feat: AI-iness density pre-check for adaptive pass strength (v2.6.0)#98
Open
adelaidasofia wants to merge 1 commit intoblader:mainfrom
Open
feat: AI-iness density pre-check for adaptive pass strength (v2.6.0)#98adelaidasofia wants to merge 1 commit intoblader:mainfrom
adelaidasofia wants to merge 1 commit intoblader:mainfrom
Conversation
Adds a density pre-check section that counts Tier 1 AI tells per 100 words before rewriting, then selects light/mixed/full pass strength: - Light (<1 tell/100 words): Tier 1 only — preserves human-first voice - Mixed (1–2 tells/100 words): default, Tiers 1–2 at calibrated strength - Full (3+ tells/100 words): all 29 rules — for AI-first text Solves the over-correction problem on journals, rough drafts, and meeting notes where the current all-or-nothing pass strips authentic voice. Tested in production on personal journal corpus (~2,400 entries). Closes blader#93
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this does
Adds a density pre-check that counts Tier 1 AI tells per 100 words before rewriting, then selects pass strength automatically:
Why
The current behavior is all-or-nothing: every rule runs on every text. This works well on AI-first drafts but over-corrects on human-first writing — personal journals, rough drafts, and meeting notes where fragments, first-person voice, and colloquialisms are intentional voice choices, not AI tells.
The density check makes humanizer safe to run on any input without worrying about destroying authentic voice.
What defines Tier 1
For the pre-check, Tier 1 = the dead giveaways: patterns 1 (significance inflation), 3 (-ing phrase pile-up), 4 (promotional language), 7 (AI vocabulary words), 8 (copula avoidance), and 20 (chatbot artifacts). These are the highest-signal, lowest-false-positive tells.
Testing
Tested in production on a personal journal corpus (~2,400 entries over several years). Light pass mode stops the "humanizer made my writing worse" failure mode cleanly on human-first content. The mixed/full threshold of 3 tells/100 words comes from empirical observation — most fully AI-generated paragraphs hit this easily; human-written paragraphs rarely do.
Closes #93