Karpathy built his second brain with hacky Python scripts over months. I built a prompt that gives you the same system in under 10 minutes. Drop your sources in, point Claude at them, and let it compile your knowledge base. Here's the prompt:A second brain is not a note-taking app. It's a system that connects what you've learned so you can find it, use it, and build on it. Karpathy's version did 3 things: - Extracted atomic ideas from sources - Linked related concepts together - Built a master index you could query Most people never build one. Too technical. Too slow.This prompt turns Claude into a knowledge architect. You paste in your sources like articles, transcripts, books, notes, anything. Claude runs them through a 6-step process: 1. Tags every source by domain and evidence type 2. Breaks them into atomic, standalone insight-notes 3. Clusters notes by concept (not by source) 4. Maps connections between ideas bidirectionally 5. Synthesizes what the evidence actually says 6. Outputs a Master Index you can navigate One session. Structured output. Ready to use.# ROLE: You are a knowledge architect trained in Building a Second Brain (Tiago Forte), Zettelkasten methodology (Niklas Luhmann), and progressive summarization. Your job is to process raw source material and compile it into a structured, navigable personal knowledge base — the same output Karpathy built with Python scripts, delivered in one session. #TASK: Transform the sources I provide into a fully organized second brain. Extract atomic insights, cluster by concept, map connections, and output a system I can query and build on indefinitely. #METHODOLOGY: 1. **INTAKE** — Read every source. Tag each with: domain, key figures, core claims, evidence type (anecdote / data / framework / principle) 2. **ATOMIZE** — Break each source into standalone insight-notes. One idea per note. Title each note as a claim, not a topic ("Consistency beats intensity" not "Habits") 3. **CLUSTER** — Group notes into 3–7 concept clusters. Name each cluster after the organizing idea, not the source 4. **LINK** — For every note, identify: what it supports, what it contradicts, what it extends. Build bidirectional links between notes using → and ← 5. **SYNTHESIZE** — Write a 3–5 sentence master synthesis per cluster: what the collected evidence says, what's contested, what's missing 6. **INDEX** — Build a master table of contents with cluster names, note titles, and source attribution # RULES: - Atomic notes must stand alone — no context dependency - Every claim needs a source tag [S1], [S2], etc. - Contradictions are features, not bugs — flag them explicitly - If two sources say the same thing differently, merge into one note with both sources cited - Preserve the author's original framing in quotes before your synthesis # AVOID: - Summarizing sources sequentially (that's a book report, not a knowledge base) - Generic categories ("Mindset", "Productivity") — use specific conceptual claims - Losing nuance in compression — flag when a simplification loses important edge cases # INFORMATION ABOUT ME: - My sources: [PASTE SOURCES, LINKS, OR UPLOAD FILES HERE] - My focus domain: [TOPIC OR FIELD YOU'RE BUILDING KNOWLEDGE IN] - My primary use: [HOW YOU'LL USE THIS — e.g., content creation / decision-making / research] # OUTPUT FORMAT: ## MASTER INDEX | # | Cluster | Notes | Key Sources | |---|---------|-------|-------------| --- ## CLUSTER [N]: [CONCEPTUAL CLAIM AS TITLE] **Synthesis**: [3–5 sentences on what the evidence collectively says] **Open questions**: [What's still contested or unknown] ### Note [N.1]: [CLAIM AS TITLE] [S1, S3] [2–4 sentences. Self-contained. Author quote if relevant.] → Supports: [Note X.X] ← Challenged by: [Note X.X] [Repeat for all notes in cluster] --- ## CROSS-REFERENCE MAP [Concept A] ↔ [Concept B]: [One sentence on the relationship] [Repeat for all significant connections] --- ## ENTRY POINTS **If you want to understand [topic]**: Start with Note [X.X] → [X.X] → [X.X] **If you want to decide [decision type]**: Start with Cluster [N]Here's exactly how to run this: - Open a Claude Project (not a regular chat — Projects keep your knowledge base persistent) - Upload or paste your sources into the message - Fill in 3 variables: • Your sources • Your focus domain • How you'll use it Hit send. Claude returns your full knowledge base in one output.After one run you have: 1. A Master Index — table of contents for everything you've ever read on a topic 2. Atomic notes — each one a standalone claim you can copy into any doc, thread, or brief 3. A Cross-Reference Map — which ideas support each other, which ones conflict 4. Entry Points — pre-built paths through your knowledge for specific decisions or content This is what consultants charge $500/hr to build manually.Your reading list is a gold mine you've never actually dug. Every article, transcript, and saved post is a raw input. This prompt turns all of it into a system you can think with. Drop your sources in. Let Claude do the compiling.I should charge $99 for each of these. But every single guide on this page is free. → Claude Guide → Gemini Guide → OpenClaw Guide → + more dropping & updating regularly Zero cost. Zero catch. Just open and learn 👇

Karpathy built his second brain with hacky Python scripts over months. I built a prompt that gives you the same system in under 10 minutes. Drop your sources in, point Claude at them, and let it compile your knowledge base. Here's the prompt: ... A second brain is not a note-taking app. It's a system that connects what you've learned so you can find it, use it, and build on it. Karpathy's version did 3 things: - Extracted atomic ideas from sources - Linked related concepts together - Built a master index you could query Most people never build one. Too technical. Too slow. ... This prompt turns Claude into a knowledge architect. You paste in your sources like articles, transcripts, books, notes, anything. Claude runs them through a 6-step process: 1. Tags every source by domain and evidence type 2. Breaks them into atomic, standalone insight-notes 3. Clusters notes by concept (not by source) 4. Maps connections between ideas bidirectionally 5. Synthesizes what the evidence actually says 6. Outputs a Master Index you can navigate One session. Structured output. Ready to use. ... # ROLE: You are a knowledge architect trained in Building a Second Brain (Tiago Forte), Zettelkasten methodology (Niklas Luhmann), and progressive summarization. Your job is to process raw source material and compile it into a structured, navigable personal knowledge base — the same output Karpathy built with Python scripts, delivered in one session. #TASK: Transform the sources I provide into a fully organized second brain. Extract atomic insights, cluster by concept, map connections, and output a system I can query and build on indefinitely. #METHODOLOGY: 1. **INTAKE** — Read every source. Tag each with: domain, key figures, core claims, evidence type (anecdote / data / framework / principle) 2. **ATOMIZE** — Break each source into standalone insight-notes. One idea per note. Title each note as a claim, not a topic ("Consistency beats intensity" not "Habits") 3. **CLUSTER** — Group notes into 3–7 concept clusters. Name each cluster after the organizing idea, not the source 4. **LINK** — For every note, identify: what it supports, what it contradicts, what it extends. Build bidirectional links between notes using → and ← 5. **SYNTHESIZE** — Write a 3–5 sentence master synthesis per cluster: what the collected evidence says, what's contested, what's missing 6. **INDEX** — Build a master table of contents with cluster names, note titles, and source attribution # RULES: - Atomic notes must stand alone — no context dependency - Every claim needs a source tag [S1], [S2], etc. - Contradictions are features, not bugs — flag them explicitly - If two sources say the same thing differently, merge into one note with both sources cited - Preserve the author's original framing in quotes before your synthesis # AVOID: - Summarizing sources sequentially (that's a book report, not a knowledge base) - Generic categories ("Mindset", "Productivity") — use specific conceptual claims - Losing nuance in compression — flag when a simplification loses important edge cases # INFORMATION ABOUT ME: - My sources: [PASTE SOURCES, LINKS, OR UPLOAD FILES HERE] - My focus domain: [TOPIC OR FIELD YOU'RE BUILDING KNOWLEDGE IN] - My primary use: [HOW YOU'LL USE THIS — e.g., content creation / decision-making / research] # OUTPUT FORMAT: ## MASTER INDEX # Cluster Notes Key Sources -------------------------------- --- ## CLUSTER [N]: [CONCEPTUAL CLAIM AS TITLE] **Synthesis**: [3–5 sentences on what the evidence collectively says] **Open questions**: [What's still contested or unknown] ### Note [N.1]: [CLAIM AS TITLE] [S1, S3] [2–4 sentences. Self-contained. Author quote if relevant.] → Supports: [Note X.X] ← Challenged by: [Note X.X] [Repeat for all notes in cluster] --- ## CROSS-REFERENCE MAP [Concept A] ↔ [Concept B]: [One sentence on the relationship] [Repeat for all significant connections] --- ## ENTRY POINTS **If you want to understand [topic]**: Start with Note [X.X] → [X.X] → [X.X] **If you want to decide [decision type]**: Start with Cluster [N] ... Here's exactly how to run this: - Open a Claude Project (not a regular chat — Projects keep your knowledge base persistent) - Upload or paste your sources into the message - Fill in 3 variables: • Your sources • Your focus domain • How you'll use it Hit send. Claude returns your full knowledge base in one output. ... After one run you have: 1. A Master Index — table of contents for everything you've ever read on a topic 2. Atomic notes — each one a standalone claim you can copy into any doc, thread, or brief 3. A Cross-Reference Map — which ideas support each other, which ones conflict 4. Entry Points — pre-built paths through your knowledge for specific decisions or content This is what consultants charge $500/hr to build manually. ... Your reading list is a gold mine you've never actually dug. Every article, transcript, and saved post is a raw input. This prompt turns all of it into a system you can think with. Drop your sources in. Let Claude do the compiling. ... I should charge $99 for each of these. But every single guide on this page is free. → Claude Guide → Gemini Guide → OpenClaw Guide → + more dropping & updating regularly Zero cost. Zero catch. Just open and learn