karpathy

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.

MWins

Project Ideas

Ok. I'm going to list off some ideas for projects. You will have to determine if any particular idea is good enough to include in a portfolio. These aren't creative ideas. They likely already exist. Some are way too advanced while others are simplistic.

I will recommend to post any project you make to github and make a github project page for it. Explain in as much detail as possible how you made it, how it can be improved etc. Document it.

If you pick an advanced idea, setup a development roadmap and follow it. This will show some project management skills.

Another piece of advice for those who are design challenged. Use different front end frameworks and use different themes for those frameworks to provide appealing designs without looking like yet another bootstrap site.