Mind

One point per passage, coloured by the source it came from and placed by how alike the material actually is — not sorted into folders by type. Lines join passages your twin reads as being about the same thing. Orbit it and zoom in to explore the shape.

Where to find it

In the old app, the Map tab on a personal twin’s Knowledge page (Sources · Lists · What it knows · Map). Only a personal twin has it there; in the new app, every agent does (below). A compact version also sits inside the What it knows tab — press Map there for the same view in miniature, with a link through to the full page. The map is requested only after you open the Map tab or press Map, so the ordinary dashboard stays light. Click a point here and its source appears below it; that’s your own dashboard view, not the shared one below.

In the new app

The map is on each agent’s Knowledge page, under Map beside List, for every kind of agent. Only you see it.

Each point is a passage your agent can quote, placed near passages that mean something similar. A ring is a question it couldn’t answer, drawn beside the source its reply leaned on. A question whose reply used none of your sources has no ring: it’s listed beside the map and never placed by guess.

To answer from the map, press a ring, or Answer on a question in the column beside it, then type and press Teach it. ⌘↵ (Ctrl+↵ on Windows and Linux) saves too. The line then says Learned and when your agent starts using it, the ring turns into a violet point and the count goes down. Answers join your agent at the next train, or at once on its first build.

Topics

Topics group your sources by where they came from — a site’s section, or what you added by hand. Each topic is labelled on the map with its unanswered count, and pointing at one lights its passages. Group by in Sources reads Sources or Topics.

Questions it answered

Press a point to see the source, and the questions its answers cited in the last 30 days. A row opens that conversation in Activity; View source opens the source itself. Only you see these questions.

On a phone, Map opens as a list of topics; Show the map draws it below. The map draws in once on your first view each session, and not at all if your device asks for reduced motion.

Drag to turn the map, and use Zoom in and Zoom out to move closer; scrolling never zooms, so the page keeps scrolling. On a phone the map stays still. Sources filters which sources are drawn, with Group by to switch between single sources and their topics. Passages as a list (under More) shows the same passages as a list, which also works where the browser can’t draw the map.

Knowledge in the sidebar shows every agent’s map as a card, with its unanswered count. One map of the whole library is a Preview and isn’t live yet.

Built from what’s live

The map is your twin’s live index, not a running list of everything you’ve added. A new source appears on it at your next train, the same moment it starts answering questions — nothing here updates before that. Before a first train the page says Nothing to map yet and offers Go to Train.

A very large library is drawn as a stable sample of itself, capped at 800 passages so the projection stays responsive, and the count above the map says so: 800 of 12,431 passages.

Note

Watch it grow: play back the order your sources actually joined, settling into today’s positions rather than replaying old ones.

You can switch between Map and List. The list is a keyboard- and screen-reader-friendly way to inspect the sampled passages when WebGL is not available. Selecting a passage shows its source in an in-place preview drawer on desktop, or a bottom sheet on a phone, so Mind stays open while you check the bounded excerpt. The source detail is fetched only after you press View source; Open original source remains available when you do want the source page in a new tab. For a large library, List keeps the page compact with collapsible sections and keeps the expanded rows inside a bounded scroll area. The first List screen shows up to 12 sources or 8 groups, with Show all when you need the complete set; search can jump to a source or group without opening everything. With View by: Sources, origins contain compact source rows and one source opens at a time. With View by: Groups, each origin group opens to show its passages.

The compact Map option under Knowledge appears only when the twin has a published, non-held source to display. A first source waiting for training keeps the text view and its Train path clear; a twin with an existing map can keep using it while another source waits for the next train.

In the owner view, View by: Sources | Groups changes the legend and filter lens without changing the map’s positions or the passages your twin can use. Sources filters individual source pages; Groups combines pages using the same origin groups as the Sources panel, such as a site or Added by hand. Click a legend row to show or hide it, or shift-click to focus on one. Clicking the last active row restores everything. This is a local visual filter: it does not add database records, call an AI model, retrain the twin, or change its answers.

Sharing it

Off by default, on your public page only. Turn it on and a visitor can orbit and zoom your map — its shape and its mix of source types, never a title, an address, a word of content, or a click-through to the source itself. A source you’ve held back or marked confidential never appears, on or off.

Was this page helpful?

View Markdown