Karpathy Shares LLM Wiki — An 'Idea File' Approach to Organizing AI Knowledge

Andrej Karpathy's Knowledge Management System for AI

Former Tesla AI director and OpenAI founding member Andrej Karpathy has shared his personal "LLM Wiki" — a curated collection of AI/ML knowledge organized as what he calls an "idea file."

What Is an Idea File?

An idea file is a personal knowledge management system where you capture:

• Technical concepts and intuitions
• Paper summaries and key insights
• Mental models and analogies
• Open questions and hypotheses
• Code snippets and experiments

Unlike traditional note-taking, idea files focus on connecting concepts rather than just recording facts.

Why Karpathy's Share Matters

Karpathy is one of the most influential figures in AI education, known for:

• His Stanford CS231n course (deep learning)
• The "Let's build GPT from scratch" YouTube series
• Clear, intuitive explanations of complex ML concepts
• His role at Tesla (Autopilot) and OpenAI

When someone at this level shares their personal knowledge organization system, it provides valuable insight into how top AI researchers structure their thinking.

The LLM Wiki Structure

While the full content wasn't publicly detailed, the concept of a personal LLM wiki aligns with a growing trend in the AI research community:

1. Second brain systems — External tools for augmenting memory and thinking
2. Personal knowledge graphs — Connecting related concepts across domains
3. Living documents — Continuously updated rather than static

Practical Takeaways

For AI practitioners and researchers:

• Start building your own idea file — Even a simple collection of insights compounds over time
• Focus on understanding, not memorization — Idea files work because they capture why, not just what
• Connect across domains — The most valuable insights often come from interdisciplinary connections
• Share selectively — Public sharing creates accountability and potential collaboration

The Bigger Picture

The concept of personal AI knowledge bases is becoming increasingly important as the field evolves rapidly. With papers, models, and techniques emerging daily, having a structured system for organizing and connecting knowledge is becoming essential for staying current.

Source: Andrej Karpathy (X/Twitter), Hacker News