which app did i see it in?

you remember the sentence almost exactly. a number, a name, a line of a spec. what you cannot remember is where it was, and on a mac that matters, because there is no search box that covers everything. asking "which app did i see it in" is really asking which of a dozen separate search indexes to open first.

this is a method rather than a list of tools: narrow by time, then by app, then by words. it is faster than it sounds, and it works often enough to be worth learning. current as of september 2026.

why is there no single search box?

every app keeps its own store, and almost none of them hand it to the system. that is the whole problem in one sentence.

spotlight indexes files, plus a few first party databases such as mail and messages. what it does not give you is search across apps. everything else is an island. slack, notion, linear, figma, discord, whatsapp, your terminal scrollback and every browser tab each maintain a private index with its own syntax, its own retention and its own idea of what counts as a match.

so a phrase you remember perfectly can be sitting in an app whose search you never open, and there is no query that reaches it.

start with time, not words

pin down the day before you pin down the phrase, because time narrows the search far more aggressively than keywords do.

most people can place a memory within a couple of days if they anchor it to something else: a meeting, a deploy, a flight, a conversation. once you have the day, the list of candidate apps collapses from twenty to about three.

three checks that recover the day quickly:

1. your calendar. what was on the schedule that week, and what were you preparing for.
2. browser history by date, scrolled rather than searched. recognition is far more reliable than recall.
3. your own outbound messages. what you sent that day usually reveals what you were reading that day.

a 40 hour week is 2,400 minutes, and a year of it is 124,800. narrowing to one afternoon removes about 99% of the haystack before you type a single word. a busy slack workspace can hold 100,000 messages a year on its own, and 10,000 of them may be in channels you never read.

how do i search slack properly?

slack search operators are the difference between finding it and giving up. the ones worth memorising are few.

• from:@person narrows to one sender.
• in:#channel narrows to one channel, including channels you have muted and forgotten.
• before: and after:, each followed by a date, bracket the window.
• has:link finds the messages that contained a url, which is most of what people are actually hunting for.

combining from: with after: beats keyword guessing almost every time when you need to find a message and cannot recall its wording. you usually remember who said it and roughly when far better than you remember their exact phrasing.

where it runs out: slack matches words, not meaning, so a paraphrase returns nothing. it does not search inside images, screenshots pasted into a thread, or the audio of a huddle. and on some plans older history is not visible to you at all, which is a billing setting rather than a search failure, so check the plan before concluding the message is gone.

what about mail, messages and notes?

apple's first party apps are the good case, because spotlight can reach into them and their own search is decent.

mail searches sender, subject and body, and the smart mailbox for a date range is underused. messages searches text and, since the live text era, can surface text inside images you were sent. notes searches body content including handwriting on ipad.

where it runs out: attachments are hit and miss. a pdf attached to a mail thread may or may not be indexed depending on whether it was downloaded. and none of them know anything about the other twenty apps.

the apps whose search will not save you

certain content is unsearchable by construction, and recognising that early saves you twenty minutes of trying.

• anything screenshared on a video call. to zoom, meet and teams that slide was video, so no text exists.
• terminal scrollback. it lives in a buffer that dies with the window.
• design tools. figma and miro search layer and file names far better than they search the words inside a frame.
• pdfs you opened but never saved. if it never hit disk, spotlight never saw it.
• anything inside a screenshot you did not keep.

that last category is bigger than people expect, and it is where most failed searches end up.

which app searches what, side by side

the table is the fast version of everything above.

the elimination method, in order

work down this list rather than searching at random, and stop at the first hit.

1. name the day. calendar, history, sent messages.
2. name the medium. was it typed at you, published, or on screen during a call. that alone eliminates most apps.
3. search the one likeliest app with an operator, not a keyword.
4. search your screenshots folder for that date. sort by date, scan, do not search.
5. ask the person, if a person was involved. thirty seconds, high hit rate.
6. give up deliberately and reconstruct the fact from source rather than burning an hour.

step six is not defeat. an hour spent failing to find a number you could recompute in five minutes is the more expensive outcome.

why does this keep happening?

searching is the most common thing people ask software to do, and personal search is the part nobody built.

per anthropic's economic index for may 2026, the top work task in sampled conversations is searching electronic sources for information at 4.95%, and searching reference materials is second at 3.74%. together that is 8.69% of everything. personal ai assistant use sits at 2.86% and knowledge retrieval and enterprise search at 3.61%.

work is 43.36% of classified conversations in that dataset, and conversation and meeting intelligence is 0.26%, which is close to nothing in a working world that runs on calls.

every one of those searches runs against a corpus someone else assembled. the corpus of what you personally looked at does not exist unless something recorded it, which is why the answer to "which app was it" is so often "an app whose index cannot help you".

the large vendors are converging on the same gap from different sides. microsoft rebuilt recall for windows with biometric gating after its 2024 reception. openai, anthropic and google each shipped memory into chatgpt, claude and gemini during 2026, but those systems remember your conversations with them, not your afternoon in figma.

apple has moved toward the present tense version of this, shipping siri ai with onscreen awareness on june 8, 2026, which reads your screen now rather than archiving what it read last month.

where remynd sits

remynd removes the which app question by indexing across apps rather than inside one. it records the focused window on a mac, runs ocr locally through apple's vision framework, and keeps that index on your machine, so a phrase you saw in slack, a pdf and a screenshared deck all end up in the same searchable place.

the scope, stated exactly: capture, ocr and storage stay on your mac by default, recordings default to 30 days of retention, you can exclude specific apps or sites from capture entirely, and your history is read only in the app. asking a question sends the retrieved slices to a cloud model, so this is not a claim that nothing ever leaves the machine.

one caution before you install anything in this category. rewind was the mac original, and in december 2025 meta acquired the company and shipped an update that disabled screen and audio capture. ask where the archive lives and whether you can export it, of every option including ours, as covered in what happens to your data when an ai app shuts down.

for the broader version of this problem, read how to find something you saw on your screen. for the mechanism that makes screen text searchable at all, read what is ocr.

download remynd free for mac → free to download, runs locally, no card.