTA models as soon as the GPT-4o, Claude 3.5, and OpenAI o1 series are launched. With the new progress of the access model, Monica’s professional search, DIY Bot, Artifacts writing Mini programs, memory and other functions have also been loved by users. Monica presents different interaction forms and functions in websites with different functions such as YouTube, Twitter, Gmail, and The Information to adapt to user needs in specific scenarios and update the personalized AI experience of hundreds of websites.
In 2024, the number of Monica users will double to 10 million. At the same time, it maintains considerable profits and ranks first among similar overseas products.
Monica’s strong performance confirms one thing:
To the extreme, it is both TPF and PMF, and ultimately leads to user value.
Monica Home Page
Photo source: Monica
Manus may have continued the Monica team’s thinking. Xiao Hong said in an interview with media person Zhang Xiaojun that products cannot have only one form of chat robots. Agents will be a new form and require new products to undertake them.
He got inspiration from the AI programming products cursor and Devin. According to Geek Park, the former is mainly copilot mode, while the latter is autopilot mode. The latter is more in line with human needs. Agents should also be like Devin, face the general public and truly be ai-led in execution. But the problem in the past was that models weren’t smart enough.
But the ability to encapsulate scenarios based on the model may be an advantage for the Monica team. Xiao Hong said that there are not many Agent product teams at present because it requires a lot of composite capabilities. For example, the team has to engage in chatbot, AI programming, browser related (because they all run on browsers), and have a good understanding of the boundaries of the model. Have a good perception of what level it has developed to today, what level it will develop to next, and so on.
“There are not so many companies with these capabilities at the same time, and companies with these capabilities may be doing a very specific business at hand, but we happen to have some classmates who have time to do this together.” he said.
Why did Monica do it? He concluded,”First of all, I think we are relatively lucky. Second, to a certain extent, if everyone goes to reasoning today, may it give more time for startups? How far can the model predict capacity spillover go?”
He believes that Agent is still in its early stages. First, the Agent is still in the planning stage and has not yet reached the execution stage in the physical world; second, the capabilities of the large model are still developing, and everything is unpredictable.
“I definitely didn’t know that Agents could be discovered in this way. It’s an unknown thing.” he said.
What is intriguing is that Monica, who “doesn’t know how to make an Agent”, has now made a product that makes the entire AI community feel explosive.
Manus may not be the ultimate AI Agent, but it has undoubtedly once again raised people’s expectations for AI by an order of magnitude after the explosion of DeeoSeek.
Welcome to join the official social community of Shenchao TechFlow
Telegram subscription group: www.gushiio.com/TechFlowDaily
Official Twitter account: www.gushiio.com/TechFlowPost
Twitter英文账号：https://www.gushiio.com/DeFlow_Intern
