Loading page…
A personality-quiz aggregator built by someone who had never made a website — under ₩3 billion in cumulative revenue from ad units on JSON files served as static pages
monetised entirely by display ads (Kakao AdFit first, later Google AdSense) · one test hit #1 on Naver real-time search and generated enough press coverage to break the free hosting
He rates his own coding ability at the start as 0.5 out of a 70-point scale, did not know React existed, used hard-coded JSON files as the database, and hosted the whole thing on free GitHub Pages — the site still runs on that original skeleton five years later.
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
In November 2020, Lee Si-yong could not build a website. He had used Python a little, for data analysis, in Jupyter notebooks — step-by-step interactive scripts. He had never produced a finished program of any kind. He describes his own level with a scale he invents on the spot: if being able to handle front end, back end and DevOps alone is 70 points and a legendary engineer is 100, he was at 0.5. His interviewer confirms the picture with a memory of the two of them stuck on reading a file, unable to move a file pointer backwards, not knowing that `seek(0)` existed. The idea came from a gap he and a business partner noticed during the first COVID winter. Cosmetics and fashion brands were each publishing their own branded MBTI personality tests to pull traffic. Nobody had aggregated them. "There was no MBTI psychological-test platform that gathered many people's content in one place, like Coupang or Amazon," so they decided to build one — a marketplace of tests where brands could later be listed as tenants. He built it in about a month, from late November through past Christmas 2020, commuting between his apartment and the Starbucks in front of it, scraping together whatever web information he could find. The single most useful resource was Opentutorials' free 생활코딩 (Life Coding) courses by the instructor known as egoing. He wanted no theory — just something that ran on his own machine. He did not know what deployment was. The site launched formally in January 2021. His partner handled marketing and planning; he handled development. Two or three months in, one test broke through, and the business began.
There was no launch push. K-Test opened in January 2021 and Lee Si-yong simply kept publishing tests, one after another, for two to three months while nothing much happened.
The quick-reference card — acquisition channels, replication playbook, and risk map — is members-only.
Join the Unicorn Club — $5/moCancel anytime, in one clickPaste into Claude Code / Codex and get a working version of this product end-to-end.
Two production-grade prompts per idea — a full build spec and an SEO growth plan — plus 10 tailored build specs a week. Members only.
Join the Unicorn Club — $5/moCancel anytime, in one clickEvery claim traces back to a listed source — open any of them and check the number yourself.
Data credibility: First-person founder interview in Korean, conducted by a friend who says on camera that he has known the subject since he was young and watched the business grow from the bottom — and who explicitly raises the credibility problem himself, noting that people who never actually succeeded often present as if they did in order to sell courses. No dashboard, ad-network screenshot, contract or financial statement appears at any point. The revenue figure kept in the original currency because the transcript never converts: cumulative revenue for the single business unit built on his first software is stated verbally as 'roughly under ₩3 billion' (30억 언더), explicitly separated from the other lines the corporate entity ran, and explicitly gross rather than net. He notes software costs are near zero but declines to give corporate margins. Other self-reported details: he wrote 97-98% of the code, the remaining 2-3% was consultation with other developers whose code never entered the product, he was 100% self-taught with no bootcamp or tutoring, development began late November 2020, launch was January 2021, the breakout came two to three months later via a personal colour test that reached #1 on Naver real-time search and started going viral on Twitter with heavy overseas traffic, the stack moved from GitHub Pages to S3 plus CloudFront around March 2021 and later to Vercel, and JSON files served as the database throughout the early period. He left the company at the end of December 2025 and at the time of recording (2 February) was running Inflearn and corporate training courses under the name 코딩베어 (Coding Bear) while planning his next solo-developer product. Region is South Korea: Naver, Kakao AdFit, Inflearn, 생활코딩 and Korean-language tests place the business unambiguously. The transcript is truncated mid-answer, so the account of later monetisation and strategy is incomplete and we have not filled the gaps.
The first money was display advertising, and the choice was made by his own skill ceiling rather than by strategy.
The flywheel is the schema. Lee Si-yong standardised the shape of a test before he had any tests, following advice he took from the free Korean course he learned on: assume this could become a hundred or a thousand items and design for it.
The silent period ran from late November 2020, when he started building with no idea how a website was made, through the two or three months after the January 2021 launch when tests went out and nothing caught.
The rest of this breakdown — 10 more sections — is in the Unicorn Club, with every section of all 1,335 ideas: first customers, first dollar, the flywheel, and the silent stretch.
Join the Unicorn Club — $5/moCancel anytime, in one click