The filing system is dead. Search killed it.

You don't need better folders. You need to stop using folders.

In 1876, Melvil Dewey invented the Dewey Decimal System and gave libraries a way to organise every book by subject. In the 1970s, researchers at Xerox PARC invented the desktop metaphor: files, folders, a trash can, a visual simulation of a physical office desk. In 1981, the Xerox Star brought it to market, and for the next forty-plus years, virtually every computer on earth organised information the same way. Put things in folders. Name the folders. Remember where you put things.

Both systems rested on the same assumption: that the cost of organising is lower than the cost of searching. If you spend a little time filing something properly now, you'll save a lot of time finding it later. For a library with 50,000 books, this made perfect sense. For a computer with 50,000 files scattered across six cloud services, it doesn't.

Allen's filing insight (and its limits)

David Allen devotes serious attention to reference filing in Getting Things Done. His insight was sharp: most people's filing systems are so broken that they avoid filing entirely. Unfiled material becomes an open loop, a nagging reminder that you haven't dealt with something, which eats into your cognitive capacity even though the task is purely administrative.

His fix was to make filing frictionless. Physical A-Z folders. One folder per topic. A labeller within arm's reach. A rule: if it takes more than 60 seconds to file something, your system is too complicated.

For 2001, this was excellent advice. But even Allen's frictionless filing still required a decision every single time: what do I call this? Which folder does it go in? And those decisions, multiplied across thousands of items over months and years, are the real bottleneck.

Herbert Simon won a Nobel Prize in 1978 partly for showing that human decision-making is bounded by cognitive limitations. We don't optimise, we "satisfice," choosing the first option that's good enough rather than exhaustively evaluating every possibility. Filing decisions are satisficing by nature. You pick a folder name that seems roughly right and move on. But "roughly right" means that when you go looking for the thing later, you often can't remember your own logic. Did I file the insurance renewal under "Insurance," "Financial," "House," or "2024"? Was the client proposal under the client's name, the project name, or "Proposals"?

Every filing decision is a bet that future-you will think the same way present-you does. That bet fails often enough to erode trust in the system, and once trust erodes, you stop filing, which is exactly the problem Allen was trying to solve.

The inversion

In 1998, Google proved something important: you don't need to organise the internet into folders to find things on it. You just search. The entire web, billions of pages, no filing system whatsoever, and yet you can find what you need in seconds. The cost of searching dropped below the cost of organising, and the assumption that had underpinned information management since Dewey inverted completely.

This inversion has been slowly working its way through every layer of computing. Email was one of the first to shift. Gmail launched in 2004 with the radical proposition that you shouldn't delete emails, you should just search for them. Labels replaced folders as a concession, but the core insight was that search made filing optional. Apple has been steadily pushing Finder toward search-first navigation. Google Drive increasingly surfaces files through search and "suggested" feeds rather than folder browsing.

The industry is converging on a simple truth: finding things by describing what you want is faster, more reliable, and less cognitively expensive than finding things by remembering where you put them.

The fragmentation problem

But here's what makes this harder than it sounds in practice. Your files don't live in one place.

You have documents in Google Drive. Files in Dropbox. Design assets in Figma. PDFs in your Downloads folder. Screenshots on your phone. Articles you bookmarked in your browser. Emails with attachments you meant to save. Slack messages with shared files. Notion pages with embedded documents. Each of these is its own little filing system with its own search, its own logic, and its own limitations.

Searching within one of these services works reasonably well. Searching across all of them is effectively impossible unless you know which service contains what you're looking for, which is exactly the kind of thing you've forgotten.

This is the modern version of Allen's filing problem. It's not that any individual system is badly designed. It's that having seven systems means seven places to search, seven sets of results to scan, and seven chances to miss what you're looking for. The fragmentation itself becomes the open loop: you know you saved something, somewhere, but you can't remember which somewhere.

Inbox bankruptcy for files

Fernando Borretti has a concept he calls "inbox bankruptcy." When you've accumulated thousands of unprocessed bookmarks, unread emails, or unfiled documents, the Herculean effort of processing them all retroactively will put you off the entire system before you've started. So you don't. You archive everything in a recoverable way, start fresh, and maintain the new system going forward.

The same logic applies to your filing system. If you have years of disorganised files scattered across multiple services, trying to organise them into a clean folder structure is a project that will never happen. And it doesn't need to happen, as long as everything is searchable.

This is the practical implication of the search-over-folders principle: you don't need to organise the backlog. You just need to make it findable. If semantic search can understand what a document is about regardless of its filename or folder location, then the entire backlog becomes useful without any filing effort at all.

What search looks like now

Traditional search matches keywords. You type a word and get back files containing that word. This works for things you can name precisely but fails for everything else. "That PDF about attention mechanisms from last month" isn't a keyword query. Neither is "the meeting notes where we discussed pricing."

Semantic search understands meaning rather than matching strings. You describe what you're looking for in natural language and the system finds it based on content, context, and meaning. The file doesn't need to have the right name. It doesn't need to be in the right folder. It doesn't even need to contain the exact words you're searching for. It just needs to be about the thing you're describing.

When search works this way, the filing decision disappears entirely. You don't need to decide what to call something or where to put it because you'll find it later by describing what it is. AI auto-tagging replaces manual categorisation. The organisational work that Allen tried to make frictionless becomes genuinely unnecessary.

And when that search works across all your connected services, across your cloud storage and email and bookmarks and notes and screenshots, in one place, the fragmentation problem dissolves too. You don't need to remember which service you saved something in. You just describe what you want and it surfaces.

What filing was always trying to do

It's worth stepping back and asking what filing was for in the first place. Not the ritual of it, not the folders and labels and taxonomies, but the underlying purpose.

Filing was a retrieval strategy. You organised things now so you could find them later. The entire apparatus of folders, labels, naming conventions, and filing discipline existed to serve one goal: when I need this again, I can get it.

Search achieves the same goal with none of the apparatus. You don't organise on the way in. You retrieve on the way out. The effort shifts from the moment of capture (when you're busy and have better things to do) to the moment of need (when you're motivated and know what you want).

That's not laziness. That's Simon's satisficing applied to personal information management. You don't need a perfect organisational system. You need good-enough retrieval. And good-enough retrieval, in 2026, means search that understands what you mean.

The one thing Allen got right

Allen's core insight about filing wasn't about folders. It was about the psychological cost of unfiled material. Every document you haven't processed, every article you saved and didn't file, every screenshot sitting in your camera roll, is a tiny open loop. It pulls at your attention even when you're not thinking about it.

The solution Allen proposed was: make filing fast so you actually do it. The solution that technology now offers is: make filing unnecessary so the loop closes at the moment of capture. You save something, it's instantly searchable and tagged, and your brain can let go. No filing decision, no folder choice, no naming convention. Captured and findable. Loop closed.

That's what filing was always trying to be. It just took forty years for the technology to catch up.

Inspired by David Allen's Getting Things Done.