Manus, an “agentic” AI platform that launched in preview final week, is producing extra hype than a Taylor Swift live performance.
The head of product at Hugging Face known as Manus “the most impressive AI tool I’ve ever tried.” AI coverage researcher Dean Ball described Manus because the “most sophisticated computer using AI.” The official Discord server for Manus grew to over 138,000 members in only a few days, and invite codes for Manus are reportedly promoting for hundreds of {dollars} on Chinese reseller app Xianyu.
But it’s not clear the hype is justified.
wonderful https://t.co/TfeV9QZ1d0
— jack (@jack) March 9, 2025
Manus wasn’t developed totally from scratch. According to studies on social media, the platform makes use of a mix of current and fine-tuned AI fashions, together with Anthropic’s Claude and Alibaba’s Qwen, to carry out duties resembling drafting analysis studies and analyzing monetary filings.
Yet on its web site, Monica — the Chinese startup behind Manus — offers a number of wild examples of what the platform supposedly can accomplish, from shopping for actual property to programming video video games.
In a viral video on X, Yichao “Peak” Ji, a analysis lead for Manus, implied that the platform was superior to agentic instruments resembling OpenAI’s deep analysis and Operator. Manus outperforms deep analysis on a well-liked benchmark for common AI assistants known as GAIA, Ji claimed, which probes an AI’s means to hold out work by looking the online, utilizing software program, and extra.
“[Manus] isn’t just another chatbot or workflow,” Ji stated within the video. “It’s a completely autonomous agent that bridges the gap between conception and execution […] We see it as the next paradigm of human-machine collaboration.”
But some early customers say that Manus is not any panacea.
Alexander Doria, the co-founder of AI startup Pleias, stated in a submit on X that he encountered error messages and limitless loops whereas testing Manus. Other X customers identified that Manus makes errors on factual questions and doesn’t persistently cite its work — and infrequently misses info that’s simply discovered on-line.
Deep Research completed in below 15 minutes. Unfortunately, Manus AI failed after 50 minutes at step 18/20!
It was performing fairly well-I used to be watching Manus’ output & it appeared wonderful. However, working the identical immediate a second time is a bit irritating because it takes too lengthy! https://t.co/bGtmOI65CP
— Derya Unutmaz, MD (@DeryaTR_) March 8, 2025
My personal expertise with Manus hasn’t been extremely constructive.
I requested the platform to deal with what appeared to me like a fairly simple request: order a fried hen sandwich from a top-rated quick meals joint in my supply vary. After about ten minutes, Manus crashed. On the second try, it discovered a menu merchandise that met my standards, however Manus couldn’t full the ordering course of — or present a checkout hyperlink, even.
Manus equally whiffed once I requested it to e-book a flight from NYC to Japan. Given directions that I assumed didn’t depart a lot room for ambiguity (e.g. “look for a business-class flight, prioritizing price and flexible dates”), the very best Manus may do was serve up hyperlinks to fares throughout a number of airline web sites and airfare engines like google like Kayak, a few of which had been damaged.
Hoping the subsequent few duties could be the allure, I instructed Manus to order a desk for one at a restaurant inside strolling distance. It failed after a couple of minutes. Then I requested the platform to construct a Naruto-inspired combating recreation. It errored out half an hour in, which is once I determined to throw within the towel.
Honest opinion after making an attempt Manus AI for the final three days, right here’s the nice and the dangerous.
Good:
– The analysis it does on the web and the studies it generates are unimaginable….
