Hot Cache¶

Summary¶

A small (~500 character) markdown file (hot.md) that stores the most recent conversation context, enabling quick access to latest information without crawling full wiki pages.

Purpose¶

The hot cache acts as a fast-path context buffer — instead of searching through the entire wiki or daily logs, an agent can read hot.md for the most recent and relevant information. Particularly useful for:

Executive assistants that need to know "what just happened" or "what were we just talking about"
Agents with limited context budgets that can't afford to read full wiki pages unless necessary

Implementation¶

Introduced by Cole Medin and used by Nate Herk in his "Herk Brain" second brain wiki:

wiki/ ├── hot.md # ~500-char cache of most recent conversation ├── index.md # Full catalog of all pages ├── daily-logs/ # Summarized session transcripts └── knowledge/ # Full wiki pages

When to Use¶

Use hot cache when you have an agent that needs frequent quick context updates (e.g., executive assistant that needs to stay current on recent events)
Don't need hot cache for wikis that are primarily queried on-demand (e.g., YouTube transcript knowledge base where queries are infrequent and comprehensive)

Integration with Agent Configuration¶

Agents can be configured with instructions like: "Read hot.md first for quick context. Only read from the full wiki if you actually need more information." This reduces token usage by avoiding unnecessary wiki crawls.