We're not just AI content transporters | WAVES New Wave 2025

This is a new era for China's venture capital and investment industry. Currently, the Chinese venture capital market is not only at the turning point of the cycle bottoming out but also in the deepening period of structural transformation. In the new ecosystem dominated by policies and with high concentration of state - owned assets and capital, only by conforming to the trends and making flexible adjustments can one capture definite opportunities amidst uncertainties.

From June 11th to 12th, at the Liangzhu Culture and Art Center in Hangzhou, the 36Kr WAVES New Wave 2025 Conference, themed "New Era", gathered top investors in the venture capital field, founders of emerging enterprises, as well as scientists, creators, and scholars deeply involved in technology, innovation, and business. They jointly discussed cutting - edge issues such as AI technological innovation, the wave of globalization, and value re - evaluation, dissected their views on business ideals and the future world, and together explored, searched for, and moved towards the "New Era" of China's venture capital.

On the afternoon of June 11th, a discussion on the fields of AI and content was held at the entrepreneurs' venue. The guests participating in the discussion were Yang Xuan, the senior content director of 36Kr and the host of the podcast "Business WHY Sauce", Zhuang Minghao, the host of the podcast "The Art of Slaying Dragons", Ryan from Spark Lab, Yuca, a partner at Mindstream Capital and the curator of the self - media "Z Potentials", and Zhong Tai from Agent Universe.

The scene of the round - table discussion

The following is the transcript of the dialogue, organized by 36Kr:

Yang Xuan: Thank you all for staying until the last session. I'm Yang Xuan, the person in charge of 36Kr's original reports. Meanwhile, I also run my own podcast - "Business WHY Sauce".

When my colleague asked me if I wanted to host this session, I said yes. Because I'm an old media person who survived the previous media era by a fluke. Now, it seems that a new wave is coming. Will our generation be left behind again? I said I must come to feel the impact of the young people.

Before the official start, let me explain why I think so. I was a media person in the era of print media. Back then, I was a reporter covering venture capital for "CBN Weekly". Unfortunately, this part of the field was disrupted by 36Kr. Eventually, I left my print media job and joined 36Kr. Fortunately, I proved that the methodology of traditional print media is still effective in the Internet or mobile Internet era. 36Kr's original reports still hold significant influence in the industry. However, now it seems that the AI era has arrived, and AI can already create content. Some friends say that the newly emerging media might be the 36Kr of the AI era. Are you guys going to be disrupted? So, this is a very crucial conversation for us.

Now, please introduce yourselves one by one.

Zhuang Minghao: I'm Zhuang Minghao, the host of the podcast "The Art of Slaying Dragons". I'm also a former investor and now engaged in strategic investment. I've had a long - standing connection with 36Kr. The company where I worked in the early years was an investor in 36Kr. The proportion of 36Kr employees I know is only second to that of my own company. Now, I'm running a podcast called "The Art of Slaying Dragons", focusing mainly on technology, venture capital, and AI. Thank you!

Ryan: Hello, everyone. I'm Ryan. I'm currently involved in Spark Lab, providing an ideal startup environment for young entrepreneurs. I serve Spark by offering funds, mentor support, and a founders' community to help them turn their works into great companies. When it comes to 36Kr, I "grew up" reading 36Kr when I first started my business. Today, my perspective in this panel is mainly from that of an entrepreneur.

Yuca: Hello, everyone. I'm Yuca. I have two identities. My main job is as an investment and financing advisor for entrepreneurs, and I'm a partner at Mindstream Capital. My other identity is the curator of the self - media "Z Potentials", which focuses on young people and globalization. The birth of "Z Potentials" is closely related to AI. In 2022, several of us curators noticed the transformation brought by generative AI. During this period, we saw many young people, especially as I've served many post - 2000 founders who received a lot of investment. We observed that young people are closely integrated with the technological wave, and these startups are global from day one. So, we hoped to create a new content platform by combining the topics of youth and globalization. Based on this, we launched "Z Potentials".

Our model is different from most media. Almost all of us work part - time. There isn't a single full - time media person in this account. We should probably thank AI for our model because we complete almost all our content through the collaboration of AI and Feishu. So, it's very different from the process of traditional media.

Zhong Tai: Hello, everyone. I'm Zhong Tai, working on a project called "Agent Universe". Since 2023, we've been focusing on Agents, conducting research and development on them. Agent is commonly translated as "agent" or "intelligent agent", but we also think of it as "secret agent". We believe that in the future, Agents will outperform ordinary humans in various fields, just like secret agents. We assume that the future will be a universe where humans and Agents coexist. Based on this assumption, we've made many explorations on Agents. At the same time, we're also running a science and technology media account to popularize relevant products and technologies, which is called "Agent Universe".

Yang Xuan: Let's do a small survey on the spot. Please raise your hands if you've heard of or read "Agent Universe"? (Quite a few). How many of you have subscribed to "Z Potentials"? (Also quite a few). Are there any readers of "AnYong"? Please raise your hands. (AnYong wins). Have any of you subscribed to the podcast "The Art of Slaying Dragons"? (Not bad). Are there any listeners of "Business WHY Sauce"? (There are also some).

Next, please let Zhong Tai and Yuca surprise us. Since 36Kr is now a relatively mature institutional media with about hundreds of employees. Can you two talk about the number and composition of your teams?

Zhong Tai: We have 6 people offline, and there are also some online partners. However, most of them are engaged in Agent technology development. Only 2 people are fully dedicated to writing articles.

Yuca: We have zero full - time employees. We can basically achieve daily updates, which is equivalent to the output of an institutional media. But we have no full - time staff. It's 5 part - time employees, plus interns, along with AI and Feishu.

Yang Xuan: Yuca mentioned 5 part - time employees. Since these 5 partners are mainly engaged in FA and investment, he said they are part - time. But actually, does it mean that you mainly adopt an AI - production model?

Yuca: Yes.

Yang Xuan: Okay, I think your cost structures are very different.

Can you two talk about your revenue from media business? Or, if you're too busy to take on media revenue, how often do you decline such opportunities?

Zhong Tai: We're all post - 2000 startup teams. Everyone is either a recent graduate or still a student. We're relatively lacking in business capabilities. However, since we started focusing on Agents early, and Agents have become very popular this year, many business opportunities have come to us actively.

Yang Xuan: What's the proportion of opportunities you've declined?

Zhong Tai: About half. For some opportunities that we can't handle, we'll recommend them to other partners.

Yang Xuan: Next, I'll ask a question that people in the content production industry are more concerned about. You produce a large amount of content with AI or through an open collaborative relationship. Can you talk about how you specifically use AI to write articles? Give us some practical insights.

Yuca: Readers who follow "Z Potentials" will notice that there is a column at the beginning of each article. This column represents a Standard Operating Procedure (SOP), which is about how humans and AI collaborate. For example, for overseas translated articles, we have a process. How to use AI to crawl the article and select it. After selection, if it's a video, which AI tool, such as Tongyi, Claude, or ChatGPT, can better convert it into a preliminary manuscript. After getting the preliminary manuscript, how to use AI to proofread and polish it, and then how to format it with the help of AI. So, we basically have such an SOP for the collaboration between humans and AI. The column at the beginning of the article represents an SOP. That's why there are columns in front of our articles. Many people think we're obsessive - compulsive, wondering why we must put two words like "In - depth" or "Express" at the beginning of each article. In essence, they all represent a process.

Yang Xuan: Can you talk about the production efficiency? For example, how long does it take to produce a certain column?

Yuca: We're different from traditional media. In traditional media, due to limited human resources, for example, one person may only write 4 articles a month. But we think that if we use AI to implement SOP, we can assign 20 topics at once, and these 20 topics should run through the process simultaneously. As long as the tasks are completed, an infinite number of tasks can be run in parallel. As long as you have enough selected topics at the front - end, the back - end supply should be infinite. So, it keeps running without a strict upper limit. Although we've already published a lot, there are still hundreds of unpublished articles in our back - end library. It's a 24/7 operating process.

Yang Xuan: Our team has an account called "Intelligent Emergence", which reports on the enterprise AI industry and writes a lot of content. Our reporters often say that as soon as we publish an article, "Z Potentials" reposts it very quickly (laughs). How do you do this?

Yuca: Going back to the definition of columns, each column has a different mission. For example, the core mission of the "In - depth" column is to carry in - depth articles. The person in charge of the column needs to organize the SOP and the collaboration between humans and AI. The mission of the "Express" column is timeliness. The person in charge of this column needs to monitor various media through a database, and then use AI to quickly summarize and publish the content.

Yang Xuan: Zhong Tai, how do you use AI to create content?

Zhong Tai: Besides the main account "Agent Universe", we're also running a matrix of accounts. We call it "Agent Native MCN". Most existing MCNs sign real - life bloggers, but we're trying to see if it's possible to use Agents to operate official accounts on WeChat and Xiaohongshu from end to end. Since AI is quite mature in generating pictures and text now, our strategy is similar to creating a virtual person. We need to write a large number of prompts. We'll spend a lot of effort on positioning an account in the early stage, determining its logo, name, tone, style, and formatting. After the positioning is determined, we'll match it with some information sources and then let AI generate content in segments. We're verifying this approach, and we'll replicate the successful experiences to the main account "Agent Universe".

Yang Xuan: When I heard Ryan talk about this, I wondered if some reporters or those doing profile reports would lose their jobs. Because our reporters have also done similar things. You go out to interview a founder, but don't want to write the article. You just give the interview transcript to AI with some prompts and let it write the article. Reporters can do this, and founders can also do it themselves. So, it's a relatively big challenge for media practitioners.

Yang Xuan: Does Zhuang Minghao have anything to share? I was just joking with him. He's been very concerned about AI recently, produced many AI - related programs, and written many PPTs, but it seems he did them manually (laughs).

Zhuang Minghao: Last night, there was news that Tencent Music was acquiring Himalaya. As an observer of this industry for many years, I thought I had to write about it. Taking this case as an example, AI helped me improve efficiency in data collection. For example, when I searched for the company's financing history, the valuation of the last round before listing, and the investors. Since the top - tier search tools can directly search SEC documents and Himalaya has released a prospectus, I basically knew where to search and what the possible results would be. I could even ask about whether the investors in the last round had a preferential liquidation right, which would directly affect the distribution of the acquisition funds. I knew these questions, and I knew that traditional search engines might not be effective, but AI could help. However, AI's help ended there. All the subsequent expression and content planning still had to be done by myself.

When I mentioned this in my fan group last night, including when I said I didn't have time to make a PPT. They asked why I couldn't use AI. I said it was because it was a matter with a long historical evolution. There were many past events that only I knew, and there were many momentary emotions that were only in my mind. AI couldn't provide those emotions and past events. So, as a pure individual content creator, when creating such content, AI can only provide basic assistance, and it can't do the rest.

The same goes for PPTs. For example, if I'm going to give a talk about what AI is to my fifth - grade son's classmates, I can use AI to create a PPT without any prompt - word skills. A simple sentence can do it. But if I need to produce a podcast episode about the industry overview of the AI industry in the first half of 2025, especially for my own podcast, I really can't rely on AI. The entire logical framework, structure, meme pictures, and even the joking parts all need to be designed by myself. It's very difficult to achieve the desired effect with AI.

Yang Xuan: I also tried it recently. I did an interview with a Silicon Valley tycoon who underwent blood - transfusion treatment and turned it into a Q&A interview transcript. As traditional media people, our aesthetic is to save time for all audiences. For example, we need to greatly compress what the interviewee said to make it look elegant and not waste the readers' time. This is the standard of traditional media. Based on my past experience, it would take about 3 working days to do such a compilation, which is quite labor - intensive. The other day, I gave some requirements to AI and asked it to compress and refine the content according to my requirements. Since AI has many hallucinations, I also had to supervise whether it was writing correctly. It took about a day to finish this task. So, I think it improved the efficiency by more than half. This is my experience.

As you can hear, as a traditional media person and Zhuang Minghao, we have relatively high requirements for our own content. It's quite difficult to use AI to create content. I'll pose a question to the younger ones. If your articles are written by AI, how do you ensure the style and distinctiveness? Or do you think this question is important?

Zhong Tai: I don't think it's that important. I fundamentally believe that in the future, most content will be generated by AI. Users won't be able to tell the difference and may not even care whether it's written by AI. They only care about the result. Whether it's produced by humans or AI doesn't matter. Maybe our creativity lies in positioning. After shaping an account, how it grows and what content it creates is up to AI. Then we'll think about whether we can create another successful account in a different category. In the long run, there is an opportunity to create the next "Toutiao". The combination of generation and recommendation can make information more personalized, which is very valuable in various application scenarios. For example, in education, for the same introductory materials on economics, a professional college student may see a lot of literature citations and very professional content. But if it's given to a kindergarten child, it will be presented in a picture - and - text format, and may even imitate the voice of his favorite cartoon character. This is what large - scale models can achieve. Both generation and distribution can be done by Agents.

Yang Xuan: Do you mean that your creativity and human initiative are mainly focused on the character setting of AI accounts?

Zhong Tai: We'll spend a relatively large amount of effort on positioning. How it grows and what content it writes