Where Did You Put It?

Machines name files so they never collide. Humans name them so we can find them. Folders, file-naming conventions, Johnny Decimal, and tags all chase one thing, meaning, and each one cracks in its own place. Part two of a series on naming, addressing, and actually finding things.

Machines, we said last time, won identity and integrity and threw meaning over the wall. Humans do the opposite. We barely think about uniqueness at all. We name things so that a person can look at the name, or the place it sits, and know what it is and where to find it. Meaning is the whole game.

The catch is that meaning, organized by hand, has failure modes every bit as reliable as a colliding counter. They are quieter, because they feel like your own fault rather than the system’s. They are not your fault. They are baked into each approach. A handful of them dominate everyday life: folders, the names we give the files themselves, Johnny Decimal, and tags. Each one chases meaning, and each one cracks in a predictable place.

Folders: meaning by location

The folder tree is how almost everyone organizes files, and it works by location. A thing’s meaning is its path: Clients / Acme / 2026 / Contracts. Read the path, know the thing. It is intuitive, it is hierarchical, and it is universal. Nobody needs a manual to use a folder.

It also carries the deepest crack of any scheme here, and it is the one from the last post wearing a friendly face. A folder path is a location, and locations are fragile.

Start with the everyday version: where does the Acme renewal contract go? Under Clients/Acme/Contracts? Finance/2026? Legal/Active? Taxes/2026? Every one of those is defensible, which is exactly the problem. The document has several true facets, and a folder forces you to pick one home for it. So you either duplicate it, and now there are two copies and one of them will go stale, or you pick a single home and accept that anyone who looks where you did not put it will simply not find it. Multiply that across a team and the same kind of document ends up scattered across four different logics, one per person, none of them wrong and none of them shared.

Folders also have no rule. Nothing stops anyone from nesting eight levels deep, or creating a Misc folder, or a New Folder (2), or the immortal Final_v2_FINAL_actually-final.docx. The structure is whatever each person felt like on the day they made it, and it drifts the moment two people share a drive. At small scale this is fine, even pleasant. At organizational scale it becomes the thing everyone complains about and no one fixes: nobody can find anything, so the search box becomes the only navigation that works. And reaching for search because you cannot navigate is not a feature. It is the system admitting it failed at its one job.

And folders have no concept of the version. Copy a file into a second folder because it honestly belonged in both, and you now have two files that will drift apart, with nothing in the system aware that one is becoming stale. The tree cannot tell you that two paths point at the same logical thing, or which of them is authoritative. It knows locations, and a location has no opinion about truth.

Folders nail meaning for one person on a good day. They have no answer for scale, none for shared discipline, and none for the plain fact that a thing can honestly belong in more than one place.

Naming conventions: putting the meaning in the name

If a folder puts the meaning in the location, the other obvious move is to put the meaning in the name itself. This is the world of naming conventions: a team agrees that files will be called something like 2026-06-02-acme-renewal-contract, a date in front for sorting, then a short readable slug that says what the thing is. Plenty of organizations live and die by a scheme like this, and the disciplined ones treat it as seriously as they treat code style.

It is a real improvement on the bare folder in one specific way: the name travels with the file. Move 2026-06-02-acme-renewal-contract anywhere and it still announces what it is and roughly when it happened. The meaning is no longer trapped in a path that shatters when you reorganize. It rides along inside the filename, self-describing, sortable, and readable by a person at a glance. This is the closest any everyday human scheme comes to a name that carries its own meaning.

But it inherits the human-scheme weaknesses wholesale. The name is not unique: two people will both produce acme-contract and the system will not blink. It is mutable: rename it for tidiness and every link and reference to the old name breaks, the exact fragility a URL has. And it lives or dies on discipline. A convention is only as good as the least careful person following it, and the day someone drops acme final.docx next to 2026-06-02-acme-renewal-contract, the scheme has a hole in it. Conventions hand you a self-describing name, which is no small thing, and then leave uniqueness, durability, and enforcement entirely up to you.

Johnny Decimal: a discipline on top of the chaos

The smartest popular response to folder chaos is Johnny Decimal. It is not a new technology; it is a convention, and it descends directly from the Dewey Decimal system your library has used for more than a century. The idea is to impose a small, fixed, numbered map on top of ordinary folders.

The library lineage is worth dwelling on, because a library is the original solution to addressing a huge pile of things a human has to find by hand. A call number puts every book at exactly one spot on exactly one shelf, and the numbering is built so that related books sit next to each other. It scales to millions of volumes. But look at what a library actually needs to function: trained staff who learned the classification, and a separate catalog that translates a subject you care about into the number where it lives. The address gets you to the shelf; the catalog tells you what is on it. Johnny Decimal is that idea shrunk to fit a person instead of an institution, and it carries the same buried requirement, the address is only useful standing next to the legend, and the legend is a thing a human has to keep.

You divide everything into at most ten areas (10 to 19, 20 to 29, and so on), split each area into at most ten categories, and number every item inside them. A thing lives at an address like 12.04. The address is short, speakable, and stable. You can say “it is in 12.04” out loud and a colleague can go straight to it, which is something you cannot do with a folder path or a hash.

Johnny Decimal fixes the two worst folder problems on purpose, and it is worth being generous about how well it does it. The ten-by-ten limit is not a weakness, it is the entire point: the map is small enough to hold in your head, and human working memory tops out at a handful of things, so a map you can memorize is a map you will actually use. The governing rule is the old one, a place for everything and everything in its place, enforced as one home per thing, which kills the many-homes ambiguity that wrecks plain folders. And the number is a durable handle: rename the human label on 12.04 and the address still points at the same thing, so you recover a little of the stability machines enjoy. For a person, or a small team willing to learn the system, Johnny Decimal is genuinely excellent. This is not a takedown.

But the ceiling is real, and it is arithmetic. Ten areas times ten categories is one hundred categories, and that is all the room the scheme offers. For a tidy personal setup or a small business, a hundred is plenty. For a large or fast-growing organization, or a domain that is naturally deep rather than wide, you run out of room, and the shallow two-level depth that makes the map memorable is the very thing that makes it refuse to fit a complex corpus. The scheme also runs entirely on human discipline. Someone has to design the map up front. Everyone has to file correctly, forever. And the meaning of 12.04 still lives in a legend you maintain by hand, so if the legend drifts or gets lost, the addresses stop meaning anything. One more limit, easy to miss: your 12.04 has nothing to do with my 12.04. The scheme is local, never global. Johnny Decimal trades folder chaos for a map you can memorize, and pays for it with a hard scale ceiling and a permanent dependence on a human keeping the legend honest.

Notice, too, what a newcomer sees when they land on 12.04 cold: a number they cannot interpret without the legend. Johnny Decimal made the system legible to the person who learned it. It did not make the thing self-describing. Hold that thought.

Tags: the rebellion against one home

If the deepest problem with folders and Johnny Decimal is that both force one home per thing, the obvious rebellion is to stop forcing it. That is tagging. Do not place the document, label it. The Acme renewal can be acme, contract, 2026, legal, and renewal all at once, and you find it later by any of those facets. Gmail did this to email and called them labels. The PARA method (Projects, Areas, Resources, Archives) is a lighter, four-bucket version. Zettelkasten, and tools like Obsidian and Roam, push it all the way, linking atomic notes into a web with almost no hierarchy at all.

Tagging works because it matches how people actually think. A thing is rarely about one subject. It sits at the intersection of several, and tags let it live at that intersection instead of being filed under whichever single facet you happened to prioritize that day. There is no filing paralysis, because you never have to decide the one true home; you describe the thing and move on. For retrieval by question, “show me everything tagged contract and 2026,” it is powerful, and it is closer to how a search-first generation expects to work.

The most ambitious version of this deserves a moment. The Zettelkasten, the slip-box method the sociologist Niklas Luhmann used to write an improbable amount, drops hierarchy almost entirely. Every note gets an address and links to other notes, and the structure is meant to emerge from the web of links rather than being imposed up front. Obsidian and Roam are its modern descendants, and for a single thinker building a body of ideas it can be genuinely magical. It is also intensely personal: the structure that emerges lives in one person’s associations, and it rarely survives contact with a team, or with someone trying to find one specific thing on a deadline. It optimizes for serendipity, not retrieval, which is a fine trade for a writer and a poor one for an organization.

The cost is vocabulary entropy, and it arrives on a schedule. Tags multiply, and almost no one governs them. Is it empathy, empathy-training, Empathy, or EmpathyTraining? Over a year, with a few people tagging independently, you get all four, and now a single concept is fractured across four labels and reliably findable under none of them. Tags also dissolve the idea of a canonical home. When everything is labeled and nothing is placed, “where is the real one?” stops having an answer, and you fall back entirely on search to reassemble what the structure no longer holds. The failure is not hypothetical: it is what an ungoverned tag system looks like in the wild, a blog with dozens of tag pages, most of them thin near-duplicates of one another, none of them a real destination. Tagging trades the rigidity of one-home-per-thing for the slow entropy of a vocabulary nobody owns.

Why do folders fail when your team grows?

What humans decided

Folders, naming conventions, Johnny Decimal, and tags are different bargains, but in the end they strike the same trade. Humans optimized for meaning, the one job the machine schemes threw away, and they won it. Every one of these names is something a person can read, reason about, and walk through. That is real, and it is precisely what UUIDs and hashes could never give you.

And every one of them paid for that meaning with the other two jobs. None provides identity in the machine sense: folders collide and scatter, Johnny Decimal addresses are local rather than global, tags fracture into near-duplicates. None provides integrity at all; nothing here can prove a file is what it claims to be. And each one assumes a disciplined human in the loop, someone to design the map, file consistently, maintain the legend, govern the vocabulary. The meaning is genuine, but it is fragile, and it lives in a person’s head or a hand-kept index, not in the artifact itself.

So human naming is the exact mirror image of machine naming. Machines gave us names that are unique and verifiable and unreadable. Humans gave us names that are readable and navigable and neither unique nor durable nor, most tellingly, self-describing to anything that had not already learned the system. Point a newcomer at a beautifully kept Johnny Decimal corpus and, without the legend, it is just numbered boxes. Point an AI at it and you get the same blank stare, for the same reason.

Faced with all of this fragility, the natural human instinct is to stop doing it by hand. If folders sprawl and tags drift and even a good numbered map needs a person to maintain it, then surely the answer is to buy a system that manages the organizing for you, that enforces a structure, tracks every version, controls who sees what, and lets you search across the whole thing at once. That is exactly what document management systems promise. It is also where a different, and considerably more expensive, set of problems begins.

That is the next post.


The prior post in this series: What’s in a Filename?, on how machines name.

Aaron Lamb Co-Founder, Hexaxia Technologies