Skip to main content

Use caseThink

Open a hard question on the canvas with your agent

One card with the question on it, and the AI app you already use adding branches beside you — options, objections, the angle you had not thought of, drawn where you can pull on them.

You start with
A question you keep going round on, and a canvas in MindMax with that question on it.
Time
An afternoon, or as long as the question deserves — you can leave it and come back to the same board.
Your part
The question, the branch you choose to pull on, and the judgement on which objections actually land.

What you'll end up with

  • A map, not a transcript: the options on one board, with the contradictions drawn between them.
  • Cards the agent added, signed with its name, so you can still tell your thinking from its suggestions.
  • Branches you keep pulling on — pick one, expand it, ask again, without losing the rest of the board.
  • A session you can re-read afterwards in MindMax, call by call, instead of scrolling a chat log.

A chat window flattens a question into one thread. A canvas keeps it in the shape you can act on.

From one question to a board you can explore

  1. Put the question on a canvas

    You

    One card is enough: should we rebuild this service, is this offer worth moving for, what is actually blocking this project. A blank canvas works; so does one you have already started.

  2. Connect the app you think with

    You

    Settings → External AI → Connect. The canvas in front of you is part of what a connected app can see, so there is nothing to copy, paste or describe.

  3. Ask for what a chat window would have flattened

    You ask · MindMax answers

    AskRead the canvas I have open. Add the options I have not considered, then red-team each one. Link the cards that contradict each other.

    Four kinds of thing come back: cards at the top level, cards growing out of an existing card, relations between cards, and points added under a card that already exists.

  4. It reads the board before it adds to it

    Cards, the points under each one, who wrote them, and the relations already drawn — the app gets the outline of the board, not a screenshot of it. Nothing it adds is a replacement for what is there.

  5. Pull on the branch that interests you

    You

    This is the part a conversation cannot do. Leave eleven branches alone and open the twelfth; come back tomorrow and the board is where you left it. Each expansion is a fresh question, not a re-explanation.

  6. Borrow one of MindMax’s own methods

    You ask · MindMax answers

    AskApply a decision matrix to the three options on this canvas and show where they actually differ.

    MindMax’s named thinking methods can be read one at a time, with the card shape each produces, and applied onto a canvas — so the structure comes from a method rather than from whatever shape the model improvises.

  7. Have it check your own answers

    Write your answers on the question cards, then ask the app to mark them. Verdicts and notes land on the cards you wrote, which is a cheaper way to find out you were wrong than finding out in the meeting.

Why it works here

The structure survives the session

A long conversation ends with the answer buried in the middle of it. A canvas ends with the answer on a card, the objections next to it, and the relation between them drawn — which is what you actually re-read a month later.

How the canvas works →

Your cards are not up for rewriting

An outside app may revise a card it wrote itself; on MindMax’s cards it can only add points. So the branch you wrote stays your branch, and what it contributes sits under it rather than replacing it.

What an outside app can touch →

The thinking is billed to your app

No MindMax model runs on this path, so no MindMax credits are spent. A long method on a big board is tokens in your Claude or OpenAI account — the same order as talking it through for an hour.

What MindMax charges for →

Also used here: Guide: connect an AI app · Thinking methods, one by one · All the tools it can call

Where it gets hard

  • A canvas is not a document

    It will not hand you the memo. Branches, options and verdicts are thinking material; the writing still happens in a document, with the board open beside it. If you want the memo, ask for it after the board is settled.

  • It sees the board in front, not every board

    Only the canvas you have open is in reach. Another canvas in the same folder is not, unless you open it, and an app cannot hop between boards to assemble a bigger picture on its own.

  • Big boards get expensive

    Every card and point is context, and a method applied over forty cards costs accordingly. Prune before you ask for a whole-board pass, and ask for one branch at a time when the board is large.

  • Signed cards are the audit trail

    Cards carry the name of the app that wrote them. That is how you tell its suggestion from your own note a week later — and why editing a card by hand to agree with it is a bad habit.

  • Not every host handles diagrams well

    Claude Code and Codex can drive a full method end to end. A smaller chat client will do the same work in smaller pieces, more slowly, and may need the canvas re-read between steps.