Verified against ChatGPT · 2026-08-12
Draft one scene using goal-conflict-disaster instead of a string of things happening
Writes a single scene structured around a concrete scene-level goal, real opposition, and a setback, grounded in specific sensory detail rather than generic description.
The prompt
Ready to copy — highlighted parts are example details you can swap.
You are drafting a single scene from a story. Structure it around goal, conflict, and disaster — not a sequence of events that happen one after another without a driving question. POV CHARACTER Tomas, a junior surgeon. WHAT THE POV CHARACTER WANTS IN THIS SCENE SPECIFICALLY Convince the attending surgeon to let him perform the incision himself for the first time. WHAT'S STANDING IN THE WAY The attending, who has legitimate reasons to be cautious after a resident's error last month cost the hospital a malpractice claim. SETTING AND SENSORY ANCHOR Pre-op scrub room, the smell of surgical soap, the fluorescent hum, cold water running longer than it needs to. TARGET LENGTH 500-700 words. Before drafting, confirm the scene-level goal is concrete and achievable-or-failable within this one scene — not a book-length goal restated, but something specific the character could plausibly get or not get before the scene ends. Open the scene already in motion toward that goal, skipping arrival and setup. Let the opposition in The attending, who has legitimate reasons to be cautious after a resident's error last month cost the hospital a malpractice claim. act with its own logic — it should be actively working against the goal, not passively present as an obstacle the character simply pushes past. Ground the scene in the sensory anchor given, but sparingly: one or two specific, concrete details woven into the action, not a paragraph of standalone description before anything happens. End the scene on a disaster relative to the stated goal — the character does not get what they wanted, or gets it in a way that creates a worse problem than the one they started with; a scene that simply achieves its goal cleanly has nowhere to go and kills momentum for whatever comes next. WHAT NOT TO DO Do not resolve the scene's tension in the character's favor without complication — a clean win is rarely dramatically useful mid-story. Do not let the scene wander into unrelated exposition or backstory; if it doesn't serve the goal-conflict-disaster spine, cut it. OUTPUT FORMAT 1. The scene at target length. 2. One line stating the specific disaster and what new problem it creates going forward.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Left to its own devices, GPT-5.1 tends to write scenes as a sequence of plausible actions loosely connected by 'and then,' because without an explicit driving question a scene has no internal test for what belongs in it and what doesn't — goal-conflict-disaster (a structure from the Dwight Swain school of scene craft) gives the model a concrete filter: every line either advances the stated goal, embodies the opposition, or sets up the disaster, and anything else is noise. Requiring the goal to be scene-scale and achievable-or-failable within the scene, rather than a restatement of the character's book-length ambition, matters because an underspecified goal produces a scene with no real tension to track — the reader (and the model generating the prose) has nothing concrete to be uncertain about. Instructing the opposition to act with its own logic rather than as a passive obstacle addresses a frequent weak-scene pattern where the antagonist or obstacle exists only to be overcome, with no legitimate reasoning of its own — scenes read as more genuinely tense when the opposing force has a real, defensible position, which is why the example asks for the attending's caution to be grounded in a specific prior incident rather than generic authority-figure resistance. The instruction to end on disaster rather than a clean win directly counters GPT-5.1's tendency toward satisfying, resolved endings even at the scene level, which is appropriate for a scene that closes out a story but actively undermines momentum mid-narrative — a scene that gives the character exactly what they wanted removes the question that was supposed to pull the reader into the next one.
What you get back
"You're not ready," Dr. Ansari said, not unkindly, already turning back toward the sink. Tomas kept his hands under the water a beat longer than necessary, the cold finally starting to bite. "I've done forty of these in sim." "Sim doesn't bleed." She handed him the retractor instead of the scalpel, which was, he understood slowly, its own kind of answer...
Verified against
ChatGPT GPT-5.1 · 2026-08-12
Changelog
- 2026-08-12 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
