Partner Dialogue: Rhapsody of AI Super Applications | 2025 AI Partner Conference
The year 2025 marks the beginning of the explosion of AI applications. As the global AI competition enters the "China Moment," a profound technological transformation is quietly reshaping the industrial landscape. At this critical juncture, the industry faces core questions: How can we bridge the gap between AI technology and large - scale applications? Where will the next disruptive AI super - application be born?
On April 18th, the 2025 AI Partner Conference, hosted by 36Kr, grandly kicked off at the Modu Space in Shanghai. The theme of this conference is "The Super App is Coming," focusing on the disruptive transformation of AI applications in all industries. The conference is divided into two major chapters: "The Super App is Coming" and "Who Will Be the Next Super App," covering seven topics such as "Growing Up in the AI World" and "In 2025, Focus on Super Apps in AI." It includes more than 10 keynote speeches, 3 round - table discussions, and two sessions for releasing the lists of excellent AI case enterprises, deeply analyzing how AI technology reconstructs business logic and reshapes the industrial landscape, and exploring the infinite possibilities brought by AI super - applications.
36Kr specially planned a Partner Dialogue at the 2025 AI Partner Conference, a round - table discussion titled "Rhapsody of AI Super - Applications." Bai Shuang, the founder and CEO of Fuzhiliu; Chen Haici, the general manager of Shanghai Big Model Ecosystem Development Co., Ltd.; Ji Kaiming, the person in charge of the intelligent industry team at Yunxiu Capital; Hu Xiuhan, the founder and CEO of NieTa; Qiao Haixin, the founder of Kotoko AI; and Xu Minwen, a partner at TTC. These six industry experts discussed "AI super - applications."
The scene of the round - table discussion on the theme of "Rhapsody of AI Super - Applications" at the 2025 AI Partner Conference
The following is the transcript of the Partner Dialogue, edited by 36Kr:
Bai Shuang (Host): Thank you all for staying until the end! I'm Bai Shuang, the host. It's a great honor to invite the person in charge of Modu Space, representatives from investment institutions, and enterprise founders today. You will share your insights from different perspectives. First, please introduce yourselves one by one.
Chen Haici: Welcome to Modu Space! This is China's first innovation ecosystem community focusing on the development of the big - model industry, and also an incubator and accelerator in the field of big models. Since its establishment in September 2023, in just one and a half years, more than 110 big - model enterprises have settled in, forming an upstream - downstream industrial ecosystem. I'm very honored that 36Kr chose this place to hold this grand event, and I also look forward to more big - model entrepreneurs starting their journey here.
Ji Kaiming: Thank you for the invitation from 36Kr! I'm Ji Kaiming from Yunxiu Capital. Yunxiu Capital was established in 2015 and is an investment bank focusing on the technology industry. Its business is mainly divided into two major sectors: First, private equity financing. So far, it has served nearly 400 transactions with a total scale of over 100 billion RMB, covering multiple technology fields such as semiconductors, artificial intelligence, and robotics. It has helped 6 clients go public on the Science and Technology Innovation Board, the Hong Kong Stock Exchange, and the Growth Enterprise Market, and has served unicorn enterprises such as Moore Threads and Unitree Robotics. Second, fund business. Yunxiu established the "Yunxiu Hard - Technology Industry Fund" with a scale of 500 million in 2023, focusing on early - stage technology innovation investment and mainly deploying in the fields of semiconductors, intelligent manufacturing, and AI. I'm in charge of the intelligent industry team, mainly focusing on the AI and robotics tracks.
Hu Xiuhan: Hello everyone. I'm Hu Xiuhan, the founder of the NieTa product. Our company is called See Concept, established in December 2022, and it was one of the earliest enterprises to settle in Modu Space. The APP was officially launched after the Spring Festival last year, and it has been more than a year now. NieTa is the world's first open - source character co - creation community. Users can create content such as pictures, texts, videos, animations, and mini - games around characters and participate in co - creation interactions. The goal is to cultivate character IPs, similar to the "Github of IPs." Currently, NieTa adds about 400,000 new original characters every month, with over one million monthly active users. Some top - tier characters have formed co - creation communities of 10,000 - 20,000 people, and its daily generation volume ranks among the top in domestic consumer - oriented startup applications.
Qiao Haixin: Hello everyone. I'm Qiao Haixin, the founder of Kotoko AI. Our product, BISIT, is a social interaction platform centered around virtual character agents. It is planned to be launched on Steam this summer and will be expanded to multiple platforms in the future to create a gamified experience. Users can create original virtual characters (OCs) on the platform and build gamified scenarios for the characters. Through multi - agent technology, automated interactions and growth between characters can be realized, providing users with a highly personalized companionship experience. This new content interaction mode combined with AI is still an innovative attempt in the industry. If everything goes smoothly, you can experience our product at the Steam Next Fest. We are also an enterprise settled in Modu Space. Thank you for the invitation from the organizer!
Xu Minwen: Hello everyone. I'm Xu Minwen, a partner at TTC. TTC was established in 2022 and is an AI - driven intelligent recruitment platform. Its main service targets are Chinese technology innovation enterprises, especially those in the AI field. We are committed to using AI technology to disrupt the traditional headhunting model. Through three major digital infrastructures, namely the interaction system, the candidate system, and the client system, we can achieve precise matching between talents and positions. At the same time, we have also developed an AI toolchain internally, such as the "Xiaomai Assistant," to improve the efficiency and delivery quality of consultants' work. Recently, we will also launch an end - to - end recruitment agent that covers the entire process of an enterprise from demand definition to interview and employment. We hope to solve the problem of resource mismatch in traditional recruitment through the in - depth collaboration between AI and humans.
Friends who follow our official account may have noticed the slogan "For AI talents, choose TTC." The reason is that the next competitive barrier in the AI industry is data, and the verticality of data determines the depth of the barrier. In the early stage of the product, we chose to focus on the AI talent market to help Chinese technology innovation enterprises find suitable local talents, which is a highly potential incremental market.
Bai Shuang (Host): Next, let's talk about the industry changes. Mr. Chen Haici mentioned that more than a hundred enterprises have settled in Modu Space. What obvious changes have occurred to the settled enterprises in the past year and a half? Have they shifted from big - model R & D to application implementation?
Chen Haici: The changes are mainly reflected in two aspects: the identities of entrepreneurs and the directions of industry applications. In the early stage, most of the entrepreneurs who settled in were technology - driven. They joined the AI wave with their excellent technical abilities. However, with the maturity of big - model technology, more and more resource providers have started their own businesses independently. For example, Wind, a data service enterprise, directly incubated a team to explore AI applications in the face of market competition.
In terms of industry applications, the development path of enterprises in Modu Space clearly reflects the evolution of big - model capabilities. In the initial stage, most enterprises focused on basic applications such as chatbots. At that time, I was also worried about the survival space of these projects. However, with the improvement of big - model capabilities, especially the breakthrough in reasoning ability, more and more enterprises in niche fields have emerged. For example, with the development of image - generation technology, related tracks such as LIBIVE have emerged. In the financial field, new enterprises settle in every week. Some are doing vertical agents, while others are deeply involved in optimizing underlying models. These changes are due to the AI equalization promoted by DeepSeek, which has lowered the technical threshold and allowed more people to participate in AI entrepreneurship.
Financial founders come from various backgrounds, including those with a pure technical background, financial practitioners, and even investors themselves. This just shows that the popularization of AI technology has broken down industry barriers, allowing people with different backgrounds to find entry points.
Bai Shuang (Host): Indeed, in my interviews, I also found that in the early stage, most founders were technical experts, while now business experts and industry veterans have become the main force in entrepreneurship. They bring business know - how into the AI field, which makes up for the key link in technology implementation. Mr. Ji Kaiming, you are engaged in both FA and investment business. From the perspective of the capital market, have you also observed that the investment focus has shifted from big models to AI applications? Is there any emphasis in the application field?
Ji Kaiming: The feeling is very obvious. This year, the big - model industry has entered the elimination stage, while AI applications have entered an explosion period. Looking back at the cloud - computing era, the value distribution between infrastructure (Infra) and applications is usually 3:7, with applications accounting for the majority of the value chain. However, currently, there are still few AI applications that have truly achieved large - scale commercialization globally. The emergence of DeepSeek has greatly lowered the threshold for entrepreneurship. On the one hand, the number of founders in the application layer that we have contacted has increased significantly. On the other hand, the overall number of startup enterprises has increased by 2 - 3 times compared with previous years, including entrepreneurs from senior executives of large enterprises, serial entrepreneurs starting their second businesses, and teams from university laboratories. Since the beginning of 2025, the atmosphere of AI entrepreneurship in China has been extremely hot.
The role of DeepSeek in boosting industry confidence is very crucial. It has not only improved the maturity of models but also promoted the development of the open - source ecosystem, allowing more entrepreneurs to iterate quickly based on existing technologies.
Bai Shuang (Host): The rapid iteration of technology is both an opportunity and a challenge for entrepreneurs. New models and applications appear every day. For example, Munas claims to be able to do everything. Will this affect your product directions? What's your opinion, Mr. Hu Xiuhan?
Hu Xiuhan: The development of technology has two - fold impacts on us. First, the maturity speed and cost changes of different modalities (such as text, pictures, videos, programming, etc.) will affect the evolution rhythm of the product. For example, we are evaluating whether web - front - end technology will be more widely adopted than videos, and then adjusting our ecological layout accordingly. Second, we pay more attention to the acceptance speed of new technologies by ordinary consumers. After all, technology ultimately depends on human use. Only when ordinary users start to widely apply and create new usage scenarios can a real Super App be born. At present, although technology is updated frequently, we prefer to find the "fixed points of demand" - that is, assuming that the technology is mature, the core needs that users still have, and design the product framework around these needs.
Bai Shuang (Host): Mr. Qiao Haixin, your product has an innovative positioning. What specific impacts does technological change have on you?
Qiao Haixin: Since the beginning of 2023, the development of models and applications has shown a "double - helix upward" trend. Take the social entertainment field where we are located as an example. Early products were mostly simple human - machine conversations (such as chatbots), and then gradually added images and scenario - based interactions. Now, we are exploring social simulations with multi - agent collaboration. Each step of evolution depends on the support of underlying technologies, especially cost and inference speed. For example, multi - agent simulation requires frequent calls to the token API. If the cost is too high, it will be difficult to achieve large - scale implementation. For instance, the paper about Stanford Town in 2023 was very popular. Many people found that after making it interesting, they couldn't bear the cost of running 1000 CCUs simultaneously.
Bai Shuang (Host): Mr. Xu Minwen, as an AI recruitment platform, how has the emergence of DeepSeek affected your business?
Xu Minwen: Our CTO is very excited about the iteration of models such as DeepSeek. From a business perspective, although the improvement of model capabilities has not significantly changed the user experience, we have benefited greatly from open - source technology and cost reduction. Previously, limited by cost, the resume - matching function was only available to 10 consultants. Now, with the maturity of technology, we have enabled all employees to use it, and the efficiency has been greatly improved. In addition, although large enterprises lead in technology, we firmly believe that industry - specific giants will emerge in each era. In the field of AI recruitment, TTC hopes to be the enterprise leading the transformation.
Bai Shuang (Host): The rapid iteration of technology makes many people worried: Will the upgrade of big models make application enterprises lose opportunities? The capital market has also become more cautious about investing in AI applications. Mr. Chen Haici, where do you think the "safe zone" for entrepreneurship is, especially for application - type enterprises?
Chen Haici: This question has been a concern since the establishment of Modu Space. Artificial intelligence has intensified the Matthew effect, but through observing the development of enterprises, we have discovered some patterns. First is the "blind spot of large enterprises." In the past, it was often said to avoid markets worth tens of billions because large enterprises tend to flock to them. However, with the change in the economic environment, large enterprises have also started to move down - market. Therefore, entrepreneurs should not only look for fields with a small market scale but also pay attention to tracks where large enterprises are difficult to respond quickly due to mechanism and resource limitations.
Second is the "destination setting." Entrepreneurship does not necessarily have to aim for listing on the stock exchange as the only goal. Many Silicon Valley enterprises aim to achieve excellence within 2 - 3 years and be acquired. With this mindset, entrepreneurs can focus on single - point breakthroughs and realize value during the window period when large enterprises react slowly. Take the game industry as an example. Large enterprises are used to using resources to build 3A masterpieces and are reluctant to try innovation easily. However, small teams can stand out through minor gameplay innovations (such as combining the gameplay of Genshin Impact with light social interaction). The same is true in the AI era. Entrepreneurs should seize the innovation blind spots of large enterprises.
Bai Shuang (Host): Mr. Qiao Haixin, your product has an innovative positioning. How did you find the entry point? What's your understanding of the "safe zone"?
Qiao Haixin: Our entry point is to explore a brand - new user experience empowered by big models, which was impossible to achieve with previous technologies. For example, using AI to integrate the discrete content created by users is difficult to achieve with traditional models. Regarding the "safe zone," I think the ability to define products is crucial. Entrepreneurs should not only focus on the next step but also think about the direction for the next two to three steps to avoid falling into homogeneous competition. In addition, the core barrier of consumer - oriented products is not technology but network effects and community stickiness. Although these soft factors seem intangible, they are the key to determining the vitality of the product.
Bai Shuang (Host): Do you agree with this view, Mr. Hu Xiuhan?
Hu Xiuhan: I agree that soft factors are barriers, but I also want to emphasize that these factors are not simply dependent on operation but need to be achieved through technical design and mechanism construction. For example, the recommendation algorithm is essentially a technical mechanism that can affect the community atmosphere and content ecosystem. Entrepreneurs need to use technical elements to design unique product logic, which is exactly the blind spot that large enterprises tend to ignore. In addition, with the change in the industry landscape, the consumer - oriented market, which was considered difficult to break into in the past, may present new opportunities due to the "stress response" of giant enterprises' competition and policy adjustments. Young entrepreneurs should dare to break free from empiricism and find entry points in the "blind spots under the lamp."
Bai Shuang (Host): Mr. Xu Minwen, has anyone ever asked if platforms like Boss Zhipin are doing the same thing as you?
Xu Minwen: This question is often asked. However, the core of TTC is the in - depth collaboration between AI and humans. Traditional recruitment focuses on the hard - matching of resumes and positions, while we use AI to modularize and structure seemingly scattered soft information (such as candidates' abilities and job requirements) to achieve precise matching. More importantly, we focus on using AI to amplify the 5% key decision - making power of consultants, which is the barrier formed by the combination of technology and service. In addition...