Manus AI Agent shockingly released: Chinese team leads the general intelligent agent revolution

68 0 0

On March 6, 2025, China’s AI field ushered in a milestone event – the world’s first general AI Agent product “Manus” created by the team of Xiao Hong, an alumnus of Huazhong University of Science and Technology, was officially released. This product surpassed OpenAI’s similar models with the SOTA performance of the GAIA benchmark test, and caused industry shocks in terms of technical architecture, scenario application and commercial attention. Even the second-hand invitation code was once sold at a sky-high price of 100,000 yuan. Its breakthrough significance lies not only in technical performance, but also in redefining the paradigm of human-machine collaboration.

Core capabilities: the revolution from “answering questions” to “delivering results”

The core value of Manus lies in its full-process autonomous execution capability. Unlike traditional conversational AI, Manus can independently complete the planning, disassembly and execution of complex tasks, such as:

Resume screening: automatically decompress files, analyze candidate qualifications, generate Excel tables with rankings, and optimize subsequent task forms according to user preferences;
Stock analysis: crawl data from companies such as NVIDIA and TSMC, write Python code for visual analysis and deploy shareable websites;
Real estate research: combine budget, security and educational resources to generate New York home purchase reports, and even call map APIs to verify community information.

Its asynchronous processing capabilities allow users to continue to execute tasks through cloud virtual machines after turning off the device, greatly improving efficiency.

In the authoritative GAIA benchmark test, Manus showed a layered breakthrough: the accuracy of basic tasks reached **86.5%** (close to 92% of humans), intermediate tasks 70.1%, and advanced tasks 57.7%, especially in multi-step reasoning and cross-tool calls far exceeding similar products. For example, in Kaggle competition-level data analysis tasks, Manus’s results are as complete as 98.2%.

Application scenarios: “Productivity reshaping” from enterprises to individuals

The explosive popularity of Manus stems from its cross-domain adaptability:

Enterprise services: replace 80% of resume screening work, automatically generate financial reports and contract review results, and reduce labor costs by 45%;
Financial investment: real-time capture of stock data and generation of interactive charts to provide decision support for institutions;
Educational innovation: convert papers into popular science PPTs, or customize HTML animation courseware for teachers;
Life management: plan travel routes, generate health advice, and even analyze physical examination reports.

The market’s expectations for Manus have shifted from technical verification to commercialization. On the day of release, A-share AI concept stocks collectively hit the daily limit, and the capital market regards 2025 as the first year of AI Agent commercialization.

Technical architecture: multi-signature system and open source ecology

Manus’s technological breakthroughs stem from three major innovations:

Multi-signature system (Multisig): Through multi-model collaborative decision-making (such as Claude 3.5, DeepSeek, etc.), dynamically schedule task processes, and achieve a 3-fold increase in parallel reasoning capabilities;

Hierarchical agent architecture: plan agent disassembly tasks, execute agent call tools (Python/browser/API), verify agent cross-verification results, and simulate the human “thinking-execution-review” link;

Open source strategy: plans to open source reasoning modules within the year to attract developers to build vertical solutions such as medical and financial services, and promote the formation of intelligent body ecological standards similar to Kubernetes.

Controversy and future: Has the “critical point” of general AI arrived?
Although Manus is regarded as the dawn of AGI (general artificial intelligence), controversy still exists:

Technical originality: It is accused of relying on large model API scheduling rather than underlying breakthroughs, and its architecture is similar to Anthropic “Computer Use”;

Business risks: If giants such as OpenAI internalize similar functions, the value of the middle layer may be weakened;

Ethics and costs: High-level tasks still require human intervention in moral judgment, and a single complex analysis costs hundreds of dollars in computing power.

However, the founder Xiao Hong’s continuous entrepreneurial genes (he once created the tens of millions of products Monica) and open source strategy have injected Manus with ecological expansion potential. As Xiao Hong said: “The shell has its uses” – technology integration can also create user value. If Manus can verify the irreplaceability of vertical scenarios (such as cross-border tax automation) in the next 6-12 months, it may truly usher in a new era of human-machine collaboration.

This AI Agent revolution ignited by Manus is not only a technological leap, but also a subversion of the productivity paradigm.