Manus AI is breaking records, but also breaking itself
Matt V
March 11, 2025
MANUS AI
Manus, an “agentic” AI platform, launched in preview last week and quickly became a viral sensation.
Its invite-only access, social media buzz, and bold claims from early users drove massive interest.
Some compared it to OpenAI’s advanced research tools, and a research lead even called it the next step in AI evolution.
Manus isn’t built from scratch; it uses existing AI models like Anthropic’s Claude and Alibaba’s Qwen.
The company behind it, The Butterfly Effects, says it can handle complex tasks like drafting reports, analysing financial data, and even buying real estate.
Early tests showed it outperformed OpenAI’s Deep Research on some AI benchmarks.
Despite the excitement, real-world testing has revealed some major issues:
Users reported crashes, errors, and delays in execution.
It struggled with simple tasks like booking flights or ordering food, often failing to complete them.
Some results were incorrect, missing key details, or lacking proper citations.
A Manus spokesperson acknowledged these issues, stating that the beta phase is meant to identify and fix problems before a full release.
Why the Hype?
Several factors contributed to Manus’ rapid rise. Limited invites created demand, with some selling for thousands of dollars.
Chinese media portrayed it as a major AI breakthrough, adding to the hype further. This, along with some influencers sharing misleading content, such as a fake demo video, created huge amounts of interest.
Not all exclusivity is a flex
While Manus may improve, it currently struggles to meet expectations.
The Butterfly Effect says updates are on the way, but for now, the technology isn’t quite as advanced as the hype suggests.
Manus is giving “straight-A student who panics during the test” energy.