Method

Learning that works, not learning that feels good

Most flashcard tools are built to feel good — points, streaks, leaderboards, a little hit of reward for every tap. Students enjoy them. Weeks later, the content is gone. Edubba is built to work instead, on one of the oldest, most tested ideas in learning science, enforced properly, with nothing left for a student to game.

Every card lives in one of five boxes, and every topic set tracks its own boxes independently — strength in one area is never masked by weakness in another.

One test at a time

A session only tests the lowest box a set currently has cards in. If a student has cards in boxes one, two, and four, only box one is tested this time. Nothing else appears until it's clear.

Every session has a single, honest purpose: strengthening the weakest point in that topic, rather than a shuffled mix that lets a student coast on what they already know.

Forgetting is part of the design

After a session, the set locks. Nobody can re-test or grind through it again to force cards up early. It reopens after a forgetting period your department sets — a simple "standard" or "cramming" preset — and that gap lengthens automatically as cards climb the boxes.

Sets don't change once they're live

Once a topic set is released to students, it's fixed. No cards get added, removed, or edited into a live set. If content needs to change, a new set replaces the old one — so nothing is ever half-learned by design.

One answer, always

Every card has exactly one correct answer, set once at creation — not a list of accepted alternatives. Learning a fact is like learning a route: drive the same road every time and it becomes automatic. Multiple accepted phrasings split that route instead of building it.

Enforced, not requested

The locks, the box logic, and the one-topic-at-a-time rule are enforced on the server, not the screen a student is looking at. There's no version of Edubba where a student can refresh their way past a lock. The system holds the line so the teacher doesn't have to.

See how this holds up: Read the specification See the evidence