FSRS (Free Spaced Repetition Scheduler)
PlatformDefinition
A modern spaced repetition algorithm used in Claude Architect Lab to schedule concept reviews. Tracks four parameters per card: stability (how long memory persists), difficulty (how hard the card is), retrievability (current recall probability), and due date. More accurate than older algorithms like SM-2.
Example Usage
The FSRS algorithm schedules each concept review at the optimal interval to reinforce memory just before forgetting.
Why It Matters for the CCA-F Exam
Understanding FSRS helps you use the platform's review queue strategically: honest ratings (especially using "Again" when you genuinely struggled) feed accurate parameter updates, which produce better-timed intervals and higher retention on exam day. Treating every review as a formality defeats the algorithm's ability to distinguish stable from fragile knowledge.
In Depth
FSRS — the Free Spaced Repetition Scheduler, now at version 5 — is the open-source algorithm that powers the Claude Architect Lab review queue. Where older schedulers relied on crude heuristics, FSRS models your memory mathematically using four per-card parameters that are updated every time you rate a review.
The four parameters
- Stability (S) — how many days until retrievability drops to 90%. A card with S = 10 will be scheduled roughly 10 days out; a card with S = 60 won't reappear for two months.
- Difficulty (D) — a 1–10 score capturing how intrinsically hard the card is for you. Cards you consistently fumble receive a higher D and are spaced more conservatively.
- Retrievability (R) — the probability, expressed as a percentage, that you can recall the card right now. FSRS targets a configurable desired retention, typically 90%, and schedules the next review at the exact moment R crosses that threshold.
- Due date — derived from S and R; the scheduler solves for the interval that keeps R at or above the target.
Why FSRS beats SM-2
The older SM-2 algorithm (used in early Anki versions) used a fixed ease multiplier that drifted toward extreme values over time, producing intervals that were either far too long or painfully short — a phenomenon Anki users know as "ease hell." FSRS replaces the ease multiplier with a continuous stability curve fitted to large-scale recall data, so intervals adapt smoothly regardless of your rating history. In head-to-head studies, FSRS achieves target retention more accurately while requiring fewer total reviews.
How it works inside the platform
Every concept in the CCA-F review queue — from spaced repetition theory itself to agentic loop mechanics — gets its own FSRS parameter set. When you answer "Again," stability resets sharply and difficulty rises slightly. "Hard" advances the card but tightens the interval. "Good" is the standard successful recall path. "Easy" extends the interval beyond the default and nudges difficulty down. Each rating feeds the FSRS equations to recompute S and D, then derives the next due date automatically.
Because FSRS tracks retrievability continuously, the platform can also surface a real-time "recall probability" for any card — useful when you want to self-assess readiness before the exam date rather than waiting passively for the queue.
Frequently Asked Questions
Does FSRS penalize me for pressing "Easy" too often?
Not as a penalty, but pressing "Easy" on cards you don't actually find easy lowers their difficulty score and pushes their next review further into the future — meaning you'll see them less often. Over time, stability for those cards may be inflated relative to your true retention, and they can appear overdue just before the exam. Rate honestly for best results.
What is the default target retention and can I change it?
The platform defaults to a 90% target retrievability — the industry-standard setting that balances retention depth against total review burden. At 90%, roughly 1 in 10 reviews will be a retrieval failure, which is actually desirable: the small difficulty of near-misses strengthens long-term encoding. If you have a fixed exam date, you can adjust target retention upward (say, 95%) to get more frequent reviews at the cost of more daily cards.
How is FSRS different from the Leitner box system?
Leitner boxes group cards into discrete buckets (Box 1 = review daily, Box 5 = review monthly) with fixed, bucket-wide intervals. Every card in Box 3 waits the same number of days regardless of how well you recall it. FSRS computes a unique interval for every card based on that card's individual stability and difficulty history, producing far finer-grained scheduling — especially valuable for a 150+ concept corpus like the CCA-F review queue where some topics are genuinely harder than others.