The story
01Why it exists
Eastern Armenian resources tend to come one at a time: a grammar here, a word list there, a reader somewhere else. A learner has to do the bridging. Hayeren treats those bridges as the product, so the word you meet in a story is the same word in the dictionary, in tomorrow’s review and in the grammar note that explains its ending.
02One word, everywhere
Armenian inflects heavily, so a reader full of surface forms is useless without a way back to the lemma. The lexicon keeps that link: every inflected form resolves to a stable entry with its senses, frequency, pronunciation and valency, and that same entry answers the reader, the dictionary search, the review queue and the authoring tools.
The working dataset holds about 188,000 Eastern Armenian entries and 713,000 inflected forms. Those are dataset counts, not users.
03Rules live in one place
Morphology, spaced-repetition scheduling, mastery and permissions sit in a shared core package rather than in each client. The web app and the Expo app both use it, so a review scheduled on the phone follows the same rules as one scheduled on the web.
04Live notebooks that can’t lose your work
Notebooks are stored in Postgres; that copy is always the one that counts. Only when a page is shared with someone who can edit it does a live session open, run by Hocuspocus inside the app. Every ten minutes a scheduled task folds live sessions back into the stored copy. A reader, or someone working alone, never opens a session at all.
Until 26 October 2026 the live layer was Liveblocks and sign-in was Neon Auth. Both now run on the server: Hocuspocus for live editing, Better Auth for accounts.
05Where it is going
Classrooms: shared notebooks, reusable course material and coursework for teachers, on top of a language layer that improves as more sources are added.