The race for AI dominance is starting to look more and more like a digital Cold War between the US and China. While the US has long led the way with companies like OpenAI and Google, China is now making serious moves to close the gap. Earlier this year, the launch of DeepSeek already shook up the sector—and now there’s Manus: an AI agent that not only responds to commands but can also perform tasks independently. I finally got the chance to test this tool and share my experience in this new article.

These days, we’re seeing weekly updates from major AI platforms. Faster, smarter, with improved reasoning and source referencing. I previously wrote about my mixed experiences with AI agents like GPT Operator and DeepSeek. More recently, I saw countless intriguing use cases pop up featuring the new Chinese AI-hype: Manus. In my view, Manus is no longer just a chatbot—it’s an AI that can think, plan, and execute without constant human supervision.

Manus positions itself as a direct challenger to OpenAI’s Operator and DeepResearch. Some have even called its launch a “second DeepSeek moment,” not just because of its sharp user pricing but also due to the quality of its output. But is this really the next big breakthrough—or just smart marketing?

Jack-of-all-Manus

Manus is an AI agent developed by the Chinese company Butterfly Effect. What sets the tool apart is its so-called multi-agent architecture. Instead of one AI model trying to do everything, Manus uses a combination of specialized sub-agents. This allows tasks to be split and handled more efficiently.

In theory, Manus could plan an entire trip: search for flights, book hotels, compare prices, and even create a travel itinerary. Or assist a recruiter by analyzing résumés, ranking candidates, and drafting interview questions. This is a fundamentally different approach from classic chatbots like ChatGPT, which still require human supervision to complete most tasks accurately.

Much like DeepSeek—proving that high-quality AI models can be built with a fraction of Western budgets—Manus aims to make a similar impact in the AI-agent category. Which, in my view, is still the biggest AI trend of the year.

Why the Hype Around Manus?

Manus isn’t just technologically interesting—it’s also a marketing masterpiece. The first demo videos went viral instantly, showing off an AI that could independently manage complex tasks.

It struck a chord: within days, the Manus community grew to over 180,000 members on Discord, with invite codes selling online for thousands of dollars.

That exclusivity—invite-only access—helped build the hype even more. Just like with DeepSeek, it created the impression that Manus was a gamechanger, available only to a select few. On top of that, the geopolitical rivalry between China and the US plays a strong role. Many Chinese users see Manus as a symbol of technological independence and a direct response to OpenAI and Google.

But the real question remains: does it actually work, and does it deliver on its promises?

Three Types of Tests with Manus

To test the promises of Manus, I put the platform through three real-world tasks I’m currently working on.

1. Booking a Vacation

I’m planning to go hiking in Iceland this June. But I have some specific preferences: a different hike each day, specific flights, a certain type of lodging, and side activities like spa visits. Manus was able to find flights and list some good hotels, but couldn’t complete the bookings. It also missed key details like baggage fees for my hiking gear.

2. Creating a Marketing Campaign

For one of my businesses, I asked Manus to set up a complete social media strategy, including ads and audience analysis. The results were surprisingly impressive: Manus analyzed competitors, created a posting schedule, and even generated ad copy. But after a review, some suggestions turned out to be unrealistic or based on outdated data. Bummer!

3. Automating a Recruitment Process

For a large event I’m organizing, I wanted Manus to help select from volunteers who submitted applications. While the AI gave solid suggestions, a deeper look revealed that some rejections were unfair. The system struggled with nuance in work experience and favored keyword-heavy résumés over actual qualifications.

Execution Falls Short

Manus is great at structuring and planning tasks efficiently, but in my opinion, its execution still leaves much to be desired. It’s not a fully autonomous AI, but more like a clever assistant that can take over parts of tasks—but still needs human oversight.

The tool struggles with reliability. On various forums and group chats, I saw users reporting that the AI would get stuck in infinite loops or generate incorrect information. This is a major issue for applications where precision matters—such as financial analysis for crypto trading.

Speed is another weak point. While OpenAI’s DeepResearch completes tasks in seconds, Manus often takes minutes. I tested this a few times, and for more complex tasks, it took quite a while to generate a usable result.

There’s also a lack of transparency. Butterfly Effect gives little detail on how the AI actually works. It’s not a fully new tool either, but a so-called “wrapper” built on existing models like Anthropic’s Claude and Alibaba’s Qwen. How much of it is truly innovative remains unclear.

And then there’s the issue of privacy and security. Just like with DeepSeek, Manus raises concerns about data protection. Western companies will likely be hesitant to grant a Chinese AI access to sensitive business information—especially given China’s strict regulations on data collection and state control. Not to mention the recent backlash surrounding DeepSeek.

Will We Keep Hearing About Manus?

Manus AI has the potential to usher in a new era of autonomous AI agents—but it’s not quite there yet. The technology is promising, but far from flawless. It feels like a rough diamond that still needs a lot of polishing before it can truly compete with established players.

If Butterfly Effect improves its infrastructure, increases reliability, and becomes more transparent about how Manus works, it could become a serious contender in the AI race—especially at its current price point. Also because it’s far easier to use than GPT’s Operator. Until then, Manus remains a fascinating experiment—with tons of potential, but also plenty of work to be done.