Most people use AI like a slightly smarter Google.
• Ask a question, get an answer, close the tab, forget everything
• Meanwhile, a small group of people is quietly building self-improving knowledge bases that compound every week
• And those people will outcompete everyone else in research, products, and careers
This article shows you exactly how to build one - using Obsidian, Claude Code, Skills, and plain markdown. No PhD required. Just a system you can set up this weekend.
---
## What You're Actually Building
Forget "taking notes". You're building a living wiki that an LLM writes and maintains for you in the background.
The idea comes from Andrej Karpathy (ex-OpenAI, Tesla AI) who recently shared that a large fraction of his LLM usage has shifted - less into writing code, more into manipulating knowledge stored as markdown files.
Here's the full loop:
• You dump raw data (articles, papers, repos, transcripts, images) into a raw/ folder
• An LLM compiles that into a structured wiki in wiki/ - markdown files with concepts, backlinks, summaries
• You view and navigate it in Obsidian with graph view
• You ask complex questions - the LLM reads your wiki and writes answers as new files
• Periodically, an LLM "health check" cleans inconsistencies and proposes new articles
• Every query files back into the wiki, making it smarter over time
The mental shift: LLM as compiler and librarian - not chatbot.
---
## Step 1 - Make Capture Stupidly Easy
Generated by Thread Navigator
Press ⌘ + S to quick-export