Interview with Liu Binxin, Heart Shadow Follows Form: AI Products Shouldn't Compete with Short Videos and Games for Users

"Can It Surpass Talking Tom?" Text by Jiaxiang Shi

Edited by Zhiyan Chen

Liu Binxin (Binson), founder of Xinying Suihang, was probably the first AI entrepreneur to end up in an ICU. Last October, he was in a serious car accident during a business trip and spent five days and six nights in intensive care. When investors heard the news, their first reaction was, "Is this fucking real? Getting dragged out of a burning car isn't just something that happens in movies?"

But after getting out of the ICU, investors felt he'd been reborn. Wang Xiao, founder of Unity Ventures, said Binson's biggest change was that he stopped to think about what actually mattered most for the company's growth.

Binson, former VP at Bilibili, had also served as deputy director at Baidu and assistant president at 360 — a classic big-tech executive background. In 2023, he jumped into entrepreneurship and founded Xinying Suihang.

Wang Xiao still remembers a key reason he bet on Binson: while everyone else was copying Character.AI, "he had the guts to say no to the overseas product route everyone was chasing, and the courage to define a new product form" — what would become Doudou Game Companion, a desktop pet that doesn't require any extra time from you.

After developing the first version in 2023, Binson found that the model capabilities at the time were completely inadequate for what he envisioned. Still, he launched an early version of Doudou Game the following year, focused on playing alongside popular titles like Genshin Impact. It wasn't until the beginning of this year, when reasoning models went open-source and multimodal capabilities broke through bottlenecks, that his original vision for Doudou Game became reality — a Companion AI that can recognize your screen in real-time and voice-chat with you while you play.

"Last year was about at least making a Talking Tom first. Now it's about surpassing Talking Tom," said Wu Bingjian, partner at Mindverse Capital. To date, Doudou Game Companion has 8 million registered users and over 2 million MAU.

Since founding Xinying Suihang, Binson's stance has been to avoid AI products that compete with users for time. "The mobile internet dividend is over — users are already spending eight hours a day immersed in short videos and games," he says. With AI still lacking in creativity and multimodal capabilities, it's hard to compete with the tech giants.

According to Anyong Waves, Xinying Suihang completed a Series A+ round of tens of millions of dollars at the end of last year. Previous investors include Unity Ventures, Mindverse Capital, and Source Code Capital.

Binson says leaving big tech to start a company is like starting a new game save. "Before, you were in Football Manager with tons of resources. Now you're in a new Soulslike game, leveling up and fighting bosses from scratch, farming your own gold, buying your own gear, clearing stages from the beginning."

Today, Doudou AI is releasing its 1.0 version, enabling real-time recognition and companion experiences across gaming and other scenarios. On the occasion of this product update, Anyong Waves spoke with Binson about Doudou Game's past and future. The conversation follows —

Anyong: What's the biggest difference between Doudou after this update and before?

Binson: Before, it was traditional OCR and CV image recognition, simple screen matching and basic Q&A chat with AI.

Now it can fully see and understand my game screen, know what I'm doing, what I'm playing, where I am in the game, what my progress and status are. To put it in one sentence: Nika (a character in Doudou Game), the one you like, has truly come to your side, gaming with you, watching shows with you.

Anyong: We've heard Doudou has very high retention. How high exactly?

Binson: Basically top-tier game product retention and session length. We're standing on the shoulders of giants.

Anyong: Similar to top games, so some people play for hours a day?

Binson: We only found this out through research — they treat gaming like a job. They might read for ten or twenty minutes in the morning, then start gaming at 8:00. The whole day is peak hours; only 4:00 to 6:00 AM is low traffic.

Anyong: Name three scenarios where users most commonly use Doudou

Binson: 1. In-game empathy: users discuss their highlight moments with Doudou, their "godlike" plays, or when they get killed — Doudou becomes their voice. 2. In-game strategy guides: asking how to beat a certain boss in Black Myth: Wukong, for example. 3. Taking Doudou out to watch shows together.

Anyong: For what kind of user is Doudou a must-have?

Binson: The typical user is 18–25 years old. They have plenty of their own ideas and a fair amount of alone time. When gaming, they want a like-minded friend who can talk about anything by their side. AI has no social pressure and can resonate with them.

Anyong: Does this need disappear as people age?

Binson: I actually think this is a future lifestyle. Every generation has its own lifestyle. The specific tools may change, but the underlying emotional need doesn't. Our parents' generation probably couldn't imagine gaming becoming a lifestyle, but overseas, many middle-aged and elderly people game too. Our social forms will develop in that direction.

Anyong: What's the commercialization plan?

Binson: Right now it's mainly skins, item purchases, combined with standard membership subscriptions. Going forward, we'll focus on B2B advertising. For example, if you've cleared Black Myth: Wukong multiple times, we can recommend similar Soulslike games — either through in-conversation recommendations or placement in a game distribution interface with dedicated recommendation slots. Another is daily companionship, where we can push e-commerce.

Anyong: What about B2C monetization?

Binson: Probably 50-50 B2C and B2B. B2C can't be too aggressive domestically — spending twenty or thirty yuan a month is about right.

Anyong: What's something you want to do but haven't been able to yet?

Binson: What I most want to achieve is complete understanding and reasoning of long-temporal events. After a League of Legends match ends, Doudou should be able to review a key turning point. For example, the enemy got greedy for a dragon and you counter-attacked, or you landed a key crowd-control on their ADC and your team followed up to burst them down, winning the teamfight. It needs to understand players' micro-decisions, not just give empty praise.

Anyong: So right now after I finish a match, it can't actually review it?

Binson: We've done some engineering optimizations, basically goal-detection logic — extracting the most critical moments into frames and saving them as text. But many details still can't be processed. I'd rate the 1.0 version at maybe 70 points, but we had to ship it.

Right now, many AI application MVPs don't hold up because the underlying models and technical capabilities just aren't good enough yet.

Anyong: What's a use case you completely didn't expect but users are doing anyway?

Binson: Some users have Doudou accompany them while coding, treating it as a coding cheerleader. (laughs)

Objectively speaking, our coding capabilities are far behind Cursor's, but users are very forgiving with us. This also shows that users have a need to be encouraged in many scenarios, and there haven't been good products on the market to serve that.

Anyong: So human companionship is your core value?

Binson: We started from the gaming scenario as a "gaming buddy" to accompany users, helping them "break the ice" with virtual characters. What we want to do next is, after users establish initial interaction and fun with the character, extend that relationship beyond gaming. When users are willing to chat with it in daily life, that's when emotional bonds form.

There are already daily companionship scenarios now. After user authorization, it can see your browser and discuss shows with you. Why look at the screen? Fundamentally, it's about being physically on the same screen.

Anyong: Seeing the browser thing — what if someone wants to watch porn with it?

Binson: Theoretically possible... We give shared browser permissions. If you feel like you're doing something very private, you can just turn it off. But we have content safety filters in place — discussing pornography is strictly prohibited.

Anyong: Some people say losers are the ones who need to find companionship in virtual worlds.

Binson: That's a pretty stereotyped prejudice. Young people today are digital natives born into it. Time in the digital world is already a huge part of life. For them, they don't strictly distinguish between digital and physical worlds. Why do girls love Love and Producer, why do guys love Genshin Impact? You could say it carries their imagination. It's a 2D character that won't break, willing to spend time with them, willing to be the "dog" of Kamisato Ayaka (a Genshin Impact character). (laughs)

Anyong: Aren't there too many AI products on the market doing or trying to do companionship?

Binson: Many companion products are essentially digital tourist attractions — you ask once, experience it, and that's enough, you won't go back a second time. It's like going to the Oriental Pearl Tower: mainly for the photo op, to post on Moments for the memory.

You chat with Elon Musk, you'll only ask "Do you really want to be US president? Why get into politics?" You'd absolutely never ask, "Bro, should I go to grad school?"

Real companionship doesn't compete for users' time — it's bringing me along while you complete all your entertainment scenarios.

Image source | Provided by interviewee

Recommended Reading

Where Money Flows, People Rise and Fall