X Long-Form Post Writer

What it does

Takes a topic and a contrarian angle and produces a finished long-form X (Twitter) post or thread in an authentic founder/CEO voice. The output is not free-form; it sits on a fixed narrative spine: Hook (1-2 lines, contrarian claim or surprising stat) → Setup (2-3 lines of fast credibility) → Sections (each running problem → what actually happened → lesson/fix) → ASCII diagram (at least one, inside a code block) → Uncomfortable truth (the insight most people avoid) → Payoff (was it worth it? yes, and here's why).

Voice rules are strict: short declarative sentences, short paragraphs, no corporate speak, no emoji in body text, no "I'm excited to share." Numbers must be real — fabricated metrics are banned; if there is no real number, it writes without numbers.

A second engine is an AI-humanizer pass: before release, the draft runs a 24-pattern checklist. It hunts 45+ banned words (delve, leverage, robust, seamless, tapestry, pivotal, underscore...), "this is not X, it's Y" negative parallelism (the #1 AI-slop tell), em-dash overuse (max 1 per 200 words), forced rule-of-three, Title Case headings and emoji decoration. It ends with a 0-100 human score; 90+ ship it, 50-69 significant rewrite, 0-49 full rewrite.

When to use it

For founder-led content, thought-leadership threads, turning a debugging or decision story into a viral lesson, a post-launch "what we learned" post, or defending a contrarian view with proof. Especially when you have a real metric or incident and want to separate it from generic LinkedIn-style content.

Method / frameworks

• 6-part narrative spine: Hook → Setup → Problem/What-happened/Lesson → ASCII → Uncomfortable Truth → Payoff.
• Mandatory ASCII diagram: system architecture, before/after, flow, hierarchy or block-character (█▓░) metric chart — under 40 chars wide, parseable in 3 seconds.
• 24-pattern AI-humanizer rubric: banned-word list + structural pattern hunt + 0-100 score.
• Analytics feedback loop: pull similar-post performance before drafting, prefer hooks/structures that have worked; measure against a baseline cohort after publishing, patch the formula only when a candidate wins.
• Thread logic: >1500 chars → split into standalone-valuable numbered tweets (2026 sweet-spot is 7-12 tweets).