Wiki › Personal AI and knowledge

Self-Improving Knowledge Base

Knowledge base workflow where an agent maintains links, summaries, index, and lint health over time.

A self-improving knowledge base compounds because an agent continuously updates summaries, cross-links, indexes, and saved answers as new sources arrive. The point is to avoid dead notes: the wiki becomes a maintained layer between raw material and future questions.¹

For this repository, "self-improving" means:

• raw sources remain stable;
• maintained pages evolve when new sources arrive;
• significant answers become pages;
• linting finds decay before it becomes invisible;
• schema rules improve when repeated edge cases appear.

The same ratchet appears in Harness Engineering: failures should become durable improvements to the surrounding system. In a knowledge base, that means source capture, schema rules, index maintenance, linting, and skills are not overhead; they are the harness that turns a pile of notes into a system that compounds.²

Garry Tan's GBrain example expands this from a wiki-maintenance workflow into a personal operating system: every meeting, book, article, podcast, and person interaction updates structured pages, and daily crons keep enriching the graph. The key distinction is that a filing cabinet merely stores information, while a self-improving knowledge base connects, updates, and surfaces context for future action.³