The launch of Manus AI by Chinese AI startup Monica has created a buzz similar to the shockwaves caused by the rise of deepfake technology. However, users are now questioning whether its capabilities have been exaggerated.
Monica has marketed Manus AI as a general-purpose AI agent capable of autonomous decision-making and actions, positioning it as a step beyond traditional chatbots.
However, beta testers with access codes have reported that while Manus’s autonomous coding abilities are impressive, its performance as a virtual assistant is limited and frequently prone to errors.
On March 5, Monica announced the launch of Manus, claiming it had achieved state-of-the-art (SOTA) performance in the GAIA test, a benchmark used to evaluate artificial general intelligence (AGI). Monica further stated that Manus outperformed OpenAI’s advanced model, Deep Research.
In a promotional video, Manus demonstrated its capabilities in various tasks, including planning a trip to Japan, analyzing Tesla’s stock price, reviewing Nvidia’s financial statements, guiding real estate purchases in New York, generating property analysis reports, and screening and categorizing resumes.
This has led Forbes to dub it the “second Deepfake moment.”
However, beta testers are skeptical about Manus being a fully developed AI agent.
They have described it as a hybrid of OpenAI’s Operator, Anthropic’s YouComputer, and Alibaba’s Q1, but with frequent errors and repeated tasks.
TechCrunch published an article titled “Manus is Not the Second Deepfake Moment,” highlighting user reports that highlight Manus’s limitations and frequent basic mistakes.
The publication also noted that Manus does not use Monica’s proprietary AI technology, and its technical details have not been disclosed.
Monica’s product manager, Zhang Tao, acknowledged that the current version of Manus is still in its early stages and does not yet meet the intended capabilities of the final product.
He explained that the beta release was meant to share initial results and gather feedback on potential issues. Monica plans to gradually expand access, prioritizing user experience as it works through the issues.
Manus sparked additional skepticism shortly after its release when access was restricted to invite-only beta codes.
Despite a surge of interest, only a small number of codes were distributed, and at one point, they were reportedly sold for more than 50,000 yuan (approximately $6,890). This led to accusations of scarcity marketing.
In response, Zhang Tao explained that the servers were set up to match the demo’s scale, resulting in limited capacity. He added that the invite-only system during the testing phase was unavoidable and that they had not anticipated such high demand.