The Model
Knowledge is a spectrum, not a switch
You don't "learn" something — you move it through stages. Most people loop in Stage 1 forever and mistake familiarity for understanding. Each step in the protocol pushes knowledge one stage deeper.
1Exposure
2Structure
3Retrieval
4Application
5Transfer
← "I've seen this" ——————————— "I own this" →
Reading, watching, browsing
Organizing into a mental model
Pulling it from memory unaided
Using it to build or solve
Teaching, analogizing, combining
The trap: Your brain can't tell the difference between "I recognize this" and "I know this." Recognition feels like knowledge but decays in days. That's why you can re-read something and feel like you already know it — while being unable to explain it without looking. Every step below forces a different type of production, which is the only reliable signal that encoding actually happened.
The Protocol
Five steps. One topic. Repeat for everything.
Do these in order. Each step depends on the previous one. The whole cycle takes about 90 minutes spread across 2–3 days — the spacing is a feature, not a bug.
This is what you're already doing. Ask Claude to create a single-page HTML site that gives you the spatial, visual overview of a topic. The key is that it's not just text — it's organized with hierarchy, groupings, and visual relationships that mirror how the concepts actually relate.
Prompt to Claude
"Create a single-page HTML site that builds my intuition for [TOPIC]. Organize it visually so I can see how the key concepts relate to each other. Use hierarchy, grouping, and color to show relationships. Include: the core mental model, the 5–8 most important concepts, how they connect, and one concrete example of each. Make it something I can glance at and immediately orient myself."
Output: an .html file you save and can revisit
Immediately after Step 1 — while the structure is fresh — ask Claude to generate Anki-style flashcards from the same content. The critical thing is card quality. You don't want "What is X? → Definition." You want cards that test relationships, causation, and edge cases. These cards are what prevent the structure from decaying over the next week.
Prompt to Claude
"Based on the intuition page you just made, generate 15–20 Anki flashcards. Rules: (1) No cards that test pure definitions — only test relationships, causation, or 'why' questions. (2) Use these formats: 'Why does X happen instead of Y?', 'If you changed Z, what breaks?', 'What's the relationship between A and B?', cloze deletions on causal chains. (3) Order them from foundational to advanced. Output as a two-column table I can paste into Anki."
Output: a table or CSV you import into Anki
Wait at least a few hours (ideally sleep on it). Then, without looking at your HTML page, explain the topic back to Claude and tell it to interrogate you. This is the Feynman technique on steroids — Claude will find the exact gaps in your understanding. The discomfort you feel when you can't answer is the learning happening. Lean into it.
Prompt to Claude
"I'm going to explain [TOPIC] to you from memory. Your job: interrupt me when I'm wrong, vague, or hand-waving. Ask follow-up questions that expose gaps. Don't be nice about it — I want to find where my understanding is shallow. After I'm done, give me a scorecard: what I nailed, what I got wrong, and what I skipped entirely. Here goes: [YOUR EXPLANATION]"
Output: a gap analysis of your actual understanding
Now use the knowledge. Ask Claude to give you a small, concrete challenge that requires applying what you've learned — not reciting it. For technical topics this might be building something. For conceptual topics it might be analyzing a scenario, making a prediction, or solving a case study. The key: Claude gives you the problem, you attempt it before asking for help.
Prompt to Claude
"Give me a concrete mini-challenge that requires me to actually apply my knowledge of [TOPIC]. It should take 15–30 minutes, require me to make decisions based on the concepts (not just recall them), and have a clear 'done' state. Don't give me hints up front — I'll attempt it first and come back to you if I get stuck."
Output: a solved challenge that proves you can use the knowledge
Write a short note (in Obsidian, naturally) that explains the topic in your own words, connects it to things you already know, and identifies what surprised you or what you'd tell a friend. Then have Claude critique it. This forces transfer — you're not just retrieving the original structure, you're rebuilding it in your own mental vocabulary and linking it to your existing knowledge graph.
Prompt to Claude
"Here's my synthesis note on [TOPIC]. Critique it: Is anything wrong? Is anything important missing? Where am I oversimplifying? Also — suggest 2–3 connections to other fields or concepts that would deepen my understanding. [YOUR NOTE]"
Output: a refined Obsidian note that becomes a permanent knowledge node
Why This Works
The underlying principles
⚡
Production over consumption
Every step after the first requires you to produce something — an explanation, a solution, a note. Your brain encodes what it produces, not what it receives. Reading is necessary but never sufficient.
🔄
Desirable difficulty
The protocol deliberately creates friction. Explaining without notes is hard. That difficulty is the signal that real encoding is happening. If it feels easy, you're probably in the recognition trap.
📐
Spaced retrieval
Steps are spread across days on purpose. Forgetting a little, then retrieving, strengthens the memory trace far more than cramming everything into one session. Anki automates this long-term.
🧩
Multiple encoding modes
Visual (HTML page) → verbal (interrogation) → procedural (challenge) → written (synthesis note). Each mode creates a different neural pathway to the same knowledge, making it more robust and accessible.
🔗
Connection to existing knowledge
The synthesis note forces you to link new knowledge to what you already know. Isolated facts are fragile; connected facts are durable. Your Obsidian graph is the external mirror of this internal linking.