Xiao Hong (Red): The Man Behind the Autonomus Genral Agent Manus

Manus is a new AI agent developed by the Chinese startup Monica, claiming to be the world's first fully autonomous AI agent. It's designed to handle complex tasks independently after an initial user prompt, such as sorting résumés, analyzing stock trends, and generating interactive websites. Currently, Manus is in a private testing phase, accessible by invitation only.

Xiao Hong is truly exceptional! Clear-minded, shrewd, and remarkably, he maintains both the right mindset and passion while drawing on years of accumulated experience. If I were an investor, I'd back him without hesitation.

I listened to Zhang Xiaojun's exclusive conversation with Manus CEO Xiao Hong in one sitting—fascinating throughout. Xiao Hong speaks plainly and honestly, without mystification or pretense, yet his insights are remarkably precise. He has crystal-clear awareness about positioning himself and his products, about the ecosystem, about the relationship between foundation models and applications, and about the future. As a 10-year entrepreneurial veteran, he maintains his own principles, staying true to himself ("be yourself"). While he fully understands that tech giants like ByteDance will quickly catch up in the large language model agent space, he believes there will always be room for application enthusiasts like himself. He consistently holds model companies in high regard, particularly highlighting how DeepSeek has brought unexpected positive factors to the application ecosystem.

The Man Behind Manus

Xiao Hong (Red) is the CEO of Manus, the recently viral large language model autonumous agent that has become China's new star in the AI industry following DeepSeek. Listening to Xiao Hong's interview feels like reading "startup notes" for the new era of AI applications. This founder, born in the 1990s but already with 10 years of entrepreneurial experience, not only clearly articulates the development trajectory of large language model applications but also explains in plain language to entrepreneurs: in this AI revolution, even if you're not a "model powerhouse" like DeepSeek or OpenAI, you can still find your place.

From Chatbots to Agents: The Evolution of AI Applications

The evolution of AI applications has been as dramatic as the shift from flip phones to iPhones:

Jasper Era: Homework-copying mindset, "Please fill in your target audience and theme~"
ChatGPT Era: Conversational instructions, "Hi, what can I help you with?"
Monica Era: Context awareness, "I see you're reading this article, would you like me to summarize it for you?"
Cursor Era: Vertical coding agent, "I don't just chat, I can write code to solve your problems!"
Agent Era: Asynchronous planning and execution, "Let me break down this task, complete it step by step, and report results as they come~"

Isn't this just like the evolution from "feature phones" to "smartphones"? Xiao Hong discovered a clear main thread behind this evolution: increasingly aligning with ordinary people's habits while expanding capability boundaries in values.

"The New Andy-Bill Law": How Application Companies Consume Model Capabilities

Xiao Hong proposed "the new Andy-Bill law": no matter how powerful the models created by model companies, application companies can consume these capabilities and transform them into user-perceivable value. This is the positioning that large model application pioneers should adopt.

This mirrors how Intel (Andy Grove) would provide more powerful chips, and Microsoft (Bill Gates) would consume that computing power with more powerful operating systems. Now, model companies provide stronger reasoning capabilities, and application companies are transforming them into intelligent agents capable of writing code, calling APIs, and planning execution.

Xiao Hong even half-jokingly offers a startup tip: "Predict what the next model capability will be, build your application around it, and wait for that model capability to launch. When that model capability improves, you'll win at the starting line you foresaw!" As an excellent product expert, he once used this strategy to succeed in his first startup.

The Agent Is Alive! The "Aha Moment" of Large Model Applications

One astounding scene from the interview is Xiao Hong describing their "aha moment" while testing the Agent in development:

They gave the Agent Manus a task to analyze how many animals appeared at a specific timestamp in a YouTube video. The Agent not only opened YouTube but also decided by its own choice to use fastfoward button to improve efficiency, precisely located the specified time, and then analyzed the screen content to provide an answer.

Xiao Hong's reaction: "You truly feel like you're creating a life."

Isn't this like a real-life prequel to "Westworld"? Except that today's Agents are still exploring the digital world, not the physical one.

"Think in Terms of Game Theory, Not Logical Reasoning"

Xiao Hong's summary of entrepreneurial thinking is insightful: don't use logical reasoning ("Baidu has the best algorithm engineers, so Baidu will definitely do recommendations well"), but instead use game theory thinking ("because a certain player joins, the entire game rules change"). Logical reasoning doesn't account for ByteDance's success (e.g. in TikTok), but game theory thinking can accommodate new players (like Liang Wenfeng for DeepSeek and Xiao Hong for Manus).

It's like chess—not simply deducing "if I make this move, my opponent will definitely make that move," but considering "because I made this move, my opponent might change their entire strategy."

With this mindset, even in a competitive environment dominated by giants, entrepreneurs can find their opportunities—not through linear extrapolation (which would only lead to the pessimistic conclusion that "everything is an opportunity for giants"), but by becoming variables that change the rules of the game.

In other words, Sam Altman's vision of top model companies crushing everything is at most only half truth. The space where models and applications each play to their strengths will likely co-exist for a long time.

Xiao Hong's Golden Quotes

In the interview, Xiao Hong offered several quotable lines worth hanging on entrepreneurs' walls:

- "When you realize you're innovating and leading, you should be more aggressive, super aggressive"
- "You should understand it using the most crazy imagination, better not to short it"
- "Don't put application companies and model companies in opposition"
- "Being yourself is most important, rather than being reactive"

Finally, facing the rapidly developing future of AI, Xiao Hong concludes by quoting Jensen Huang: "What would happen in the next few years that would surprise you?" Huang's answer was: "Basically nothing."

In other words, in the AI era, don't be too surprised by whatever crazy things happen—even if your Agent starts using fastforward key to watch YouTube videos and answer questions. After all, the awakening of agents is only just beginning!

[Related]

The Agent Era: The Contemporary Evolution from Chatbots to Digital Agents