Karpathyのセカンドブレインを完全解説 — コード不要で構築する方法

Karpathy dropped a post last week about building a personal knowledge base with AI. 41K people bookmarked it. But most of them will never actually build one. Here's why you should, and exactly how to do it this weekend. PS: Want the skill that builds this system for you? Grab the free LLM Knowledge Base skill. ------------------------------------- Why Most Business Owners Are Sitting on a Gold Mine of Unused Knowledge You already have everything you need to build a competitive advantage. It's sitting in your bookmarks, your notes app, screenshots you forgot about, articles you saved and never re-read, meeting notes from six months ago. The problem isn't collecting information. You're already doing that. The problem is you can't find any of it when you need it. A personal knowledge base fixes this permanently. You dump everything into one place, point an AI at it, and it organizes the mess into a searchable wiki that gets smarter every time you use it. No special software. No database. Just folders and text files. The Setup (Takes 5 Minutes) Create a project folder anywhere on your computer. Inside it, make three subfolders: raw/ - Your junk drawer. Articles, notes, screenshots, meeting transcripts, bookmarks, research. Everything goes here. Don't organize it. That's the AI's job. wiki/ - Where the AI writes the organized version. Summaries, connections between ideas, topic pages. You never edit this by hand. outputs/ - Answers, reports, and research the AI generates when you ask questions against your knowledge base. That's the entire architecture. Three folders. The One File That Makes Everything Work Create a file in the root of your project called CLAUDE.md (or AGENTS.md, the name doesn't matter). This is your schema file. It tells the AI what the knowledge base is about and how to organize it. Here's a starter template: # Knowledge Base Schema ## What This Is A personal knowledge base about [YOUR TOPIC]. ## How It's Organized - raw/ contains unprocessed source material. Never modify these files. - wiki/ contains the organized wiki. AI maintains this entirely. - outputs/ contains generated reports, answers, and analyses. ## Wiki Rules - Every topic gets its own .md file in wiki/ - Every wiki file starts with a one-paragraph summary - Link related topics using [[topic-name]] format - Maintain an INDEX.md that lists every topic - When new raw sources are added, update the relevant wiki articles ## My Interests [List 3-5 things you want this knowledge base to focus on] This file is the equivalent of giving a new employee a training manual on day one. Without it, the AI just guesses at what matters. With it, every output is structured exactly how you want. Fill Your Raw Folder (10 Minutes) This is where people stall. They create the folders and stare at an empty directory. The answer: dump everything. Copy-paste articles into .md or .txt files. Export notes from whatever app you use. Save screenshots. Paste in meeting notes, research, project docs. Don't rename anything. Don't clean it up. I keep 15+ raw source files across my content pipeline. Clipped articles, competitor breakdowns, analytics reports. None of it is organized by hand. Tell the AI to Build Your Wiki Open Claude Code, Cursor, or any AI tool that can read your files. Point it at your project folder and say: "Read everything in raw/. Compile a wiki in wiki/ following the rules in CLAUDE.md. Create an INDEX.md first, then one .md file per major topic. Link related topics. Summarize every source." Then walk away. Let it work. When it's done, you'll have a wiki folder full of organized articles with connections you didn't see, summaries of things you forgot you saved, and an index that makes everything searchable in seconds. The Compounding Loop This is where the system becomes valuable. Once your wiki has 10+ articles, start asking questions: "Based on everything in wiki/, what are the three biggest gaps in my understanding of [topic]?" "Compare what source A says about [concept] vs source B. Where do they disagree?" "Write me a 500-word briefing on [topic] using only what's in this knowledge base." Save the answers back into the knowledge base. Every question makes the next answer better. That's the loop. Monthly Health Check (Non-Negotiable) Once a month, tell the AI: "Review the entire wiki/ directory. Flag contradictions between articles. Find topics mentioned but never explained. List claims not backed by a source in raw/. Suggest 3 new articles to fill gaps." This catches errors before they compound. If the AI writes something slightly wrong and you save it back, the next answer builds on that mistake. The health check is your quality control. You Don't Need Obsidian Half the internet is pitching Obsidian plugins for this. You don't need them. A folder of .md files and a good schema file will outperform a fancy tool stack 90% of the time. I've watched people spend more time configuring their note-taking app than actually building their knowledge base. Stop shopping for the perfect tool. Start building. Three folders. One schema file. An AI that maintains everything. That's the whole system. Pick your topic. Create the folders. Drop in what you already have. Let the AI do the rest. PS: Want the skill that builds this system for you? Grab the free LLM Knowledge Base skill (free).