ReMyndhow to find something you saw on your screen
you saw it. you are sure of it. a number, a name, a link, a line in someone's slide. you closed the tab and moved on, and now you need it back and it is nowhere.
the single biggest thing people use ai for is searching for information. per the anthropic economic index, "search electronic sources, databases, or repositories for information" is the number one work task in ai conversations globally, at around 5% of all usage. but every one of those tools searches the open web or a corpus someone else assembled. none of them can search what you already looked at.
here is every real method, roughly in order of how much effort it costs you, and where each one runs out. current as of september 2026.
why no ai tool can just answer this
the numbers make the gap plain, and they come from the anthropic economic index for may 2026.
searching electronic sources is 4.95% of sampled work conversations and reference searching another 3.74%, which is 8.69% of everything. meanwhile knowledge retrieval and enterprise search sits at 3.61% of request topics, personal ai assistant at 2.86%, and conversation and meeting intelligence at 0.26%. the demand is retrieval; the supply of a retrievable personal archive is close to zero.
every large vendor is now building toward that gap from the side it controls. microsoft put recall into windows, rebuilt after its may 2024 reception with biometric gating and on device encryption, which we assessed in is microsoft recall safe. apple shipped siri ai with onscreen awareness and personal context on june 8, 2026, though as we covered in apple intelligence on mac it reads your screen in the present tense rather than archiving it. openai shipped a background memory system on june 4, 2026 and anthropic reworked claude's into editable entries on july 10, 2026.
none of them can search what merely crossed your display. the options you actually have, in the order worth trying them:
- browser history, if it was a web page in the last 90 days.
- spotlight, if you saved the file.
- your screenshots folder, which is bigger than you think.
- the app's own search, if you know which app it was.
- a colleague, if they were there.
- screen memory, if none of the above and you were recording.
how do i search browser history properly?
most people search browser history wrong. they hit cmd+y, type a word, find nothing, and give up.
two things to know. first, chrome and safari search page titles and urls, not page content. if the thing you remember was in the body text, the title probably does not contain it. search for something that would plausibly be in the title instead: the site name, the product, the section.
second, use the date. if you can place it within a day or two, scrolling the history for that afternoon is often faster than guessing keywords. you recognise a page when you see it far more reliably than you can describe it.
where it runs out: roughly 90 days of retention by default. only pages loaded in that browser. nothing from a desktop app, a pdf, a screen share, or a message.
what does spotlight search on a mac actually cover?
people reach for spotlight expecting mac search everything and get file search instead.
spotlight indexes file contents, not just names, so a phrase inside a document or a pdf is genuinely findable. and modern macos runs live text over images, which means text inside a screenshot you saved can surface too.
that last part is better than most people realise and worth testing before you go further.
where it runs out: it only knows about files. the thing you are looking for has to have been saved as one. the overwhelming majority of what you look at in a day never becomes a file.
the screenshot folder you forgot about
if you are the kind of person who screenshots things reflexively, your screenshots folder is an accidental archive. sort by date, jump to the week in question, and scan. combined with live text, it is searchable.
where it runs out: it only contains what you thought to capture at the time. which is the whole problem. the things you need later are the things that did not feel important enough to screenshot.
ask the person
genuinely underrated. if the thing came from a conversation, a slack message or a call, asking "what was that link you sent me" costs thirty seconds and works.
where it runs out: it does not scale, it depends on someone else's memory being better than yours, and it is unavailable for anything you found on your own.
why do all of these fail in the same way?
look at the pattern. every method above searches something you deliberately kept: a page you loaded in a specific browser, a file you saved, a screenshot you took, a message someone else sent.
but the thing you are trying to find is almost always in the other category. the number that was on screen for eleven seconds during a call. the apartment listing you closed. the config value in a doc someone shared. the api response you glanced at in a terminal.
you did not save any of it, because at the time it did not feel like it mattered. that is not a discipline failure. nobody can predict which detail they will need three weeks later, and a system that requires you to try is a system that will fail.
the methods, side by side
each option covers a different slice, and none of them covers the one you usually need.
| method | what it covers | how far back | searches text on screen |
|---|---|---|---|
| browser history | pages loaded in that browser | roughly 90 days | no, titles and urls only |
| spotlight search mac | files you saved | as long as the file exists | yes, via live text, saved images only |
| screenshot folder | screenshots you took deliberately | forever | yes, if you remember to look |
| app search | one app's own content | varies | within that app |
| asking a colleague | what they happened to notice | unreliable | no |
| screen memory | everything captured, all apps | as long as you keep it | yes, via ocr |
the first five have the same hole. they cover what you deliberately saved or what one vendor happened to store. finding a closed tab from july, or a figure on a slide someone shared for 40 seconds, is outside all of them.
the category that solves it
screen memory tools invert the model. instead of asking you to decide what to keep, they record continuously and make everything searchable after the fact.
the mechanism is straightforward. the app captures your screen as a stream of frames. ocr reads the text out of each one, so a picture of a slack message becomes findable by the words in it. that text goes into a local index. then you search it, or ask an ai that answers from it.
the thing that makes this work is that ocr step. it is why screen memory can find text that was never a file, never synced, never saved: text inside a video call, inside a screenshared deck, inside an app with no export.
how do i find a closed tab after history has aged out?
to find a closed tab past the roughly 90 day window, history will not help and neither will reopening recently closed tabs.
what sometimes works: your browser profile sync on another device, a bookmark you made and forgot, the link pasted into a slack thread, or the page's own appearance in your email. what reliably works is having recorded the screen at the time, which is the only method that does not depend on you having taken an action.
a warning from the one that went away
before you install anything, ask where the archive lives, because this category has an unusually bad record.
rewind defined it on the mac. in december 2025 meta acquired the company and shipped an update that permanently disabled all screen and audio capture on december 19, 2025, 14 days after the announcement, with 7 markets losing access immediately. the full pattern is in what happens to your data when an ai app shuts down.
what to check before you install one
a tool that records your whole working life deserves more scrutiny than most software.
where does the archive live. on your machine, or on a company's servers. there is a real difference and the default matters more than what is possible.
can you get it out. rewind ai, the app that made this category popular, stopped recording entirely in december 2025 after its company was acquired. anyone who could not export their archive did not own a memory. they owned a subscription.
which model sees it. some tools hardcode a cloud model. others let you point at something local so inference never leaves the machine either.
trained on, or sold. two separate promises. you want both in writing.
where remynd fits
remynd records your mac, runs ocr on device, keeps the archive local by default, and lets you hit ⌘K to search or ask across everything you have seen and said. search over your own history runs against the local index. asking the ai a question sends the retrieved slices to a cloud model, and there is no local model option today. optional cloud backup is encrypted and exports to your own s3 bucket.
it is the version of this built for someone who read the section above and cared about the answers. if you are weighing it against the alternatives, the rewind ai alternatives page has the full field.
the most common thing people search their own history for is not a lost file. it is the answer to "what did i actually do this week", which is why how to write your weekly update is really an article about recall.
download remynd free for mac → free to download, runs locally, no card.