Andrej Karpathy published his personal LLM 'idea file' as a GitHub Gist, revealing how a top AI researcher organizes evolving knowledge.
Andrej Karpathy shared a GitHub Gist titled 'LLM Wiki – example of an idea file,' offering a rare look at how he personally tracks and structures LLM-related knowledge. The gist functions as a living document — a curated, opinionated knowledge base rather than a traditional wiki. No major product or model announcement accompanies it; the signal is methodological, not technical. It surfaces a workflow pattern increasingly relevant to anyone building or reasoning about AI systems at speed.
This isn't a technical breakthrough — it's a process artifact. Karpathy's idea file shows how a world-class AI researcher externalizes fast-moving LLM knowledge into a structured, queryable format. For developers drowning in papers, tweets, and changelogs, the methodology matters more than the content. The real value is seeing what categories he chooses to track and how he handles knowledge decay in a field that obsoletes itself monthly.
Clone Karpathy's gist structure into your own Obsidian or Notion vault this week and populate it with the last 10 LLM developments you've been meaning to organize — it forces prioritization faster than any reading list.
Go to gist.github.com and open karpathy/442a6bf555914893e9891c11519de94f
Tags
Also today
Signals by role
Also today
Tools mentioned