TODAY'S SIGNAL
Andrej Karpathy — the man who coined "vibe coding" — just shared a new AI workflow that's gone massively viral: instead of using LLMs to write code, he's using them to build a personal knowledge base. Raw sources go in, a structured, interlinked wiki comes out — automatically written and maintained by an AI, no manual note-taking required.
WHY IT MATTERS
Most people use AI like a smarter search engine — ask a question, get an answer, close the tab. Nothing carries forward. Karpathy's system flips that entirely. Drop articles, papers, repos, and datasets into a folder. An LLM compiles them into a living wiki of summaries, backlinks, and concept articles — his own has grown to 100 articles and 400,000 words. No RAG pipelines, no vector databases. Just markdown files that an AI writes, maintains, and updates. He called the follow-up distribution method the real signal: he didn't share the code. He shared an "idea file" — because in the agent era, you share the idea and let each person's AI build it for them.
THE TAKE
This isn't a productivity tip. It's a shift in what LLMs are actually for. We've spent two years treating AI as an answer machine. Karpathy is using it as knowledge infrastructure — something that compounds over time instead of resetting every conversation. The people who figure this out early are going to have a significant advantage over everyone still starting every chat from zero.
THE NUMBER
400,000 — the number of words in Karpathy's personal AI-maintained knowledge base on a single research topic. That's longer than most PhD theses. Built without writing a single word manually.
Want the full breakdown of how the system works — and what it means for how you use AI? Read the full article on Analytics Drift.