Folders, in the digital sense, began life as an ingenious skeuomorphic metaphor. A digital stand-in for physical folders and filing cabinets pioneered by Xerox and first seen in their Alto machine. The metaphor helped early computer users map the new digital world onto something they already understood.

After reading The Science of Managing our Digital Stuff it became clear that, despite advances in organizational tools, there’s a deeper reason the folder has persisted.

Using fMRI scans it was seen that a spatial retrieval task uses a ~150 million year old region of the brain whereas search uses a newer, ~3 million year old region. The research suggests that generating search terms is cognitively demanding because it requires linguistic resources in working memory. In contrast, spatial navigation is semi-conscious and allows the user to keep their working memory on the main task which eases switching costs incurred by search actions.

Humans remember where things are in relation to other things.

That’s why people still recall where a file lives, not just what it’s called or some of its metadata. Folders leverage our ancient navigation system.

Nonetheless researchers, designers and developers kept making progress in more efficient retrieval patterns. Real-time search, fuzzy matching, and tagging made it possible for a file to “live” in many places at once, or even be left unorganized but retrieved on demand.

In theory, hierarchy should have been dead.

But despite these advances, most people kept their folders and in multiple HCI studies, even when participants were trained how to search effectively 80%+ still defaulted to a folder structure when asked to recall a file.

Creative design patterns continually emerged (e.g. Cmd + K-style search bars) but the majority of users stayed grounded in ‘inefficient’ spatial organization.

In the AI-native era, we were promised the true end of folders. With semantic retrieval and intelligent organization, the assistant would first organize and then retrieve anything we desired without any pre-planned or rigidly imposed structure.

Yet as we shift from passive assistants to agentic workflows, folders might be given further life-extended properties with a new purpose. They’re no longer simply organizational buckets but scoped contexts.

A folder now defines what an agent can see, read, or write. It sets the scope for reasoning and the boundary for permissions.

Folders are dead, long live folders