Skip to content

Add pattern #30: debunking-pose headings#116

Open
AnthonyDavidAdams wants to merge 1 commit intoblader:mainfrom
AnthonyDavidAdams:add-debunking-pose-headings
Open

Add pattern #30: debunking-pose headings#116
AnthonyDavidAdams wants to merge 1 commit intoblader:mainfrom
AnthonyDavidAdams:add-debunking-pose-headings

Conversation

@AnthonyDavidAdams
Copy link
Copy Markdown

Summary

Adds a 30th pattern to the humanizer skill: debunking-pose headings.

The existing 29 patterns target prose. Headings often slip through editing passes because editors (human and AI) treat them as structure rather than copy. A common AI tell is a heading that poses against an implied wrong version of the topic: "actually," "the real," "that lands," "the truth about," "the long game," "demystified." The pose announces everyone else has it slightly wrong, but I'll give you the truth — without delivering anything different than a plain heading would.

This is pattern #27 (Persuasive Authority Tropes) operating at the heading level, and a sibling of pattern #1 (Significance Inflation): the heading promises a reveal it cannot keep.

I discovered this when I ran the skill across a 24-chapter book in parallel. The agents cleaned the prose beautifully but left every "actually / the real / that lands / the long game" title intact. Two reasons: (a) the skill's 29 patterns are written for prose with prose examples, and (b) editors instinctively leave headings alone. So this pose lives in a gap between the two.

Changes

  • SKILL.md: New ### 30. Debunking-Pose Headings section after pattern Feature: SOTA Tiered Architecture for AI Detection Skill #29 with phrases-to-watch, problem statement, and before/after heading examples. Asks editors to audit headings as a separate pass.
  • README.md: New row in the Style Patterns table, pattern count updated 29 → 30, version history entry for 2.6.0.
  • SKILL.md frontmatter: version: 2.5.1 → 2.6.0, description updated to include "debunking-pose headings."

Test plan

  • Read the new pattern feat(skill): implement tiered architecture (v3.0.0) #30 entry — does the before/after capture the tell?
  • Spot-check whether the phrase list is the right set (suggestions welcome — I leaned on what surfaced in the book I tested it on)
  • Confirm version bump is appropriate (minor: additive pattern, no breaking change)

Happy to revise if any of the wording or framing doesn't fit the project's voice.

The existing 29 patterns target prose. Headings often slip through human
and automated editing passes because editors treat them as structure
rather than copy.

A common AI tell is a heading that poses against an implied wrong
version of the topic: "actually," "the real," "that lands," "the truth
about," "the long game," "demystified." The pose announces *everyone
else has it slightly wrong, but I'll give you the truth* without
delivering anything different than a plain heading would.

This is pattern blader#27 (Persuasive Authority Tropes) operating at the
heading level, and a sibling of pattern blader#1 (Significance Inflation):
the heading promises a reveal it cannot keep.

Pattern blader#30 asks editors to audit headings explicitly as a separate
pass, lists the typical pose phrases to look for, and provides
before/after examples drawn from real chapter titles.

Bumps version 2.5.1 → 2.6.0.
Updates README pattern count 29 → 30 and adds row 30 to the Style
Patterns table.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant