Under the Hood

The learning science behind LeapLearn

Every interaction pattern, topic archetype, and AI pipeline decision is rooted in decades of cognitive science research — Bloom’s taxonomy, the Feynman technique, spaced retrieval, and desirable difficulties.

0

Interaction Patterns

0

Topic Archetypes

0

AI Pipeline Passes

FSRS

Spaced Repetition

Try a Live Pattern

These are real interaction patterns from inside the app — some are fully interactive right here. Each tab shows the pattern, what the AI evaluates, and when it’s used.

"You've learned it — now prove you can explain it."

Explain compound interest as if teaching a curious friend who has never invested before. No jargon allowed.

Audience: a smart 15-year-old who asks “but why?”
Tier 3 — Explain it

What gets evaluated

Core mechanism explained clearly
At least one concrete real-world example
Compounding effect or "why it matters" addressed

Used when

The final challenge of every concept. Learners explain in their own words — the Feynman technique. Most powerful consolidation pattern in the system.

31 Interaction Patterns

Every concept gets the right kind of thinking — recall to build hooks, analysis to test understanding, creation to lock in deep fluency. Mapped to Bloom’s taxonomy tiers: Recall → Apply → Create.

Recall

6 patterns

Build mental hooks. Low-friction, recognition-heavy patterns for first exposure.

Match & Connect

Pair terms with their meanings.

Odd One Out

Spot the item that breaks the pattern.

Complete the Analogy

A is to B as C is to ___.

Fill the Gap

Drop the right words into a sentence.

Label the Diagram

Tap the correct label onto each part.

Prior Knowledge

Connect what you already know to anchor new ideas.

Analysis

15 patterns

Real understanding, not mimicry. Apply rules, find flaws, predict outcomes.

Predict & Reveal

Guess first, then learn why.

Misconception Trap

"Most people think..." — then the truth.

Transfer & Apply

Use the concept in a brand-new context.

Socratic Dialogue

Guided questions that surface your reasoning.

Sort & Classify

Group items into the right categories.

Estimate First

Slide to a guess before seeing the number.

Sequence Builder

Drag steps into the correct order.

Sequence Order

Restore shuffled steps to the right sequence.

Spot the Difference

Compare two versions and find what changed.

Spot the Error

Find the flaw in reasoning, data, or a process.

Question the AI

Narrow down a hidden concept with yes/no questions.

Hierarchy Build

Assemble a hierarchy — what belongs under what?

Evidence Match

Connect supporting evidence to the right claims.

Area Estimate

Estimate a spatial or quantitative area.

Variable Impact

Change one variable and see how it ripples.

Creation

10 patterns

Deep fluency. Teach it back, build a system, defend a position.

Teach Back

Explain it like you're tutoring a friend.

Simulate & Explore

Move sliders and watch outcomes change live.

Sandbox Play

Open-ended exploration. No wrong answers.

Build the System

Assemble components into a working whole.

Take a Side

Pick a position and defend it with reasoning.

Perspective Switch

Same scenario, multiple viewpoints — synthesize.

Draw Your Prediction

Sketch how you think it works, then compare.

Time Traveler

"What if X had gone differently?" — counterfactuals.

Case Study Choice

Make real decisions in a scenario, then see outcomes.

Concept Map

Map how concepts connect to each other in a web.

9 Topic Archetypes

Every topic is classified by the AI into one of 9 archetypes. A procedural topic gets ordering and simulations; a narrative topic gets timelines and perspective switching. The patterns always fit the subject.

Conceptual

Abstract ideas and theories

e.g. "What is gravity?"

Procedural

Step-by-step processes

e.g. "How photosynthesis works"

Systemic

Cause-and-effect systems

e.g. "How interest rates move markets"

Narrative

Stories, history, case studies

e.g. "The fall of Rome"

Linguistic

Vocabulary and language patterns

e.g. "Spanish past tense"

Creative

Open-ended, design-driven topics

e.g. "Color theory in branding"

Kinesthetic

Physical / hands-on processes

e.g. "How to throw a curveball"

Spatial

Diagrams, anatomy, architecture

e.g. "Parts of a cell"

Quantitative

Numbers, magnitudes, statistics

e.g. "How big is a black hole?"

5-Pass AI Content Pipeline

Every topic goes through 5 sequential AI passes before a learner sees it.

Pass 0

Pre-classification

Fast Haiku call — determines tentative archetype, domain, and scope to select the right Pass 1 prompt variant.

Pass 1

Structure extraction

Sonnet extracts 3–7 sub-concepts with titles, summaries, explanations, key facts, misconceptions, and curiosity hooks. Wikipedia/DDG enrichment grounds facts.

Pass 2

Topological sort

Kahn's algorithm reorders concepts so prerequisites come before dependents — guaranteeing a smooth, dependency-aware learning path.

Pass 3

Archetype classification

Sonnet assigns a primary + up to 2 secondary archetypes. Keyword hints from concept text serve as tie-breakers.

Pass 4

Interaction generation

Opus/Sonnet generates 3 interactions per concept (one per Bloom tier) + 1 microCheck + 2 practice MCQs. Depth gate, archetype filter, and grounding scan all run here.

Seen enough?

The best way to understand it is to try a session.

Start Learning Free