| name | quiz |
| description | Interactive spaced repetition quiz on architecture concepts. Uses SM-2 scheduling (Spacing), mixes categories (Interleaving), and asks WHY questions (Retrieval Practice). Use when: 'quiz me', 'quiz', 'test my knowledge', 'spaced repetition', 'review session'. |
Quiz — Spacing + Interleaving + Retrieval Practice
Structured quiz session using the SM-2 algorithm. Combines three proven learning techniques.
Learning Principles
Spacing: Reviewing at increasing intervals (1d → 6d → 15d → 35d) beats cramming by 200-400%.
Interleaving: Mixing topics (web, auth, infra, React in one session) builds stronger connections than blocking (all React, then all auth).
Retrieval Practice: Answering WHY questions from memory, not recognition. The struggle to recall is what strengthens the memory.
Instructions
- Read concepts from
~/.learn-while-code/data/concepts.jsonl
- Find concepts due for review:
- Status
encountered — highest priority (never quizzed)
nextReview <= today — SM-2 scheduled review- Prioritize concepts with higher cross-project frequency (
~/.learn-while-code/project-map.json)
- Interleave: don't pick 3 concepts from the same category; mix categories
- Select up to 5 concepts for this session
- For each concept:
a. State the concept name and category
b. Read the matching pattern from
~/.learn-while-code/data/patterns.jsonl
c. Ask the whyQuestion — don't show the explanation yet
d. Wait for the user's answer
e. Show the correct answer (2-3 sentences from explanation field)
f. Ask user to self-rate: "How confident are you? (1=no idea, 2=wrong but recognized it, 3=correct with difficulty, 4=correct easily, 5=trivial)"
g. Update concept via SM-2 algorithm, append to quiz-history.jsonl
h. If Obsidian sync is enabled, sync the concept note AND run syncAllIndexes() to update category/project index notes
- Show summary after all questions
Summary Format
## Quiz Complete
| Concept | Rating | New Status | Next Review |
|---------|--------|------------|-------------|
| Express Middleware | 4/5 | understood | 2026-06-14 |
| Zustand State | 2/5 | quizzed | 2026-06-09 |
| JWT Validation | 3/5 | quizzed | 2026-06-14 |
Streak: 3 days | Total quizzes: 15 | Avg confidence: 3.2/5
SM-2 Quality Mapping
| Rating | Meaning | SM-2 Effect |
|---|
| 1 | Blackout, no idea | Reset interval to 1 day |
| 2 | Wrong, but recognized the topic | Reset interval to 1 day |
| 3 | Correct with difficulty | Interval grows: 1→6→15→35 days |
| 4 | Correct with minor hesitation | Interval grows faster (higher ease) |
| 5 | Trivial, instant recall | Maximum growth, ease factor increases |
Gotchas
- Use the learning profile. The SessionStart hook injects what the user has mastered vs. weak. When explaining correct answers, connect to concepts they already understand — "you know Express middleware; this is the same layering idea applied to auth."
- Never quiz on
unseen concepts — they haven't been encountered yet.
- If no concepts are due: "All caught up! Keep coding — new patterns are detected automatically."
- Maximum 5 questions per session to prevent quiz fatigue.
- Match the user's language.
- The SM-2 ease factor starts at 2.5 and adjusts based on performance. Consistently low ratings make a concept appear more often.
