Das Startup-Projekt eines ehemaligen Vizepräsidenten von Feishu hat erneut Finanzierungen erhalten. Er hat auch sieben kontroverse Aussagen über die AI-Anwendung gemacht | 36 aufstrebende Perspektiven

Emergence is a key phenomenon in the wave of generative AI. When the scale of the model expands to a critical point, AI will exhibit human - like intelligence, being able to understand, learn, and even create.

"Emergence" also occurs in the real world. With the silicon - based civilization on the verge of breaking out, entrepreneurs and creators in the AI field are using their wisdom and minds to light up the long journey towards achieving AGI.

At the juncture of the alternation of old and new productive forces, "Intelligent Emergence" launches a new column called "36 People of Emergence". Through conversations with key figures in the industry, we will record the new thinking at this stage.

Text by | Deng Yongyi

Editor | Su Jianxun

It's quite difficult to interview Yubo. He not only has his own unique world and a Jike account, often expressing some very "human - touch" viewpoints. If you search for podcasts about Yubo's interviews, the historical records will fill up your screen.

His experiences have been narrated on different occasions many times. In 2008, he joined Alibaba, where he worked in technology and developed an internally well - regarded document product called "Yuque". He managed a team of hundreds of people and later served as the vice - president of the Feishu product at ByteDance.

After founding the new company YouMind, Yubo led a team of more than a dozen people to develop AI tools for creators, aiming to provide a one - stop AI Creation Studio for various types of creators, including those who take notes, write articles, produce podcasts, make videos, and write novels.

Recently, YouMind officially completed its second - round financing, with a total financing amount exceeding ten million US dollars.

After leaving a large company to start his own business, Yubo has repeatedly stated that he dislikes the label of "senior executive of a large company". Some people directly call him a "rebellious high - level employee".

However, Yubo is a great person to talk to. He is willing to express himself and has a sincere attitude, just as his signature on the social platform says: "An honest person."

△The initial interface of YouMind, starting creation from an input box. Source: YouMind 

The pressure was on us. After listening to and reading most of Yubo's interview contents, we met him in Hangzhou. The starting point of our conversation was the more fundamental motivation for product development: "You graduated from the physics department, and your previous work experiences were more focused on the development of efficiency products. Why do you want to create a product for creators?"

"No one has ever asked me this question." Yubo said, and then he told us in a shy tone: "You may not expect that I got a full score in the college entrance examination Chinese composition, and I even published a poetry collection in college."

Yubo is very active in public expression. When he was at Alibaba, he was regarded as a "mouthpiece" by many people because he dared to tell the truth on the intranet. After founding YouMind, he has been in high - frequency communication with creators to understand their creative environments and habits.

In 2023, shortly after ChatGPT became extremely popular, products like Rewind.ai were once people's imagination of the prototype of general artificial intelligence (AGI). It could remember all of a person's information and thus give more personalized answers.

But after chatting with hundreds of creators, Yubo wanted to ask a question: "Do humans really need to remember so much information?"

He believes that the "forgetfulness" of AI is a very good characteristic.

This also determines the positioning of YouMind. It is neither a knowledge base like Evernote or Notion, nor an efficiency - oriented tool focusing on collaboration like DingTalk or Feishu Docs.

YouMind focuses on more time - sensitive, project - based creative scenarios. "For example, many creators interview guests and write articles. They usually start preparing only two or three days in advance. YouMind is designed for this kind of situation." Yubo gave an example.

There are many similar examples.

For instance, if a Xiaohongshu creator sees the Apple Worldwide Developers Conference (WWDC) in 2025 and wants to write a Xiaohongshu note to summarize it, what they need to do is very simple:

Open YouMind, create a new creative board (Board), and enter "Latest information about WWDC 2025";

Wait for the AI to search for and organize the information, and then enter the Board workspace;

In the right - hand workspace, users can choose to view various information saved and organized by the AI. During the process, they can take notes, search for more information, modify reports, etc., and complete the creation while learning.

△Source: YouMind

To put it simply, YouMind doesn't want to be a stuffed backpack that crams all information in. Instead, it wants to be the "paper and pen in the AI era", directly handing the tools to users and enabling them to start creating immediately.

The "cold start" time for creation will be greatly shortened. Whether it's making a PPT or writing a note, if a creator has difficulty starting, YouMind's AI assistant can quickly provide ideas, help find information, write drafts, match illustrations, etc., and accompany users in the creative process.

However, on the contrary to the restraint in product design, YouMind has ambitions for content forms.

Yubo summarized YouMind's evolution path in eight characters: "All things turn into manuscripts, and manuscripts give birth to all things." In the future, various forms of content, including text, pictures, audio, and video, will definitely be able to be converted into each other. What AI application entrepreneurs need to do is just to follow the continuous evolution of the underlying models.

In the just - passed June, YouMind was updated to version "0.3", providing an integrated AI Agent experience for finding information and writing drafts. Meanwhile, the first version of the iOS App has also been launched on the App Store (search for YouMind to download). Next, YouMind will soon launch multi - modal related functions.

In 2025, the base large - model manufacturers are still at the forefront of the storm, while the booming embodied intelligence companies seem to be in a "Great Leap Forward" momentum. Choosing to start a business in a relatively niche track, Yubo maintains his honest side, often making "bold remarks" in his observations of AI applications. For example, it's normal to spend several years finding the Product - Market Fit (PMF); and many great companies didn't have a promising start in their business.

We have condensed Yubo's observations and viewpoints into a collection:

Entrepreneurship is a high - risk endeavor, and it's normal to spend 2 - 3 years finding PMF

In my definition, AI is still there to assist humans.

The core of YouMind is actually to use AI to lower the threshold of creation, enabling users who have a yearning for creation, have written something before but then stopped, to start creating again.

However, the upper limit and style of creation are still determined by humans. For example, when I use it, it will be in Yubo's style. In the past, you had to do everything by yourself, but now it's like having a junior editor or a small assistant.

For example, you can ask YouMind to match two or three pictures for you based on the content of the current document, and then the AI will start to select pictures for us. After that, the AI can even insert the pictures with one click and change the layout to the Xiaohongshu style, allowing the AI to handle the repetitive and non - creative parts.

Our version 0.1 was similar to a bookmark or a clipping plugin. Through the browser plugin or the iOS version, users could collect videos and articles to YouMind with one click.

Version 0.2 can be regarded as an Office or a Word, but this Word comes with resource management and AI assistance, similar to an IDE (Integrated Development Environment) in programming. What we've created is an ICE (Integrated Creation Environment).

Version 0.3 is the newly launched version, which comes with an Agent function, aiming to help users solve a major difficulty in creation: how to start.

In the future, we'll move towards multi - modality. I'm very certain about the direction of "manuscripts giving birth to all things". Next, multi - modality is something we'll definitely implement.

Many people say that AI companies can find PMF within a few months, but most of these claims are false.

I prefer to study products with a long - term development cycle, such as Figma, Notion, and Linear. Well - known unicorn enterprises, even companies with a valuation of over one billion dollars, have all gone through a long - term development process. It takes at least three to four years, and sometimes even seven or eight years, to find PMF. On average, it's about 3.5 years.

Is "creator tools" really a viable AI entrepreneurship opportunity?

I'd like to ask in return: What is entrepreneurship? In my understanding, "creation" is more important than "business".

Entrepreneurship, in a sense, has been misinterpreted by the Internet. Opening a small restaurant downstairs is also a form of entrepreneurship.

As we all know, entrepreneurship is a high - risk endeavor. Sometimes, it's a matter of taking a gamble. For example, no one was optimistic about Xiaohongshu at the beginning.

The core of my current entrepreneurship is to support the team and serve users well, achieving a healthy business cycle.

Agent is not the future, at least not for users

YouMind started as a bookmark, allowing users to collect videos and articles they saw with one click through the browser plugin. In the latest version 0.3, the Agent function was launched.

However, we don't want to be a vertical - specific Manus. Instead, we're more interested in doing something similar to what Gamma does.

We want to solve the "blank page problem". Many times, when you open a PPT page, the most difficult part is the blank page. Many people will get stuck there, not knowing what to write. What Gamma does is that you describe the result you want, and it will first find a lot of information for you and then write a first draft.

In essence, the work produced in this way may not be a finished product, but it gives you a starting point for creation.

Currently, "Agent" (intelligent agent) is very popular. In my opinion, in fact, Agent is a concept for VCs.

Whether it's the Agent in Feishu or DingTalk, these tools are used for automating task processing. They combine the current large - language model technology and can automatically execute some tasks. Nowadays, this kind of technology can be repackaged as "Agent".

I think Agent will eventually become a technical concept. It's a middleware or a technical means.

These concepts aren't really important. What users care more about is what AI can do for them and what goals they can achieve in specific scenarios.

Agent isn't the future, at least not for users. How would users know what an Agent is?

Don't focus on user profiles, focus on usage scenarios

YouMind has had paying users since its launch. Currently, the pricing is $20 per month, and in China, it's 200 yuan per year. The price will soon be adjusted upwards.

We hope to filter out precise paying users from the very beginning. From version 0.1 to the newly updated version 0.3, each new version has attracted a group of users willing to pay for the new features. The first - wave users were mostly attracted by the clipping function. They hoped that the mobile - phone version would be launched soon because a lot of reading happens on mobile phones, and they wanted to save all the information they needed to one place.

However, we found that many users just collected a lot of information but never read it again. After launching version 0.3, we attracted a second - wave of users who are already in the process of creating regularly, professional creators.

Professional creators are our top - tier paying users. At the same time, many users who want to try content creation have also paid for our service.

Actually, we don't spend much time researching user profiles. This is also influenced by Liang Ning's book "Real Demands". He said that in this era, the most important thing isn't to study user profiles but user personas.

We mainly focus on what users are using our tools for and where their needs lie.

As long as a user uses our tool to write a newsletter, it doesn't matter whether they're an old man or a college student.

In essence, a C - end product should attract a group of people with the same cultural background, rather than a single - type of user.

Our approach is based on cultural differences. First of all, our product is targeted at the C - end market and is global from day one.

Our product design is more in line with the new European and American design style. Typical users may like the Apple style, with a relatively simple layout. These users are concentrated in the United States, Europe, and Singapore, and there are also some in China.

However, our product has difficulty attracting Japanese users. This is because many Japanese websites have very dense layouts, and they have a higher requirement for the information density of products.

The tone of many products already determines the size of the user group.

△Source: YouMind

Knowledge management is a false proposition, humans don't need to remember all information

The fact that AI can "forget" is actually a very good characteristic.

In essence, human attention is limited. This is a theory in positive psychology. Human attention should be more focused on the present and the future, which is definitely more important than the past.

Currently, there are many products that claim to be a "Second Brain", enabling AI to remember all of your information and retrieve it whenever you need it.

For example, the Rewind AI project, which was very popular a few years ago, aimed to record everything you saw. However, in essence, this is a case of "garbage in, garbage out". If you take a picture of a cloud one day, is it really important to remember that cloud? You may forget it tomorrow.

Perhaps only some specific agents, such as the watch on your hand that needs to record your health information, really need this kind of information.

In the design of YouMind, this is also a realistic and differentiated choice, considering both user needs and costs. It's difficult for models to truly remember all information. If you have a