What is Manus AI?

China's go-to-market application layer takes some bold risks.

Good Morning,

There’s been some hybe about Manus AI, out of China.

You can read about it here, here or here.

But what are we talking about really? We are talking about an era where Claude wrappers also take on Chinese characteristics.

If 2025 was supposed to be the year of AI agents, we have Salesforce flopping big-time, and a bunch of options outside of just Replit. We have more demos and prototypes of agents that don’t and cannot do what they claim. All respect to the Chinese startup Monica.

Manus, an “agentic” AI platform that launched in preview last week is going viral mostly in China, and that means it’s being pushed on X too.

Officially launched in early March 2025, it is hailed as the world’s first fully autonomous AI agent.

Deirdre Bosa of CNBC on Manus:

The product is already drawing widespread comparisons to last month's viral DeepSeek moment as well as a lot of criticism, and I find it all fairly endearing as teams in China experiment on the application layer.

The Butterfly Effect 🦋

But SOTA?

Manus AI has reportedly achieved state-of-the-art performance on the GAIA benchmark, a respected measure for evaluating AI’s ability to perform real-world tasks. But this has not been independently verified.

Beijing-based AI startup Monica.im (the company’s Chinese name is 蝴蝶效应, translated as Butterfly Effect).

Key Points:

• According to Maginative, Manus functions as a truly autonomous agent, independently analyzing, planning, and executing complex tasks from start to finish.

• Early testers report it completes weeks of professional work in hours, with authentic agentic capabilities that go beyond existing AI assistants.

• The system uses a multi-agent architecture, breaking down complex tasks into components handled by specialized sub-agents.

It’s a Claude wrapper with a little help from Qwen (Alibaba) models.

• Some casual thoughts follow, with a healthy dose of skepticism and tidbits on the Founder.