MaKe｜MaHui Member LynkSoul's AI Product Gains 1 Million Users in 10 Days After Launch

One million new users in ten days... The AI gaming companion product "Doudou AI" made waves among gamers and the broader market the moment it launched, and the company behind it is MaHui member LynkSoul. Founder Binson Liu recently sat down for an in-depth media interview, walking through the product's development details and underlying logic.

• This article was originally published by Jiemian News, reported by She Xiaochen, edited by Wen Shuqi.

After stepping down as vice president of Bilibili, Binson Liu embarked on his first-ever startup journey.

On August 28, at a conference hosted by his "old employer" Baidu, Liu announced that Doudou AI had surpassed 9 million users and simultaneously released its overseas 1.0 version, officially entering North America, Japan, and other regions. By this point, two years had passed since he founded LynkSoul. Just the week prior, the company had launched Doudou AI 1.0. During its testing phase, the product had already accumulated 8 million users; within ten days of official launch, that number grew by another million.

Recently, Binson Liu spoke with Jiemian News in an exclusive interview, diving deep into the development and thinking behind Doudou AI.

The concept behind Doudou AI as an AI gaming companion is straightforward — the AI plays the role of a gaming buddy, offering players strategy guides, post-match analysis, voice chat companionship, and real-time conversation. When using the product, the IP-character-based AI functions somewhat like a desktop "digital pet," but considerably smarter.

Doudou AI's user interface

But delivering genuine emotional value through a companion product is no small feat. Liu told Jiemian News that the biggest technical challenges for an AI gaming companion boil down to three things: first, getting the AI to actually see the player's screen and in-game situation; second, understanding the player's emotions; and third, remembering the player's experiences. Accordingly, Doudou AI 1.0's core innovations center on real-time perception, long-term memory, and seamless companionship — with "perception" being the toughest technical hurdle to crack. Liu explained that this tests the model's ability to understand both breadth and depth, meaning perception of the screen itself and the player's surrounding context.

In reality, from late last year through early this year, progress on multimodal models fell short of the team's expectations. Liu felt that previous approaches relying on OCR (optical character recognition) for screen matching didn't give AI a genuine understanding of what was on screen. So LynkSoul spent over six months researching VLM (vision language model) technology. Doudou AI's real-time perception feature now employs VLM to break through the AI perception layer, enabling real-time recognition and companionship across gaming, video, work, and study scenarios.

LynkSoul was founded in April 2023, when market attention was fixated on large language models and domestic AI companion startups were scarce. But by the second half of 2023, numerous companion products had flooded the market. Doudou AI's development predated this wave, yet its launch and marketing remained relatively low-key, with an extended testing period before official release.

In Liu's view, the primary reason was simple: "the technology wasn't good enough yet." Constrained by model development and numerous technical bottlenecks that couldn't be solved overnight, Doudou AI remained in beta testing — yet still accumulated 8 million users. Liu sees this as proof of intense user demand and anticipation. It wasn't until August this year that he judged the product had reached maturity, with qualitative improvements in performance, prompting the official 1.0 release.

As for why he chose to build an AI companion product, Liu's answer comes as no surprise. With previous stints at Baidu, Qihoo 360, and Bilibili, he brings extensive "big tech experience" in commercial product development. He doesn't shy away from summarizing that in the previous wave of internet growth, he was among those who "rode the wave."

After leaving Bilibili, Liu began exploring AI entrepreneurship. On one hand, he sensed gamers' demand for "companionship." On the other, he firmly believes AI companion products should start as software, with eventual cross-platform account integration. In his view, hardware timelines stretch longer and iteration slows.

In just over two years since founding, LynkSoul has completed four funding rounds. Regarding the intersection of AI and gaming, investors typically ask about market size and monetization — and plenty of them "don't get" the value of gaming companionship.

On the choice of entrepreneurial direction, Liu wants to emphasize something beyond the obvious: his read on technology. He argues that building products in the AI era is fundamentally different from the Mobile Internet period — you can't rely on simple A/B testing to determine product direction. When technology remains immature and is still evolving, a product's current state doesn't represent its future potential. Therefore, "predicting" where model capabilities are headed becomes crucial. "When model capabilities reach where you predicted, your business succeeds — you've taken a major step forward."

Returning to the product itself, since testing began in June last year, Doudou AI's game coverage has expanded from open-world and anime-style games to SLG and beyond. Liu believes AI companionship holds value for any game meeting three conditions: first, significant stretches of relatively repetitive, low-information-density time; second, game segments where players' emotional value needs amplification; and third, scenarios where players need strategy guidance.

Currently, Doudou AI's user base skews 18–25 years old, with roughly 70% male. Liu notes that as more IP characters are added, the female user ratio is rising. The official IP selection includes original characters with anime aesthetics as well as popular virtual streamers and content creators from Bilibili.

This year, Doudou AI has accelerated its marketing push. The company's revenue currently comes from skins, virtual items, and subscription memberships. But looking further out, Liu sees greater monetization potential in advertising — for instance, having AI characters make recommendations to users, spanning games, e-commerce, and even food delivery ads.

Doudou AI's IP selection. Image source: official website screenshot

2025 has been dubbed the breakout year for AI applications. AI companionship has become a white-hot赛道, with Robopoet recently completing a multi-million RMB angel round, drawing industry attention. In the Agent space, following Manus, sector-specific Agent products have been emerging one after another.

Doudou AI's ambitions extend beyond gaming. Going forward, it hopes to expand into TV watching, online shopping, even studying and work. Liu's confidence stems from what he calls gaming's "electronic prison" quality — chatting and socializing during those two to three hours in a closed environment comes naturally, making gaming an effective gateway. Conversations can pivot from games to life and TV dramas. Doudou AI's data shows that non-gaming usage already accounts for half of total user time.

Yet it must be acknowledged that the market hasn't offered many correct answers on how AI applications can successfully monetize. This is a question every entrepreneur must grapple with, Liu included.

How to make money comes down to how the company positions its AI application. One thing is clear: LynkSoul has no intention of making Doudou AI a pure utility product. "Because pure utilities get migrated away, lack emotional connection, and are easily replaced," Liu argues.

But Doudou AI does have utility attributes — what Liu calls the "hook." "For many users, strategy guides are their first touchpoint with the product. On that foundation, it becomes something you can chat with, gradually forming an emotional connection, and naturally expanding into other scenarios." A simple analogy would be Douyin — once you have the community, users progressively move into shopping, video editing, and so on. For now, though, Doudou AI isn't overthinking monetization; Liu's priority is scaling the user base first.