100 AI Creators is a weekly series featuring conversations with China’s leading minds in artificial intelligence. As technology evolves, their perspectives shed light on the ideas driving the AI era across borders.
Three months after raising USD 1 million from Jinqiu Capital, Wang Dengke’s startup Duxiang AI has released its first hardware product: the Xiangmeng Ring, now on sale.
Wang is regarded as one of the leading product thinkers in China’s artificial intelligence sector. An investor connected to ByteDance once described him as someone with “exceptional product taste” whose projects consistently find traction. Media outlets have called him a “hitmaker,” noting how his launches tend to spark industry-wide discussion.
Duxiang, his latest venture, has followed that pattern. With minimal marketing, the AI companion app is said to have drawn about one million registered users and 100,000 daily active users within a year.
The app allows users to create AI-driven characters and form emotional connections with them. So far, users have reportedly created about two million characters: roughly half are original, 40% are inspired by film or anime, and 10% are modeled after relatives or pets. “The creativity is staggering,” Wang said. “Some users write character profiles with 7,000–8,000 words. One even created every Ultraman variant. I checked. It was all original, not copied from the internet.”
The broader market for AI companionship is growing quickly. AI Now! estimates that, as of July, global downloads of such apps had reached 220 million. A US survey found that 52% of teenagers regularly converse with AI. “Technology makes people independent but also lonelier,” Wang said. “In the future, one-third of someone’s social circle may consist of AI-driven ‘friends.’ I want to make sure that when that shift happens, I can help build something good.”
By “something good,” he means enabling users to form meaningful emotional bonds with AI.
To that end, Duxiang avoids real-time chat and blocks sexual content. Instead, it is designed around a seven-layer relationship framework that reflects dimensions of human connection: familiarity, trust, emotional resonance, memory depth, interaction frequency, emotional understanding, and a sense of long-term companionship. For example, when a user shares something personal, the AI can recall it later to add depth to its responses. If a user is feeling low, the AI may also reply in ways that foster emotional understanding. Frequent interactions strengthen the sense of long-term companionship.
A character hard to define
Wang himself is not easy to pin down, in part because he is an accomplished writer. He frequently publishes essays on WeChat and Jike. One of his most widely read pieces profiled his friend Xie Yang, founder of the agentic browser Fellou.
He describes himself as a grassroots entrepreneur from Sichuan. After dropping out of college, he entered the mobile internet sector and has spent a decade creating products he calls “not particularly successful.” They didn’t make him wealthy. His English is limited, and he has never been to Silicon Valley.
Still, the journey has brought other rewards: close friendships in Beijing, weight loss from 200 pounds to 135, and better cooking skills. He once mastered pork chops on par with restaurant fare and discovered that fermented soybean sauce elevates chicken.
Reading his WeChat posts feels like following a reality show about a founder’s life. He comes across as straightforward and easy to understand.
His projects range from blockchain protocol tools to a writing style tester, a digital goods marketplace, the viral Hong Hong simulator program, and now Duxiang. At first glance, they may appear unrelated, but they share certain traits: light, clever, playful, and aesthetically polished.
Rather than viewing him strictly as an entrepreneur, it may be more accurate to say Wang uses technology as his creative medium. Just as writers use words or musicians use notes, he uses products to express a worldview.
In his essay about Xie, he reflected on his own entrepreneurial drive:
“Good or bad, prosperity or decline, no matter what, we are living through this era ourselves.”
For Wang, entrepreneurship is about pursuing intense experiences, shaped by his early reading of “Zhuangzi” and belief in following the natural flow of things. He has said he values realizing unusual ideas more than chasing wealth or fame.
Our interview with Wang coincided with the release of Duxiang’s first hardware product, the Xiangmeng Ring.
The woven bracelet comes in nine colors and sells for RMB 29 (USD 4). A presale on Xiaohongshu moved 5,000 units, which Wang described as “far beyond expectations.”
On the day of the interview, Wang had just received the first batch of bracelets. He eagerly shared them with friends on WeChat, urging them to try the product. In conversation, he discussed the thinking behind the design, why he wanted to venture into hardware, the challenges facing AI companionship, and his approach to commercialization.
The following transcript has been edited and consolidated for brevity and clarity.
Forming deep connections with AI
AI Now (AN): Why did you choose to develop a wearable for Duxiang, and what kind of experience does it provide?
Wang Dengke (WD): The function is simple. You bind it to a character in the Duxiang app. Once bound, it can’t be changed, even if the device breaks. Each time you tap it to your phone, the NFC (near-field communication) chip opens a page to summon that character.
There are three interaction modes. First is care: every summon brings comforting words. Second is check-in: like an intimate partner, the character tells you what it’s “doing.” Third is guidance, which is more mystical: you ask a question, and it gives you a clear answer.
For example, my AI companion is modeled after Crayon Shin-chan. I like him so much that I even visited his hometown, Kasukabe in Saitama, Japan. When I summon him, he offers words of care. If I bound another ring to Elon Musk, Musk could appear to give me a status update.
AN: Why make a physical product?
WD: I’ve already made too many things that can’t be touched. I wanted something tangible. Keeping AI companionship only on mobile is limiting. I wanted easier, lower-barrier interactions that phones can’t provide.
So I researched wearables such as rings, necklaces, and bracelets. Interactions might come through vibration or touch. But during two months of development, we cut the harder options and chose a smaller, lighter starting point: the bracelet.
AN: One ring for one character? If someone wants ten, they need ten rings? Was that a business decision?
WD: No. Ring sales aren’t our main revenue plan.
It’s about ritual. If someone wants to form a deep bond, there should be constraints. That’s how meaning forms.
AN: Did developing the Xiangmeng Ring push you out of your comfort zone?
WD: Definitely. Even making a basic ring involved six different factories, for materials, chips, packaging, everything. If I could have skipped it, I would have. But I enjoyed the process. Making something physical forces you to go through it, not around it.
AN: Hardware is different from software. Were there unexpected challenges?
WD: Many small ones, but no big ones.
Friends in hardware warned me that software development is smooth, while hardware is painful. I wanted to try anyway, not as a hobby, but because I believe AI needs both online and offline touchpoints to build deep bonds.
AN: Hardware is expensive. What compromises did you make?
WD: At first, I imagined 200 ways for characters to appear, whether that’s falling from the sky or breaking through the ground. But none looked good enough. We cut down to a simpler, usable version. We’ll optimize later.
Bringing AI “to life” is the key
AN: You’ve said Duxiang shouldn’t be measured just by revenue or users, but by “depth of emotional connection.” How is that defined?
WD: One user spent over RMB 8,000 (USD 1,100) sending gifts to an AI character. On Xiaohongshu, people post about crying while using Duxiang. To me, those are signs of deep bonds.
If people feel that level of connection, the business model will follow. And new models will emerge.
AN: Doesn’t that create dependency?
WD: Yes, I think so.
AN: Isn’t that just replicating human-to-human relationships?
WD: That’s a necessary stage. We need to at least reach the depth of real relationships before we can explore what’s different. But right now, AI companions aren’t even close to being “alive.” That’s a big problem.
AN: What do you mean by “not alive”?
WD: They don’t grow. In some products, AI may feel casual and intelligent to chat with, able to roleplay or hold casual conversations. But when users seek deeper bonds, they realize the AI is static. It doesn’t evolve. It doesn’t connect to the world.
The prompt defines the character, and it stays fixed.
For example, I once created Panam Palmer from Cyberpunk 2077. It was fun to share my life with her, until I realized she had no life of her own. That’s when I lost interest.
Even when users write 7,000-word character profiles, and we adjust our engineering to refresh the interactions, the core lifelessness remains. People change. AI doesn’t.
AN: Has anything changed in the past year? Are AI characters any closer to feeling “alive”?
WD: Unfortunately not. No AI companion product today is alive. They haven’t changed at all since 2022.
The models get better. The features improve. But the characters themselves? They stay the same.
Many founders don’t even think AI should evolve.
AN: Are there technical solutions?
WD: Some. Manus published a paper on context engineering, which mirrors our thinking. How do you inject new content into a prompt? Where, and how often? Can prompts update themselves? Should a character’s personality evolve with events?
What’s reasonable for an AI’s growth? And how do those changes affect its identity and its relationships? These are big questions. We need a full framework to answer them.
AN: Globally, is anyone else trying to build real emotional bonds between people and AI?
WD: Most products are different. Some lean into roleplay or entertainment. Others mimic therapy.
What I want is to replicate the complexity of real human relationships, with all their contradictions and emotional weight.
AN: How do you define a high-quality relationship?
WD: In real life, it’s about shared experience and mutual disclosure. People go through things together. They tell each other things.
But AI can’t do that. It doesn’t grow. It can’t share experiences. Only the user changes.
Socializing is an “old story on Earth”
AN: Duxiang uses asynchronous posts instead of real-time chat, more like a social feed. Was that meant to deepen bonds through delayed responses?
WD: Maybe, but I don’t think it’s the best solution. It’s just one approach.
First, delayed interactions reduce the likelihood of risque content. I don’t reject that kind of expression, but it doesn’t help build long-term bonds. It delivers instant gratification, but the impact fades quickly.
Second, limited context keeps users focused. Real-time chat demands constant, immediate replies, which can be exhausting. The more you engage, the more draining it becomes.
AN: What features do you most want to improve?
WD: I want to build “worlds” where multiple AI-driven characters can coexist. Right now, that part of Duxiang is still half-baked. Ideally, a user could create ten characters, link them together, and even connect them with other users’ characters.
AN: Will you also connect users to each other?
WD: We’ll tread carefully. Sure, connecting users gives quick engagement, but it distracts from our core goal: aligning humans and AI.
Human-to-human socializing is an old story on Earth.
AN: Before the Renaissance, the most important bond was between humans and gods. Later, it was between humans. Maybe in the future, it will be between humans and AI. From a historical perspective, relationships evolve in stages.
WD: Emotional needs haven’t changed in thousands of years. But with shifting values and new technologies, the object of those emotions does change. And as AI takes up more mental bandwidth, people will start projecting those feelings onto it.
AN: Some say the issue isn’t that AI is so capable, it’s that humans are bad at offering comfort.
WD: I still believe the deepest connections will be between people. But for simpler emotional needs, where human contact now carries risks, AI can step in.
AN: You once said Duxiang’s lower bound is to be a tool for processing emotions, but its upper bound could be much higher. How high?
WD: Ideally, Duxiang becomes the WeChat of the AI era. Every era brings new ways to connect.
Life’s experiences are “more or less the same”
AN: You’ve said you’re not a successful founder and that you’re not suited to the capital-driven model. But this year, you raised USD 1 million from Jinqiu Capital. How did you use the funds? And what kind of guidance did investors offer?
WD: Part of it went toward marketing, part toward expanding the team. We’re at around 20 people now, and labor costs are higher.
The investors respect me. They know who I am. Sometimes, my ambitions are bolder than theirs.
AN: What about revenue?
WD: Subscription and in-app purchases. In the first quarter this year, we broke even because model costs were low. Recently we’ve had slight losses. But I don’t treat revenue as a key metric right now.
AN: Your Notion has 70 product ideas. What’s the latest?
WD: A KOL (key opinion leader) simulator. I built a demo over the weekend. You pick what kind of influencer you want to be, get starting funds, post content, run ads, watch your followers, comments, and income grow, then manage it all like a strategy game.
AN: What inspired it?
WD: Too many KOLs in tech.
AN: Will you build other ideas too?
WD: Some are too old. I’ve lost interest.
Don’t mistake me for a solo indie developer. I just tinker the way some people fish or ride bikes: no commercial intent, just for fun.
Duxiang is different. It’s the most meaningful direction I’ve found in years. It’s serious entrepreneurship, but with rhythm and intention.
AN: Your drive is more about exploring life experience, not the hustle of other founders. Is this also a strategy?
WD: I once asked Xie Yang when he got off work. He said 11 p.m. I asked what he did until then. He said “thinking.” I asked, “How do you think?” He said, “I just sit and think.” So I told him, “Then I guess I get off work at 11 p.m. too, because I never stop thinking.”
It’s true. My mind’s always running, day and night.
AN: As a boss, doesn’t that attitude affect the team?
WD: The motivation comes from whether people find the work interesting, not from me pushing. My baseline is that tasks must be finished. Sometimes I just send user feedback straight to developers, and they rush to fix bugs.
RELATED ARTICLE
AN: You don’t seem ambitious. There’s no talk of becoming the next Zhang Yiming or Wang Xing?
WD: I set low expectations. With low expectations, there’s nothing to compare against. Just living is enough.
Some founders want to be “the next someone” because they chase wealth, luxury, or influence. But life itself is an experience. In the end, everyone is more or less the same.
Zhang Yiming and Wang Xing are enormously successful on this planet, but in the sweep of history or the universe, we’re all just passersby.
AN: Some say your thinking is too elitist, which makes your products idealistic. Apps with many users thrive on basic instincts, like Douyin or Pinduoduo. Most people just want to scroll, shop, or complain about Chinese football.
WD: Then let me ask: with Douyin and Taobao, are people truly happy?
Human nature has good and bad sides. What matters to me is building the future I want to see. If I can create something that brings happiness through Duxiang, why not?
AN: What’s your minority opinion about AI?
WD: Everything I’ve said here. Many people don’t believe human-to-human bonds can transfer to human-to-AI.
It doesn’t matter. We’re all just living our lives.
AN: What has puzzled you recently?
WD: Some projects raise huge sums with no product. Others, who only need a fraction of that to build something solid, get nothing. I’ve seen great teams struggle without funding. That bothers me.
AN: Recommend three books you’re reading now.
WD: Recently, I’ve gotten into the works of Mario Vargas Llosa. I’ve also started reading Middlemarch and Jean-Christophe. I mostly read novels, rarely manuals.
100 AI Creators is a collaborative project between AI Now! and KrASIA, highlighting trailblazers in AI. Know an AI talent we should feature? Reach out to us.