- New agentic AI tool Manus is being hailed as more capable than previously available agents
- Others are less impressed, citing crashes and hallucinations, while experts call for caution
- A use case gallery shows Manus being used to plan trips and analyse stocks
Manus, a new ‘agentic’ AI tool from Chinese startup Monica, has entered into invite-only preview, and, according to some users, “[redefines] what’s possible” for automated AI workflows to achieve.
That’s according to one Twitter user who asked Manus to ‘code a threejs game where you control a plane’ and “ended up concluding that Manus is the most impressive AI tool I’ve ever tried”.
Manus’ own use case gallery (via TechCrunch) lays out some of the possibilities, such as planning a trip to and producing a personalised guidebook for a trip to Japan, accessing a database to sort information from it into tables, analyzing stocks in a human-readable dashboard, and scheduling job candidate interviews without intervention from a real person.
What are AI agents?
AI ‘agents’, though powered by large language models (LLMs), differ from AI writers in that they utilise systems to automate a workflow based on a user’s prompt, rather than produce a single piece of output.
Manus isn’t offering anything especially new, though it is notable that, up until now, perhaps, AI agents haven’t had the best reputation; known for struggling with factual information, and basically being non-functional.
Others, however, aren’t as impressed. TechCrunch’s Kyle Wiggers gave a cohesive account of their experience, reporting crashes, and failures on tasks ranging from simple (booking a table at a restaurant) to the convoluted (“Then I asked the platform to build a Naruto-inspired fighting game)”.
Comparisons have also been made with DeepSeek’s LLMs, though these don’t quite add up. Manus and DeepSeek might both hail from China, but DeepSeek is not an agent, and Manus is not open source, and indeed has no release schedule. That said, an open source version, OpenManus, is in active development.
Manus coming from a Chinese startup has caused unease, with some circles, Axios notes, worrying China is gaining ground in a so-called “race for AI supremacy”.
Experts doubt Manus
Two industry figures told TechRadar Pro their doubts surrounding agentic AI still remain. Corpora.ai’s CEO, Mel Morris, said that “Manus’ demonstrations showcase the potential of Agentic AI as a significant step forward. However, granting autonomous AI agents like Manus the ability to perform independent actions raises serious concerns.”
“We have already witnessed numerous examples of generative AI and so-called reasoning models producing flawed outputs. If given autonomy over high-stakes tasks – such as buying and selling stocks – such imperfections could lead to chaos. The ability to compare stock fundamentals and trading activity in minutes is vastly different from successfully navigating the complexities of the market.”
Similarly, Copyleaks co-founder and CEO Alon Yamin said that ,“while innovation in AI can drive remarkable efficiency, it also amplifies concerns around transparency, data security, and responsible oversight. Manus’ cloud-based operations surface several critical questions about data governance: Who controls the information? Where does it reside? How is the data protected?”
“As AI systems become more autonomous, ethical development and regulatory alignment must be prioritized. AI governance solutions that help organizations detect and mitigate AI risks remain critical, ensuring advancements like Manus AI are deployed responsibly and transparently.”