Google Cloud recently introduced the Open Knowledge Format (OKF), and the surprising part is how simple it is: just Markdown files in a folder, with YAML frontmatter carrying type, title, description, tags, and timestamps. The files connect through standard Markdown links, creating an implicit graph. Google even recommends an index.md at each level for navigation.

If you use Obsidian, this feels familiar — it mirrors vault architecture with frontmatter, backlinks, and folder-based organization. Google openly acknowledges the inspiration: Obsidian, Notion, Hugo, and the LLM-wiki pattern that traces back to Andrej Karpathy’s idea of AI-readable Markdown documentation.

The real insight isn’t that Google invented something new. It’s the validation that the thing a lot of us have been doing anyway is the right foundation. Plain text lasts. It avoids vendor lock-in and the migration pain that comes with proprietary formats.

Forte’s framework, in parallel

Around the same time, I started reading Tiago Forte’s Building a Second Brain. It presents the CODE methodology:

• Capture content that resonates
• Organize by actionability using PARA (Projects, Areas, Resources, Archives)
• Distill notes down to their essentials
• Express knowledge through creation

It embodies David Allen’s principle that the mind is for having ideas, not storing them.

These developments feel complementary. Forte’s framework optimizes for human cognition; Google’s specification supports agent traversal. Both use Markdown as the substrate.

Where it gets hard

I’ve experimented with wiring AI agents into my personal Obsidian vault. It works reasonably well for text queries, but it doesn’t scale cleanly. PDFs are cumbersome. Images aren’t inherently searchable. And agents navigate very differently from how a human explores by intuition.

Larger vaults need explicit structure — master indices that clearly state what exists and where to search. The emerging challenge is building infrastructure where Markdown-based knowledge systems stay reliable for autonomous agents as they grow: indexing speed, non-text representation, and explicit navigation paths replacing the contextual familiarity a human has.

The synthesis

Three perspectives converge here. Obsidian established the foundational architecture. Forte articulated the human discipline required. Google provided an agent-focused specification. The unexplored territory is bridging them — turning a collection of Markdown files into a system that agents can reason through dependably.