write with ai 5 ai writing prompts workflow

Why this is in the vault

Five reusable AI-prompt patterns for writing production (idea generation, tone-rewrite, hook analysis, quick research, sequential-pass editing) — kept for the one technique with real production reuse, not as a wholesale workflow adoption.

The core argument

The issue frames "10x writing efficiency" as five discrete prompt patterns to run against ChatGPT/Claude/Perplexity, each solving one stage of the writing pipeline:

1. Idea generator — a two-step mega-prompt: AI first proposes 10 verb-led, outcome-focused subtopics; after confirmation, it generates exactly 3 numbered, quantifiable sub-questions per subtopic ("3 tips for X," "5 steps for Y"). The claimed insight is that forcing numbered-list framing produces more concrete ideas than open-ended prompting.
2. Writing upgrader — seven one-line tone-swap rewrite prompts (persuasive, informative, vivid, comedic, urgent, emotional, concise) run against an existing draft to quickly test alternate voices.
3. Hook analyzer — reverse-engineers a high-performing piece of text: explain why the hook works in ≤35 words, convert it into a mad-lib template with goal-oriented placeholders (e.g. {Desired Outcome}), then propose one application.
4. Research assistant — a minimal single-line Perplexity pattern ("top lessons/tips/frameworks from [expert/book] about [topic]") for fast synthesized research instead of manual digging.
5. Editor — a copyeditor-persona prompt that runs three sequential, confirmed passes: structure first, then content specificity/POV, then grammar — rather than one flat "fix this" pass.

The issue also name-drops a "2-Year Test" and a "4A Framework" without defining either, and promotes the author's own "Start Writing Online 5-Day Challenge" and "AI Writing Skool" course.

Mapping against Ray Data Co

Weak-to-medium, and mostly weak: four of the five techniques (idea generator, tone-swap, hook analyzer, research assistant) are generic content-ops prompts, not newsletter-craft-specific, and don't map to anything in the Sanity Check production pipeline as-is. The one piece worth naming is technique 5 — the sequential structure → content/POV → grammar pass order — which is a cleaner decomposition than a flat single-pass edit and could sharpen how /draft-review sequences its own checks (hook strength → tangible-vs-abstract → CopyThat patterns → voice consistency currently run more as one combined pass). Technique 3 (hook-templating from someone else's high-performing hook) brushes directly against the no-derivative-Sanity-Check-pieces rule and should not be adopted for SC hook-writing — sources are evidence, not a template to reskin.

Related

• [[2026-06-07-write-with-ai-ogilvy-writing-rules-claude-skill]]
• [[2026-01-18-write-with-ai-blind-prompt-tests-llm]]
• [[feedback_no_derivative_sanity_check_pieces]]

⚠️ Sponsorship

Self-promotional: the issue's closing CTA promotes the author's own paid "Start Writing Online 5-Day Challenge" and "AI Writing Skool" course/membership. No third-party sponsor. Bias implication: the five techniques are framed to make the free content feel incomplete without the paid follow-on, so treat the "10x efficiency" framing as marketing copy, not a measured claim.