Most literature workflows do not compound.
You collect PDFs in folders. You annotate some of them. You ask an LLM questions over the corpus. It gives useful answers, but every answer is rebuilt from scratch. The system does not become more organized after you use it. Your understanding improves, but your knowledge base does not.
I wanted something different.
I wanted a literature workflow that behaves more like a compiled system:
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.
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.