Cards that come back right on time

Every card you learn gets a due date, and the due list on your Journal is Riff asking for it back. The timing is not fixed: the Pace you chose in onboarding, from Easygoing to All in, sets how hard the schedule pushes.

Two layers of research sit underneath. The old layer is the spacing effect: a meta-analysis of 317 experiments shows that memories last longer when reviews are spread out and the gaps grow over time, instead of being crammed together.

The new layer is the scheduler itself. Riff’s review timing builds on a memory model from recent machine-learning research, published at a leading data-science conference and fitted on over two hundred million real review logs. Per card, it keeps estimating how fast your memory of it fades, then asks for a review just before forgetting would win.

Pace is the honest half of the deal. Aiming for higher retention means more reviews per day; a gentler target means fewer. The algorithm optimizes the timing. You choose the workload it optimizes for.