Xinying Sui Founder Liu Binxin: Building AI Products That Don't Compete for Users' Time | Xingxing PORTFOLIO
Build AI Products That Don't Compete for Users' Time
In a recent interview at the AIGC Industry Summit, Liu Binxin, founder and CEO of Xinyin Suiying, shared four contrarian insights: How can AI create joy? How should we respond to shifts in industry technology and user acceptance of AI? How do we genuinely solve user needs and let time be our ally? How can a deep understanding of multimodal AI help us seize the next wave of opportunity?
The following is an exclusive interview with Liu Binxin at the AIGC Industry Summit. This article is republished with permission from QbitAI (ID: QbitAI), author: Aifeisi.
Beyond using AI as an efficiency tool, could it also bring people some happiness?
A thought from over two years ago set Liu Binxin, founder and CEO of Xinyin Suiying Technology, on his own entrepreneurial path:
Use AI technology to address the social pain points in gaming scenarios. Position the product around "companionship" rather than competing for users' time, making AI a gaming partner that understands players' emotions.
Moreover, the MVP model commonly used in the mobile internet era no longer holds in the AI age. We can now do more to predict which scenarios will become genuine necessities in the future, and which technologies can meet those needs.
Facing two major industry shifts — "breakthroughs in large model technology making complex reasoning possible" and "rapidly growing user acceptance of AI" — Xinyin Suiying has consistently anchored itself to its own unique core value.
To fully capture Liu Binxin's thinking, QbitAI has edited and organized his remarks without altering their original meaning, hoping to offer you more inspiration.
The China AIGC Industry Summit, hosted by QbitAI, is a cutting-edge AI industry conference. Over 20 industry representatives participated in discussions. It drew more than 1,000 offline attendees, 3.2 million+ livestream viewers, and 20 million+ total impressions.
Key Topics
- "Using AI to create joy" — this is what we hope to do.
- Don't compete with users for their time; accompany them instead.
- Advances in large models have made the once inconceivable possible, improving the experience.
- Collaborate with ecosystem partners upstream and downstream to form a holistic, all-encompassing wrap that fully satisfies user value.
- Multimodal understanding may be the key driver of the next wave of growth and new scenarios/applications.
"Using AI to Create Joy"
Thank you, everyone. I'm delighted to be here at the QbitAI conference to share with you all.
Over two years ago, when I saw Transformer and ChatGPT, I wondered: beyond efficiency tools, could we use AI to bring people some happiness? "Using AI to create joy" — this is what we hope to do.
On another front, when looking for scenarios to apply AI, I no longer wanted to compete for users' time. From my mobile internet experience, I saw that users' time was already saturated — eight hours already spent online, with games to play and short videos to watch. I believe we shouldn't compete with users for time, but rather accompany them. This was the original idea behind Doudou Gaming Companion.
At the same time, we believe that as AI technology evolves, truly meeting users' real needs still requires starting from the users themselves, connecting with their experiences.
We found that Bilibili's young users share one very typical trait: they all love gaming. Gaming is a social topic for young people. In the gaming process, a major pain point is finding friends to play with. Many games are single-player, played alone. In this process, could there be friends to accompany them and provide more emotional value? This was our starting point.
Simply put, the core of Doudou Gaming Companion is entering existing scenarios rather than creating new ones, satisfying existing needs rather than inventing new ones. To grow big, you must first start small.
The core function of Doudou Gaming Companion is having characters with various personalities and styles in the app. Beyond chatting and singing with you, more importantly, they can game with you. "They" can see your gameplay, empathize with you, celebrate your highlight moments, or roast your clumsy moves, giving us extra emotional value during gaming.
What's the underlying logic of this need? It's satisfying the perfect social, showing-off, and achievement needs within gaming scenarios.
People all like to feel valued. For example, in an arcade, when you play Street Fighter, there's always a crowd of friends around, eager to try or cheering for you, making the game feel captivating. You want your gaming achievements to be shared and recognized in real time, bringing rich emotional value and doubling the joy.
Two Major Shifts in the Environment
From product launch to now, the environment for building AI products has undergone some significant changes.
First is the shift in industry technology, particularly at the large model foundation level. Advances in large models have made the once inconceivable possible, improving the experience.
For example, strategy search — AI search used to have various hallucination issues, requiring all kinds of RAG for assistance. Before reasoning models existed, writing prompts for RAG was quite an art. The order of RAG results, the sequence of prompt text — these mattered because there was frequent interference. Back then, more background knowledge wasn't always better.
Now with reasoning models, we can throw in everything we search for without worrying too much about ranking or order. The model can reason and answer well on its own. This is a very obvious change.
The new version of strategy search that Doudou Gaming Companion is about to launch achieves over 90% accuracy, basically becoming usable. The same applies to AI search in the industry — before reasoning models, people felt AI search wasn't very useful. After reasoning models, AI search became very effective. This is a huge change, technology driving massive transformation.
The second change is user acceptance of AI. People are increasingly being exposed to AI. Early last year it was Doubao, later it was DeepSeek.
Actually, AI penetration in China is very high overall, higher than in many overseas regions. People have gradually grown accustomed to using AI and AI scenarios. This is a very good thing — people are more willing to accept and explore these things. I think this is also good for those of us building products and applications.
Genuinely Solving User Needs, Letting Time Be Our Ally
From one-hit wonders to evergreen products, the key factors in maintaining user stickiness and long-term advantage mainly include several points:
First, the most important thing is still to genuinely and authentically solve user needs, starting from user pain points, truly giving them an end-to-end complete experience. Even if only part of it can be solved by AI, the non-AI parts must also be done, giving everyone a truly end-to-end complete experience. This matters most.
Second, it's important to let time be our ally. As technology develops and the industry changes, more and more technologies and applications keep stacking up, riding the wave of the era and delivering more features that exceed expectations. This is quite important.
Doudou Gaming Companion maintains user stickiness through several approaches —
We have quite a few character designs. Users' daily chats and interactions with these characters generate data that forms emotional bonds.
With this rich chat data, plus the ability to see the entire screen during companion scenarios, AI can understand much about user behavior, habits, and preferences, forming precise user profiles.
Based on these profiles, we continuously train the AI characters on our platform to make them more personalized, even more than users know themselves.
Additionally, on the ecosystem level, with increasing UGC and PUGC data accumulation, our collaboration with ecosystem partners upstream and downstream — including OEM manufacturers and gaming partners — enables us to form a holistic, all-encompassing wrap that fully satisfies user value.
Next Wave of Opportunity: Multimodal Understanding
For the key drivers of the next wave of growth and new scenarios/application opportunities, I personally think one direction worth watching is multimodal understanding.
Many people are working on multimodal generation — text-to-image, text-to-video. But multimodal understanding also has huge opportunity. Understanding here means comprehending the entire video, not just what objects appear in a single frame.
We can imagine a scenario: if AI watches a 90-minute soccer match with you, discussing everyone's performance, the coach's tactical arrangements, team strategies; or watching an NBA game, playing a game, then reviewing how that Honor of Kings match went, discussing whether the team fights were good or bad — these would all be very interesting scenarios.
The mobile internet and MVP model no longer hold in the AI age. Using the smallest model and shortest path to test whether a need or product is right — this is very likely to produce changes and deviations. In the AI age, we can do more to predict which scenarios will be genuine necessities in the future, and which technologies can meet those needs.
Proceed along this path, let time be your ally, and there's huge opportunity.
Identifying consensus opportunities is very difficult — this is what big companies do. We startups shouldn't do this.
(Could you share a little about which new scenarios and features Doudou might explore, given the significant growth in multimodal technology in 2025?)
Discussing a game with you based on the match — say League of Legends or Honor of Kings — talking about positioning, voicing your little thoughts or roasting you a bit. Many gamers would find this quite interesting.
Originally they keep their mics off in-game. They usually don't like playing against bots, feeling there's no communication with a machine. Including the original Dota AI, it was purely machine behavior during gameplay, players couldn't direct it.
But now, with Doudou Gaming Companion's understanding-based reinforcement learning, we can truly achieve AI teammates who strategize with you. No need for four opponents to flame their noob teammates for dragging them down — the AI fully gets your little thoughts. In Minecraft, the AI companion helps you build castles together, knowing your goals. You just need to provide the design ideas, and it handles the repetitive work.
Our original intention is to give players more experiences like this, and we're actively developing such features.
Heart Capital was founded in 2022 as an early-stage venture capital fund focused on technology and digitalization in China. The team is primarily composed of Yan Han, founding partner of Lightspeed, core investors, a chief financial officer, and seasoned investors from industry. The team's past investments include Series A investments in Xpeng Motors (NYSE: XPEV, 09868.HK), Full Truck Alliance (NYSE: YMM), as well as FinVolution Group (NYSE: FINV), RoboSense (02498.HK), Baichuan, Manman Cold Chain, Dedao, World Logistics, Micro-nano Starry Sky, LandSpace, Lanhu, Starfield, and others. Rooted in China with a global outlook, Heart Capital is committed to finding true value in non-consensus. It respects the value of "people" and champions the potential of the "heart," looking forward to accompanying more young Chinese entrepreneurs to strengthen China and go global.
