A scroll-driven essay · illustrative data
We got very good at producing them and never got any better at keeping them findable. This page is about the gap between those two facts.
Scroll. The figure on the right responds.
01 — Volume
Not the ones with owners and version numbers. The other ones. A plan. A summary. A comparison table. A chart someone asked for in passing on a Tuesday.
Each dot is one of them. A month of ordinary work for one person.
02 — Reads
Sort the same dots by how many times they were ever opened after the moment they were made. The distribution is not a bell curve. It's a cliff.
Most of what you produce is read once, by the person who asked for it, on the day it appeared.
03 — Decay
The second read doesn't fail because the work was bad. It fails because by the time somebody wants it again, nobody can say where it is.
The document didn't die. It became unaddressable, which turns out to be the same thing.
04 — Location
Almost nowhere with a front door. They live in scrollback, in a downloads folder, as an attachment inside a one-to-one message, on a laptop that belongs to someone who has since changed teams.
None of those are storage. They're sediment.
05 — The shift
Ten years ago the expensive part was making the thing. That constraint did the filing for us: if a document cost two days, it got a name, a folder and an owner.
The making is now the cheap part. Volume went up by an order of magnitude and the shelf stayed exactly the same size.
06 — The shape of a fix
Nothing exotic. Render the thing on arrival instead of downloading it. Give it a stable address. Know who is allowed to open it. Then stay put.
Four properties. Every one of them existed in 1996.
Every part of the pipeline that used to be slow has been optimized, except the last one. We can generate a well-structured, genuinely useful document in the time it takes to describe it — and then we hand it to someone as an attachment, which is the 1997 solution, or paste it into a chat window, which is worse.
The result is a strange asymmetry. The marginal cost of producing an artifact has collapsed toward zero while the marginal cost of finding one again six weeks later has held completely steady. Whenever those two curves diverge, you get accumulation without retrieval. Which is a fancy way of describing a landfill.
A document nobody can locate is indistinguishable from a document nobody wrote.
The uncomfortable part is that this is not a hard technical problem. It's a defaults problem. The places our files land were designed around a filing-cabinet metaphor at a moment when the most valuable thing in the cabinet stopped being a file and became a small piece of software — something that renders, responds, and needs an address rather than a filename.
If opening it requires an application, a plugin, or a "preview not available", it's already lost.
One URL that keeps working. Not a path inside a tree only you can navigate.
Access as a property of the artifact, not of the folder it happens to be sitting in.
Survive the sender changing laptops, teams, or employers. Most things in this list do not.