How FeynmanFSRS works
Flashcards test whether you recognize something. This tests whether you can explain it — which is what exams, teaching, and actually using knowledge demand.
The loop
- A card's front is a prompt: “Explain comparative advantage.” Write your explanation from scratch, or hit 🎤 and say it out loud — explaining aloud is the original Feynman move.
- Flip. You see your explanation next to the reference answer and its key-points checklist — the three to five things a complete explanation must contain.
- Tick what you actually covered, then grade yourself. The checklist is what keeps self-grading honest: the question is “did I hit these points”, never “did that feel good”.
- FSRS schedules the next review for right around when your understanding is predicted to fray. Every explanation you write is saved to the card's history timeline — over months you can watch yourself get clearer, or catch yourself getting vaguer.
The four buttons
- Again — you missed the core of it. The card comes back within minutes.
- Hard — you got there, but slowly or with real gaps in the checklist.
- Good — you hit the key points. This is the normal grade; most reviews should be Good.
- Easy — effortless and complete. Use it sparingly; it pushes the card far out.
Keep decks small — this is by design
An explanation takes one to three minutes against five seconds for a flashcard. So this is a small-deck tool: the twenty to fifty core concepts of a course, not ten thousand vocabulary items. A healthy deck reviews in about fifteen minutes twice a week. Keep using Anki for facts — FeynmanFSRS sits next to it, for understanding.
Writing good cards
- Prompts start with Explain (or Why / How), one concept per card. “Explain comparative advantage”, not “Define comparative advantage”.
- The reference is the explanation you wish you'd write: short, plain language, intuition first. Markdown works.
- Key points are checkable claims, not vibes — “uses opportunity cost, not absolute cost”, not “understands the idea”.
- If AI generation is configured, it can draft the reference and checklist when you create a card — edit before saving. Grading stays yours; there is no AI in the review loop.
Keyboard shortcuts
| ⌘/Ctrl + ↵ | Flip the card (show the reference) |
| 1 / 2 / 3 / 4 | Grade Again / Hard / Good / Easy |
| Space | Grade Good |
| U | Undo the last review |
Around the loop
- Catch up (free) — came back to a pile of due cards? The deck page offers to spread them evenly over a few days. Only due dates move; scheduling stays honest.
- Study ahead (free) — nothing due but you have time? Bring the next few scheduled cards forward. Early reviews are honest: the scheduler works from the real gap since your last look.
- Focus (free) — a toggle on the review screen hides everything but the card.
- Exam mode (Pro) — set a test date on a deck and nothing schedules past it; the deck shows predicted recall on the day.
- Import from Anki (Pro, BETA) — bring an .apkg, pick up to 30 cards, and AI converts them into explanation cards you review before saving.
- Simulator (Pro) — Stats projects your next 90 days of workload and stable-knowledge growth from your real scheduling state.
The hub, and getting your data out
The hub is where decks are shared: publish a snapshot of a deck (prompts, references, key points — never your history), copy other people's decks with fresh scheduling, and leave comments under your username. And your data is never held hostage: every deck exports as JSON (full fidelity, history included), Markdown, or CSV from its deck page — everything at once from settings.
One honest caveat: FSRS is calibrated on recall grades, and explanation self-grades are noisier. It still works — the scheduler only ever sees the four buttons — but treat the intervals as good, not clinical.