The recently unveiled Manus, an “agentic” AI platform that entered preview last week, is creating as much excitement as a Taylor Swift concert.
Victor Mustar, head of product at Hugging Face, hailed Manus as “the most remarkable AI tool I have ever used.” Dean Ball, an AI policy researcher, referred to Manus as the “most advanced computing system leveraging AI.” The official Discord group for Manus amassed over 138,000 members within days, and reports indicate that invitation codes for Manus are being traded for thousands of dollars on the Chinese resale platform Xianyu.
However, it remains uncertain if the enthusiasm surrounding Manus is warranted.
— jack (@jack) March 9, 2025
Manus was not built entirely from the ground up. Social media reports indicate that the platform combines existing and fine-tuned AI models, including Anthropic’s Claude and Alibaba’s Qwen, to carry out tasks like drafting research papers and evaluating financial disclosures.
Nonetheless, on its site, Monica— the Chinese startup behind Manus— presents several ambitious examples of what the platform allegedly can achieve, including investing in real estate and developing video games.
In a popular video shared on X, Yichao “Peak” Ji, a research lead for Manus, suggested that the platform exceeded capabilities of agentic tools such as OpenAI’s deep research and Operator. Ji asserted that Manus scored better than deep research on a well-known benchmark for general AI assistants called GAIA, which evaluates an AI’s skill in executing tasks online, utilizing software, and more.
“[Manus] isn’t simply another chatbot or workflow tool,” Ji declared in the video. “It’s a wholly autonomous agent that connects thought and action… We envision it as the next evolution in human-machine collaboration.”
Nevertheless, some initial users are reporting that Manus is not a cure-all.
Alexander Doria, co-founder of AI startup Pleias, mentioned in a post on X that he faced error messages and infinite loops while experimenting with Manus. Other users on X pointed out that Manus provides erroneous answers to factual queries and does not consistently reference its sources—and frequently overlooks easily accessible information online.
Deep Research completed the task in less than 15 minutes. Unfortunately, Manus AI failed after 50 minutes at step 18 out of 20! 😑 It showed promise—I was monitoring Manus’ output and it appeared solid. However, re-running the same prompt was frustratingly slow!
— Derya Unutmaz, MD (@DeryaTR_) March 8, 2025
My own experience with Manus has not been overwhelmingly positive.
I tasked the platform with a seemingly simple request: to order a fried chicken sandwich from a well-rated fast-food restaurant nearby. After approximately ten minutes, Manus crashed. On the second attempt, it managed to find a menu option that aligned with my request, but it was unable to finalize the ordering process or even provide a checkout link.
Manus faltered similarly when I requested a flight booking from NYC to Japan. With clear instructions that seemed straightforward (e.g., “search for a business-class flight, focusing on price and flexible dates”), all Manus could deliver were links to fares from several airline websites and fare search engines like Kayak, many of which were not functional.
In an attempt to turn things around, I asked Manus to book a table for one at a nearby restaurant. After a few minutes, it gave up. I then requested that the platform create a Naruto-inspired fighting game, but it errored out after half an hour, prompting me to finally give up.
After using Manus AI for three days, here are my thoughts—both positive and negative.
Pros:
– The research it conducts online and the reports it generates are extraordinary.
– The ability to execute tasks through behind-the-scenes scripting is impressive.
– The plans it…— Ashutosh Shrivastava (@ai_for_success) March 9, 2025
We have contacted Monica for a response and will update this article if we receive one.
So, if Manus isn’t delivering on its technical promises, why did it gain such rapid popularity? Several reasons played a role, including the exclusivity resulting from a limited number of invites.
Chinese media quickly hailed Manus as a groundbreaking AI; QQ News referred to it as “the pride of domestically developed products.” Meanwhile, AI influencers on social media contributed to misinformation regarding Manus’ features. A widely circulated video depicted a desktop program, purportedly Manus, taking action within multiple smartphone applications. Ji confirmed that this video was not an actual demonstration of Manus.
Furthermore, other influential AI accounts on X tried to draw parallels between Manus and the Chinese AI firm DeepSeek — comparisons that are not necessarily based on truth. Unlike DeepSeek, Monica did not create in-house models and while DeepSeek made many of its technologies publicly accessible, this is not the case with Monica—at least not yet.
To be fair to Monica, Manus is still in its early access phase. The company asserts that it is working to enhance computing power and resolve issues as they arise. However, at this stage, Manus appears to be a situation where the hype exceeds the actual technological advancements.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
