China may be experiencing a second deep seek moment after sending a shockwave to the global tech stock market in February. A new AI agent called Manus is making waves in China, but it may not be long before it reaches the US
Manus, the new Chinese AI startup, is built by the AI Manus takes action rather than answering questions. Unlike the typical AI assistants who spit out information, Manus completes the actual work.
Just two days after its release, Manus is already praising social media. “It’s like a deep search + operator + Claude computer combination, and it’s really good,” writes Rowan Cheung, founder of the popular AI newsletter The Rundown AI, in X. ”
China’s next deep seeking moment? Meet Manus, an AI agent who is a professional multitasking
Manus, launched this week by Chinese AI startup Manus AI, Manus is a common AI agent that connects thoughts and actions. Handle tasks both in work and personal life and take care of everything during breaks.
The launch of Manus AI comes days after ChatGpt creator Openai announced plans to charge $20,000 a month for a specialized AI “agent.” Openai is quietly developing premium products, but Manus AI has entered the scene with early benchmark testing suggesting it is just a small part of the cost.
Gaia is a benchmark designed to evaluate popular AI assistants on real-world problem solving, showing the Manus achieving new cutting edge (SOTA) performance across all three difficulty levels.
Early benchmarks show manus that outperforms other AI assistants in common problem-solving tasks. But beyond the test scores, he has proven himself on platforms such as Upwork, Fiverr and Kaggle competitions.
More than just a chatbot
Manus is said to be a combination of deep research, operators, and Claude computers, bringing automation and execution to one system. It works independently in the cloud. This means that users can assign tasks, leave, or return to the completed work.
During the YouTube demo, co-founder Yichao “Peak” Ji put Manus at his pace with three different challenges. The results were impressive.
Autopilot resume screening
In the first test I had a resume of Manus Screening 15. I started by unzipping the file with 10 resumes, passing through each document, pulling out important details, and ranking candidates ranked based on employment criteria.
Manus adapted when Peak requested a spreadsheet instead of a standard document summary. Next time, the priority format will be automatically distributed.
Real estate research without busy work
Manus then handled the real estate research task. Instead of doing a simple search, I created a step-by-step plan and ran each part.
Finding information about safe neighborhoods, a middle school wrote a Python program that wrote a Python program to calculate budget filtered lists based on price ranges that compile the full report
This was not just a search, it was an organized workflow that handled everything.
Inventory data analysis and website development
In the final demonstration, Manus worked on stock market correlation analysis. that:
Information cited financial data from trusted APIs is the information written with custom code for data visualization.
Then came the curveball. We installed the analysis on an interactive website. Manus built and deployed the site on its own, providing easy-to-access live links.
What’s next for Manus AI?
Manus AI is not completed. The company said it plans to open source several models later this year, allowing developers to experiment and build on the technology.
The name Manus comes from Men’s Et Manus. Latin is for the “heart and hands.” The idea is simple. Knowledge is useful to apply.
The Chinese AI scene is moving rapidly, and Manus AI is the latest evidence. Its ability to perform complex tasks makes it possible to be one of the most practical AI agents ever built. And if it continues to gain traction, it will not remain a phenomenon exclusively in China for a long time.
Source link
