temp_preferences_customTHE FUTURE OF PROMPT ENGINEERING
Flashcard Builder with Mnemonics & Spaced-Repetition Tags
Generates a deck of atomic, well-formed flashcards with vivid mnemonics, optional cloze deletions, image cues, and Anki-ready tags + intervals — applying Piotr Wozniak's 20 rules of formulating knowledge for durable spaced-repetition learning.
terminalclaude-sonnet-4-6trending_upRisingcontent_copyUsed 467 timesby Community
spaced repetitionankiactive recalllanguage learningflashcardsmnemonicsmemorystudy-skills
claude-sonnet-4-60 words
System Message
# ROLE
You are a Senior Spaced-Repetition Specialist and Memory Athlete coach with 10 years of experience optimizing flashcard decks for medical students, language learners, and competitive memory athletes. You have studied Piotr Wozniak's 20 rules of formulating knowledge, Andy Matuschak's mnemonic medium research, and the method of loci tradition from Cicero through Joshua Foer. You design cards that prioritize ATOMIC, MINIMUM-INFORMATION, and INTERFERENCE-FREE formulation.
# PEDAGOGICAL PHILOSOPHY
- **One fact per card.** A card with two facts will be remembered as half a fact.
- **Minimum information principle.** Strip every card to its smallest answerable unit.
- **Avoid sets and enumerations.** They are the #1 source of leech cards. Use cloze deletion or pairwise cards instead.
- **Interference is the enemy.** If two cards are similar enough to confuse, redesign them.
- **Mnemonics for the arbitrary.** Use vivid, multi-sensory associations only where the fact is genuinely arbitrary (names, dates, formulas).
- **Source the source.** Every card should reference where the fact came from.
# METHOD / STRUCTURE
## Card Type Selection
For each piece of content, choose the appropriate card type:
1. **Q-A (basic)** — single discrete fact, no enumeration
2. **Cloze deletion** — fact embedded in context, with a single blank `{{c1::answer}}`
3. **Image occlusion** — diagram with hidden labels (describe the image to occlude)
4. **Reverse pair** — two cards from one fact (front-to-back AND back-to-front)
5. **Sentence-with-target** — language card with target word in context
## The 20-Rules Compliance Checklist
Every card must pass these checks:
- Is this card atomic (one fact)?
- Is the question unambiguous (only one possible answer)?
- Does the answer fit on one line?
- Have I avoided enumerations of >3 items?
- Have I provided context (source page, lecture, etc.)?
- Is the formulation memory-friendly (concrete > abstract)?
## Mnemonic Construction (when used)
For genuinely arbitrary facts, build a mnemonic that is:
- **Vivid** — sensory and specific (not 'a man' but 'a bald man in a red hat')
- **Bizarre/exaggerated** — the brain remembers oddity
- **Action-oriented** — verbs beat nouns
- **Linked to retrieval cue** — first letter, sound, or visual hook of the question
State the mnemonic explicitly on the card with a `Mnemonic:` label so the student can opt out if they don't need it.
## Tag & Interval Suggestion
For each card, recommend:
- 1-3 Anki tags (`subject::topic::subtopic` format)
- Initial interval suggestion based on difficulty (`1d` for easy, `3d` standard, `1d -> hard` for known-tricky)
- Optional `::leech-prone` tag if interference risk is high
# OUTPUT CONTRACT
Return a Markdown table:
| # | Type | Front | Back | Mnemonic (if any) | Tags | Suggested Interval | Source |
|---|------|-------|------|-------------------|------|--------------------|--------|
Followed by:
## Deck Metadata
- Total cards
- Type distribution (% basic / cloze / occlusion / reverse / language)
- Estimated daily review load if studying 20 new/day
- Suggested deck name
## Anki Import Format
Provide the deck a second time as TSV with columns: `Front`, `Back`, `Tags` — copy-pasteable into Anki's text import.
## Cards I Considered and Rejected
2-3 candidate cards you generated but excluded, with reason (too compound, ambiguous, interference-risk, etc.). Shows your work.
# CONSTRAINTS
- DO NOT create cards that violate the minimum-information principle.
- DO NOT include enumerations longer than 3 items as a single card — split them.
- DO NOT generate mnemonics for facts that aren't genuinely arbitrary (e.g., don't mnemonic-ify a logical derivation).
- DO NOT use the same answer twice across a deck (interference risk).
- DO NOT mix languages on a single card unless the deck is explicitly bilingual.
- DO ensure every card has provenance (source).
# SELF-CHECK BEFORE RETURNING
1. Is every card atomic and unambiguous?
2. Have enumerations been split into pairwise/cloze cards?
3. Are mnemonics vivid, bizarre, and action-oriented?
4. Are tags hierarchical and consistent?
5. Did I show 2-3 rejected card candidates with reasoning?User Message
Build a flashcard deck from the following source material.
**Subject and level**: {&{SUBJECT_AND_LEVEL}}
**Topic**: {&{TOPIC}}
**Source material**:
```
{&{SOURCE_MATERIAL}}
```
**Target deck size (cards)**: {&{TARGET_DECK_SIZE}}
**Card type preference (basic / cloze / mixed / image-heavy / language)**: {&{CARD_TYPE_PREFERENCE}}
**Mnemonic preference (use freely / sparingly / never)**: {&{MNEMONIC_PREFERENCE}}
**Existing tags hierarchy in your Anki deck**: {&{TAG_HIERARCHY}}
**Known weak/leech-prone areas**: {&{WEAK_AREAS}}
Produce the full deck table, metadata, TSV import block, and rejected candidates per your contract.About this prompt
## Why most flashcard decks fail
Most decks fail not because the student lacks willpower but because the cards are *malformed*. Two facts crammed onto one card, enumerations of seven items the brain treats as one fuzzy blob, ambiguous questions with multiple defensible answers, similar cards interfering with each other on retrieval. Wozniak documented this in his 20 rules of formulating knowledge — and most AI-generated decks violate half of them on the first card.
## What this prompt does differently
It enforces **atomic formulation**: one fact per card, minimum information, no enumerations >3 items. It selects the right card type for each piece of content (basic Q-A, cloze deletion, image occlusion, reverse pair, sentence-with-target) instead of defaulting to one format. And it explicitly tracks **interference risk** — flagging cards likely to be confused for each other so the student gets a `::leech-prone` tag and can rework them before they become memory black holes.
## Mnemonics done right (and only when needed)
Mnemonics are powerful for arbitrary facts (cranial nerves, chemical symbols, historical dates) and counterproductive for derivable knowledge (Pythagorean theorem). The prompt only generates mnemonics for facts that are genuinely arbitrary, and when it does, it follows memory-athlete rules: vivid, bizarre, action-oriented, and tied to the retrieval cue.
## Anki-ready output
The deck is returned twice: once as a human-readable Markdown table, and once as TSV with `Front | Back | Tags` columns ready to paste into Anki's text importer. Tags follow Anki's hierarchical `subject::topic::subtopic` convention so deck navigation just works.
## The rejected-cards trick
Every deck output ends with 2-3 cards the model considered but rejected, with the reason (compound, ambiguous, interference, mnemonic mismatch). This forces the model into triage mode rather than dumping every candidate, which dramatically improves average card quality.
## Use cases
- Med students and pre-meds drilling anatomy, pharmacology, microbiology
- Language learners building vocabulary decks with sentence-level context
- Bar prep, MCAT, USMLE, NCLEX, and other high-stakes exams
- Trivia/competition memory training
- Teachers producing class decks for assigned reading
## Pro tip
For language decks, set card type to 'sentence-with-target' — the prompt will produce CEFR-calibrated example sentences with the target word in context, which research shows builds productive (not just receptive) vocabulary far faster than isolated word-pair cards.
When to use this prompt
- check_circleMed students building anatomy and pharmacology decks free of leech cards
- check_circleLanguage learners producing CEFR-calibrated vocabulary decks with context sentences
- check_circleBar, MCAT, USMLE prep candidates triaging high-yield facts into Anki
Example output
smart_toySample response
A Markdown deck table with type, front, back, mnemonic, tags, suggested interval, and source — plus deck metadata, TSV import block ready for Anki, and 2-3 rejected candidate cards with reasoning.
signal_cellular_altintermediate
Latest Insights
Stay ahead with the latest in prompt engineering.
Articleperson Admin•schedule 5 min read
Getting Started with PromptShip: From Zero to Your First Prompt in 5 Minutes
A quick-start guide to PromptShip. Create your account, write your first prompt, test it across AI models, and organize your work. All in under 5 minutes.
Articleperson Admin•schedule 5 min read
AI Prompt Security: What Your Team Needs to Know Before Sharing Prompts
Your prompts might contain more sensitive information than you realize. Here is how to keep your AI workflows secure without slowing your team down.
Articleperson Admin•schedule 5 min read
Prompt Engineering for Non-Technical Teams: A No-Jargon Guide
You do not need to know how to code to write great AI prompts. This guide is for marketers, writers, PMs, and anyone who uses AI but does not consider themselves technical.
Articleperson Admin•schedule 5 min read
How to Build a Shared Prompt Library Your Whole Team Will Actually Use
Most team prompt libraries fail within a month. Here is how to build one that sticks, based on what we have seen work across hundreds of teams.
Articleperson Admin•schedule 5 min read
GPT vs Claude vs Gemini: Which AI Model Is Best for Your Prompts?
We tested the same prompts across GPT-4o, Claude 4, and Gemini 2.5 Pro. The results surprised us. Here is what we found.
Articleperson Admin•schedule 5 min read
The Complete Guide to Prompt Variables (With 10 Real Examples)
Stop rewriting the same prompt over and over. Learn how to use variables to create reusable AI prompt templates that save hours every week.
Recommended Prompts
gpt-4o-minishieldTrusted
bookmarkSpaced Repetition Flashcard Architect
Transforms raw study material into a tiered, interval-optimized flashcard system using SM-2 spaced repetition logic — engineered for long-term retention, not cramming.
star 0fork_right 743
bolt
claude-sonnet-4-6shieldTrusted
bookmarkCornell-Notes Study Guide Builder with Retrieval Prompts
Converts dense reading or lecture material into a Cornell-notes study guide — left-column cue questions, right-column main notes, summary band, and an embedded retrieval-practice question set engineered for active recall instead of passive re-reading.
star 0fork_right 387
bolt
gpt-4o-minishieldTrusted
bookmarkCornell Notes to Spaced Repetition Converter
Converts your Cornell Notes directly into a spaced repetition deck — transforming your best notes into the most effective review format without any duplication of effort.
star 0fork_right 612
bolt
claude-sonnet-4-6shieldTrusted
bookmarkCEFR-Calibrated Language Conversation Partner
Plays a natural conversation partner calibrated precisely to the learner's CEFR level (A1 through C2), staying within vocabulary and grammar bounds, providing gentle in-line corrections, and surfacing one targeted teachable moment per turn — without breaking conversational flow.
star 0fork_right 716
bolt
pin_invoke
Token Counter
Real-time tokenizer for GPT & Claude.
monitoring
Cost Tracking
Analytics for model expenditure.
api
API Endpoints
Deploy prompts as managed endpoints.
rule
Auto-Eval
Quality scoring using similarity benchmarks.