Think of a deck like a subject folder. Each deck tracks its own:

• Card due dates (scheduled by SM-2, FSRS, or IRT)
• Per-card stability and ease values
• Session history and accuracy trends
• KPI gate performance (accuracy, retention, response time, fatigue)

Tags let you cross-reference cards across decks. Use tags like chapter-3, difficult, or exam-ready to filter your review queue. Tagged searches work in the Statistics and Decks views.

Click ▶ Start Session in the top bar or open a specific deck and press Study Now. The session header shows:

• Deck name and algorithm badge (SM-2 · Ease 2.4 / FSRS · S=7d / IRT · θ=1.2)
• Cards remaining and a thin progress bar
• Session KPI gauges (accuracy, retention, response time, fatigue)

• Space — flip card (front → back)
• 1 — Again    2 — Hard    3 — Good    4 — Easy
• Ctrl+D — toggle dark/light theme
• Esc — close settings or dialogs

After the last card, a summary panel shows four KPI gates. Green = gate passed, amber = borderline, red = gate failed:

• Accuracy ≥ 75% (Good + Easy) / total responses
• Retention ≥ 70% reviews where recall was maintained
• Avg Response ≤ 6 000 ms mean think-time per card
• Fatigue Drop ≤ 15% accuracy decline from first-half to second-half

Failing a gate adapts the next session's card count or difficulty distribution automatically.

Light Mode

Clean, bright interface with colorful backgrounds perfect for daytime study sessions.

Dark Mode

Easy on the eyes with proper contrast ratios, ideal for evening study or low-light environments.

High Contrast Mode

WCAG AAA compliant design with maximum contrast for users with visual impairments.

Auto Theme

Automatically switch between light and dark modes based on your system settings or time of day.

Text Size

• Small: Compact view for more content on screen
• Medium: Standard size for optimal readability
• Large: Enhanced readability for comfortable viewing
• Extra Large: Maximum readability for accessibility

Font Families

• System: Your device's default font for familiarity
• Inter: Modern, highly legible typeface
• Roboto: Google's clean, friendly design

Intensity Control

Adjust the intensity of shimmer animations on cards and UI elements. Set to low for subtle effects, or high for more pronounced visual feedback.

Toggle On/Off

Completely disable shimmer effects to reduce visual distraction or improve performance on older devices.

Flip Effects

Choose between different card flip animations when revealing answers:

• Classic Flip: Traditional card flip animation
• Fade Transition: Smooth crossfade between question and answer
• Slide Reveal: Answer slides in from the bottom
• Instant: No animation for fastest response

Speed Settings

• Slow: 0.5s transitions for relaxed study pace
• Normal: 0.3s balanced speed for most users
• Fast: 0.1s quick transitions for rapid reviews

Deck Loading

Customize loading indicators when opening decks or syncing data. Choose from spinner, progress bar, or skeleton loading styles.

Card Loading

Control how new cards appear during study sessions with smooth entrance animations.

365-Day Learning Calendar

Visualize your study consistency with a GitHub-style activity heat map showing your daily review counts over the past year. Darker colors indicate more intensive study days.

Study Streaks

Track your current streak and longest streak of consecutive study days. Maintain momentum and build lasting study habits.

Ease Factor & Stability Histograms

Visualize the distribution of SM-2 ease factors and FSRS stability values across your deck. Cards bunched at low stability indicate overdue reviews.

Forgetting Curves

Live SVG chart showing R(t) = e^(-t/S) at four stability levels (3 d, 7 d, 15 d, 30 d). See exactly when each cohort of cards crosses the 90% retention threshold.

IRT Ability Estimate

Track your θ (theta) ability score over time. A rising θ means cards are getting easier relative to your skill — the platform will introduce harder material automatically.

Alex Chen

Member since: January 2023

Total Reviews: 8,932

Overall Accuracy: 87%

Current Streak: 23 days

Study Goal: 50 cards per day

Preferred Study Time: 7:00 PM - 9:00 PM

Weekly Summary

• Cards reviewed this week: 347
• New cards learned: 42
• Average session length: 18 minutes
• Most studied deck: Japanese Vocabulary
• Peak performance day: Wednesday

Study Preferences

• Daily review goal: 50 cards
• New cards per day: 10
• Review ahead limit: 4 days
• Notification time: 7:00 PM
• Auto-sync: Enabled

Full Profile Export

Export your complete study data including:

• All card history and statistics
• Deck configurations and settings
• Learning progress and achievements
• Study session logs and analytics

Available formats: JSON, CSV, PDF Report

What you can export

• JSON (full backup) — all decks, cards, scheduling history, session logs, settings
• CSV — card content only (front, back, tags), compatible with any spreadsheet app
• PDF Report — formatted study report with statistics charts, streak history, and KPI summaries
• Anki .apkg — export to Anki-compatible format for use in the original Anki desktop app

Access exports from Profile → Export Options or from the three-dot menu on any individual deck.

Study data is saved to localStorage in your browser. For cross-device access, export a JSON backup and import it on the second device. Auto-sync to cloud storage is on the roadmap.

Tip: Set a weekly reminder to export a JSON backup. The file is typically under 2 MB even for large decks.

Ebbinghaus Memory Model

Retention decays exponentially: R(t) = e^(-t/S) where t is time since last review and S is the card's stability (in days). FSRS estimates S from your response history; SM-2 approximates it via ease × interval multiplication.

Interval Scheduling

SM-2 adapts intervals based on each response:

• Again (Red): Card resets — shown again in 1–10 minutes
• Hard (Orange): Interval × 1.2 — slight increase, ease penalized
• Good (Green): Interval × ease factor — standard progression
• Easy (Blue): Interval × ease × 1.3 — fast-track for well-known cards

Memory Stability Model

FSRS (Free Spaced Repetition Scheduler) treats memory stability S as a learnable parameter per card. It predicts the probability of recall at any future time using R(t) = e^(-t/S), then schedules the next review at the point where R(t) crosses your target retention threshold (default 90%).

FSRS is more accurate than SM-2 at the extremes — very easy cards get longer gaps, very hard cards get shorter ones, without the ease "hell" problem.

Difficulty & Ability Estimation

IRT models both question difficulty (b) and learner ability (θ) on the same logistic scale. The probability of a correct response is:

P(correct) = 1 / (1 + e^(-(θ - b)))

This allows the platform to select cards that are appropriately challenging — not too easy (boring) and not too hard (discouraging) — by targeting items where P(correct) ≈ 0.7.

The standard two-sided card. The front holds a question, prompt, or term; the back holds the answer, definition, or explanation. Works for vocabulary, facts, formulas, and concepts.

Example front: "What does the mitochondria do?"
Example back: "Produces ATP via cellular respiration (oxidative phosphorylation)."

Fill-in-the-blank format. Wrap the tested word or phrase in {{c1:: }} markers. The cloze is hidden on the front and revealed on the back.

Example: "The powerhouse of the cell is the {{c1::mitochondria}}."
Becomes: "The powerhouse of the cell is the _______."

Use c2, c3… for multiple blanks in one card — each generates a separate review.

Overlay rectangles on a diagram or image to hide labels. Great for anatomy, maps, circuit diagrams, and labelled charts. Each occluded region becomes its own testable question.

Render syntax-highlighted code blocks on either side. Supports JavaScript, Python, Rust, SQL, and 50+ languages. Ideal for API signatures, algorithm snippets, and regex patterns.

Use :: in a deck name to create a hierarchy. For example, Biology::Chapter 3::Cells appears as a nested tree in the sidebar. Sub-decks inherit the parent's algorithm setting unless overridden.

Each deck has an Options panel:

• New cards/day — throttle introduction rate (default 10)
• Max reviews/day — cap the review queue to prevent overload
• Algorithm — SM-2, FSRS, or IRT per deck
• Retention target — FSRS uses this to compute next-review intervals (default 90%)
• Leech threshold — cards rated "Again" more than N times get flagged for manual review

• Empty cards: Scan and delete cards with missing front or back
• Duplicate check: Find near-duplicate fronts across all decks
• Reset progress: Clear scheduling data for a card or entire deck (useful after re-learning a topic from scratch)
• Archive: Move inactive decks out of the main view without deleting them

Sync Problems

Issue: Cards not syncing between devices

Solution: Check internet connection, log out and back in, or force sync in settings.

Performance Issues

Issue: App running slowly or lagging

Solution: Disable animations, reduce text size, or clear app cache.

Display Problems

Issue: Text too small or hard to read

Solution: Increase text size, enable high contrast mode, or switch to dark theme.

How often should I study?

Consistency is key. It's better to study 15 minutes daily than 2 hours once a week. Aim for daily reviews to maintain optimal retention.

What's the difference between "Hard" and "Again"?

"Again" means you got it wrong and need to see it soon. "Hard" means you got it right but struggled, so the interval increases only slightly.

Can I study without an internet connection?

Yes — all three scheduling algorithms (SM-2, FSRS, IRT) run entirely in your browser. Only the AI Insights panel requires a network call to Groq · LLAMA-3.

How do I backup my data?

Study data is stored in localStorage. Use the Export feature (Profile → Export Options) to download a JSON backup. Import it on any browser to restore your history.

Which algorithm should I use?

Start with FSRS — it outperforms SM-2 for most users by modeling memory decay more accurately. Switch to IRT if you want fully adaptive difficulty selection. SM-2 is the most battle-tested if you prefer a proven baseline.