Stable, long-lived information about a project that remains relevant across multiple LLM sessions. Important because static context is often large and needs indexing. It includes all your relevant existing code, the vast majority of which never changes, and can also include coding standards, architecture documents, long-lived administrative prompts, API documentation, and large bodies of data such as issue trackers, databases, and logs. Often retrieved via RAG.
— Vibe Coding, Gene Kim & Steve Yegge