Memory has rules. We follow them.

Everyone forgets on a curve. We schedule against it.

100 75 50 25 0 % RECALL 0 3 6 9 12 15 18 DAYS SINCE LAST SEEN →
Day 1recall ~90%

Right about here, while recall is still strong. A timely review helps the next return move further out.

An interactive forgetting-curve instrument. Recall starts at 100% on the day you last saw a word and fades over the following days, and flique aims to bring each word back while recall is still strong, around 90% in this illustration (marked in green), so the next review can build on it. The low end of the curve is the forgetting it prevents.

Three ideas do the work.

Recall, not recognition

Spotting a word in a lineup proves nothing. Producing it does.

Whole sentences

Fragments are the first thing you forget.

Corrections that teach

Miss an accent and it says accent. Then you fix it once.

Wrong by one accent isn’t wrong.

Type the answer

"repeat": the vous form

You typed repetez

Almost, check the accents

répétez

Type it right once. That’s the whole penalty.

The receipts.

Six decisions, each with the paper behind it. They studied memory, not us.

The sharpest scheduler there is

FSRS, the open algorithm that beats every classic scheduler in public benchmarks.

That’s the why. The what is over here.