Oris
Paused at Phase 0. Open-source gamified learning — the part worth keeping is the data model.
Paused. Last commit August 2026. Foundation scaffolded, nothing live.
Duolingo × Brilliant, open source, with the flagship course being Russian for Uzbek speakers — something a lot of people here need and nobody built properly.
The idea I’d keep
Language apps built elsewhere assume your interface language, your first language and the language you’re learning are obvious. Here they never are. So Oris keeps three axes apart:
| UI locale | language of the interface — uz |
| Source (L1) | the language a course teaches from — uz |
| Target (L2) | the language being taught — ru |
Course identity is (subject, L1, L2). ru-from-uz and ru-from-en are
different trees sharing a skeleton. Non-language courses have L2 = null.
Sounds like bookkeeping until you build it the other way and find out “Russian” is four products depending on who’s asking.
The other decision: exercises are JSON validated against a versioned schema, not code. That’s what makes “anyone can author a course” real instead of a line in a README.
Why it stopped
maymun answers the same audience by generating the lesson instead of authoring it. I couldn’t work on both honestly, and the one needing 200 hand-written exercises lost.
The roast says it plainer: I don’t finish things, I promote them to interesting ideas. Fine — but at least now it’s written down what it lost to.
It comes back if maymun’s voice test fails. Then authored content wins and Oris was right all along.
comments