Research note
The screenshot problem
Screenshots are now a primary medium of personal record-keeping. They're also the least legible thing a computer stores: an array of pixels with an autogenerated name and a timestamp.
The gap between how easily they're made and how poorly they're indexed is the widest in personal computing.
01 · Asymmetry
One keystroke to make, no metadata at all.
Every other saved artifact arrives with something: a document has a title and an author, a link has a URL and a page title, an email has a sender and a subject. A screenshot arrives with a date and nothing else.
Its entire meaning is inside the image, in a form that plain file search cannot reach.
- No title
- Screenshot 2026-03-04 at 11.42.png describes nothing.
- No source
- The app or site it came from is usually not recorded.
- No text
- Everything readable is pixels unless something recognises it.
- High volume
- They accumulate faster than any other save type.
02 · Loss
The volume makes manual description impossible.
The obvious fix, name and file each one, cannot work at the rate screenshots are produced. Any solution that scales must be automatic, and it must run at the point of capture rather than in a cleanup session that never happens.
This is why screenshots are the clearest test case for a memory layer. If a system can make screenshots findable, the easier formats follow.
Solve screenshots and you've solved the general case.
03 · Fix
Read the pixels, keep the moment, connect the neighbours.
Text recognition converts the image's content into something searchable. A short description of what the image appears to show gives it a subject. The arrival time gives it a place in the timeline, which is how people date their own memories. Adjacency to other saves reconstructs the episode it belonged to.
None of that requires the person to do anything, which is the only property that matters here.
Next thought
Recognising the text is only half of it. Retrieval has to accept vagueness too.
Read: search is not memoryIf this idea interested you…
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.