How it works
The algorithm knows when you’ll forget.
flshcrds schedules with ts-fsrs, an implementation of FSRS-6 — learning steps 1 m → 15 m → 8 h, one 10-minute relearning step, target retention 90% by default. Everything on this page runs that exact scheduler, not a mock.
Recall falls on a curve.
From the second you learn a card, your chance of recalling it drifts down — steep at first, flatter with age. That is the orange line falling.
Your target is the tripwire.
You choose the retention target — 90% by default. FSRS books the next review for the exact moment recall touches that line — late enough to be worth it, early enough to still be there.
Every pass stretches the gap.
A successful review resets recall to 100% and makes the memory more stable — so the gaps grow: 1 h → 4 h → 1 d → 4 d. Skipping one drops you onto a dashed curve.
The same card, reviewed across growing gaps — 1h, 4h, 1d, 4d — drawn with the same FSRS-6 forgetting curve the app schedules with. Every dashed line runs to the edge: the retention you would be left with had that review never happened.
Simulator — drive one card yourself.
the app’s exact schedulerlearning steps 1m → 15m → 8h · target retention — default 90%
LEARNING / STEPS
Three same-day steps — 1 min → 15 min → 8 h — before a real interval.
Difficulty and stability already exist and move with every answer, but due times come from the steps, not from the model. The point is a clean recall before the card leaves for days.
WHAT EACH RATING DOES
The highlighted state follows the card in the panel; the others are one tap away.
1 review
TAP TO SIMULATE — THE CLOCK JUMPS TO EACH DUE DATE
REVIEW LOG
- 1Goodday 0 · new → learning · +15 min