China’s Manus AI Challenges US Tech With Autonomous Capabilities

China’s Manus AI is making waves in the AI industry, challenging US tech firms with its new general AI agent.

Developed by Wuhan-based startup Butterfly Effect, the tool claims to be the world’s first truly autonomous AI agent, capable of performing tasks like resume screening, itinerary planning, and stock analysis with minimal supervision, as reported by Bloomberg.

Manus launched a preview version last week, quickly gaining attention with a viral demonstration video. Co-founder and chief scientist Yichao Ji described it as “truly autonomous,” contrasting it with other AI agents that require more guidance, as reported by Bloomberg.

The tool has drawn comparisons to DeepSeek, another Chinese AI breakthrough that disrupted the market earlier this year by offering high-quality AI models at a fraction of the cost US competitors spend.

MIT notes that the hype surrounding Manus has led to a rush for access, with its Discord community surpassing 186,000 members. However, only a small fraction of users on the waitlist have received invite codes. Despite the excitement, initial reviews have been mixed.

Derya Unutmaz, a professor at the Jackson Laboratory, praised Manus for delivering “great quality outcomes” but noted that “it takes longer than OpenAI’s Deep Research to process the tasks,” reports Bloomberg

Others have pointed out performance issues, including slow response times, occasional crashes, and factual mistakes. “Manus is actually a half-finished product,” said Yiran Chen, an electrical and computing engineering professor at Duke University, as reported by Bloomberg. He suggested that the company is pushing the product early to attract investors.

Manus functions differently from chatbots like ChatGPT and DeepSeek, which are based on single large language models. Instead, it integrates multiple AI models, including Anthropic’s Claude 3.5 Sonnet and Alibaba’s fine-tuned Qwen, and employs independent AI agents that operate autonomously, as reported by MIT.

It also features a “Manus’s Computer” window, allowing users to monitor its progress and intervene when necessary.
MIT Technology Review tested Manus on tasks like compiling a list of tech reporters, finding real estate listings, and nominating candidates for an innovation award.

While the AI initially cut corners—admitting it got “lazy” when listing journalists—it improved with more feedback, producing a more comprehensive and accurate output.

Butterfly Effect, the company behind Manus, has raised over $10 million, according to Bloomberg, but it remains unclear how much of its AI is built from scratch versus refined from existing models. Unlike DeepSeek, Manus has not published technical papers or open-sourced its code.

Despite its limitations, some users believe Manus has the potential to redefine AI automation. “I haven’t seen anything like Manus,” said Ashutosh Shrivastava, a Bangalore-based developer who used it to build a website and a game, as reported by Bloomberg. However, its long-term success will depend on improvements in reliability and transparency.