Manus, an AI agent developed by Chinese startup Butterfly Effect, has generated significant buzz in the tech world. Described as “the first general AI agent” capable of autonomously executing complex tasks, Manus has been hailed by some as China’s second “DeepSeek moment” – referring to the earlier breakthrough when Chinese AI model DeepSeek R1 outperformed Western competitors.
What is Manus?
Manus operates as a multi-agent system that divides complex tasks into manageable components and assigns them to specialized sub-agents. According to its developers, it can research, analyze data, generate reports, automate workflows, and even write and deploy code without continuous human supervision.
The system was developed by Butterfly Effect, a Chinese company with offices in Beijing and Wuhan. The team is led by Xiao Hong, a 33-year-old entrepreneur and graduate of Wuhan’s Huazhong University of Science and Technology. Yichao “Peak” Ji (photo above) serves as co-founder and chief scientist.
Unlike some competitors, Manus wasn’t built entirely from scratch. According to reports on social media, the platform uses a combination of existing AI models, including Anthropic’s Claude 3.5 Sonnet and fine-tuned versions of Alibaba’s Qwen models. The team is reportedly testing an upgrade to Anthropic’s newer Claude 3.7 model.
Mixed reviews from early users
Early testers have reported varied experiences with the system:
- Rowan Cheung, founder of The Rundown AI newsletter, claimed Manus successfully created and deployed a biography website about him with “100 percent accuracy” and developed a full AI course generating eight chapters of content.
- Former Googler Bilawal Sidhu called Manus “the closest thing I have seen to an autonomous AI agent” in a YouTube review, praising its ability to research locations and extract industry insights.
- Andrew Wilkinson, co-founder of startup acquirer Tiny, wrote that Manus analyzed 20 CEO job applicants and provided detailed assessments of each.
However, not all feedback has been positive:
- Alexander Doria, co-founder of AI startup Pleias, reported encountering error messages and endless loops during testing.
- TechCrunch writer Kyle Wiggers described multiple failures when asking Manus to perform seemingly straightforward tasks like ordering food delivery or booking restaurant reservations.
- Some users noted that Manus makes factual errors and doesn’t consistently cite its sources.
Limited availability fuels speculation
Access to Manus remains highly restricted through an invitation-only system. Zhang Tao, Manus AI’s product partner, admitted in a social media post that the team “underestimated the enthusiasm of the public response” and had only planned server resources for demonstration purposes.
This scarcity has created a secondary market, with invitation codes reportedly selling for thousands of dollars on Chinese reseller app Xianyu. The exclusivity has also fueled speculation about whether Butterfly Effect is using scarcity marketing tactics to generate hype.
How does it compare to Western competitors?
Manus is entering a competitive field of AI agents, including OpenAI’s Deep Research and Operator features, Google’s AI tools, and Anthropic’s Computer Use mode. All of these systems aim to take control of a user’s computer or programs, moving cursors and typing to perform actions within software.
Butterfly Effect claims Manus outperforms OpenAI’s Deep Research on the GAIA benchmark for general AI assistants, which tests an AI’s ability to carry out work by browsing the web and using software. However, independent verification of these claims remains limited due to restricted access.
Unlike DeepSeek, which developed its own foundation models and made many of its technologies openly available, Butterfly Effect relies on existing models and hasn’t yet released its technology publicly. The company has stated that it plans to release some of its models to the public later this year.
A spokesperson for Manus told TechCrunch: “As a small team, our focus is to keep improving Manus and make AI agents that actually help users solve problems […] The primary goal of the current closed beta is to stress-test various parts of the system and identify issues.”
While Manus has generated extraordinary interest, particularly within Chinese media and among AI influencers, the limited access and mixed early reviews suggest that the current reality may not fully match the hype surrounding this emerging AI agent.
Sources: Forbes, VentureBeat, TechCrunch