Skip to content

Using a tool

Locking a card

Unlocked, a card fires once and closes. Locked, it stays open and every run lands in its own slot in a collecting frame.

Press Run on an unlocked card and it fires once, collapses to a pill, drops its result and closes. That is right for a one-shot and wrong for the way most work actually goes, which is several quick attempts in a row.

Click the padlock and the card stays.

  • The card stays open and idle after each run, with your prompt and settings still in it, ready for the next press
  • Results stop replacing each other. Each run lands in its own slot
  • The slots collect into a frame that grids them as they arrive
  • A short cooldown sits between presses, so a fast double-click cannot fire twice

Fire six runs at a locked card and you end with a tidy grid of six, not six overlapping images you now have to arrange.

Same prompt, different model. Lock, then swap the pill between presses. You end with one prompt read several ways, side by side, which is how you pick an engine for a specific job rather than in the abstract.

Same model, different prompt. Lock, then edit one clause between presses. The grid becomes a record of what each change did, and because the card never closed you never retyped the parts you kept.

Click the padlock again and the card goes back to closing after its next run.

The frame you collected stays on the board as an ordinary frame. It stops auto-gridding, so you can rearrange inside it, and it keeps everything it gathered.

Lock again and a fresh frame starts. That is deliberate: a session is a line of enquiry, and the boundary between two of them is worth keeping.

Two cases turn lock on without you asking.

A tool can ship locked by default, when it only makes sense as a series.

More often you will meet it through variables. Binding a { } variable in a prompt puts the card into batch mode, which is lock mode driven programmatically: one press, N runs, all of them landing in the same collecting grid.