hraness

knowledge systems

markdown vaults, retrieval, and evidence beside the code.

the repo is the database. every project here keeps its working knowledge in markdown and git — searchable, diffable, and durable — with indexes treated as replaceable caches rather than the source of truth.

this category is about building knowledge systems agents can actually use: hybrid retrieval, typed relationships, source capture with provenance, and review workflows that keep generated notes honest.

projects

planned lessons

  1. markdown as the write path; indexes as caches
  2. hybrid retrieval: exact, text, and semantic search together
  3. capturing sources with provenance you can audit
  4. typed relationships and bounded graph context for agents

these lessons are planned, not published. each ships as a page under this category when it is written.

AI-drafted at Ben Guo's direct request as category seeds; each planned lesson is unpublished until it is written.