The prompt
Copy it whole, replace every bracketed variable, then paste it as your first message.
You are my drafting partner for short posts. I direct, you propose, I judge, I own what gets posted. Read everything before you reply.
## 1. The post
- Platform and limits: [PLATFORM AND LIMITS]
- Who I am to my readers: [WHO YOU ARE TO YOUR READERS]
- The observation, in my words: [THE OBSERVATION]
- The moment it came from: [THE MOMENT]
- Batch, if any: [BATCH]
## 2. My register
Two past posts I was happy with, or my voice profile. Match rhythm, paragraph length, and directness. Do not reuse their phrases.
[YOUR REGISTER]
## 3. Rules
- The post delivers the observation through the moment. The moment is real and stays as I described it; do not add people, dialogue, numbers, or drama I did not give you.
- Write in paragraphs of two to four sentences. Not one sentence per line. Not a single block either.
- The first line is a plain sentence about the moment or the observation. It is not a hook, a question, a bold claim, or a cliffhanger.
- No numbered lessons, no "3 things I learned", no "Here's what nobody tells you", no "Unpopular opinion", no "Let that sink in", no "Agree?", no "Thoughts?", no "Repost if".
- No closing question to drive comments. End on the observation or on the moment. If the post has a point, the reader gets it without being told.
- Do not universalize ("We all...", "Every leader..."). It happened to me; keep it mine.
- No hashtags unless the platform limits say so. No emojis unless my past posts use them.
- Do not invent facts, names, figures, or quotes. Mark anything missing with [VERIFY].
Banned anywhere: [BANNED LIST], "game-changer", "journey", "grateful", "humbled", "thrilled to announce", "I'm excited to share", "lessons learned", "takeaway", "mindset", "hot take", "the truth is", "most people", rhetorical questions, lines that are just one word.
## 4. Batch rules (only if a batch is given)
- Each post has a different shape: one opens on the moment, one opens on the observation, one opens on what I got wrong. No two posts have the same paragraph count.
- No shared phrases across posts. No shared ending move.
- Vary length within the limits.
- Label each post with its observation so I can match them.
## 5. How we work
Do not ask questions unless the moment is unclear; then ask one. Draft directly.
When I say "change only X": return only the changed sentences marked OLD and NEW.
When I say "options for the first line": three plain first lines, no hooks, labeled by what each leads with.
When I say "tells": do not rewrite. List every sentence that sounds generated or platform-shaped and say why.
When I say "shorter": cut whole sentences, list what you cut.
## 6. Begin
Draft the post, or the batch. Nothing else.- Best for
- LinkedIn posts from professionals who have actual experience to share
- Short-form posts on any text platform where you post under your own name
- Turning a conversation, a meeting, or a small failure into one honest post
- Batching a week of posts without the batch looking like a batch
- Not for
- Engagement farming, fake stories, or posts about events that did not happen to you. The prompt forbids inventing the moment
- Posting on behalf of a brand or executive without disclosure to the audience, where the platform or your context expects it
- Replies and comments. Write those yourself. They are where voice is tested most
Social platforms are where generated writing is most visible, because the tells are so consistent: the one-line hook, the line break after every sentence, the 'Here's what I learned:', the numbered lessons, the closing question to drive engagement. Readers have learned to scroll past the shape before reading a word.
This prompt treats a post the way the Newsletter prompt treats an issue: as the delivery of one observation you had. It asks for the observation and the moment it came from, in your words. It bans the platform formulas by name. It drafts in paragraphs, in your register, and when you want a batch, it enforces variation in shape so the week's posts do not look like siblings.
It will not make a post go viral. It will make it sound like a person, which on a feed full of templates is now the differentiator.
Variables
What to put in each bracket. Required ones are marked. Thin answers here are the usual cause of generic output.
[PLATFORM AND LIMITS]RequiredWhere it will be posted and any length constraints.
ExampleLinkedIn, under 180 words, no hashtags
[THE OBSERVATION]RequiredThe one thing this post says, in your rough words.
ExampleThe best engineer I ever managed asked more clarifying questions than anyone, and I spent the first year thinking that was a weakness
[THE MOMENT]RequiredThe specific thing that happened that made you notice. A conversation, a meeting, a message. Real, and told in your words.
ExampleA design review last Tuesday where she asked 'who is this for' four times and the fourth time nobody could answer
[WHO YOU ARE TO YOUR READERS]RequiredHow your followers know you and what they expect from you.
ExampleAn engineering manager who posts about management mistakes, mostly my own
[YOUR REGISTER]RequiredTwo past posts you were happy with, pasted in full, or a Voice Capture profile.
ExamplePaste two posts
[BANNED LIST]Things you do not want, beyond the defaults.
ExampleAny mention of 'leadership', emojis, the word 'unpopular'
[BATCH]If drafting several posts: the number, and one observation and moment per post. Leave empty for a single post.
Example3 posts: (1) the clarifying questions; (2) the time I skipped a one-on-one and what it cost; (3) the retro where nobody spoke first
Why it is built this way
Each block does one job. Knowing the job lets you shorten or adapt the prompt without breaking it.
Observation plus moment
The same engine as the Newsletter prompt, scaled down. The moment is the evidence, and the observation is the point. A post with only the point is advice, and advice on social reads as generated even when it is not.
Paragraphs of two to four sentences
One rule that defeats the most recognizable tell on LinkedIn, the line-break-after-every-sentence shape. Readers associate that shape with templates now.
A plain first line
Hook formulas are the second most recognizable tell. A plain sentence about what happened is, counterintuitively, what stops the scroll on a feed full of hooks.
No closing question, no universalizing
Engagement bait and 'we all' endings are the third tell, and both are the model's default ending. Ending on the moment keeps the post yours.
The banned vocabulary
Short and specific to the platform: 'humbled', 'thrilled to announce', 'let that sink in'. Each one is a word readers have learned to skip.
Batch variation rules
Drafting several posts at once produces siblings: same opening move, same paragraph count, same ending. Forcing different shapes and no shared phrases makes a batch look like a week of a person posting.
The 'tells' move
A self-audit that works better than you would expect, because the tells are consistent and the model can see them when asked to look rather than write.
Example run
A representative first response, with the input summarized. Your results will differ by model and material.
Filled for the clarifying-questions observation: an engineering manager, LinkedIn under 180 words, the design review where 'who is this for' was asked four times, and two past posts for register.
In a design review last Tuesday, the engineer I used to think asked too many questions asked "who is this for" four times. The fourth time, nobody in the room could answer, including me. For most of her first year I read the questions as hesitation. She was slower to start building than the people around her, and I measured that. What I did not measure was how often the people who started fast built the wrong thing. The review ended with the feature cut. It would have taken two sprints. She had spent four minutes.
No hook, no lesson list, no closing question, and the point lands without being stated. The writer changed 'including me' to 'me included' to match how she talks and cut 'For most of' to 'Most of'. The post is 118 words. Compare it to the shape the platform rewards and ask which one you would stop for.
Follow-up prompts
The second turn is where vibe writing happens. These are the short messages that keep the session moving.
Before posting
Tells.
When the first line is doing too much
Options for the first line.
When the post explains its own point
Cut the last paragraph. Return the post ending on the sentence before it.
When a batch looks like siblings
Posts 1 and 3 open the same way and end the same way. Change only post 3: open on what I got wrong, end on the moment, different paragraph count. Return post 3 only.
When you want to reply to a comment in your voice
Do not draft a reply. Here is the comment: [COMMENT]. Tell me in two lines what they are actually asking or asserting, and whether it deserves a reply at all. I will write it.
When it goes wrong
Symptoms you will actually see, the usual cause, and the correction that works.
The line breaks are back.
- Cause
- The platform shape is strongly associated with LinkedIn in the model's training. It returns after a few turns.
- Fix
- Paste your own post as 'mine' and say 'this paragraph shape'. Keep a master copy and re-paste it with each edit request.
The moment gained dialogue or a dramatic beat you did not supply.
- Cause
- The model made the story better. That is invention.
- Fix
- Cut it. Reply: 'The moment is exactly as I gave it. Remove anything I did not say happened.' On social, an embellished story is one screenshot away from being a reputation problem.
The post ends with 'What would you have done?'
- Cause
- Engagement endings are the default.
- Fix
- Cut it by hand, lock the ending with 'keep', and add 'no final question of any kind' to the banned list.
It sounds like a thoughtful stranger, not like you.
- Cause
- Register samples were too short or too polished.
- Fix
- Paste the post of yours that got the most replies from people who know you, even if it is rough. That is your register.
Before you publish
The prompt cannot do these. You can.
- The moment happened as written. Nothing was added for effect.
- Anyone mentioned or implied would be fine with the post. If the moment involves a colleague, consider whether they would recognize themselves and whether that is okay.
- Read the first line alone. If it is a hook, rewrite it as a plain sentence.
- Check your organization's social media policy if you post about work. Some require disclaimers, and some restrict posting about internal matters.
- If you are posting as someone else, or using AI to generate a large share of a professional presence, think about what your readers believe they are getting. The ethics lesson has a framework.
Questions
Won't a post without a hook get less reach?
Possibly, on a given day. Over months, posts that read as a person tend to earn the replies and reposts that reach depends on, while hook-shaped posts earn scrolls. Reach is also not the only cost: a feed of formula posts under your name changes what people think you are.
Should I disclose that AI helped draft a post?
If the post is substantially your observation and your moment, drafted and then rewritten, most people do not disclose and most readers would not expect it. If the assistant is generating the substance, disclosure is honest and increasingly expected. The ethics lesson has a decision framework.
Can I use it for X, Threads, or Bluesky?
Yes. Set the platform and limits, and the paragraph rule relaxes naturally for very short posts. The bans on hooks, universalizing, and engagement endings apply everywhere.
How many posts can I batch?
Three to five. Beyond that, the variation rules get hard to satisfy and the observations get thin. If you have ten observations, that is two sessions, or more likely six real observations and four you should drop.