Technology

Manus: The New AI Platform Generating Buzz

Published March 10, 2025

Manus, an "agentic" AI platform, launched in preview last week and is creating significant excitement, reminiscent of a popular concert.

The head of product at Hugging Face labeled Manus as "the most impressive AI tool I’ve ever tried." AI policy researcher Dean Ball referred to it as the "most sophisticated computer using AI." Moreover, Manus's official Discord server amassed over 138,000 members within days, with invite codes reportedly reselling for thousands on a popular Chinese app.

Despite the vast excitement, questions arise about whether the hype is warranted.

Manus was not created entirely from the ground up. Reports indicate that the platform leverages a combination of existing and refined AI models, including Anthropic's Claude and Alibaba's Qwen, to execute tasks like drafting research reports and analyzing financial documents.

On its website, the startup behind Manus, Monica, showcases some ambitious claims about the platform's capabilities, such as purchasing real estate and developing video games. Additionally, in a widely viewed video, research lead Yichao "Peak" Ji suggested that Manus surpasses other tools like OpenAI’s deep research and Operator. Ji asserted that Manus performed exceptionally well on the GAIA benchmark, which evaluates an AI's ability to complete tasks by navigating the web and utilizing software.

Manus isn’t just another chatbot or workflow,” Ji mentioned in the video. “It’s a fully autonomous agent that bridges the gap between conception and execution We believe it represents a new paradigm in human-machine collaboration.”

However, initial users have reported challenges with Manus.

Alexander Doria, co-founder of the AI startup Pleias, shared on social media that he faced error messages and loops during his tests with Manus. Other users noted that the platform struggles with factual inquiries and often fails to provide accurate references.

"Deep Research completed in just 15 minutes. Unfortunately, Manus AI failed beyond 50 minutes at step 18 of 20! While it appeared to be performing well, running the same prompt a second time was frustrating due to the lengthy time it took," tweeted Derya Unutmaz, MD.

My own experience with Manus was not entirely positive either.

I attempted to use the platform for a straightforward task: ordering a fried chicken sandwich from a well-rated local fast-food restaurant. After about ten minutes, Manus crashed. The second attempt yielded a workable menu item, but Manus failed to complete the order or even provide a checkout link.

When I requested a flight booking from NYC to Japan, defining parameters like "seek a business class ticket, prioritizing price and flexible dates," the best Manus managed was to direct me to various airline websites and search engines, some of which were broken.

Next, I tasked Manus with reserving a table at a nearby restaurant, which it failed to accomplish after several minutes. I even tried to prompt Manus to create a Naruto-themed fighting game. Unfortunately, it encountered errors after a prolonged period, prompting me to give up.

Overall, after testing Manus for several days, here’s my assessment:

Pros:

  • The research functionality and report generation are impressive.
  • It can execute backend scripts to perform tasks.

We’ve reached out to Monica for comments and will update with any responses.

But if Manus isn't meeting its technical ambitions, what led to its rise in popularity? Factors such as invite scarcity contributed to a sense of exclusivity.

Chinese media quickly spotlighted Manus as an AI innovation, with QQ News dubbing it "the pride of domestic products." Some AI influencers on social media propagated exaggerated claims about Manus’s abilities. Notably, a viral video purporting to show Manus’s functionalities was, in fact, misidentified, as confirmed by Ji.

Additionally, several AI accounts attempted to equate Manus with the Chinese AI company DeepSeek, despite the lack of factual basis. Unlike DeepSeek, which has developed significant in-house models, Monica's approach does not seem to follow the same trajectory. Furthermore, DeepSeek has made many of its technologies openly accessible, which Monica has yet to do.

It’s essential to note that Manus is still in early access. The company asserts it is working on scaling capabilities and rectifying instances as they arise. However, as of now, Manus appears to embody a case where hype is outpacing technological advancement.

AI, Manus, Technology