Verified against ChatGPT · 2026-07-18
Draft in Canvas and get surgical edits instead of a full rewrite every round
A Canvas setup prompt that locks each revision round to the exact section you point at, so asking for one paragraph to be sharper does not quietly rewrite the rest of the document.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Open this in Canvas so we can iterate on one persistent draft instead of regenerating a new version in the chat each round. DOCUMENT BRIEF A one-page internal memo proposing we move standups to async written updates AUDIENCE AND GOAL Engineering leads who are skeptical of losing daily verbal syncs, goal is a trial period, not full buy-in LENGTH TARGET Under 400 words, fits on one screen without scrolling Write the first full draft in canvas now. For every revision after this first draft, I will point at a specific section rather than describing the whole document again. When I do: - Edit only the part I pointed at. Leave every other paragraph exactly as it was, unless the edit has a direct knock-on effect elsewhere, for example a term you renamed that also appears later, and if so, name that follow-on edit before making it. - Do not change tone, structure, or claims outside the section I pointed at. - If my revision request is ambiguous, ask one clarifying question before editing the canvas. First revision to apply after the initial draft, if any: Make the opening paragraph state the proposal in one sentence before any justification
Customize the highlighted detailsoptional — the prompt above already works
Why this works
Canvas is a persistent editing surface separate from the chat stream specifically so a document or a piece of code can be revised in place rather than regenerated as a new wall of text every round, and it is documented to apply targeted edits, shown as a visible diff, when a revision request is scoped tightly enough for it to do so. The failure mode this prompt is built against is the vague revision request: told simply to make it better or tighten the intro, a model reverts to chat-style behavior and rewrites more of the canvas than was asked for, because nothing constrained the edit to a boundary. Instructing it to point at a specific section and edit only that section reinforces the exact interaction Canvas is designed around, where highlighting or naming a passage and requesting a change to it should produce a scoped diff, not a fresh draft. Requiring it to name any knock-on edit before making it, such as a renamed term that also appears elsewhere, prevents the opposite failure: an edit that is technically scoped to one section but silently breaks consistency with a part of the document the model was told not to touch.
Verified against
ChatGPT GPT-5.1 (Canvas) · 2026-07-18
Changelog
- 2026-07-18 — Initial publish, verified against ChatGPT GPT-5.1 Canvas.
Building this for real?
This is a free starting point. If you'd rather have what Scult builds built and running for your business, that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
