The mind map was the right shape all along

An AI mind map is worth using again because the old weakness was never the diagram — it was that a blank map only holds the angles you already thought of. When each branch can apply a named thinking model to your actual material, the map stops reflecting your habits and starts widening them.

Mind mapping has been quietly out of fashion for a decade, and the usual explanation — that the tools were clunky — has never been true. The tools were fine. Something else was wrong, and it is worth naming precisely, because it turns out to be the part that changed.

The shape was never the problem

Start with what a map actually buys you over a list, stated narrowly enough to be arguable.

Branches coexist. Three readings of the same problem sit next to each other permanently. Any linear medium destroys this by construction: ask a chat assistant about a second angle and the first one scrolls out of view, so comparing them means holding both in your head while you scroll between them. That is exactly the working memory you wanted the tool to spare you.

Position is memory you do not have to carry. Five levels into an exploration, the question "which threads have I already followed" gets answered by looking rather than by remembering. This, incidentally, is why a map that re-arranges itself is worse than one that does not — we turned down force-directed layout in MindMax for this reason. Stability beats tidiness, because the thing being preserved is your spatial memory of where you put things.

Incompleteness is visible. A short branch next to a dense one is a question you have not worked on yet. A bulleted list gives you no such signal; every bullet looks equally finished.

None of that has aged. If anything it matters more now, because the default interface for thinking with a machine turned out to be the most linear one ever built.

The problem was who fills it in

Here is the part the productivity literature was always vague about. You sit down in front of a blank centre node and fill in the branches — with what?

With your own habitual moves. Everyone has three or four they reach for: what are the pros and cons, what would this cost, what could go wrong. The map then renders those three angles in a handsome radial layout, and a narrow exploration acquires the visual authority of a thorough one. That is worse than a plain list, which at least looks unfinished.

The knowledge is not what is missing. You have heard of a pre-mortem. You know what a steelman is, and roughly what first principles means. Charlie Munger made the case for a latticework of mental models thirty years ago and the canonical lists run to a hundred entries, most of which an interested person can name.

Almost nobody runs them. Three reasons, none of which is fixed by reading another list:

1. Recall fails at the wrong moment. Knowing what a pre-mortem is does not mean it occurs to you on the Tuesday you are deciding, while you are inside the problem and already leaning toward an answer.
2. Applying one is real work. An assumption audit, honestly done, means listing every premise the argument rests on and checking each. That is an hour. Faced with an hour, you do the five-minute version in your head and arrive where you were already going.
3. The useful ones are the unwelcome ones. The models that would help most attack a position you have grown attached to — which is precisely when you are least inclined to reach for them.

So the map filled up with what you were going to think anyway, and then it dead-ended, because turning a diagram into a memo meant retyping the whole thing somewhere else.

What actually changed

Not the diagramming. The thing that changed is that a method can now be carried out rather than merely named.

That is the whole design of the canvas in MindMax. You put a question, a passage, or a document from your own folder at the centre. Then instead of typing a branch, you pick a thinking model — SCAMPER, six thinking hats, first principles, a red team pass, an option matrix — and each branch comes back with that method applied to your actual material. Recall is solved by it being a menu you open at the moment of need. The hour of grinding is solved by something else doing the grinding.

Two design consequences we would defend:

Nodes are cards, not words. A node that fits only a label cannot hold a claim, and the branch we care about is the one that says something specific enough to be wrong. It makes the map denser than a classic keyword map. That is a trade we made on purpose.

A branch has somewhere to go. It can be handed to the writing pipeline as a ready-made outline, or saved into your folder as an ordinary Markdown note. The board itself is an open JSON Canvas file in your own folder, which means Obsidian can open it and we cannot hold it hostage.

What it does not do

It does not do the judging, and that is not modesty — it is where the value sits. A wall of applied models is material, not a conclusion. Most of what comes back will be reasonable and unremarkable; the two branches that are not are the reason you did this, and finding them is yours. Someone who keeps everything the canvas produced has generated volume and will feel it later, when the draft made from it says nothing.

It also does not rescue a vague question. A red team pass on "how do I grow the business" produces vague objections to a vague plan. Sharpening the centre card is still the first move, and nothing downstream can do it for you.

If you want the mechanics: how the mind map works, or the two definitions underneath it — mind map and mental model.