How the smart queue works

You can study the same queue in different places: focused review in the web app, browse your collection when you just need a lookup, message a Telegram bot for one term on demand, or let a desktop widget rotate terms in the background. Same ranking everywhere. What changes is when something counts as a sighting.

• Web review: A focused study session. Pick a known or unknown pool, work through a ranked batch, and mark terms as you go. Setup shows a preview; each card shows why it was picked. Opening a card counts as one sighting.
• Web collection: Browse and look up terms in your collections. Opening a term card counts as one sighting for that visit. Good for reference, not a substitute for review.
• Telegram bot: Get terms delivered on a schedule, or pull one on demand with /next. Uses the same ranking as web review, one unknown term at a time. Mark known or still learning from inline buttons.
• Desktop widget: Terms rotate on your screen in the background. Rotation alone does not count as a sighting. Tap Next when you want one recorded, or open the term to read it properly.

Each term carries a history: how often I've seen it, when I last looked, how I rated it. The queue uses that to decide what comes next within the pool you picked.

• Never seen: Rises to the top until you've seen every term in the pool at least once.
• Still learning: You marked it still learning, or got it wrong on a quiz.
• Forgot: You marked it forgot or cleared it as known, ranked above still-learning terms.
• Seen 3+ times, not solid: Opened without marking it known, seeing it often doesn't mean you've learned it.
• Recently added: Terms added in the last few days get a nudge so they don't sink to the bottom.
• Not seen recently: Skip a day and a term starts climbing again. Wait longer and it climbs higher, up to about a week.

One signal lowers priority instead:

• Recently marked solid: Marked known or answered correctly on a quiz, sits out for roughly 3 days while the queue works on other terms.

That cooldown is the only scheduled behavior. Nothing else gets a future review date.

Review cards, quiz questions, and the queue preview label terms with these badges. A term can have more than one at a time.

Once I've seen every term in a pool at least once, never-seen stops applying. After that, neglected and weak terms rise, a gentle cycle with no reset button.

Once you're in, you can tune this under Settings → Review. I added presets because sometimes I want to drill weak spots, sometimes to clear never-seen terms first. Pick one, the same preset applies everywhere: web review, Telegram /next, quizzes, and delivery.

• Balanced: Default mix of new, struggling, and stale terms.
• Learn new first: Prioritize unseen terms and recently added content.
• Drill weak spots: Focus on terms you're struggling with or forgot.

Quizzes use the same preset and pool rules as review, with one extra tilt I added: struggling signals rank higher, still learning, forgot, or seen many times without marking known. Setup shows a preview; questions show badges.

This ranks what's in the active pool when I show up. I built it that way on purpose, no scheduled future reviews.

• Not Anki, no fixed schedule, intervals, or “cards due today”
• Not a notification system, I study when I want
• Not random, every pick comes from my history in that pool