Explain it once. It remembers from then on.

Most assistants meet you as a stranger every time. This one already knows you.

There are two sides to what 1Presence keeps, and it helps to hold them apart. There is your vault — the notes and files you can open, read and edit yourself, like any folder of documents. And there is its memory — the quieter thing that lets it recall the right detail at the right moment: who Maya is, that pricing is still undecided, that Tuesday mornings are yours.

The vault is where things live where you can see them. Memory is what makes a new conversation pick up as though the last one never ended. You shape both the same way — by talking.

It pays attention to the things that tend to matter later, and lets the passing stuff go. You do not curate this — it happens as you talk — but it is worth knowing the shape of it.

The throwaway lines — a passing question, a one-off aside — it lets slide. Memory is meant to be the signal, not a recording of everything you ever said.

It is tempting to picture memory as the assistant simply holding your whole history in its head and reading it all back. That is not what happens, and the real picture is the reason you can trust it.

When something is worth keeping, 1Presence writes it down — as a filed note, and as a link between the things it connects (this person, that project, this decision). Later, when you ask about something, it does not sweep back through every past conversation. It finds the few notes that are actually relevant and works from those. Ask "what do I know about the Meridian project?" and it gathers the handful of things that touch Meridian — the decisions, the people, last week’s change — and writes you a coherent picture from them.

Because memory is notes and links rather than a black box, two good things follow: it can connect what it knows (a question in one corner pulls in what matters from another), and you can see and correct all of it. What it knows is exactly what has been written down, nothing hidden.

Three kinds of memory, drawn on as one

Underneath, that written-down memory is not one pile of notes but three kinds, each suited to a different shape of thing. You never sort anything into them yourself — it happens as you talk — but knowing the shape explains why recall feels so complete.

The important part is that they are drawn on as one. A single question does not pick just one — it quietly pulls from all three at once, following the links between them. Ask "what was I working on in March?" and it walks the timeline; ask "what do I know about Meridian?" and it gathers the file you wrote, the notes it took, and the people and decisions the map connects to it — then hands you one coherent picture, not three separate lookups.

This is also why a question in one corner can pull in what matters from another: because the connections cross between the three, something you filed under one project can surface when a related person or decision comes up elsewhere. The memory feature page lays out the three in full.

You never fill in a form to teach 1Presence something, or to make it forget. You just say it, the way you would tell a colleague who is good at remembering.

“Forget that” and “remember that” are real commands — not settings buried in a menu. And you can always browse what it has kept: open the memory view to see everything filed away, and search it for any person, project or moment, even with a half-remembered name.

1. 1

   You mention something worth keeping

   In the flow of a normal chat — a preference, a person, a decision. You do not have to flag it specially; it notices the things that tend to matter.

2. 2

   It files it, and links it

   The fact is written down as a note and connected to what it relates to — the project it belongs to, the person it concerns.

3. 3

   It resurfaces when it is relevant

   Weeks later, when the topic comes round again, it brings the note back on its own — you never have to remind it.

4. 4

   You correct or remove it any time

   Tell it what changed, or ask it to forget. The update holds from then on, everywhere.

Your memory is yours. Anything you flag as sensitive is kept in your vault but never lifted into the shared picture other agents can draw on — so a personal detail does not leak into your researcher’s context. And nothing is locked in: your files are plain documents you already hold, and what it knows can be exported and taken with you if you ever leave.

When you go further and build more than one agent, memory gains a clean structure. Facts about your world — your projects, the people, the decisions — sit in shared knowledge every agent can use, while each specialist’s own working notes can stay open to the others, private to it, or walled off entirely. In a team, the same idea ringfences memory so what belongs to one person or project stays there.