Skip to content

Research note

The forgotten internet

A write-only archive is one where the rate of addition vastly exceeds the rate of retrieval, and where nothing in the design corrects the imbalance. Most personal saved-item systems are write-only within a year of being started.

The companion essay in Ideas takes this personally. This one takes it structurally.

01 · Conditions

Three properties make an archive write-only.

First, entry requires initiative: nothing surfaces on its own, so every retrieval starts with the person remembering that the archive contains something relevant. Second, retrieval is keyword-dependent, so the attempt fails when the wording is gone. Third, ordering is by save time, which is uncorrelated with relevance.

Any one of these is survivable. All three together guarantee that the archive is visited less as it becomes larger, which is the exact opposite of what an archive should do.

Entry requires initiative
Nothing comes back to you unprompted.
Keyword-dependent retrieval
Fails precisely when the words are gone.
Save-time ordering
Sorted by the least relevant attribute.

02 · Compounding

The failure accelerates.

Each failed retrieval teaches you, correctly, that searching your own archive is unreliable. So you search the web instead, and the archive's visit rate falls further while its size keeps climbing.

By the time it holds thousands of items, the archive is a sunk cost people feel vaguely guilty about rather than a resource they use.

Archives don't die from neglect. They're designed into it.

03 · Reversal

Two properties reverse it.

The first is retrieval that accepts non-lexical clues: time, appearance, source, project, meaning. This restores the possibility of finding something with the clues you actually have.

The second is initiative on the archive's side: relevant older material appearing while you work on something related. Overdone, that's a feed, and a feed is a different and worse product. Done sparingly it converts a graveyard into a memory.

Next thought

Both reversals depend on the machine understanding what you kept.

Read: the rise of personal AI memory

Shelf V3 is out now

Shelf V3 is the current build. Shelf does not use automatic updates, so download the latest build here and install it over your current version.

This is intentional. Shelf is built around a local-first, privacy-focused philosophy. An automatic updater requires additional background processes, network connections, update infrastructure, and another system with permission to modify the app. We prefer to keep Shelf’s network footprint and background activity as small as possible.

Manual updates keep the process explicit: Shelf only checks the internet when a feature actually needs it, and you decide when a new version is downloaded and installed.

Download the latest version →Version 3.0 · macOS 14+