Summary: A small, not-quite-official Google spec, OKF, for organizing knowledge as markdown files in a standard structure: an .okf/ folder whose files share a fixed layout with YAML frontmatter, can reference tables, and are token-efficient for agents. Not a replacement for the wiki model; its value is interoperability, so other people’s agents, other tools, and future models can all read the same knowledge.
Key ideas
- A shared format matters little solo but a lot for teams of 10 or 1,000 where everyone stores knowledge differently (osim-tochna—second-brain-and-llm-wiki, dana-maman).
- Model portability: today Claude, tomorrow Codex, so knowledge shouldn’t be locked to one tool (osim-tochna—second-brain-and-llm-wiki).
- Skill packs already build on the format and give features like the 3D graph view for free (osim-tochna—second-brain-and-llm-wiki, amit-bendor).
- Most agent-readable knowledge today is markdown, web pages, PDFs, and books; OKF standardizes the markdown part so it’s “ready context” for any agent (ai-engineering-podcast—company-brain, roi-zalta).