Skip to content

Pick a Voice

Voice is the most consequential choice in a composed instruction. Tone, style, and format adjust how a piece reads; voice determines who is reading it back to you. Get the voice wrong and no amount of tone or format polish will save the artifact.

This guide walks through the choice in three steps.


The voice that fits is the voice the reader expects from the person they think is talking. Before you pick anything, write down two facts:

  • Who is the actual reader? (engineering manager, product leader, customer, congregation)
  • What relationship does the writer have with them? (peer, advisor, leader, teacher)

If you cannot answer both, no voice will reliably fit. Refine the brief before composing.


Each voice in the catalog embodies a recognizable professional archetype. The right archetype is the one a real person in that role would already be using.

Reader and relationshipLikely voice
Senior engineer reading a design reviewpragmatic-architect
Product team reading a feature proposalproduct-thinker
New hire reading an onboarding guidefriendly-mentor or technical-writer
Operations team reading an incident runbookoperator
Board or VP reading a quarterly summaryexecutive
Reader of an opinion column or blogcolumnist
Congregation reading a devotionalpastoral
Practitioner reading API referencetechnical-writer
Coachee reading a development plancoach
Anyone, any context, just needs the pointdirect-communicator

Browse taxonomy/voices/ for the full list. The one_liner field at the top of each ENTRY.md is designed to fit on a card so you can scan it fast.


Once you have a candidate, read the matching example file under examples/vertical-slices/async-standups/voice-<id>.md. The example is the voice rendering the same situation that every other voice also renders, so you can compare the candidate against its neighbors.

Two questions to ask while reading:

  1. Does this voice sound like the person you would want writing to your reader?
  2. Does the reader, in your mental model of them, find this voice credible?

If both answers are yes, you have your voice. If either is no, look at the candidate’s confusable_with field and check the entries listed there. The confusable entries are almost always the better fit when the first choice felt close but off.


When you are unsure between two voices, prefer the more domain-specific one. pragmatic-architect beats direct-communicator when the audience is technical, even though direct-communicator is technically valid. The specific voice carries more useful constraints into the composition.

When the reader is mixed, pick the most demanding reader. A doc that has to work for both engineers and executives should usually be voiced for the executive - engineers can read across, executives often will not.

When the topic is emotionally weighted, voice matters more than usual. pastoral and pragmatic-architect are not interchangeable on a topic like organizational restructuring; the voice signals what kind of conversation this is.

When you are writing to yourself or one peer, you may not need to specify a voice at all. The composition is a control surface; not every axis needs to be filled.


If you find yourself wishing for a voice the catalog does not have, two things to try:

  1. Re-read the closest candidate’s pairs_well_with field. Sometimes the gap is not the voice itself - it is the tone or style pairing.
  2. Propose a new entry. See How to Add an Entry. The catalog is designed to grow.

The Phase 1 target is fifteen voice entries; current count is ten. Real gaps in the catalog become future entries.