The Mind
What your workspace remembers about your work and the people in it, where that memory comes from, and how to correct or delete any of it.
Checked against the running product on 7 min readMarkdown
On this page
Most tools file what you write and never read it again. The Mind is the part of The Margin that reads it back. From the notes, cards and conversations you were already making, it works out who and what keeps coming up, and it uses that to answer questions no single note could.
What it holds#
- Hubs
- the people, projects, organizations and agents that keep turning up. Mention Sarah in a note and again on a card, and both lead to the same Sarah. Names, under Mind in the sidebar, lists them.
- Memories
- durable facts about you, each with an importance and each traceable to the note, card or conversation it came from. "I do not eat meat" is kept. "Can you make me a board for the trip" is a request, and is not.
- Connections
- links between things. This note is about that project; these two cards describe the same work under different names.
The Mind page draws all of it as a map. A [[wikilink]] you type yourself
draws a line on that map straight away, because a link you made on purpose is
better evidence than one the engine inferred.
The button at the top of the Mind page, Connect ideas (or a count such as 3 to connect), opens Possible links: pairs of things that look related, each with a line on why. Connect draws the link and Not now sets the pair aside, so nothing is linked until you say so. Opening the list costs nothing; Find more looks again and uses a weave credit.
How it learns#
- New and edited content is picked up in the background, in batches, about every fifteen minutes.
- An edit is read again, so the Mind does not keep a stale copy. A memory that was true in March and wrong by August does more harm than a gap.
- Recall matches on meaning as well as on the words. Ask "did anyone say anything about the wedding" and it finds the note that discussed the ceremony without using the word.
Reading and correcting what it remembers#
Everything the Mind durably believes about you is listed in Settings→Your Brain, under "What your Margin remembers about you". It is written as plain sentences, grouped as things you have shared, how you work, and patterns it has noticed. Any line can be edited, or forgotten. Forgetting deletes it, and recall stops using it at once.
That list is the whole of it. A wrong entry you cannot see misleads every answer that touches it, which is why the list is there to read.
Whose memory it is#
Memory belongs to the workspace it was learned in. What Margin learns in the family workspace stays in the family workspace, and every member can read it on Settings→Your Brain, because the household's assistant answers all of them from it. Your personal workspace's memory is yours alone.
In a shared workspace each fact keeps who taught it, shown under it as "Taught by" and their name. Only that person, or an owner or admin of the workspace, can edit or forget it; everyone else sees whose it is and has nothing to press.
About you#
Some facts are about you rather than about any workspace: you like short answers, you are vegetarian, you do not take calls before ten. Those live in About you (Mind→About you, also linked from Settings→Your Brain). About you follows you into every workspace, Margin Intelligence reads it in each one, and nobody else can see it, a workspace owner included.
- Add, reword or remove lines yourself, up to 100 of them, each one short sentence (500 characters at most). Each line says where it came from: added by you, saved by Margin Intelligence, or saved by a connected assistant.
- Tell Margin Intelligence in your own workspace and it keeps the fact in About you. In a shared workspace it asks once, with two buttons: For the household (or For the team) keeps it in that workspace's memory, and Just for me keeps it in About you.
- It holds facts about you and nothing else. A workspace's notes, cards and lists never go into it.
- Workspace exports leave it out. Download on the About you page gives you your own copy.
- Assistants you connect over MCP can read and change it, for you only
(
list_about_me,add_about_me_fact,update_about_me_fact,forget_about_me_fact). - On the phone app, About you is in the Mind's menu.
For a person or a project, open it on Names, the middle tab at the top of the sidebar (Mind→Names), which lists the names the Mind has noticed in your writing. It is not the list of people in your workspace, which is People. A name's sheet shows the short profile Margin keeps, where it appears and what it connects to, and a field to teach Margin about it directly. What you write there is treated as fact, and only the workspace owner and admins can change it.
Keeping things out of it#
Four controls, lightest first.
| Control | Where it is | What the Mind does with the item |
|---|---|---|
| Weave privately | A note's or folder's menu; Private weaving in a board's settings | Still learns from it, but files its suggestions in a collapsed Private group, never leads recall with it, and leaves it out of every export and memory pack |
| Hide from memory | A note's menu | Never embeds or recalls it. The note stays in your notes and your own search |
| Move to Vault | A note's menu, a board's settings, a checklist, a whiteboard | Seals it: hidden on screen and from the assistant while your Vault is locked |
| Put away | A learned star on the Mind map | Seals everything that star was learned from, takes it out of memory, and deletes what was learned from it on the spot |
Weaving privately is a curtain. The Vault is the lock, and Your data says exactly what it does and does not encrypt. Put away is undone from the Vault panel in Settings→Your Brain, one Bring back per item. See Curating the Mind.
In a shared workspace the Mind is scoped the way your data is. Each person's recall reads only what that person can open, so a card on a board you are not on never reaches your answers, whether it comes up in the background or when Margin Intelligence looks something up in memory. The Mind cannot become a way to learn things about the people you live with that they never told you.
Bringing your memory from another assistant#
If another assistant, such as ChatGPT, already knows you, you can bring that over. Mind→Bring your memory (also in Settings → Your Brain) gives you a prompt to paste into the assistant you use. Paste its answer back and each fact is listed with its kind: a preference, a person, a project, a routine or a fact. Uncheck or fix anything, then save. Nothing is kept until you do.
Facts Margin already knows start unchecked, and anything that names something you hid from memory stays out. Saved facts go into this workspace's memory, so in a shared workspace the other members can see them; a personal workspace is the best place for them. Each import is listed on the same page with Undo this import, which removes exactly what it added.
A ChatGPT data export does not include what ChatGPT remembers about you, only your chats, so the prompt is the way in. If your export does carry a memory file, the page reads it.
Taking it with you#
Settings→Your Brain exports everything the Mind has learned about you, as JSON to import into another Margin or as Markdown to read. The same page brings a JSON file back: near-matches update in place, so importing your own file twice does not duplicate anything. A memory pack is a frozen copy of your file that you share with one person, who imports it into their own workspace.
Anything sealed in your Vault stays out of every file and pack, and so does anything woven privately. More on exports in Your data.
What it costs#
The Mind is on every plan, and so is exporting it. What varies is how much background work it does for you each month. Weaving has its own allowance, separate from the assistant's, so indexing your writing never eats the budget you wanted for questions.
Weave credits a month
25
150
300
150
50
Both allowances and what is left of them are in Settings→Billing. Reading and correcting the Mind in the app are yours on every plan. The paid part is letting an agent you connect from outside write to it, which comes with the plans that allow outside agents at all. See Plans and limits.