Liangpei wants users to leave fast—that's exactly how it makes money
Liangpei (良配) sells memberships that can't be renewed; users leave once they match. The faster they go, the fewer tokens the platform burns—exactly opposite to most AI products profiting from retention.
The argument · tap a timestamp to hear it
Traditional matchmakers price by what you can pay, not what service costs
Zeng Xinxun notes that among 150 couples who registered at Shenzhen's Civil Affairs Bureau, fewer than 5 percent succeeded through professional matchmakers. Traditional matchmakers don't price based on the service itself, but first probe your willingness to pay and credit card limits—the same service sells for 20,000, 30,000 or 50,000 yuan depending on how much you can afford. More egregiously, some agencies pressure young women to take loans, using the pitch ‘it's better to find a husband than buy a handbag’ to package dating as an investment.
— Zeng XinxunGrassroots teams lack resources to compete with well-funded rivals
Zeng ran a campus food delivery platform in university that generated over 10 million yuan in annual revenue at peak, but was completely shut out when Ele.me raised 80 million dollars and rapidly expanded to 200 cities. He concluded this wasn't a product or execution problem, but that grassroots teams lack the resources to race against opponents already backed by capital markets—which is why he spent 8 years before this venture deliberately ‘accumulating equipment’ (technology, team, industry credentials).
— Zeng XinxunThree hours to raise 15 million required eight years of prep
A week after resigning as CTO of Kimi, Zeng met with investors; by chance at a hackathon café, he talked for 15 minutes with someone from Jujin Capital, who immediately called to verify his Kimi background and set up a meeting with Xu Xin the next day. In three hours of negotiation, Xu Xin proactively offered to provide the full 15 million yuan representing 20 percent equity in one payment, with zero protective clauses in the agreement. Zeng says this wasn't luck, but the accumulated technology, team, and industry credentials he'd built over 8 years all materializing at that moment.
— Zeng XinxunFour structural flaws prevent dating apps from helping you pair up
Zeng identifies four categories of problems in existing dating products: platforms mix users with different needs to boost DAU, with profiles limited to 100-150 characters creating severe information shortage; an endless stream of candidates creates the illusion of abundance, making people constantly compare and undervalue the person in front of them; and the platform's business model rewards longer user stays, giving it no incentive to actually help you find someone, putting user and platform interests fundamentally at odds.
— Zeng XinxunSingle-match design escapes the user-platform prisoner's dilemma
Liangpei designed the EV1 matching mechanism, which doesn't offer a continuous candidate pool but shows one person at a time. Zeng believes previous products didn't dare try this because match success rates were too low—EV1's risk directly exposes you to the 1-minus-success-rate failure probability; Liangpei only dared build EV1 because AI collected two to three thousand characters of context, raising match accuracy high enough to support it.
— Zeng XinxunOnly non-renewable memberships make platforms want you to succeed
Liangpei's core memberships are non-renewable: the 2,000 yuan ‘guaranteed marriage membership’ keeps the payment if users marry within three years, otherwise fully refunded; there's also a 1,000 yuan perpetual membership paid in one lump sum. Zeng explains the design works because the platform burns tokens daily for each user, and the longer users stay the higher the cost, so the platform naturally wants users matched quickly and gone—the opposite of most software that profits from renewal and time spent.
— Zeng XinxunAcquisition cost rose from a few yuan to over 50
In 2018-19 doing campus delivery, acquiring a user cost just a few yuan, but today in the two-sided market Liangpei faces, 50 yuan acquisition cost isn't enough. In the first week after launch, Zeng watched growth curves daily with anxiety, because the earliest users trusted the product most but experienced the smallest sample size and matching pool, until a few articles went viral and user scale picked up.
— Zeng XinxunBetting everything on one path beats hedging between many options
Zeng says at every major inflection point in his life he chose only one path—applied to only Southern University of Science and Technology for college, picked one place when job hunting and went there, and for this venture only does dating, something he experienced personally, rather than the billion-dollar-market spaces investors suggested. He believes that with multiple choices you fall into decision paralysis, efficiency drops lowest, and going all-in actually provides clarity.
— Zeng XinxunIn their own words · checked verbatim
It creates the illusion of endless candidates—you think you have many backups, but you're also someone's backup. You always think the next will be better, so you never treasure anyone. The platform wins. The longer you stay, the higher its revenue.
他给你一种 候选者 源源不断的这种错觉 你以为你有很多的 所谓的在引号的备胎 你也是别人的备胎 你总觉得下一个会更好 你就不会珍惜任何一个人 最后受益的就是平台 你留着时间越长 他的收益越高
Zeng Xinxun0:00
Once it entered the capital market, it quickly expanded nationwide, launching in 200 cities simultaneously.
它现一步的进入了资本市场 很快他们就开始全国铺开 200个城市一起发展
Zeng Xinxun21:15
In the end he'd scraped together 8.9 million, and then he said let me round it up to 15 million for you.
他最后凑整凑了有890万 他说给你凑个整吧1500万吧
Zeng Xinxun40:24
You're back in the cycle I believed in from the start: spending time, spending money, spending your emotional value to validate a wrong choice.
你就回到了我最开始相信的那种循环里面了 花时间、花钱、花自己的情绪价值 去验证一个错误选项
Zeng Xinxun45:29
For me, a company burning tokens daily for each user, the less time you spend, the fewer tokens I burn.
对我是一个要每天给你烧token的一个公司 你走的越少我的token就烧的越少
Zeng Xinxun51:36
I prefer to solve one problem at a time, cleanly. Once I finish, it's solved—one hole, one turnip.
我会更喜欢确定的解决一个问题 这样我做完了 他就解决了一个问题 一个萝卜一个坑
Zeng Xinxun56:39
Figures
| Jujin Capital stake | 20% | 39:23 |
| Liangpei registered users | 18,000 | 54:39 |
| Liangpei paying users | over 1,000 | 54:39 |
| Follow-up successful matches (small sample) | 5 of 30 successful, 20 actively progressing | 52:37 |
| Guaranteed marriage membership cost | 2,000 yuan (full refund if unmarried after 3 years) | 50:36 |
| Perpetual membership cost | 1,000 yuan | 50:36 |
| Online acquaintances in Shenzhen survey | 15% | 2:00 |
| Family/friend referral rate in Shenzhen survey | 30% | 2:00 |
Glossary
- EV1 matching / one-shot expected value matching
- Showing one candidate at a time instead of a continuous pool, relying on higher per-match success rate to make it work.
- T-degree / Zhang Yiming-style steepest ascent
- Yang Zhilin's internal concept for finding the path closest to your final goal with maximum force behind it.
How to listen
Entrepreneurs and investors focused on AI application business model design, protective clauses, and early-stage funding pacing.
0:00-2:00 the opening about gender polarization can be skipped, as it has little connection to the product content that follows.