Mind maps — how the ideas connect

The same video read by meaning rather than by the clock: a branching map you can fold, search and jump into.

The same video, read a different way: not what happens when, but how the ideas fit together. Headings give you the shape; underneath them sit whole sentences saying what was actually covered — so the map is worth reading on its own, not a contents page you have to go back to the video to decode.

This is not the Overview

Both are trees over one video, so the difference is worth being exact about. An Overview is organised by the clock: it answers “what happens when, and where do I skip to”. A mind map is organised by meaning: it answers “how do these ideas relate”.

That last row is the whole reason both exist. An Overview that merged those three moments would be wrong — you'd lose the ability to skip to the second one. A map that split them would be wrong too — you'd never see that they were the same idea.

The levels do different jobs

This is the part that makes a map worth having, so it is worth being explicit about:

• The top two levels are headings — short noun phrases, two to six words. A heading tells you

  where you are

  .

• Below them the map writes in sentences — fifteen to thirty-five words, with the specifics in them: the names, the numbers, the mechanism. A leaf tells you

  what was said

  .

A map made only of noun phrases — “Attention heads”, “Training data” — is a table of contents. It tells you what was covered and nothing about what was said, so you still have to go back and watch, which is the thing the map was supposed to save you. The two levels of heading are there to be scanned; the sentences underneath are there to be read.

Opening one

1. Turn Mind map on in Study → Settings — it's off by default, like everything except notes and quizzes.

2. On any video row press the Study chip and pick the map icon. It opens in the same box the Overview, Transcript and Quiz use, so nothing on the card moves until you ask.

3. The panel opens on a Map this video button, not on a map. Pressing Study never spends anything by itself, and the button says what building one costs before you press it.

It opens as a list, fully open — the sentences are the point, so nothing hides them by default. Fold a branch to quiet it, search to find an idea (its parents come with it, so a hit three levels down still reads in context), and press a timestamp to jump: seeking the player on the video you're watching, or opening it at that second anywhere else.

Diagram swaps that list for the map itself, right there in the card — drag to move around it, scroll to zoom, fold a branch by its handle. It is the same nodes read a different way: the list is for reading what was said, the diagram is for seeing how it all hangs together. Folds carry across, and switching costs nothing.

The diagram, and the full-width tab

The video's name sits top-left as a filled pill; the top-level headings run in a row beside it; and each one owns a column that flows straight down, indented a step per level and wired with square elbows in that branch's colour. Drag to pan, scroll to zoom, fold any branch by its handle, and download it as a PNG.

The ⤢ button opens exactly the same diagram in its own tab, full width — worth it for a map with a lot of branches, where a sidebar column is simply narrow.

Where the timestamps come from

The same discipline the quizzes use: the model has to copy a verbatim line of the captions for each moment an idea is covered, and we find that line in the transcript ourselves. The number you press is the one where the quote actually occurs — never the model's own guess — so a timestamp can't point past the end of the video.

• A quote we can't find in the captions loses its timestamp, not its branch. The idea is still in the video; we just can't promise where.

• Some branches carry no timestamp at all, and that's correct rather than missing — they're ideas the video builds up to without ever stating in one place.

What it costs

Questions