Is anyone else eagerly waiting for Manus to save the day? My inbox is buried under 5,800+ unread emails, and my direct messages are constantly overflowing. I feel like I’m drowning in information and want to go on strike.
Overflowing Emails and DMs
Now, I can’t wait to rub my hands and wait for Manus’s application code! Come and save me, Manus!
On March 6, 2025, AI development ushered in a major moment — Manus was released! This is an autonomous AI agent that can independently complete complex real-world tasks without direct human guidance. It was created by the Chinese startup Monica.
The name Manus comes from the Latin “Mens et Manus” (head and hand), which coincides with the motto of MIT, symbolizing the perfect combination of creativity and execution.
Xiao Hong — China’s “Ultimate AI Warrior”
To understand Manus, you first need to know its founder — Xiao Hong, widely recognized as the “Hexagonal Warrior in AI” for his exceptional technical expertise and commercialization skills.
Image source: Sina Technology Monica Founder Xiao Hong
Born in the 1990s, Xiao Hong is already a seasoned entrepreneur:
2015: Founded Nightingale Technology, launching WeChat tools Yiban Assistant and Weiban Assistant, serving 2M+ business users. The project was later acquired by a unicorn in 2020.
2022: Founded Butterfly Effect and launched the AI browser plug-in Monica — initially a ChatGPT-for-Google tool. It quickly amassed 10M+ users, becoming a top AI assistant worldwide.
Xiao Hong actually started raising funds for Manus back in 2017, selling 10% equity for 3M RMB (USD 410K). At the time, many investors were skeptical. But through persistence and innovation, Manus has emerged as the world’s first general AI agent, redefining what AI can achieve.
What problem does Manus solve?
We often say that your knowledge and vision shape your life. Many people set goals — like saving $10,000 for a trip — but struggle to create a clear, actionable plan. That’s where Manus comes in!
As an AI agent, Manus doesn’t just give advice — it independently plans and executes complex tasks, delivering real results. Here’s what makes it a game-changer:
1. Autonomous Execution — No need for constant prompts. Manus completes tasks independently.
2. Multitasking Mastery — It can handle multiple tasks at once, even unpacking compressed files automatically.
3. Smart Task Breakdown — When tester Ma Ning asked Manus to “explain Bernoulli’s principle to a 4-year-old”, Manus didn’t just generate text — it created an interactive webpage with balloons, airplanes, and bubbles, plus mini-games to make learning fun. In contrast, ChatGPT or DeepSeek would only provide a text-based explanation.
Technical Advantages
Manus uses a Multiple Agent architecture, leveraging tools like code writing, web browsing, and app operation within a virtual environment to complete tasks. In the GAIA benchmark*, it even performs better than other AI products, such as OpenAI.
* The GAIA benchmark is a test that measures how well an AI agent can plan, execute, and complete real-world tasks independently — without constant human guidance.
Is Manus Better Than General Large Language Models?
Manus and general large language models (LLMs) like DeepSeek and ChatGPT aren’t the same kind of product.
As serial entrepreneur Fu Sheng aptly puts it, LLMs are the core of intelligence — like a deep-thinking brain. Manus, on the other hand, enhances AI’s usability by adding a “shell” to this powerful brain, enabling it to seamlessly connect with websites, tools, and the real world.
Serial entrepreneur Fu Sheng
Here’s a simple analogy:
— LLMs are like Stephen Hawking (the famous physicist behind A Brief History of Time) — brilliant, deep thinkers but limited in interacting with the world. — Manus is like Hawking’s wheelchair, empowering him to communicate and act freely.
A Brief History of Time Stephen Hawking
What makes Manus unique? It bridges the gap between intelligence and execution, turning AI’s knowledge into action.
Is Manus the World’s First General AI Agent?
This claim is a bit tricky. Large language models (LLMs) like DeepSeek and ChatGPT are “general” by nature, while AI Agents like Manus act more like specialized templates built on human experience, enhancing LLMs’ capabilities in specific areas. Since each field requires different templates, it’s nearly impossible for AI Agents to cover all domains and achieve true “generality.”
In the future, we’ll likely see multiple specialized agents — much like humans, who share similar intelligence but develop unique professional skills through different training. So agents are destined to be experts, not jacks-of-all-trades.
What’s your take on Manus? Share your thoughts below!
For more insights on global cutting-edge tech and innovations, follow “KellyOnTech”!
In this era of rapid AI advancement, do you fear being left behind? Today, I’ll introduce a powerful tool — the Scenario Maturity Assessment — to help you stay ahead.
This is my key method for evaluating whether AI-enabled technology companies are worthy of investment. It is not only suitable for entrepreneurs and investors to identify opportunities, but also helps confused parents plan for their children’s future and and take charge in the AI era!
What Is the Scenario Maturity Assessment Method
The concept of scenario maturity was introduced by Zheng Yan, Chief Expert of Huawei Cloud AI Transformation. This framework evaluates opportunities across three critical dimensions:
1. Business Maturity: A well-defined and stable payer exists. Clear ownership and accountability are established. Process rules are transparent and actionable. User touch points are well-defined and measurable.
2. Data Maturity: Existing data enables a cold start for the scenario. Business data flows continuously, updating and generating feedback. Operations inherently serve as data annotations. Knowledge data is systematically governed.
3. Technology Maturity: Existing technology is capable of realizing the scenario effectively.
How to use the Scenario Maturity Assessment Method
The current U.S. President, Donald Trump, once hosted the popular reality show The Apprentice. In its first season, contestants were tasked to sell lemonade. I’ll use this project to demonstrate how the Scenario Maturity Assessment method can be applied.
1. Business Maturity: Can your lemonade business make money? Is there demand? (Are there thirsty students or office workers nearby?) Who is in charge? (Are you running it alone or with a team?) What’s your sales strategy? (Will you set up a stall, push a cart, or use other channels?) How will customers find you? (Are you located near a bus stop, a school, or another high-traffic area?)
2. Data maturity: Do you know how to make delicious lemonade? Do you have a recipe? (How much lemon and sugar should you use?) Are you tracking sales? (How many cups did you sell today? What flavours are most popular?) Will you refine it based on feedback? (If customers prefer sweeter lemonade, should you add more sugar?)
3. Technology Maturity: Do you have the right tools? Do you have a juicer? (Or are you squeezing lemons by hand?) Do you have a measuring cup? (Or are you eyeballing ingredient proportions?) Do you have a cooler? (Or are you selling lemonade at room temperature?)
By assessing these three areas, you can determine how mature your lemonade business is. The more developed each aspect is, the higher the chances of success!
Scenario Maturity Analysis: Home-Based Elderly Care Humanoid Robot Market
Let’s use the Scenario Maturity framework to evaluate the home-based elderly care humanoid robot market.
Business Maturity Who pays? Who decides? Families with elderly care needs are the primary buyers, with purchasing decisions typically made by adult children or the elderly themselves. However, high costs remain a barrier for many families. As technology advances and production scales up, prices are expected to decrease, making these robots more accessible.
What can robots do? Elderly care scenarios are complex and varied, with differences in habits, schedules, and home layouts. Robots currently handle simple tasks like companionship and medication reminders well, but complex tasks (e.g., assisting with bathing or stairs) require further process optimization and safety improvements. How do users interact with robots? Most interactions happen via voice commands or a mobile app, allowing users to check the weather, play music, call family members, or monitor health data. However, voice interaction technology still needs improvement in accuracy and semantic understanding to better meet user expectations.
Data Maturity
Where does the data come from? Currently, data is limited, relying mainly on simulations and small-scale testing, which differ from real home environments. As more robots enter households, they will collect extensive real-world data, such as elderly living habits, health metrics, and interaction records, enabling smarter robot performance. How is the data used? Robots can track real-time data like heart rate and movement patterns, transmitting it for analysis. This helps detect health issues early and adapt services to better meet the elderly’s needs. How is data labeled? Each household is unique, making standardized labeling difficult. A flexible framework can allow personalized labeling—tracking task completion, user satisfaction, and care routines to improve robot performance. How is data security ensured? Elderly care data is sensitive, requiring strict privacy protection. Secure collection, storage, and usage practices must comply with regulations to prevent misuse. Proper data management will also help optimize robot functionality and care services.
Technology Maturity
What can robots do now? They can navigate independently, avoid obstacles, understand basic speech, chat with the elderly, and monitor vital signs like heart rate and blood pressure. What are the current limitations? Robots still struggle with cluttered home environments, sometimes bumping into objects. They may not understand dialects and cannot perform delicate tasks like dressing or bathing assistance. What’s next? Future advancements will enhance adaptability, improving sensors and robotic “hands” for greater precision. Robots will work alongside family members and doctors to provide more comprehensive care. Interoperability challenges Robots need to integrate with smart home and medical devices, but compatibility issues exist due to different standards. Establishing unified protocols will enable seamless communication and better functionality.
The home-based elderly care robot market holds great potential, but it also faces challenges. Using the Scenario Maturity Assessment framework, we can see that while current robots need improvement in data and technology, advancements and rising demand will drive significant progress.
This method applies to any industry, offering a structured way to assess its state through three key dimensions:
Business Maturity – driven by demand and regulations.
Data Maturity – shaped by data availability and security.
Technology Maturity – defined by capabilities and innovation
Whether you’re investing, launching a startup, or planning a career, this framework provides clear insights to help you make informed decisions.
端粒（Telomere）是染色体末端的保护帽，帮助细胞保持健康，由重复的 DNA 序列和相关蛋白组成。
为了更直观地理解，不妨想象一下鞋带。端粒就像鞋带末端的塑料套，看似毫不起眼，却默默守护着鞋带的完整性，保护我们的 DNA 不受损坏或纠缠。细胞每一次分裂，就如同鞋带在日常使用中不断受到摩擦，端粒也会随之逐渐变短，恰似塑料套慢慢磨损。当塑料套被磨没了，鞋带就会从末端散开，直到没法用了。同理，随着时间的推移，当端粒变得太短时，细胞就无法正常工作。端粒与衰老和某些疾病的产生有着千丝万缕的联系。
Is it data? Algorithms? Computing power? While these elements are essential, they are not the true core of the AI era. The real game-changer is decision quality.
AI has transitioned from the lab to the real world not just because it can process vast amounts of data or execute complex calculations, but because it has fundamentally improved the way decisions are made. This leap in decision-making capability is what’s reshaping industries, transforming our daily lives, and defining the future of technology.
Go Battle: The Pinnacle of AI Decision-Making
Remember the Go* match that shocked the world in 2016? AlphaGo played against the top human player Lee Sedol. Since then, no human has ever defeated AI in the world of Go. AlphaGo’s victory is not accidental. Every move it made was a carefully calculated decision. Unlike human players, who rely on intuition, experience, and real-time judgment, AI evaluates countless possibilities, predicting dozens — or even hundreds — of moves ahead. This duel was more than the victory or defeat of a Go game; it marked a revolution in decision-making itself.
*Go, an ancient Chinese strategy game is played with black and white stones, where the player who controls the most territory wins. Known as the “King of Board Games,” it embodies art, wisdom, and deep philosophy.
Autonomous Driving: AI Decision-Making in the Real World
The improvement in the quality of AI decision-making is not only reflected on the chessboard but also in our daily lives. Autonomous driving is one of the best examples.
What is the core of autonomous driving? It is the perfect fusion of perception and decision-making.
The limitations of human driving: Humans rely on our eyes to navigate the road, but our field of vision is limited and blind spots are unavoidable. We use our ears to detect sounds, yet distractions — music, conversations, or outside noise — often distract us. More critically, human learning is finite. Once we learn to drive, few of us refine our driving skills further. Fatigue, stress, and emotions like road rage further increase accident risks.
Advantages of AI driving: In contrast, AI-driven cars are equipped with multiple cameras, radars and sensors that provide a 360-degree view of the surrounding environment without blind spots. AI doesn’t get distracted, fatigued, or emotional. Unlike humans, it learns and optimizes continuously, adapting to complex road conditions, and improving safety over time. Eventually, AI-driven cars will surpass human drivers in reliability and efficiency.
From the strategic brilliance of Go to the precision of autonomous vehicles, AI is changing the way the world works with more efficient and accurate decisions. AI is transforming decision-making across every domain. As AI systems become more advanced, making faster and more accurate choices than ever before, they are redefining how the world operates.
As AI decision-making seeps into every corner of life, the real question is: How can we find our place in this wave? In this AI-driven revolution, are you ready?
Special Notes
Over the past 20 years, I have had the privilege of working with founders, investors, and technical teams from technology companies worldwide, particularly in China, the United States, Canada, Singapore, the United Kingdom, the Netherlands, Germany, Japan, and beyond. Given the diverse national and cultural backgrounds, misunderstandings are sometimes inevitable, especially when it comes to perceptions of China.
To bridge this gap, starting this year, I have decided to regularly share insights from thought leaders in China’s tech industry, offering cutting-edge perspectives and unique viewpoints.
My aim is to foster deeper mutual understanding and break down unnecessary barriers. After all, in an era of rapid AI advancements, humanity must work together to navigate the complexities of intelligent technologies and jointly build a brighter future.
Liang Ning Introduction
The above content is compiled based on Liang Ning’s understanding of the AI era.
Liang Ning is an outstanding business thinker in China’s technology field. In her 20s, she worked with Academician Ni Guangnan to develop China’s domestic chips and operating systems. Later, she founded a well-known outdoor travel portal, Green Man. After the website was acquired by Tencent, she served as Tencent’s product manager. After leaving in 2014, she focused on in-depth research of enterprises.
