Overview

Spaced repetition produces 74% better retention than cramming (meta-analysis: 317 experiments, 839 assessments), with molecular mechanisms (CREB, MAPK) explaining exactly why spacing is biologically mandatory. Yet 99.9% of users fail: only 0.1% of Duolingo users complete a course, and education apps have the lowest retention rate (1.76%) of any mobile category. The paradox reveals fundamental conflicts between learning science and business models—engagement metrics demand daily sessions while optimal learning requires strategic difficulty. Modern algorithms like FSRS predict forgetting with impressive precision, but prediction accuracy ≠ learning outcomes. The real failure isn't the math; it's the recognition-production gap (flashcard mastery doesn't create conversational fluency), review burden exponential growth, and the 140-year adoption failure despite robust evidence. Successful learners treat SRS as 10-30% of study time within holistic systems emphasizing comprehensible input and production practice. Full research report: https://research.yuda.me/podcast/episodes/algorithms-for-life/ep1-spaced-repetition/report.md

Key Timestamps

• 0:00 - Introduction: The 99.9% Failure Paradox
• 3:00 - Molecular Switches: CREB and the Biology of Memory
• 7:00 - MAPK: The 45-Minute Temporal Window
• 10:00 - Hippocampal-Cortical Transfer: The Long-Term Archive
• 13:00 - Algorithm History: From SM-0 to SM-2
• 15:00 - FSRS: Modern Machine Learning Scheduling
• 18:00 - Prediction vs. Learning: The Critical Gap
• 21:00 - Anki vs. Duolingo: Two Opposing Philosophies
• 25:00 - Business Models: Engagement vs. Learning
• 29:00 - The 140-Year Adoption Failure
• 33:00 - Recognition-Production Gap: Why Users Can't Speak
• 37:00 - Polyglot Integration: SRS as Supplement
• 41:00 - Key Takeaways: Trust Desirable Difficulty

Sources

Sources for Algorithms for Life: Ep. 1, Spaced Repetition

Research Tools Used

• Perplexity (Academic & Official - automated)
• Grok (Real-Time & Regional - manual)
• GPT-Researcher (Industry & Technical - OpenAI GPT-5.2 - automated)
• Gemini Deep Research (Strategic & Policy - automated)

Verified Sources by Tier

Tier 1: Meta-analyses, Systematic Reviews, Official Statistics

Tier 2: RCTs, Large Studies, Government Reports

Tier 3: Case Studies, Industry Reports, News

Notes

• Research compiled: 2025-12-15
• Sources cross-validated across multiple tools
• Conflicting sources noted in research/p3-briefing.md