Duolingo's founders never used Duolingo
A 500-day streak, one real phone call in Chinese, and the dogfood test every learning product fails or passes.
// trace: where this idea came from
- ↳ video diary @ 23:43 (la llamada que rompió el streak)
- ↳ video diary @ 24:03 (la red flag de los fundadores)
I kept a Duolingo streak of about 500 days. Nearly finished the Chinese course. Then I got on a call with an actual Chinese speaker and couldn’t say anything ▸ 23:43. Two years of daily use, zero usable language.
Later I learned the detail that reframed everything: Duolingo’s founders already spoke their two languages before building it, and haven’t learned a new one with their own app since ▸ 24:03. The app isn’t built to teach; it’s built to retain and monetize. The subreddit confirms the drift: users now pay to find out why their answer was wrong ▸ 25:02, which is the one thing a learning loop can never paywall.
¿el fundador usa su propio producto? →
The counterexamples
GeoGuessr and chess.com mint genuine world-class cracks, people who identify a country from one photo in seconds ▸ 22:46. Their shared design: guess, immediate truth, immediate score. The shorter the window between attempt and verdict, the faster the skill compounds. Streaks measure attendance; feedback builds competence.
The tests we keep
For judging any product (including ours): does the maker need what they built, and does the free tier keep the feedback loop intact? Our language game exists because I need it, and its whole architecture is instant correction. If we ever paywall the part that tells you you’re wrong, quote this entry back at us…