☀️ AI Morning Minute: Manus
The Chinese AI agent that doesn't wait for instructions
Most AI tools answer questions. You ask, they reply, you ask the next thing. Manus is built on a different idea: give it a goal, walk away, and come back to a finished result. That shift from "assistant" to "operator" is what made it the most talked-about AI launch since DeepSeek.
What it means:
Manus is an autonomous AI agent built by Butterfly Effect, a Chinese startup based in Singapore. It launched in March 2025. Unlike a chatbot, Manus doesn't just respond to prompts. It plans multi-step tasks, executes them using web browsers and code editors, and delivers complete results. You ask it to plan a trip to Japan, and instead of giving you a list of suggestions, it researches flights, builds an itinerary, and produces a finished document. It runs in the cloud, so you can close your laptop and it keeps working.
Why it matters:
Manus is built on top of other companies’ models, not its own. It uses Anthropic’s Claude and Alibaba’s Qwen, orchestrated together with custom software. That’s a significant strategy: instead of competing with frontier labs to build the best model, Manus builds the best agent on top of existing models. The plumbing is the product.
It beat established systems on the GAIA benchmark, a test designed to measure how well AI agents handle real-world tasks like research and analysis. This is what got the AI world’s attention. A small team using other people’s models outperformed OpenAI’s much larger Deep Research system on the metric specifically built to measure agent capability.
Meta acquired it in December 2025 in a deal reportedly worth $2-3 billion. That’s a big bet on the agentic AI direction. Meta said it would integrate Manus technology into Meta AI while continuing to operate the standalone Manus product. The acquisition is now under regulatory review by Chinese authorities, which adds a layer of geopolitical complication to the story.
Simple example:
You hire a travel agent the old-fashioned way. You say "plan a 10-day trip to Tokyo, I like food and history, $4,000 budget." A bad agent comes back with three websites you should check out. A good agent comes back two days later with flights booked, hotels reserved, restaurant recommendations, a day-by-day itinerary, and a packing list.
Manus is trying to be the second kind of agent, but for any task you can describe in a sentence. The pitch is simple: stop telling AI what to type. Tell it what you want done.