Show, Don't Tell: Every Claim Demonstrated Inline

What this means for you

The voice rule for Choco HQ copy is engineer-to-engineer documentation. If a claim cannot be demonstrated inline (a live widget, a real ledger slice, a token a visitor can paste and verify), the claim gets cut. Adjectives like “auditable” or “verifiable” without a paste-and-verify proof are credibility leaks, not features.

The pitch

Specs, diffs, exit codes, pinned versions. Errors quote their failure-mode code. Numbers are exact. Closer in feel to Stripe Docs, Tailscale, Linear, and Sigstore than to ChatGPT, Cursor, Notion, or AWS Console. Restraint signals confidence; loudness signals insecurity.

Who it’s for

The writer or designer about to type “powerful,” “robust,” “seamless,” or “leverages” into a Choco HQ surface.

Proof points

• Five forbidden tells called out in PRODUCT.md: no em dashes, no “delve,” no “robust,” no “leverage,” no “navigate the complexities of”
• Numbers on the page are real: 14 canonical-serialisation test vectors, 9 failure-mode gates (FM-01 through FM-09), blake3-wasm pinned to 2.1.5 because v3.0.0 broke the API, 856 tests across 6 packages at v1.0.0-rc.1
• Reference voices listed by name: Stripe Docs, Tailscale, Linear, Sigstore. Anti-references listed by name: ChatGPT, Cursor, Notion, AWS Console
• The audit token at /audit/[token] is itself a proof: the reader pastes the token, the page verifies the Ed25519 signature client-side, no server round-trip required

mindmap
  root((Voice rules))
    Show
      live fingerprint widget
      pasteable audit token
      real ledger slice
    Cut
      "powerful"
      "robust"
      "seamless"
      "leverages"
      em-dashes-everywhere
    Pin
      exact numbers
      named versions
      failure-mode codes
    Reference
      Stripe Docs
      Tailscale
      Linear
      Sigstore