Your documentation is still in your Mum's filing cabinet
The article argues that rigid folder hierarchies no longer match how people or AI find information. It shows users forage for clues and AI retrieves by meaning, not location, and recommends multiple discovery paths—search, metadata, links and tags—to build a connected knowledge graph.
- Folder hierarchies force single "right" locations for info that actually spans multiple teams/contexts (e.g., a component accessibility decision touching design, engineering, content, support)
- AI retrieval tools already surface content by meaning/keywords rather than folder path, proving folders organize storage, not knowledge
- People actually find info by foraging—searching, skimming a couple nav levels, then asking a colleague—not by drilling through directory trees
- The fix is layering multiple discovery paths (search, tags, metadata, cross-links) to build a connected knowledge graph, mirroring tools like Obsidian's graph view and accessibility's "never rely on one cue" principle
Most teams still treat docs like old-school filing cabinets—nested folders, strict hierarchies, single “right” location. But knowledge doesn’t live that way. A single decision about component accessibility, for example, touches design, engineering, content, support. Dropping that info into one folder means everyone else either misses it or duplicates it. Researchers since the ’90s—see work on Semantic File Systems—have argued we should retrieve info by meaning and attributes, not by drilling down through directories.
AI retrieval tools already ignore folder paths. They surface a design token page because it mentions “colour contrast,” not because it sits under Foundations → Accessibility → Colour. That shift makes clear that folders serve storage, not knowledge. In practice, people search first, skim a couple levels of navigation, then give up and ask a colleague. We need multiple discovery routes—search, tagging, metadata, cross-links, semantic relationships—that reflect real human foraging patterns described by Pirolli and Card.
Obsidian’s graph view shows how ideas overlap by linking and tagging rather than forcing notes into one tree. Accessibility guidelines push the same principle: never rely on just color or shape, always layer cues. Good docs follow that model—clear headings, consistent language, rich metadata and inter-page links. Those same traits help AI retrieve relevant snippets regardless of where they’re stored.
The aim isn’t a deeper folder tree. It’s a connected web of knowledge you can reach from any entry point—browser search, a related-topic link, a tag cloud or an AI prompt. When documentation works that way, humans and machines both spend less time hunting and more time using what they find.