How it works.
Four principles. Each one built into the app. Start anywhere — they reinforce each other, and the order doesn't much matter.
“Effectiveness is a function of atomicity, active generation, effortful retrieval, and algorithmic scheduling.”
One fact per card.
If two facts each have 90% recall probability, a card testing both has only an 81% chance of success (0.9 × 0.9). Three facts? 73%. The more you bundle, the worse your scores — and the algorithm can't tell which fact you forgot.
Atomic cards solve this. One fact, one card. If you forget it, only that card gets rescheduled. Your review time goes exactly where it's needed.
See the difference
Describe the process of eutrophication.
Excess nutrients enter water, algae bloom, algae die, decomposition uses oxygen, aquatic organisms die.
Five separate steps. If you forget step 3, the algorithm can't target it — split into one card per step.
What is the first stage of eutrophication?
Excess nutrients (N and P) enter a water body.
List the inputs and outputs of a forest ecosystem.
Inputs: solar energy, precipitation, nutrients. Outputs: oxygen, biomass, runoff.
Six facts in one card. Split by component — one input or output per card.
What is the main energy input to a forest ecosystem?
Solar energy.
What is biodiversity?
The variety of life in an area.
Which level — species, genetic, or ecosystem? Pin one so the right answer is unambiguous.
What is species diversity?
The variety of species in a given area.
Explain how a waterfall forms and retreats.
Hard rock overlies soft rock, soft rock erodes, undercutting creates an overhang, overhang collapses, waterfall retreats upstream.
Five sequential steps bundled together. Pick one mechanism per card.
What causes undercutting at the base of a waterfall?
Erosion of the softer rock beneath the resistant cap rock.
In the app: Production mode links cards to syllabus points. AI coaching flags cards that test too many things at once.
You write it, you learn it.
The generation effect: you remember information you produce yourself better than information you passively read. AI-generated cards skip the most valuable step — the act of turning knowledge into a question.
AI coaching, not AI generation
Seven rules, one chip per card
Every evaluation also names why. The card gets a single chip drawn from a fixed taxonomy of seven rules. Same rule, two names — the good name when met, the bad name when broken. As you write more cards, you start to recognise the same five or six labels and your mental rubric sharpens.
What the feedback looks like
Each evaluation has three slots: the principle chip, the exact phrase from your card the AI is reacting to, and one sentence of direction. The AI never rewrites the card for you — the move is yours to make.
What are the inputs and outputs of an ecosystem?
Solar energy, heat, oxygen, biomass.
Four different items in one answer. When you flip the card you can't tell which one you remembered — split into four cards, one per item.
What is the primary energy input to an ecosystem?
Solar energy.
Question patterns
Strong patterns
- “What is the [term] for [phenomenon]?”
- “What causes [process] in [context]?”
- “[Process X] results in what change to [Y]?”
- “What distinguishes [A] from [B] regarding [criterion]?”
Weak patterns
- “Explain…” / “Describe…” / “Discuss…”
- “List the X causes/factors…”
- Vague prompts without specific interrogation
- Answers with “and” connecting distinct concepts
Tips for IB ESS & Geography
- Split system components: separate cards for inputs, outputs, processes.
- Separate “what happens” from “why it happens”.
- For case studies: test specific fields (location, cause, effect) — not “describe the case study”.
- Break processes into steps — each step = one card.
In the app: Click Get AI feedback on any card. You get a traffic-light rating, a principle chip naming the rule, the offending phrase from your own writing, and one sentence of direction.
The struggle is the point.
Active recall beats passive review. Reading notes feels productive but doesn't strengthen memory traces. The effort of retrieval is what builds long-term memory. The magic happens in the pause between seeing the question and revealing the answer.
Try the review flow
Try it out
What is the main greenhouse gas emitted by rice paddies?
Rating guide
Key tip: Be honest with your ratings. The algorithm works best when you accurately report how well you remembered. Don't be afraid to press “Again” — it's part of learning.
In the app: Revision mode shows one card at a time — question, mental effort, reveal, rate. The flow is designed to maximise the retrieval effect.
Review at the right time.
The Ebbinghaus forgetting curve: memory decays exponentially without review. Each successful review resets and extends the curve. The trick is to review just before you forget — too early wastes time, too late means relearning from scratch.
The forgetting curve
Why regular review matters
- Intervals only grow when you show up consistently.
- Missing reviews means the algorithm has stale data — it doesn't know your current memory state.
- Regular short sessions (10–15 min daily) beat cramming.
Why honest ratings matter
How intervals grow
As you master a card, you see it less. A well-known card might only appear once every few months.
In the app: The Schedule page shows upcoming reviews and assessment dates. FSRS manages intervals automatically — just show up and review honestly.
See your quality at a glance.
Once you have a few dozen cards, every card you've made paints as a single coloured tile, sorted red → amber → green. As you internalise the seven rules, the grid drifts greener over time.
See every card you've ever written, by quality.
Every card you've made paints as a single coloured tile, sorted red → amber → green. Click any red tile to open the card and rewrite it. As you internalise the seven rules, the grid drifts greener.
- The red tiles are an obvious to-do list.
- Click any tile to open and rewrite the card in place.
- The grid is per subject and can be organised by unit.
Four principles. One system.
Atomic cards, written by you, recalled with effort, scheduled by an algorithm. That's how Revise turns study time into lasting knowledge.
Get started