The Agentic Future is Already Here

Manus presentation at Aldrich Hall

Inside Tao Cheung’s Manus AI and the rise of consumer-centric agents.

This past month, I had the chance to host Tao Cheung, better known online as Hidecloud. Tao is the co-founder of Manus AI, one of the most talked-about startups in the AI world right now. During his global tour to meet users and share product insights, Tao stopped by HBS to deliver a thought-provoking presentation and fireside chat on AI’s evolving role in our lives and how Manus is shaping the consumer agent landscape in a way that feels radically different from Silicon Valley.

What Is Manus AI?

If you haven’t heard of it yet, Manus AI is an agentic productivity tool designed for individuals. Think of it as a highly intuitive, customizable AI assistant that can help you ideate, research, write, plan, and execute while remembering what you're working on.

It’s not just a chatbot or a clever plugin. As Tao put it, Manus is more like an intern. Not necessarily the HBS kind, but one who shows up, gets the job done (most of the time), and costs very little to manage.

Built by the Chinese startup Butterfly Effect, Manus combines models like Claude 3.5 and Qwen, coordinating them through multiple autonomous agents. These agents don’t just respond to prompts; they also break down goals into tasks, search the web, explain their reasoning, and ask for feedback along the way. The entire process is made visible through a transparent interface called Manus’s Computer.

Real Use Cases: Beyond Just Talking

Some of the tasks I’ve completed using Manus that stood out include:

• Ordering food from Uber Eats (this workflow was impressively automated)

• Comparing Dyson vacuum prices across platforms

• Building websites or workflows

• Generating structured documents in Word or Excel

• Researching my field project company in Kenya

Unlike other agents that operate in a black box, Manus is fully transparent. I’m able to see every step as it browses the web, encounters paywalls or captchas, and updates its approach. The interface even lets you jump in and take control at any time to provide feedback or instruction.

Manus AI’s Web Interface

“General” Isn’t Vague. It’s Foundational.

One of the most compelling takeaways from Tao’s talk was how Manus redefines what “general” AI agents actually mean.

While most startups are building vertical tools — AI for coding, trip planning, or website building — Manus takes a broader approach. Instead of building dozens of products, the team created one deeply capable system. According to Tao, Manus now covers 76% of agentic use cases from the Y Combinator Winter 2025 batch. That’s not a coincidence. It’s by design.

“General doesn’t mean vague,” Tao explained. “It means foundational.”

Unlike ChatGPT, which resets context with each chat, Manus remembers your tasks, tracks instructions over time, and allows users to follow or override the agent at any point. It’s not a conversation; it’s a collaboration.

Why Tao Chose “Manus’s Computer” Instead of MCP

A key design decision behind Manus is the use of Manus’s Computer, a virtual browser interface that keeps users in the loop. Tao shared that the team had built an internal prototype using MCP (Multi-Agent Command Protocol) to orchestrate tasks in the background. While it worked technically, they found it problematic from a user experience perspective.

“Users had no visibility or control. They had to wait until the task was completed,” Tao said.

A live view of Manus’s Computer completing a real-world task—ordering food from Uber Eats in Boston. Users can follow along, step in at any moment, and guide the agent through decisions like platform selection or item choice.

With Manus’s Computer, users can watch their agent work step by step while still using their own device. It’s the difference between watching someone build beside you versus sending your task into a black hole.

Why China’s AI Looks Different

Tao also reflected on how different China’s tech environment is from the West.

“B2B SaaS doesn’t really exist in China.”

Instead of chasing enterprise contracts, Chinese startups like Manus focus on consumer-first innovation. They ship fast, iterate publicly, and grow through community. Chinese users are more likely to experiment with AI tools across their personal lives, using them to journal, design, research, or shop.

Manus leans into this openness. The interface is intuitive, mobile-native, and emotionally resonant. The roadmap is driven by global user feedback, not just market gaps.

Final Thoughts

Tao’s visit offered a rare glimpse into the mind of a founder who isn’t just building tech; he’s building a movement.

When I asked if he’s worried about competitors copying the product, he shook his head.

“That’s why I’m on this tour,” he told me. “We’re building a creator community. We want users around the world to share how they use Manus and help shape its future.”

In a world overwhelmed by AI noise, Manus AI stands out not because it shouts the loudest, but because it listens.

If this is the future of agents, it’s already here. And it might just be built in China.

Elena Li (MBA ’26) is originally from China and most recently lived in New York City. She graduated from NYU Stern School of Business and worked in product management at Citi before coming to HBS. Elena is passionate about AI and enjoys exploring new restaurants.