Learn keeps two numbers on every flag: how many times you got it right in a row, and how many days until it shows up again. Struggle, and it comes back tonight. Nail it, and it can wait months.
The scheduler is a slimmed-down SM-2 — the same family Anki uses. This post is the exact schedule, the session mix, and the three places we diverged on purpose.
The three ratings
After the flag is revealed you tap one. Pick a button here — or a column in the schedule — to follow that rating.
Again, Good, Easy
Same colours as the session.
Knew it, needed a moment
The gap grows in moderate steps, from a day toward months.
How long until it comes back
The wait depends on the rating and how many correct answers are already stacked. Again always clears the stack.
Interval schedule
Choose a rating to trace its path.
| Stage | |||
|---|---|---|---|
| 1 First look | ~10 min | ~10 min | 4 days |
| 2 After 1 correct | reset · ~10 min | 1 day | 7 days |
| 3 2+ correct | reset · ~10 min | × 2.5 | × 4.0 |
Mastery on the dashboard
The badge is just the consecutive-correct count. Again sets it back to zero.
Repetitions → status
- 0
New / Tricky
Never seen, or the last tap was Again.
- 1
Learning
One correct answer. Still easy to lose.
- 2
Almost there
Holding. Next Good or Easy stretches the gap.
- 3+
Review
Long spacing, like Anki review cards. Still comes back.
A session is ten flags, or five the first time
Learn and Review pull from the same deck. The difference is whether new flags are allowed in.
Learn
Mix of due flags and new ones. When both pools have enough, about seven of ten are new.
Review
Due flags only. Nothing new until you open Learn again.
Extra due flags wait for the next Review. A streak ticks once per calendar day on the first rating. One tap is enough; you don't have to finish the session.
Where the numbers live
The working copy is in this browser, under spaced-repetition-v2. Sign in
and review events sync card schedules to your account, so another device can catch up.
Streak and the 14-day activity chart stay on the device.
Clear this browser, or switch devices without signing in, and local progress is gone. Sign in on each device before you treat the decks as the same.
Why this shape
SuperMemo 2 (1987) is the core. Three cuts keep a ten-flag session from turning into homework:
- 1
Fixed multipliers, no ease factor
SM-2 tweaks a per-card ease number after every rating. Learn uses 2.5× for Good and 4.0× for Easy, so the same tap always does the same math.
- 2
Three ratings, not four
SM-2 has Hard between Again and Good. Learn skips it. After the flag flips, you pick from three.
- 3
Missed flags wait ~10 minutes
Again does not shove the flag back into the same pass. Try again on the recap is one extra look at flags you missed, then you're done.
You give up a little personalisation. You keep a schedule you can finish on a commute.