← ClaudeAtlas

learnanythinglisted

10-step AI-assisted learning loop (STORM multi-perspective research → contradiction map → brief → peer review → resource curation → learning ladder → 20% core plan → active-recall exam → Feynman loop → one-page cheat sheet). Use when the user wants to learn, study, research, or master a new topic/skill/domain — e.g. "help me learn X", "I want to study X", "teach me X", "10x learn X", or invokes /LearnAnything. Not for writing articles or one-off factual questions.
flexfengfeng/LearnAnything · ★ 1 · AI & Automation · score 62
Install: claude install-skill flexfengfeng/LearnAnything
# LearnAnything — 10x Learning Loop Structured learning method combining Rahul (@sairahul1)'s "Learn Anything 10x Faster" system with Stanford's STORM multi-perspective research method (NAACL paper). Core insight: **learning speed depends on structure, not on whether you use AI**. Real learning needs: multiple perspectives, one path, testing, compression, and a feedback loop. The full loop: research a topic from clashing perspectives → distill into a brief → self-critique → pick 5 resources → build a skill ladder → learn the core 20% → get examined until you break → patch gaps with Feynman loops → compress to a one-page cheat sheet. ## How to run this skill 1. Ask the user for the **topic** (if not given) and which mode they want: - **Full loop** — all 10 steps in order - **Research only** — steps 1–4 (fast domain map) - **Study plan** — steps 5–7 - **Test me** — step 8 (they already studied) - **Feynman** — step 9 on a specific concept - **Cheat sheet** — step 10 2. Execute each step yourself using the templates below, substituting the user's topic. Steps 1–7 and 10 are outputs you generate (use web search when it materially improves accuracy). Steps 8 and 9 are **interactive** — one question at a time, wait for user answers, never dump everything at once. 3. After each step, show the output, then continue (pause where the step needs user participation). 4. Persist outputs: save each step's result as markdown into `learn/<topic-slug>/` in the working di