What your tools know about you that you have forgotten
Pick a month from two years ago. Not a special one. A month with no wedding, no move, no hospital visit. Try to name five things that happened in it.
Most people get two, and one of those turns out to be from a different year. This is not a personality flaw and it is not early decline. It is the normal shape of forgetting, and it applies to everyone reading this.
Here is the part that gets less attention. That month is not gone from the world. It is gone from you. Substantial pieces of it are sitting on servers right now, timestamped to the minute, in the accounts you are already logged into.
Go and look
This is worth doing rather than reading about. Open a few of your own accounts and look at a date you have no memory of.
Your calendar is the obvious one. Scroll it back to that month. The events are still there, and they carry more than titles: who organized them, who else was invited, how long they ran, where they were. A 4pm on a Thursday with three names attached is a fact about your life that you no longer hold and your calendar never let go of.
If you scrobble to Last.fm, your listening history goes back to the day you made the account, every track stamped with the minute you played it. That is a strange thing to have. Nobody remembers what they were listening to on an ordinary Tuesday afternoon in a specific year, and you can read it off a page.
The same is true across the rest of the pile. Steam records hours played per game. GitHub keeps every commit you pushed with its message and timestamp. A check-in app knows the name and category of the venue you walked into. A task manager knows the exact moment you ticked something off. Your streaming and watch histories know what you finished and when. Your camera roll is sorted by date whether or not you ever look at it that way.
None of this was assembled for your benefit. Each service keeps a log because a log is how the product works. Sync needs timestamps. Recommendations need history. Billing needs records. Your past is a side effect of someone else’s architecture.
But the coverage is real, and if you go looking at a specific month you have lost, the honest reaction is usually some version of: I did all that?
Why it does not feel like a memory
So the record exists. Reading it still does not give you the month back, and the reason is worth being precise about.
It is in fifteen places. No single view exists. To reconstruct one Thursday you would need to open your calendar, your music history, your commit log, your task archive, your messages, and your photos, then line up six timelines by hand. Nobody does this. Not because it is hard, but because you would have to already care about that specific Thursday to start, and the whole problem is that you do not know which Thursdays to care about.
Every one of these systems is searchable and none of them is browsable. Search assumes you know what you are looking for. You can find the meeting if you remember the project name. You cannot search for the week you started to feel better, or the afternoon a conversation changed your mind about something. The queries you would actually want to run are not made of keywords.
It is in the wrong format. A row in a listening history is data about a track. A commit is data about a diff. Read a hundred rows and you have facts, not a day. Nothing in any of these logs performs the small act of joining that turns records into a scene: that the long focus block and the 11pm commit and the album on repeat were all the same afternoon, and that the afternoon had a shape.
Nothing points you at it. These logs sit behind menus you visit for other reasons. You open your calendar to see next Tuesday, not last March. There is no moment in an ordinary week where any of these products invite you to read your own past, because that is not what they are for.
So the situation is not that your ordinary days were never recorded. It is that they were recorded in a form that no human being will ever read.
The missing operation is assembly
Notice what is not missing. Collection is not the problem, and it has not been the problem for about fifteen years. You produce an enormous amount of structured, timestamped evidence about your life as a byproduct of using normal software.
What is missing is a step that nobody was doing: take one day, gather everything from every source, and write it down as a day. One place. One timeline. Prose you can read at breakfast.
That step is small and mechanical, which is exactly why it is worth automating. And it is worth being clear about how much of it is honest work rather than invention. Assembly does not add anything to the record. It joins facts that were already true and puts them in an order a person can read. The 9:30 with two names, the four things finished before lunch, the coffee shop at 3pm, the same album twice: these are not interpretations. They are your Thursday, in one place, for the first time.
This is a different job from the one the note-taking and second-brain tools do. Those organize what you deliberately captured. Assembly works on what you never captured at all, because it was captured for you.
What you get back
The thing you get from an assembled day is not information. You already had the information, technically, spread across six accounts.
What you get is the specific experience of reading an ordinary day from your own past and finding it is not blank. The facts are dry on the page. The reaction is not. A venue name you have not thought about in a year, sitting next to a meeting you remember badly and a song you played four times, does something that none of the three does alone. The rest of the day comes back around it, unevenly, in pieces, but it comes back.
That is the whole reason this is worth doing. Not the record. The moment of reading it.
Where deariary fits
deariary does the assembly step. It connects to services you already use, currently thirteen of them (GitHub, Google Calendar, Slack, Todoist, Bluesky, Discord, Steam, Trakt, Last.fm, Toggl Track, Swarm, Linear, and a generic webhook for anything else), reads what they logged for a given day, and writes that day out as prose. Overnight, without asking you anything.
An entry looks roughly like this: the morning meeting and who was in it, the pull request that went out before lunch, the long afternoon block on one problem, the four tasks that closed, the cafe you checked into at three, the album that played twice while you worked. Below the text, the images that came in through your connected accounts and a short list of the day’s structured signals. It is plain writing. It is not trying to be literature.
Two honest limits.
The first is scope. deariary reads what your connected services record, and that is not your life. It has no idea what you thought during the long afternoon, whether the meeting went badly, or that the reason you left early was a phone call. An assembled day is the outline. You supply the rest when you read it, and on some days there will not be enough outline to supply anything.
The second is the past. Connecting an account today generates entries for the last seven days and then continues forward from there. The years already sitting in your calendar and your listening history stay where they are. deariary is not an archaeology tool for a life you already lived. It is a way to stop the next few years from ending up in the same condition.
Which is the honest pitch. The record of your ordinary days has been getting made this whole time, on machines you do not own, in a format nobody reads. Assembling it starting tomorrow costs you nothing, and in six months you will have something to read on a Tuesday you would otherwise have lost.
