Quick answer
The short version
Write down what you remember before you search anything. Then search from your strongest clue to your weakest: exact words, then subject, then time, then what it looked like, then what surrounded it.
01 · Framework
Ask yourself six questions.
Answer these out loud or on paper. Two or three answers are usually enough to find it, and the act of answering often produces the clue you'd forgotten you had.
- 01. What was it?
- Screenshot, article, PDF, image, note, website, message.
- 02. What did it look like?
- Colour, layout, visual style, an object, an interface.
- 03. What did it say?
- A phrase, a name, a headline, an error, a number.
- 04. When did you see it?
- Today, last week, during a project, around an event.
- 05. What were you working on?
- Research, client work, a product decision, a trip.
- 06. What was around it?
- Related screenshots, articles, people, projects.
What you remember
“There was a diagram explaining how something remembers. I think I saved it during the research week.”
What you get back
Two candidates from the same afternoon. One of them is it.
02 · Section
Search from strongest clue to weakest.
If you remember exact text, start there: it's the only clue that can produce a single result. If you don't, use the topic, which narrows to a set. Then time, which narrows the set. Then visual information, which is how you'll recognise the right one when you see it.
Surrounding context comes last, and it's the one that rescues genuinely hopeless cases: find anything from the same week and read outward.
03 · Section
Don't confuse location with memory.
“Which app did I put this in?” is a harder question than “What do I remember about the thing?”, and it's the wrong one to lead with, because your memory never stored the app.
It stored what the thing was about. Any search that starts from location asks you for information you never recorded.
The place is the one detail you were never paying attention to.
04 · Where Shelf fits
Shelf is built around that second question.
Shelf keeps what you save in one place on your Mac, reads it, describes it, dates it and connects it, so a query can use content, meaning, time and context at once instead of one at a time.
It won't make your memory precise. It's designed to work with the imprecise version.
Your memory is messy.
Your search system should expect that.