Building Backward
A few weeks ago, a friend sent me a link. It was a website called SpeakVlog — you watch YouTube videos with synced subtitles, tap on words you don’t know, do some flashcards. That’s it.
Then she told me the number: roughly 23,000 paying subscribers at about 40 yuan each. That’s close to 920,000 yuan a month. From a website.
I stared at the screen for a while. My first reaction was something like: really? This? It’s not a beautiful product. It’s not doing anything technically impressive. There’s no AI magic, no viral loop, no growth hack. It’s a website that helps Chinese learners study English through YouTube videos, and tens of thousands of people are paying for it every month.
I couldn’t stop thinking about it.
Even as a product manager, I’m not immune to this. I spend my day job figuring out what to build and why. But when it’s my own project, something changes. The excitement takes over. I skip the questions I’d ask at work and go straight to the editor.
Last month I made a WeChat mini-program called Little Deer ABC — a kids’ English learning tool. I vibe-coded it in a few days with my OPE agent. And when it was done, I realized I had no plan for how anyone would find it.
Knowing how to build a product and knowing how to sell one are two very different skills.
This time something shifted. Instead of opening my editor, I sat there and did something I’d never done so deliberately before: I studied why this product was already working.
Not the code. Not the UI. The business.
What is this product actually doing? It takes vlogs from YouTube — native English speakers talking about travel, food, daily routines, the kind of things you’d actually talk about in real life — and turns them into structured lessons. You’re not just watching a vlog for entertainment. You’re learning English in context: real people, real scenes, real language. It’s scenario-based learning without feeling like a classroom.
Who’s paying for this? I went to Red Note — the Chinese version of TikTok — and searched for English learning. The posts have tens of thousands of likes. The comments are full of real pain — people describing how their brain tries to translate every word and can’t keep up with the speaker, how by the time they finish translating the first sentence the speaker is three sentences ahead. These aren’t hypothetical users. They’re real people, describing a real problem, in a place where I can find them.
Is anyone already paying? SpeakVlog already proved that.
How do they find it? Red Note content. No paid ads, no App Store optimization. Just content that reaches the people who have the problem.
There’s a huge difference between knowing the right questions and actually having answers before you write code. I’d never had all of them answered at the same time before starting a project.
So I built it. I used my OPE agent and went from zero to a fully functional product in 24 hours. Login, dashboard, video player with synced subtitles, word highlighting, flashcards, dictation mode, learning records. Seven pages, pixel-perfect.
That speed wasn’t because I’m fast. It was because there was nothing to figure out. All the product decisions were already answered by the product I was learning from. I just had to execute.
When direction is clear, building is just labor.
But building backward doesn’t mean stopping at a copy.
A copy is a starting point, not a destination. The real value of building backward is that it gives you a foundation of validated assumptions. You know the market is real. You know people pay. You know how to reach them. From that foundation, you can see opportunities the original product can’t.
That’s where I am now. I’m building a second product — same validated model, completely different audience. Can the pattern of learning English through video content, distributed through Red Note, work for parents who want their young children to speak English? I don’t know yet.
Stay tuned.