Overview
- Google published the Open Knowledge Format as a v0.1 standard that defines a simple folder-and-file convention so agents can read machine‑readable knowledge without custom ingestion code.
- Practitioners have built working OKF systems that use YAML frontmatter, an index.md file, and typed markdown files for concepts, entities, playbooks, references, and systems so agents know what each file contains.
- Marie Haynes published a demonstrator 'brain' that automatically ingests and links new content into a knowledge graph, checks Google documentation daily, updates reference files when docs change, and supports query-driven synthesis and corrections.
- Her implementation runs procedural playbooks through an agent called Antigravity to draft client proposals and analyze site impacts, cutting tasks that took days down to hours and preserving the author's voice and logic.
- The broader context shows markdown-based bundles from Karpathy’s LLM Wiki to gstack point to a shift where teams may gain a durable advantage from owning structured markdown knowledge, though OKF remains early stage and needs more consumer tools to reach wide adoption.