Vol.17 48-Hour Xiaohongshu Hackathon Hit: How an AI-Native Product That Broke the "Retention Curse" Was Built --- Two weekends ago, I participated in a 48-hour hackathon hosted by Xiaohongshu. Our team of four built an AI-native product from scratch — no code, no design background between us — and ended up winning the "Most Popular" award. The product? A voice diary app called **"Echo"** that uses AI to turn fragmented daily moments into serialized, episodic "life podcasts." Think *This American Life*, but starring you. What surprised me wasn't that we won. It was that people kept using it *after* the demo. Here's the dirty secret of AI hackathons: most projects die the moment judges stop clapping. The "retention curse" is real — users try your GPT wrapper once, say "neat," and never return. We broke that pattern. Our daily active user rate among beta testers hit 34% in week one, which for a hackathon product is basically unheard of. How? Three deliberate choices we made against hackathon orthodoxy. **First, we refused to build a chatbot.** The default AI product in 2024 is still "talk to a large language model." We explicitly rejected this. Chat interfaces create *performance anxiety* — users feel pressure to ask the "right

Vol.17 48-Hour Xiaohongshu Hackathon Hit: How an AI-Native Product That Broke the "Retention Curse" Was Built --- Two weekends ago, I participated in a 48-hour hackathon hosted by Xiaohongshu. Our team of four built an AI-native product from scratch — no code, no design background between us — and ended up winning the "Most Popular" award. The product? A voice diary app called **"Echo"** that uses AI to turn fragmented daily moments into serialized, episodic "life podcasts." Think *This American Life*, but starring you. What surprised me wasn't that we won. It was that people kept using it *after* the demo. Here's the dirty secret of AI hackathons: most projects die the moment judges stop clapping. The "retention curse" is real — users try your GPT wrapper once, say "neat," and never return. We broke that pattern. Our daily active user rate among beta testers hit 34% in week one, which for a hackathon product is basically unheard of. How? Three deliberate choices we made against hackathon orthodoxy. **First, we refused to build a chatbot.** The default AI product in 2024 is still "talk to a large language model." We explicitly rejected this. Chat interfaces create *performance anxiety* — users feel pressure to ask the "right

June 11, 2026

There's a well-known rule of thumb in product circles: only when your 30-day retention exceeds 10% do you even get to talk about hitting a million DAU. But reality is brutal — data shows the global average for mobile app 30-day retention is just 6%. That means 94 out of 100 users vanish within a month.

In the AI emotional companionship space, which is already driven by curiosity and prone to "use and leave" behavior, this curse is a make-or-break line for countless products.

But this seemingly locked-in script is being rewritten by a tiny AI Native team of just a few people. NoonWake — the first investment from Yunqi Capital's Y Transformers program — has achieved 15% 30-day retention with nearly 50,000 DAU. During the 2026 Spring Festival, they even ran an extreme test: with all four full-time employees on holiday, they went full AI-automation workflow and still brought in nearly 100,000 RMB through AI alone.

In this episode, Peter (Hao Liang) from Yunqi Capital's investment team sits down with Shawn Wu, founder of NoonWake and the Y Transformers program, to reveal the golden rules for breaking the retention curse.

How did a four-person team, built after leaving a major tech company, build a cross-format product matrix in the global emotional companionship red ocean? Shawn offers a sharp and pure underlying logic: the mobile internet era demanded scale; AI demands authenticity. Getting users to tell the truth to AI — that's the real core asset for To C products.

What you'll hear in this episode

  • Why simply competing on "emotional value" creates no commercial moat, and how capturing users' "genuine, unfiltered truth" becomes the retention antidote for AI To C.
  • How the highly human-dependent, low-frequency business of general psychology and traditional mind commerce gets reconstructed into a high-frequency, scalable AI Native hardware-software closed loop.
  • With the Agent Harness architecture, how high is the productivity ceiling for AI-native teams of geeks?
  • Yunqi's "appetite" for non-consensus young founders.

Timeline

01:41 — 02:56

Connecting with Shawn: From Xiaohongshu "hot takes" to Y Transformers' first investment

How did Shawn get "fished out" by Yunqi investors?

02:56 — 05:05

4-person team on holiday, AI automation workflow earns 100K during Spring Festival

With all four full-time employees on break, multi-Agent collaborative workflows through OpenClaw automatically sent SMS to 100,000 overseas users, adjusted pricing, and ran re-engagement campaigns.

05:05 — 07:55

PMF core: Productizing general psychology — young people don't want science, they want a framework for interpreting the world

On the surface it's AI emotional value, but what's the underlying logic? As anxiety goes global, why does a "logically self-consistent and unfalsifiable" framework for interpreting the world command such massive commercial premium?

07:55 — 11:46

Core entry point: Why "authentic context" is the core asset for To C

People often perform in social companionship, but in general psychology scenarios, they're unlikely to lie to AI. Compared to centralized tags in the mobile internet era, what's the real difference with decentralized, personalized customization in the AI era?

11:46 — 13:37

Paradigm shift: Input quality matters, but "authenticity" matters more

Shawn breaks down his experience with hardware-software integration at a major tech company and social entrepreneurship: in the AI era, context is everything.

13:37 — 16:59

AI Native workflow: Humans file the bug, Agents fix it overnight

With only four people on the team, how did they complete frequent AB tests and product iterations within two weeks?

16:59 — 27:27

The essence of AI-native hardware: enabling users to "confess their truth painlessly and elegantly"

Deconstructing the "Good Luck Calendar Machine" that won second place at the Xiaohongshu Hackathon for Wanxiang Youling (Spirit of All Things).

Why won't we see an all-in-one全能 device within five years?

How does vertical-scenario hardware, through cameras, voice, and 3D holographic divination chambers and other immersive interactions, collect real-world authentic context with lower friction?

27:27 — 33:31

Global ambitions: How does the calendar format naturally fit globalization?

Chinese users get the Chinese experience; North Americans recognize astrology and tarot. How does a four-person team, without getting dragged down by supply chains, use one universal hardware piece to carry different cultural localizations for different countries?

33:31 — 44:55

The real moat is "long-term memory" that humans cannot replicate

Technology leadership has an extremely short shelf life, but the social chains and long-term Memory users leave within a product cannot be copied. How do vertical AI applications differentiate from general-purpose large models?

44:55 — 48:28

VC perspective: Why did Yunqi move fast at this moment?

Peter's investment post-mortem: An industry that has existed for thousands of years, extremely human-dependent and trust-intensive, low-frequency — how does AI reconstruct it into a high-frequency, high-payment-imagination space? Why do we only look at "human efficiency"?

48:28 — 51:41

Ultimate either-or

In the next three to five years, what's the only hard metric for AI application companies?

Check your phone first thing in the morning, or check your Good Luck Calendar Machine? In the future, will AI-native tech companies have more science majors or humanities majors? Why do flamboyant geeks pursue "steep growth curves," while rational VCs only look at "whether human efficiency has grown 100x"?

This episode's interactive

What unforgettable moments have you had opening up to AI? What kind of AI software/hardware would you want to use every day?

Share your thoughts in the comments — we're giving away 20 good-luck gifts from "Wanxiang Youling":

  • 10 Wanxiang Youling monthly passes — experience high-retention AI's exclusive companionship and listening.
  • 10 "Life K-Line Chart" guides — reframe your worldview, ease diffuse anxiety.

This episode's interactive

Within five years, will there be an AI product you want to open first thing every morning? What would it look like?

Share in the comments — we're giving away 20 good-luck gifts from Wanxiang Youling:

  • 10 Wanxiang Youling monthly passes — tell it what you dare not say to anyone else
  • 10 "Life K-Line Chart" guides — a new framework for seeing the world, a little less anxiety

About Yunqi

Founded in 2014, Yunqi Capital is an early-stage lead investor focused on digital intelligence and hard tech. We've been repeatedly named to Zero2IPO, ChinaVenture, 36Kr, and other institutions' "China's Best Early-Stage Investment Firms TOP 10" lists.

Over the past 12 years, we've accompanied more than 200 tech startups, including MiniMax, JD.com, Manycore Tech, DeepRoute.ai, Neolix, Keenon, Independent Variable Robotics, Astribot, RealMan, Noematrix, PingCAP, Zilliz, ASTRONSTONE, XTransfer, and other industry leaders.

Contact us — For pitch decks or ecosystem partnerships: community@yunqi.vc

WeChat Official Account: Yunqi Capital (ID: yunqipartners)

Xiaohongshu: Yunqi on Anfu Road

View episode transcript on Xiaoyuzhou