A second track under Basics, read one step at a time instead of one long page. Every lesson ends in a Test panel you can answer by tapping word choices or typing the answer yourself, in kana or romaji - switchable in Settings. First lesson up: Basic Phrases (everyday greetings).

Fixed a bug where quizzing on a reading in "romaji" mode could speak the raw English letters instead of the actual Japanese pronunciation.

Light/Dark/System could get stuck showing the wrong appearance after the app sat idle for a while, or on first load. It now corrects itself automatically instead of needing a trip to Settings.

New app icon, and the tab bar now has icons instead of a broken image on mobile.

The Basics tab now has real lesson content instead of a placeholder. Hiragana and Katakana primers are up first, with more lessons planned in textbook order.

Letters, words, and kanji now each track their own accuracy per account, replacing an older table that had stopped working correctly. The Stats page reads straight from this.

Selecting several categories now pulls an even mix from all of them, instead of one or two categories dominating every session.

Letters, words, and kanji now live in a proper shared structure instead of one-off tables per language. Kanji readings and meanings are tracked separately for the first time, setting up richer kanji content and additional languages down the road.

Light/Dark/System theme, account status, and flashcard filter reset now live under Settings.

Refills now respect every applied category (word and letter). Decoys can no longer come from the same row as the correct answer.

は and へ are now spoken as "wa"/"e" when used as grammatical particles, not their base kana reading.

With per-item accuracy now tracked, the next step is new ways to pick which cards come up - starting with a mode that leans on whatever you’re scoring lowest on, instead of pure random.

A Discord/Facebook-style way to message other users, post in language forums, and eventually author your own lessons is being scoped out. On hold for now - it needs its own foundation (user profiles, moderation) before work starts.

Whisper-based transcription and native audio playback are planned but not started.