Verified against ChatGPT · 2026-08-08
Turn a flat work anecdote into a story with real tension, not a timeline of what happened
Restructures a raw, chronological anecdote into a scene-based narrative built around a want and a turn, ready to drop into a talk, an About page, or an application essay.
The prompt
Ready to copy — highlighted parts are example details you can swap.
You are restructuring a true anecdote I'm going to give you, chronologically, into a short story with actual narrative tension — the kind that works in a talk, an About page, or an application essay, not a diary entry. RAW ANECDOTE (as I'd tell it out loud, unpolished) We were three weeks from a client renewal and found out our main integration partner was shutting down their API with 10 days notice. I had to tell the client before they found out from the partner directly. WHERE THIS WILL BE USED Opening anecdote for a 12-minute conference talk on incident communication. WHAT I WANTED WHEN IT STARTED I wanted to be the one who told the client, before they heard it secondhand and assumed we'd been hiding it. WHAT WAS ACTUALLY AT STAKE IF IT WENT WRONG A $400k renewal and, more immediately, the client's trust that we'd tell them bad news early rather than late. TARGET LENGTH About 300 words, readable aloud in under two minutes. RULES Find the single moment in the raw anecdote where the outcome was genuinely uncertain — not the moment I've been trained by retelling to treat as the climax, but the actual point where I didn't know what would happen next. Open the story at or just before that moment, not at the beginning of the chronology; readers don't need the setup I think they need, they need the tension. Cut every event from the raw anecdote that didn't change what I wanted or what was at stake — a true story has more events in it than a good story does, and the job here is subtraction. State the want and the stakes early and concretely, in scene, not as a summary sentence explaining why this mattered — show the moment I realized what I stood to lose, don't tell me I realized it. End on the turn itself or just past it, not on a moral or a lesson tacked onto the end; if the story is well-built the point should already be obvious without being stated. WHAT NOT TO DO Do not add invented dialogue, invented people, or invented details I didn't give you — flag any gap you had to guess-fill instead of silently inventing texture to make it read more vividly. Do not soften the stakes to make me look better in hindsight; the version where I was genuinely unsure and it could have gone badly is the version with actual tension. OUTPUT FORMAT 1. The restructured story, at the target length. 2. A one-line note on which moment you chose as the true opening and why the earlier chronology didn't need to be there. 3. A list of any details you flagged as gaps rather than invented.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
GPT-5.1's default move on a raw, chronological anecdote is to summarize it faithfully in order, because that's the safest way to preserve the facts it was given — which produces a report, not a story, since a report and a story differ mainly in where they start and what they cut. Naming the true moment of uncertainty and forcing the model to open there, rather than at the chronological beginning, works against its instinct to front-load context the way a status update would; most people's retelling of their own story has already been smoothed by hindsight into starting too early and explaining too much, so the instruction has to explicitly override the source material's own framing, not just follow it. Making stakes and want concrete and in-scene rather than summarized matters because the model will otherwise produce a sentence like 'I knew this mattered a lot,' which does the reporting work of a stakes statement without doing the story work of making a reader feel it — the difference only shows up when you require it to happen inside a moment rather than as connective narration. The explicit instruction against inventing dialogue or people addresses a specific and common failure mode: asked to make a true anecdote 'read more vividly,' models reliably add plausible-sounding texture — a remembered quote, a described reaction — that wasn't in the source, which is fine for pure fiction but turns a true story into an embellished one the moment it's presented as fact, so the prompt has to make fabrication a flagged exception rather than a silent default.
What you get back
The Slack notification came in at 4:47 on a Tuesday: our integration partner was shutting down their API in ten days. I read it twice before I understood what it meant for the renewal call I had booked for Thursday. The only real decision was whether the client heard it from me first, or from someone else's press release...
Verified against
ChatGPT GPT-5.1 · 2026-08-08
Changelog
- 2026-08-08 — 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
