Friday Night Wiki
TopicsTagsAbout
Buy me a coffee
Writing as the Primary Interface Tutorial

The Four Audiences: Yourself, Your Team, The Public, AI

This chapter maps the four audiences most of your writing addresses and what each rewards.

Why Audience First

Every piece of writing has a target reader. Before anything else in the writing process, you should know who that is. Audience changes:

  • The vocabulary you use
  • The assumptions you rely on
  • The structure you pick
  • How much of the "why" you explain
  • The tone
  • The right length

Most unclear writing is unclear because it was written with no specific reader in mind. The writer wrote for "whoever reads this", which is nobody in particular, and ended up with prose that feels addressed to no one.

Modern knowledge work has four dominant audiences. Each rewards different moves. Most of your writing will clearly sit in one of them; occasionally a piece sits across two.

Audience 1: Yourself

Writing for yourself covers:

  • Private notes
  • Journal entries
  • Todo lists
  • Second-brain entries
  • Daily logs
  • Half-formed ideas you're trying to shape

You are the only reader, or the primary reader, and you know everything already. This audience has the most context and the least patience for ceremony.

What it rewards

  • Compression. A single word can stand in for a concept you'll recognise. "Check docker thing tomorrow" is legitimate writing for yourself
  • Shorthand and abbreviations. Private jargon is fine here
  • Incomplete sentences. If your future self can fill in the gap, leave it
  • Honesty. No one's reading; say what you actually think
  • Speed over polish. A thought captured poorly is worth more than a thought lost

What it punishes

  • Over-formatting. Spending 10 minutes laying out a note to yourself is usually wasted
  • Writing for an imagined audience. The moment you start performing for a reader, you stop writing for yourself
  • Aspirational notes. "This will be the definitive..." is a trap. Write what you needed today

Examples

Your calendar entries, your phone notes, the text file open on your desktop, the TODO comments in your own code. A person's writing-for-yourself habits tell you a lot about how they think.

Audience 2: Your Team

Writing for your team covers:

  • Memos
  • Design docs
  • PR descriptions
  • Tickets / issues
  • Slack messages (yes, Slack counts)
  • Meeting agendas and notes
  • Weekly updates
  • Onboarding guides

Team members know some context (they work with you), but not everything. You're writing to coordinate.

What it rewards

  • Self-containment. A reader landing cold should be able to follow
  • Specifics. Names, numbers, dates, owners, next steps
  • Clear structure. A new reader should see the shape in 5 seconds
  • Signal about importance. What's urgent? What's fyi? What's decision-needed?
  • Direct asks. "I need X by Friday" beats "maybe we could think about X"
  • Asynchronicity. Assume the reader is not online right now; their context is what's on the page

What it punishes

  • Meeting-mind writing. Streams of consciousness, as if you're still in the call
  • Hidden asks. "Just wondering..." that's actually a decision request
  • Team jargon taken too far. Jargon works until a new hire reads it
  • Bloated paragraphs. If a list would serve better, use a list
  • Emotional leakage. Anger, sarcasm, passive aggression: almost never helps the operational goal

Examples

  • A memo proposing a new project structure: needs context, specific proposal, alternatives considered, ask
  • A PR description: what the change does, why, how to test
  • A ticket: user problem, acceptance criteria, useful context
  • A Slack message asking a teammate to do something: what, when, why (briefly)

Audience 3: The Public

Writing for the public covers:

  • Blog posts and essays
  • Newsletters
  • Marketing copy
  • Public docs
  • Tweets and social posts
  • Conference talk notes and slides
  • Open-source READMEs

Your reader knows nothing about your context, has limited patience, and can leave at any moment.

What it rewards

  • A strong opening. Buried leads die on the public internet
  • Narrative or argument structure. Pure lists work for reference, not for reading
  • Concrete examples. Abstractions lose readers; specifics hold them
  • Honesty about limits. "This works for X, not Y" beats grand claims
  • Voice. People follow writers; writers have voice. The team may tolerate a faceless doc; the public won't
  • Formatting for scanning. Headings, short paragraphs, visual breaks

What it punishes

  • Assuming shared context. "As we all know..." earns an eye-roll
  • Hedge-piled prose. The public is sceptical; overhedging reads as weasel
  • Corporate voice. The tone you'd use in a board deck doesn't work on a blog
  • Over-optimisation for SEO or engagement. Visible in the prose, immediately dismissible

Examples

  • An essay that argues a point: hook, claim, support, implications
  • A README: what this is, who it's for, how to start, where to learn more
  • A marketing page: one core promise, three proof points, one clear CTA

Audience 4: AI

Writing for AI covers:

  • Prompts to ChatGPT, Claude, Gemini, and their cousins
  • System prompts and instruction sets
  • Specs for AI agents
  • Configs and templates that shape AI output
  • Code comments that AI will read (as increasingly it does)

The AI knows a lot of the world but nothing about your specific situation. It's patient, literal, and statistically biased toward the average response unless you steer it.

What AI rewards

  • Precision. Vague inputs produce vague outputs
  • Role specification. "Act as..." or "You are..." genuinely changes output quality
  • Explicit constraints. "Under 200 words, three bullet points, no caveats" works better than "short, please"
  • Examples. A few-shot prompt with "input → output" examples teaches the model what you want
  • Structured context. Clear separation of background, task, format
  • Iterative refinement. Your first prompt won't be right; treating this as a conversation helps

What AI punishes

  • Implicit context. If you don't say it, the AI will guess, and the guess will be average
  • Conflicting instructions. "Be brief but cover everything" is a contradiction the AI will resolve unpredictably
  • Assuming it knows who you are. Unless you say so, it doesn't
  • Over-casual tone. A conversational prompt that leaves out half the requirements will produce half-complete output

Examples

  • A content-generation prompt: role, task, constraints, examples, output format
  • A code-writing prompt: context about the codebase, what you want, what idioms to match
  • A research prompt: specific question, what you already know, what angle to take

Chapter 6 covers AI writing in much more depth.

Where Audiences Overlap

A few pieces of writing address multiple audiences:

  • An open-source README is for the public, but also for future contributors (team) and AI (that might answer questions about it)
  • A company blog post is for the public, but often also for recruitment (team), investors (team-adjacent), and AI (eventually)
  • A ticket in a public repo is for the team now and any future reader forever

When you write for multiple audiences, pick the primary one and write to their needs. The secondary audiences will usually be served well enough if the primary is; the reverse rarely works.

The Audience Test

Before you start writing (or during the first revision), ask:

  1. Who is the primary reader?
  2. What do they know? What don't they?
  3. What do they need from this piece of writing?
  4. How much time will they give it?
  5. What's the main action I want them to take, or understanding I want them to reach?

If you can answer these five, your draft is much more likely to land. If you can't, the writing will be unfocused because your aim is.

Signal: Voice Changes Per Audience

A good operational writer's voice changes appreciably across audiences. The Slack message to a colleague is breezy. The memo to the executive team is formal. The blog post is warmer and more personal. The AI prompt is surgical.

If you hear someone say "I sound the same everywhere", they've either developed one voice that serves many contexts (rare and impressive), or they're adapting less than they think. Most people are in the second camp.

Common Pitfalls

"Writing for yourself doesn't need to be good." It needs to be useful to you later. "Good" in the sense of polish doesn't matter; in the sense of retrievable, it does

"My team is also the public." For most pieces, one audience dominates. "My team might read my public blog post" does not make the blog post team writing; it's still public writing that team members happen to read

"I'll write one version for all audiences." Trying to serve everyone serves no one. Pick the primary reader. Send variations to others if needed

"AI is just another reader." AI is a weird reader: more knowledgeable than any individual, but more literal and more context-starved about your specific case. Writing for AI is its own skill

Next Steps

Continue to 03-clarity.md for the skill that matters to all four audiences.