Manus is Unlikely to Be China’s Next DeepSeek Moment
Hugging Face’s head of product called Manus “the most impressive AI tool I’ve ever tried,” while AI policy researcher Dean Ball described it as “the most sophisticated computer using AI.” Meanwhile, its official Discord server quickly surpassed 138,000 members, and invite codes are reportedly selling for thousands of dollars on the Chinese reseller app Xianyu.
However, whether the excitement is warranted remains uncertain.
Manus wasn’t built entirely from scratch. Reports on social media suggest that the platform relies on a mix of existing and fine-tuned AI models, including Anthropic’s Claude and Alibaba’s Qwen, to handle tasks like drafting research reports and analyzing financial filings.
Despite this, The Butterfly Effect—the Chinese company behind Manus—claims on its website that the platform can achieve far more ambitious feats, such as purchasing real estate and developing video games.
Manus Touted as Superior to Other AI Agents, but Benchmark Claims Face Scrutiny
In a widely shared video on X, Yichao “Peak” Ji, a research lead for Manus, suggested that the platform surpasses agentic tools like OpenAI’s deep research and Operator. Ji claimed that Manus outperforms deep research on GAIA, a benchmark assessing an AI’s ability to navigate the web, use software, and complete complex tasks.
“[Manus] isn’t just another chatbot or workflow,” Ji said in the video. “It’s a fully autonomous agent that bridges the gap between conception and execution […] We see it as the next paradigm of human-machine collaboration.”
However, early users report that Manus has its flaws.
Alexander Doria, co-founder of AI startup Pleias, shared on X that he encountered error messages and infinite loops while testing the platform. Other users noted that Manus struggles with factual accuracy, often fails to cite sources, and sometimes overlooks easily accessible information.
My experience with Manus has been less than impressive.
Manus Struggles with Simple Tasks, Failing to Complete a Food Order
I gave it what seemed like a simple task: ordering a fried chicken sandwich from a highly rated fast-food restaurant within my delivery area. After about ten minutes, the platform crashed. On my second attempt, Manus successfully identified a menu item that matched my request but failed to complete the order—or even generate a checkout link.
Manus also struggled when I asked it to book a flight from NYC to Japan. Despite providing clear instructions—such as prioritizing price and flexible dates for a business-class ticket—the platform only managed to generate links to various airline websites and airfare search engines like Kayak, some of which were broken.
Hoping for better results, I asked Manus to book a table for one at a nearby restaurant. After a few minutes, it failed. I then requested it to create a Naruto-inspired fighting game, but the platform encountered an error after half an hour—at which point, I decided to give up.
A Manus spokesperson provided TechCrunch with the following statement via direct message:
“As a small team, our priority is to continuously improve Manus and develop AI agents that genuinely assist users in solving problems […] The primary goal of the current closed beta is to stress-test different aspects of the system and identify issues. We greatly appreciate the valuable feedback from everyone.”
So, if Manus isn’t quite living up to its technical claims, why has it gained so much attention? Several factors contributed, including the exclusivity created by limited invite availability.
Hype Outpaces Reality as Chinese Media and AI Influencers Exaggerate Manus’ Capabilities
Chinese media quickly labeled Manus as a major AI breakthrough, with QQ News calling it “the pride of domestic products.” Meanwhile, AI influencers on social media spread exaggerated claims about its capabilities. One widely shared video appeared to show Manus operating multiple smartphone apps, but Ji later confirmed that the clip was not an actual demonstration of the platform.
Additionally, some prominent AI accounts on X compared Manus to Chinese AI company DeepSeek, though the comparison isn’t entirely accurate. Unlike DeepSeek, The Butterfly Effect hasn’t developed any proprietary AI models. DeepSeek also made many of its technologies publicly available, whereas Manus has yet to do the same.
To be fair, Manus is still in early access. The company states that it is working on scaling its computing capacity and addressing reported issues. However, at this stage, the platform seems to be a case where hype has outpaced actual technological progress.
Updated 6:02 p.m. Pacific: Added a statement from a Manus spokesperson and corrected the misidentification of the company behind Manus.
Read the original article on: TechCrunch
Leave a Reply