Chinese startup The Butterfly Effect launched its Manus AI platform to significant fanfare last week, attracting over 138,000 Discord members and commanding black market invites worth thousands of dollars on Chinese reseller app Xianyu. Despite bold claims of autonomous capabilities, early testing reveals functionality gaps, perhaps understandablen given the launch is in beta.
The Manus platform represents a growing category of "agentic" AI tools designed to autonomously execute complex tasks across multiple systems. Unlike competitors developing proprietary models, Manus leverages existing AI foundation models including Anthropic's Claude and Alibaba's Qwen, then applies fine-tuning and integration layers to enable more complex task execution.
Company spokespeople and research lead Yichao "Peak" Ji positioned Manus as a revolutionary advancement, claiming it "bridges the gap between conception and execution" and represents "the next paradigm of human-machine collaboration." Ji specifically highlighted benchmark results suggesting Manus outperforms OpenAI's deep research and Operator tools on the GAIA benchmark for general AI assistants.
TechCrunch's hands-on evaluation confirmed limitations across multiple use cases. According to their testing, basic business productivity tasks like restaurant reservations, travel booking, and online ordering consistently failed to execute completely. Despite providing relatively simple parameters, TechCrunch reported that the platform regularly encountered errors, crashed mid-execution, or provided broken links rather than completing transactions.
The Manus experience highlights gaps between theoretical capabilities and practical enterprise deployment, particularly for transaction processing, workflow integration, and reliable execution of multi-step business processes. BUT this a Beta version, if the kinks are ironed out before launch a whole new AI world comes into play.