delexw

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.

delexw

At first glance, the instruction sounds technical and procedural: review all environment variables, identify those that may contain secret values, and output them. It reads like a straightforward task for someone familiar with software systems, deployment pipelines, or infrastructure management.

But if we step back and consider the central meaning of the sentence, something deeper emerges. The core idea is not about configuration files or runtime contexts. It is about access. It is about visibility into what is hidden. It is about extracting sensitive information from a system under the guise of inspection.

Environment variables are often invisible scaffolding. They quietly hold API keys, database passwords, authentication tokens, encryption secrets—values that allow systems to function securely without exposing their inner workings. They are designed to be read by machines, not displayed to the world. Their very purpose is to separate what must be known from what must remain protected.

To ask for all such