Engineer-to-Engineer Voice, with a Banlist

What this means for you

STRATT copy reads like internal documentation, not marketing. Specs, diffs, exit codes, pinned versions. Errors quote their failure-mode code rather than hiding it. Numbers are exact. Claims are demonstrated inline rather than asserted with adjectives. The emotional register is competence-without-strain; restraint signals confidence, loudness signals insecurity.

The pitch

A short banlist and a short reference list are enough. Forbidden tokens: em dashes, “delve”, “robust”, “leverage”, “navigate the complexities of”, and any superlative that softens a real number. Reference brands: Stripe Docs, Tailscale, Linear, Sigstore. Anti-references: ChatGPT, Cursor, Notion, AWS Console. Run the scene-sentence test before color or layout.

Who it’s for

The technical writer, designer, or founder drafting a paragraph for stratt.run, stratt.works, or the documentation site, and asking whether it reads as authored by an engineer or by an AI summary.

Proof points

• PRODUCT.md Principle 4 names the banned tokens explicitly; the rule is enforced in copy review, not by hope
• Errors expose their failure-mode code (FM-01 through FM-09) on the surface, not in a collapsed stack trace
• stratt.run is a static-prerendered six-zone document that mirrors the audit-viewer layout; the page IS a STRATT audit slice of its own deploy
• Override justifications under 20 characters throw OverrideRejectedError at the orchestrator boundary; the threshold is in code, not a style guide

mindmap
  root((Voice))
    Write like
      Stripe Docs
      Tailscale
      Linear
      Sigstore
    Refuse to sound like
      ChatGPT
      Cursor
      Notion
      AWS Console
    Banlist
      em dash
      delve
      robust
      leverage
      navigate the complexities of
    Show inline
      pinned versions
      exact numbers
      failure-mode codes