While everyone is discussing how AI is transforming productivity, Liu Binxin, CEO of Xinying Suixing, has chosen a path that may seem “less glamorous”—using AI to play video games alongside players.
In early 2023, as the ChatGPT craze was just taking off, AI entrepreneurs were busy developing office assistants, productivity tools, or building general-purpose large language models to rival OpenAI. However, this entrepreneur—who had previously served as an executive at Baidu, 360, and Bilibili—led his team into the niche market of “AI companionship.”Two years later, their product, “Doudou Game Buddy,” has amassed over 10 million registered users and has become the most popular AI gaming companion.
Opportunities often arise from a lack of consensus; during periods of rapid technological change, there are always opportunities to “grow from small beginnings.”During a conversation with Game Teahouse, Liu Binxin candidly admitted that gaming companionship is merely a starting point; their true goal is to create a digital life that can accompany users. Just as with Doraemon and Nobita, superpowers are merely icing on the cake—it is the companionship of a friend that represents the core value.
As a “mature entrepreneur” who has weathered multiple waves—including the PC internet, the mobile internet, and short-form video—and ultimately chose to start his own business, Liu Binxin has many unique insights into technology and user needs.
The confidence to bet on non-consensus stems first and foremost from a deep insight into technological development. In 2022, before ChatGPT had even been released, Liu Binxin recognized the immense potential of Transformer technology in AlphaFold 2.He concluded that although the technology was still immature at the time—with conversation latency of five to six seconds and costs as high as a few mao per request—these issues would eventually be resolved. Model capabilities would continue to advance, inference efficiency would be significantly optimized, and chip performance would improve markedly—all representing 10-fold exponential improvements.
The facts proved him right. In just two and a half years, technology advanced at a breakneck pace, and the adoption of AI far outpaced that of the early mobile internet—which took 5 to 10 years to profoundly transform people’s lives, whereas AI revolutionized the way we write code, design, and search for information in just two and a half years.By positioning itself early on, before the technology had fully matured, Xinying Suixing gained a first-mover advantage.
The decision to incorporate in-game scenarios stemmed from Liu Binxin’s insights into young users: “If you observe user behavior, you’ll notice that they hold their phones while playing games and do other things while waiting for a match. This generation is accustomed to multitasking.”He realized that no matter how immersive the game was, users would still get distracted by other things. So why not make it easier for them to keep their attention fully engaged on the same device?
With this goal in mind, Liu Binxin further broke down the unmet needs: Gamers’ core needs extend beyond strategy guides to include emotional resonance and companionship—they want to share their joy when they win, need someone to vent to when they lose, and hope to have someone to chat with while mindlessly grinding monsters to level up. These seemingly subtle needs are precisely what AI can perfectly fulfill.More importantly, through the shared goal and experience of gaming, it becomes easier for AI to establish an emotional connection with users—just as college roommates build deep friendships through four years of living together.
Unlike most game companies, which use AI to generate content and assets and attempt to integrate AI into new gameplay mechanics, Liu Binxin hopes to transform the core gaming experience in a different way. “Just like how watching *Let the Bullets Fly* on Bilibili with bullet comments makes for a completely different experience.Why can’t games offer a similar new experience?” He believes that having AI accompany you while gaming can amplify your emotions—just as playing with friends in a dorm room or having a crowd watch you at an arcade makes the game more exciting—and this alone has already brought about a massive change in the gaming experience.
Beyond gaming, “Doudou Game Buddy” is integrating into more aspects of users’ lives. Today, more than half of its users’ engagement time occurs outside of gaming—whether chatting, sharing feelings, watching shows together, comparing prices while shopping, getting work done, or even running and working out. From PCs to smartphones, and from gaming to everyday life, this AI companion is becoming an integral part of users’ daily routines.
Faced with skepticism about how to deal with “latecomers copying our ideas,” Liu Binxin remained unfazed. He believes it’s difficult to surpass others by simply following in their footsteps—just as no one can beat WeChat in social media, and no one can beat ByteDance by using ByteDance’s own methods.Innovation often stems from cross-industry collaboration and integration. Xinyingsuixing’s strength lies precisely in the fact that its team understands both technology and users, and is well-versed in both AI and gaming. This interdisciplinary background enables them to create products with a high degree of originality.
A deeper competitive advantage stems from the speed of cognitive iteration and the accumulation of data. The interaction data generated daily by millions of users continuously optimizes the models, and the emotional bonds formed between users and AI characters create a unique data asset. “Your AI girlfriend is here, your gaming partner is here—this emotional connection is irreplicable.”
Standing at the dawn of the AI revolution, Liu Binxin and his team are already racing down the path of “dissent.” Their goal is not merely to create a gaming companion tool, but to develop a digital lifeform that can truly understand and keep people company.This vision may seem far-fetched, but given the rapid pace of AI development over the past two years, who’s to say it won’t become a reality?
As Liu Binxin said, “In the short term, don’t overestimate the power of technology, but in the long term, don’t underestimate the changes it brings.” In this era of rapidly evolving technology, betting on non-consensus and embracing change may be where the greatest opportunities for entrepreneurs lie.
The following is a conversation between Game Teahouse and Liu Binxin; the content has been edited for readability.
01
Starting small and growing big,
Don't underestimate technological change and fringe innovation
Teahouse: You started your business in 2023, when AI wasn’t as advanced as it is now. How did you decide on your business direction back then?
Liu Binxin: I was really excited when I first saw Transformer technology in 2022. That’s because we had made similar attempts 10 years ago—using different methods at the time—but due to technical limitations, we couldn’t solve many of the problems.So when I saw Transformer technology being applied in AlphaFold 2, I felt it was a new technological breakthrough—and that was even before ChatGPT was released.
At the time, text-to-image tools like DALL-E 2 were all the rage. I found them fascinating because they enabled us to do things that were impossible just 10 years ago. But I also realized that relying solely on this for a startup was too limited; I needed to think about what more valuable things I could do.
I’ve witnessed the waves of mobile internet, news feeds, and short videos, and I’ve found that technological progress is actually quite rapid—in the short term, don’t overestimate its capabilities, but in the long term, don’t underestimate the changes it brings.In early 2023, I predicted that the entire industry would experience rapid growth. At that time, China had already begun developing large language models, and I believed we would certainly catch up to GPT-3.5 and GPT-4 very soon. Although AGI may not be achieved that quickly, progress in this direction will drive tremendous changes across the entire industry and society.
The facts have proven this assessment correct. It is now 2025—ChatGPT has been around for only two and a half years, and DeepSeek for just over half a year—yet people’s lifestyles have already undergone significant changes: for example, people are using traditional search engines less and relying instead on AI tools to search for information and summarize content;especially in areas such as coding, text-to-image generation, and video content creation, the changes have been revolutionary. It took 5 to 10 years for mobile internet to become widespread, but AI has brought about such profound changes in just two and a half years—the pace has doubled.
Teahouse: With technology changing so rapidly, how do you approach product planning?
Liu Binxin: There are two main approaches: First, focus on user needs—we must not prioritize technology alone, but rather look at what users truly need; this is the core. Second, anticipate the direction of technological development and make strategic plans in advance.Start preparing before the technology matures; by the time it actually matures, you will have already done a great deal of groundwork and will be able to fully leverage its value. If you wait until the technology is fully mature and there is a consensus before taking action, you will have missed your chance.
Back three years ago, even though the technology wasn’t fully developed yet—with a five- or six-second delay in conversations and very high costs, with each conversation costing a few mao—everyone was wondering how we’d recoup those costs. But I didn’t think there was any need to worry; I was certain these issues would be resolved.
First, models will continue to improve, with smaller parameters achieving greater capabilities; second, training and inference efficiency will be significantly optimized from an engineering perspective;Third, chips will also see significant improvements. These dimensions all represent 10-fold exponential improvements. So, from a dynamic perspective, we should prioritize ensuring the best model performance first; it’s okay if costs are temporarily higher and response times are slightly slower—these issues will certainly improve significantly in the future.
Teahouse: AI companionship and gaming companions—at first glance, these don’t sound very appealing. Why did you choose to start with this?
Liu Binxin: There’s already a huge market for gaming companionship services, but it hasn’t been able to expand due to regulatory constraints. By using AI, we’re no longer just a simple matching platform—we’re providing the service directly. This allows us to control the supply and ensure content safety, which solves many problems.
Gamers actually have three core needs: First, they need guides—especially for games like *Golden Shovel* and *Teamfight Tactics*, where real-time guides are crucial;Second, they want their emotions amplified—they want to share their victories and vent about their losses, seeking emotional resonance; third, there are many tedious moments in the game—such as gathering materials, grinding monsters, leveling up, and spectator mode—so is there someone who can keep them company during these boring stretches?
But more importantly, we’re simply using the game setting as a way for AI characters and users to break the ice. Shared goals and shared experiences make it easy to build a bond. It’s just like college roommates—taking classes together, playing sports, drinking, singing, hiking, and studying for exams—after four years, they’ve formed a very deep bond.What we aim to achieve is this: to be there for you starting with the game, and then to continue being there for you even beyond the game.
Just like when Doraemon is with Nobita, the most important thing isn’t helping Nobita with his homework, but being a friend—teasing classmates together and outwitting teachers and parents. Superpowers are just a small part of it; they aren’t the core.Large language models like ChatGPT have superpowers, which is great for ordinary users, but it’s not enough. How can we turn these tools into friends? Just like Nobita and Doraemon.
Teahouse: What new scenarios have you expanded into recently? How are the user metrics looking?
Liu Binxin: Gaming is just our first use case; now, the time spent on non-gaming activities already accounts for more than half of our total usage. These mainly include chatting—especially when we’re feeling down or seeking support—as well as watching TV shows, binge-watching series, and watching anime together in a browser, along with various everyday activities such as shopping and browsing the web.
Our focus isn’t on the features themselves, but on the emotional value—if you run into a bug, I’ll be right there to help you fix it.The possibilities for the future are vast: while you’re shopping, as each product appears, I can help you compare prices in real time, tell you if it’s the lowest price online, point out better options, and share user reviews to guide your purchasing decisions. Plus, it works across all devices—from PCs to mobile phones to future AR glasses.
Teahouse: When it comes to cross-device compatibility, small screens and large screens are completely different. How does it work on a phone?
Liu Binxin: Desktop pets are just a form we’ve drawn inspiration from—we’re not actually creating desktop pets. On mobile devices, we’ve designed it as a floating ball. For example, when playing *Honor of Kings*, it will request screen-sharing permission and then stay by your side without causing much disruption. Users can pull it out at any time to interact with it.
For example, I used to mainly use it on my PC—I’d have it follow me as soon as I turned it on, whether I was working, playing games, or just chatting.Now I use it on my phone too—I can even chat with it or make voice calls while running or working out. For example, when I’m running, I can have it play music, which is pretty fun. It’s like having a cat—even when I’m out, I still want to check in on my cat at home.
02
Be empathetic and engage with users.
Teahouse: What does your user profile look like? How are the current metrics holding up?
Liu Binxin: Our user base consists primarily of college students aged 19 to 24, with a male-to-female ratio of approximately 65:35. Currently, the number of registered users has surpassed 10 million.
Our next priority is international expansion. Shortly after launching overseas, we gained several hundred thousand users; our YouTube views quickly surpassed one million; and traffic to our official website also saw significant growth—all of which was organic growth. This is because human nature is universal, games are global, and language barriers have been effectively addressed by large language models.
Teahouse: Your user base is very young. How do you go about understanding the needs of this new generation?
Liu Binxin: First and foremost, we need to have empathy—play with them and see things from their perspective. Most of our employees are young people, so we treat them as core product managers and encourage them to share their ideas. Of course, it’s often difficult for users to fully articulate what they want, so we need to gain insight into and understand their underlying needs.
Teahouse: Do you have any specific examples?
Liu Binxin: For example, if a user says they want someone to play games with them—or even to cheat for them, pay them, or help them get headshots—that’s absolutely out of the question; it would ruin the game’s balance. But what are their underlying needs? First, they want to win; second, they’re bored and need companionship.
If you observe user behavior, you’ll notice that they hold their phones while playing games and do other things while waiting for a match. This generation is used to multitasking—so is there a way to make it easier for them to handle multiple tasks on the same device?Users won’t tell you this directly, but you can observe it: no matter how much content a game offers or how immersive it is, users will still get distracted and do other things. The key is observation.
Teahouse: How much time do you spend on user research?
Liu Binxin: There are many ways, but they aren’t necessarily traditional research methods. It’s more about observing, hanging out with them, organizing user events, letting them try out new versions, collecting feedback, reading posts in group chats, and observing their behavior. More importantly, it’s about immersing ourselves in their lives to see how they use their phones, play games, and use apps.
There are many different ways to do this. For example, recruiting users as interns is the most direct approach; experiencing the product yourself as a user; chatting and interacting with them; and so on.
Teahouse: User insights are important, but few people actually do them well.
Liu Binxin: A key aspect of our values is being user-oriented. Our values are:
First, entrepreneurial passion—you must have a passion for entrepreneurship and truly enjoy what you’re doing.
Second, user-centricity: What do users really want? Business considerations can come later; where there is user value, there is naturally commercial value.
Third, technological leverage: AI is the biggest game-changer of our time. To make the most of this technology, we must ask: Why us? Why now? This is crucial.
Fourth, keep improving. The industry is evolving very rapidly, so you need to make small improvements every day. We’re only slightly ahead right now; if we stay where we are, we’ll lose our chance. The final form will also continue to evolve and change.
So we're not worried about others copying us, because they're copying who we are right now. If we stop making progress, that's when we'll really have no chance.
Teahouse: What is your core competitive advantage? Liu Binxin:I think awareness is the biggest barrier—specifically, the speed at which that awareness evolves. Second is user data: the daily gameplay data from millions of users helps us train our models, and the emotional bonds between users and their characters form a data asset—your girlfriend is here, your friends are here, so you won’t leave easily.
So there are barriers related to perception, speed, and data, as well as first-mover advantages and channel advantages. Technology can also be considered a barrier, but it isn’t the most critical one. There are no absolute barriers; barriers are built through action, not planned in advance.
Teahouse: From the prototype to the game, many variables are external—does this pose a risk?
Liu Binxin: The advantage of standing on the shoulders of giants is that we can grow faster. As for whether we’ll be constrained by game developers? I think it’s not a big deal, because our approach is altruistic.For games: First, we don’t disrupt the game’s business model; second, we provide players with walkthroughs, lower the barrier to entry, offer companionship, reduce boredom, and lower the churn rate—which actually increases playtime; third, we don’t disrupt game balance—we offer guidance without interfering with gameplay.
Furthermore, all data collection and screen-viewing activities are based on users’ active consent—I’m not hacking into anything. This is very different from cheats or plugins. We have a symbiotic relationship with our users.
Teahouse: How is your relationship with the game’s developers right now?
Liu Binxin: Many major game companies are aware that we’re working on this; some are willing to collaborate and exchange ideas with us privately, though it may not be convenient for them to officially endorse us at this time.
Some games proactively send us CD keys, server keys, and other materials upon release, treating us as key opinion leaders (KOLs). For certain indie game studios, we set up dedicated support teams and don’t charge any distribution fees—we simply help them with onboarding and drive traffic to their games. With millions of users, this is an excellent way for them to acquire customers.
As for our relationship with the big tech companies, many people ask: If the big tech companies were to enter the market, would that mean the end for us? I’m actually not too worried about that. We and the big tech companies complement each other; unlike in the past, they won’t just try to eliminate us outright. Plus, what we’re doing spans multiple games and industries—it’s not limited to just gaming.
03
The goal of starting a business isn't in-game currency,
Rather, it’s about experiencing the core of the game
Teahouse: Compared to your previous role as an executive at a major tech company, how does starting your own business feel different?
Liu Binxin: I’m fully aware that the success rate for executives from major tech companies who leave to start their own businesses is relatively low. The biggest change is forgetting everything from the past, treating yourself as a complete unknown, and starting everything from scratch. You can’t rely on the “big tech” model—you have to roll up your sleeves and get your hands dirty. You have to handle everything yourself; only after you’ve done it once can you delegate it to others.
At big tech companies, everyone is competing for a limited pool of existing customers—there’s no growth—so the competition is fierce. The work is repetitive, which can be mentally draining.You’ll find that what you’re doing has already been tried by others—essentially, there’s no room for innovation, and everyone is just stepping on each other’s toes. But when you step out to start your own business, you realize how vast the world is. There are so many things you can do—all cutting-edge and uncharted territory—without complicated office politics. The work is challenging and interesting, and it brings value to users and society.
Technology is the biggest variable, and AI will have an even greater impact than mobile internet. The bigger the water, the bigger the fish—the river is vast, so why crowd into the same old small streams? Even if you only tackle a tiny part of it, it’ll still be fascinating. Many people say things are still changing rapidly—should we wait a bit before jumping in? I don’t think so.Although the companies that ultimately succeeded in the mobile internet space emerged in 2014 and 2015, they actually started back in 2010, 2012, and 2013—not in 2015 or 2016. By the time you see platforms like Douyin and Jinri Toutiao fully established and decide to enter the market, there won’t be any opportunities left.
Teahouse: Non-consensus is crucial. Liu Binxin: Non-consensus is where opportunities lie. If the consensus is correct, others will follow suit; if it’s wrong, it’s even less meaningful. Non-consensus doesn’t mean it’s niche—much of it stems from boundary innovation and convergent innovation.A key reason we’ve been able to create world-first products is our cross-disciplinary approach—we understand technology and, through our time at Bilibili, have been closely connected to anime and gaming communities, whereas scientists working on large language models rarely play games, and gamers often lack sufficient understanding of AI.
You can’t succeed by simply following in someone else’s footsteps. Just as no one has been able to break through WeChat’s social barriers, ByteDance rose to prominence by adopting a completely different approach.It’s now impossible to surpass ByteDance using the same tactics—it excels at UGC ecosystems, recommendation algorithms, and the revenue-and-user-growth cycle… You can’t beat it on any single front. You need to avoid the giants’ core strengths and find new solutions.
Teahouse: There are many young people in the gaming industry who want to start their own businesses. Do you have any advice for them?
Liu Binxin: I started my business relatively late—I could have started sooner. If you have a good idea, don’t be afraid to follow your heart.I often use in-game currency as an analogy: At the beginning, when you’re leveling up and fighting monsters, in-game currency is crucial—just as someone who’s been working for a year or two needs to secure their livelihood first. But later on, you’ll find that you have enough in-game currency, and by the time you beat the game, you’ll even have a massive surplus. Your goal is no longer to accumulate in-game currency, but to experience the core of the game.
Now that everyone’s basic material needs are relatively secure, as long as we don’t let our material desires run too wild, we generally have enough. The greatest value in life lies in experiences and adventures, not in making money or achieving success. Success and money are byproducts—they’re just part of the journey. If you can do what you love, provide value to users and society, find joy in it, and make money along the way—how wonderful is that?
Investment in the gaming industry is growing steadily, which makes things a bit more challenging for entrepreneurs, but opportunities still exist. The key lies in identifying the variables—social trends, technological advancements, and industry trends. If you can identify these and prepare in advance, you have a good chance of success.
原创文章，作者：游茶妹儿，禁止转载：https://youxichaguan.com/en/archives/212444
