Core concepts

Platforms

Presets for blog, X, threads, LinkedIn and email: what each one changes.

The platform parameter applies the conventions of where the text will be published: length, formatting, openings and endings. It changes form, never facts.

Presets at a glance

platform	Length	Formatting	Notes
`blog`	Source length ±10%	Paragraphs and links kept	Removes summary sections and restated closes unless your voice uses them.
`x`	≤ 280 characters	Plain text, no hashtags by default	Rejects sources that can't fit without dropping a protected fact (422).
`thread`	Posts of ≤ 280 characters	Returned as an array in posts[]	The first post carries the point; no “🧵 1/”-style numbering unless your voice does it.
`linkedin`	≤ 3,000 characters	Short paragraphs, line breaks kept	At most 3 hashtags, only if your samples use them.
`email`	Source length	Subject and preheader preserved	Sign-off taken from the voice, never invented.
`generic`	Source length	Formatting preserved	The default. No platform conventions applied.

LinkedIn launched a “seems like AI slop” report on 30 July 2026 and says flagged posts get about 40% fewer views. The linkedin preset avoids the patterns readers report most: one-line “hook” openers, emoji bullets and a closing question nobody asked for.

Source: LinkedIn, via TechCrunch.

Threads

Threads return an extra posts array alongside text. Fidelity checks run across the whole thread, so a fact can move between posts but not change.

{
  "text": "…",
  "posts": [
    "Most blog posts go wrong before anyone writes a word.",
    "The writer gets a topic and nothing else. Here's what we send instead: …"
  ],
  "fidelity": {
    "status": "pass"
  }
}

Choosing a preset

Use the preset for where the text will be read, and a voice built from samples published there. A blog voice with the x preset works, but a voice built from the person's actual posts works better.