Manus: Chinese team releases world's first universal AI Agent to blow up tech scene

In the early hours of March 6, it was another sleepless night in the tech world after DeepSeek, when everyone was distracted by a new program calledManusThe product of the screen. the AI circle is boiling, AI intelligent body plate soared.ManusThe release of the company has once again shaken the world with China's AI technology.

According to its team, Manus is a truly autonomous AI agent capable of solving all kinds of complex and varied tasks. Unlike traditional AI assistants, Manus not only provides suggestions or answers, but also delivers complete task results directly.

Manus comes from the Latin Mens et Manus, which means mind and hand.

Manus uses a multisignature (multisig) system driven by multiple independent models. Later this year, there are plans to open-source some of these models, particularly the reasoning part of Manus.

Video Demo Showcase

Synchronized with Manus out of the ring is a four-minute long demo demo. in these cases, Manus completes the entire process from planning to execution completely autonomously, demonstrating true Agent capabilities rather than simple assistant functions.

In the video, Ji Yichao shows three demos about Manus performing tasks.

The first task was to help us screen resumes.

Upon receiving a Zip archive containing 10 resumes, Manus acts like a human being by unzipping the file and then going through each resume page by page, noting important information in the file.

It's worth noting that Manus works asynchronously in the cloud, which means you can turn off your computer at any time and Manus will notify you when everything is done.

Also, you can give it new instructions at any time during the task. During the demo, the staff uploaded five more resumes.

After poring over 15 resumes, Manus provided ranking recommendations, candidate biographies and evaluation criteria as supporting materials, as well as the ability to create forms upon request.

And even more conveniently, Manus has its own knowledge and memory that you can teach it to "deliver the spreadsheet immediately the next time it tackles a similar task."

The second case is very much for Chinese tastes, helping us to select properties.

In this case, the user wants to buy a property in New York and requires to meet the requirements of a safe neighborhood environment, a low crime rate, and quality primary and secondary education resources, while still having an affordable monthly income.

After receiving task instructions, Manus AI breaks down complex tasks into to-do lists, matches them one by one, and even writes itself to look at a Python program that calculates affordable real estate budgets based on a user's income.

Finally, Manus consolidates all the information gathered and writes a detailed report that includes a neighborhood safety analysis, school quality assessment, budget analysis, a list of recommended properties, and links to related resources, with less hassle than a professional real estate agent, and with Manus based on the interests of the user.

In the last case, Manus demonstrates the ability to analyze stock prices.

Users asked to analyze the correlation between the stock prices of NVIDIA, Maywell Technology and TSMC over the past three years, and Manus similarly wrote its own Python code to simultaneously analyze the data, while also bringing in specialized tools related to finance for the analysis.

Ultimately, through the data visualization charts, with detailed comprehensive analysis report way, to the user feedback on the cause and effect relationship. The whole process is very much like human behavior, and more professional and meticulous.

The Manus website also provides a number of Manus application cases, covering a wide range of scenarios, from professional research to daily life, data analysis, and education,Productivity toolsand creative entertainment, among many others.

Officially, what was shown is still just the tip of the iceberg of Manus' capabilities.

(coll.) fail (a student)AI AgentBy invoking a long chain of thought and tool calls, and ultimately outputting a complete, professional result, users are beginning to marvel at the fact that AI really can do things for humans.

To ensure reproducible results, Manus was evaluated using configurations identical to its official version.

According to the official website, Manus achieved new state-of-the-art (SOTA) performance in all three difficulty levels in the GAIA benchmark, which evaluates the ability of general-purpose AI assistants to solve real-world problems.

What is Monica.im, the company behind Manus?

Monica is an All-in-One AI assistant, and its product form is slowly expanding from a browser plug-in to an app and a web page. The mainstream use scenario is that when users click on its small icon in the browser, they can directly use its access to the major mainstream models. Through the accurate understanding of the user needs of the subdivided scene, Monica has picked the "low-hanging fruit" of the big models.

Its founder, Xiao Hong (nicknamed Red), is a young serial entrepreneur, born in 1992, graduated from Huazhong University of Science and Technology. 2015, he graduated and started his own business, and his early ventures didn't go well (e.g., campus socialization, second-hand bazaar). 2016, he started a business to provide editing and data analytics tools for WeChat's public number operators, which gained millions of users and became profitable, and the final product was sold to a unicorn company in 2020. The final product was sold to a unicorn company in 2020.

Waiting for the big model wave of 2022, he officially founded Monica to focus on overseas markets through the independent developer product ChatGPT for Google Internet companyThe product has completed a quick cold start.

In 2024, at the first time GPT-4o, Claude 3.5, and the OpenAI o1 series went live, Monica gave users access to the latest SOTA models. Along with new advances in accessing models, Monica introduced features such as Professional Search, DIY Bot, Artifacts to write applets, and Memory, which are also loved by users. And Monica presents different interaction forms and functions in web pages with different functions such as YouTube, Twitter, Gmail, The Information, etc. to adapt to user needs in specific scenarios, updating hundreds of web pages with personalized AI experiences.

In 2024, Monica doubles its user base to 10 million. In the meantime, it maintains a sizable profitability and ranks at the head of its category overseas.

Monica's strong performance validates one thing:Shelling to the extreme is both TPF and PMF, and in the end it all leads to user value.

Manus' technology philosophy of "less structure more intelligence" is a bit different from the mainstream. They believe that when the data is good enough, the model is powerful enough, the architecture is flexible enough, and the engineering is solid enough, computer use, deep research, coding agent and other capabilities will emerge naturally without being designed into specific product functions.

Manus may not necessarily be the ultimate AI Agent, but it has certainly raised expectations for AI by another order of magnitude after the DeeoSeek explosion.

At present, almost all the AI track technology companies are staring at the "AI Agent", March 6, OpenAI announced that it will charge $20,000 (about 145,000 yuan) per month for AI Agent who has reached the level of doctoral degree, which is mainly oriented to the high-end needs of enterprise users, especially in the financial, medical, manufacturing and other data-intensive industries. data-intensive industries.