Many leading US AI developers are racing to develop AI Agents that can perform complex tasks on behalf of users, but have not succeeded. However, a small Chinese startup claims to have done this.

Manus AI last week released a preview of its general-purpose AI Agent tool, which can scan candidate profiles, plan travel itineraries, and analyze stocks when users provide basic instructions. The company claims its service outperforms OpenAI’s DeepResearch in some respects.

manus vcg
Manus AI once again raises questions about the US's position on the AI ​​front. Photo: VCG

While some AI agents still require some human supervision, Manus co-founder and chief scientist Yichao Ji said the AI ​​is “truly autonomous.” The demo video quickly went viral, leading to a scramble for trial codes for the tool.

The initial “fever” put Manus on the scale to compare with DeepSeek, a fellow startup that surprised Silicon Valley with its low-cost, high-performance AI model in January.

Like DeepSeek, Manus once again questions America’s leadership in AI, this time in a category that American tech companies see as a key investment area.

But according to Bloomberg , the reactions of those who have had the chance to try Manus are not the same. Derya Unutmaz, a professor at Jackson Research, praised the tool for producing “high-quality results” even though it “took longer than DeepResearch to do the job.”

Others complained that the service was too slow and sometimes crashed before completing a task, likely due to Manus’s limited computing resources. Some users also found Manus to be riddled with errors.

Yiran Chen, a professor of electrical and computer engineering at Duke University, calls Manus an unfinished product. He believes Manus wants to gain a first-mover advantage to attract investment, despite the fact that the product is not yet fully developed.

There are still many questions surrounding Manus and its product, such as whether Manus builds its own system or builds on top of other developers’ models. The first approach could cost tens of millions of dollars or more.

According to Chinese media, Butterfy Effect – the company behind Manus – has raised more than $10 million in investment.

Unlike DeepSeek and some US companies, Manus does not publish detailed reports at product launch to provide information about technology development.

Those who want to try Manus have two options: standard mode or high-effort mode, the latter of which requires more request processing time.

Yong Qian, founder of OmniEdge, said Manus “creates a to-do list, goes through each step, fixes errors along the way, and asks questions if new commands need to be added” before providing feedback.

In recent months, OpenAI, Anthropic, and other AI companies have launched AI Agents that can use a user's computer to browse the web, search online, and do multi-step tasks.

Still, some Manus users say the tool is superior to what's already on the market.

“I have tried hundreds of AI tools. I try a new one every day,” said Ashutosh Shrivastava, a software developer in Bangalore, India. “ I have never seen anything like Manus.”

(According to Bloomberg)

Manus AI, the 'existential crisis' of workers Manus AI – the world's first autonomous AI agent developed by China – is shaking up the global tech world, raising legal and ethical issues.