We have enough chatbots.
If you’re like me, your day is filled with "smart" assistants that give you a list of instructions but leave the actual clicking to you. You ask for a travel itinerary, they write a text list. You ask for a website, they give you code snippets you have to host yourself.
That changed this week.
Manus AI, now under the Meta umbrella, has officially launched with a slogan that feels like a direct attack on the current state of LLMs: "Hands On AI."
I’ve spent the last few days testing it, and the difference is jarring. Manus doesn't just talk about doing things; it takes control of the browser and does them.
The shift from talking to operating
The core philosophy here is "Less structure, more intelligence."
Most agents today are text-in, text-out. Manus is different because it’s an operator. It treats the web browser the way a human does—visually navigating, clicking buttons, filling forms, and managing tabs.
When you ask Manus to "Find a cheap flight to Tokyo and put it on my calendar," it doesn't just query an API. It literally navigates travel sites, filters results, picks an option, and then opens your Google Calendar to block the time.
It’s the difference between a colleague who tells you how to file an expense report and a colleague who just takes the receipts out of your hand and files them.
The Browser Operator
The technical heavyweight here is the Manus Browser Operator.
This isn't a simple script. It’s a vision-based system that "sees" the webpage. This allows it to handle dynamic sites where elements move around—something that usually breaks standard automation tools like Selenium.
In my testing, I watched the "Manus Browser" window (a dedicated interface) flick through pages at superhuman speed. It feels a bit like watching a ghost use your computer, but the utility is undeniable.
Building, not just coding
One of the flashiest demos is the app creation workflow.
We’ve seen coding assistants before (Cursor, GitHub Copilot), but Manus goes a step further by handling the deployment and design context. You can drop a messy sketch or a rough description into the prompt, and it builds a functional web app.
But it’s not just code. The "Nano Banana Pro Slides" tool (yes, that’s the real name) generates pitch decks that don't look like generic corporate templates. Because it understands the content of your research—thanks to its "Wide Research" capability—it fills the slides with relevant data rather than "Lorem Ipsum" placeholders.
The Meta factor
It’s impossible to ignore the elephant in the room. Manus is now part of Meta.
While some might worry about data privacy (a valid concern with any tool that has this much access to your digital life), the acquisition explains the polish. The integration across platforms—Web, Mobile, Windows, and even Slack—is tight.
This isn't a research preview anymore. It’s a consumer product backed by infinite resources.
Why this matters
We are moving past the "Chatbot Era" into the "Agent Era."
For the last three years, we've been impressed by AI that can write poetry or pass the Bar exam. But utility has often lagged behind capability. Manus represents the correction. It’s less concerned with having a philosophical conversation and more concerned with renaming 500 files or booking a dentist appointment.
It’s hands-on, and frankly, it’s about time.
Official Links
Conclusion
Manus isn't perfect. It still gets confused by complex CAPTCHAs, and giving an AI control over your browser requires a massive leap of faith. But for the first time, I feel like I have an assistant that actually assists, rather than just offering advice.
If you want to try it, start with the Browser Operator. Just don't blame me if you never want to book your own flights again.