refactor(rules): probe the still-warning rules, downgrade two - #8
Merged
Conversation
Most of the catalog reached warning before the info tier meant anything -- before no-x-no-y-frag and not-x-but-y split off the ambiguous half of their patterns, warning was just "not that one." Now that the split has real weight, the 21 rules still at warning deserved the same scrutiny those two got: probed against a 1.06M-word corpus of the same five authors (Austen, Melville, Madison, Thoreau, Emerson) used to justify prior rationale. 19 held up clean -- zero hits, or hits that were genuinely the described cadence occurring rarely in careful prose (exactly-the, rich-tapestry), consistent with no-x-no-y and em-dash-overuse, which went through this already and stay untouched here. Two didn't: `is-real-and-not` requires nothing about what follows the conjunction, so "is real, and/but/not" fires on any predicate-adjective sentence regardless of what comes next. Emerson's "my debt to my senses is real and constant" isn't a both-sidesing move, just two adjectives -- indistinguishable from the tell on surface form alone. Moves to info. `worth-naming` collapses the AI meta-signpost with the plain sense of a thing worth mentioning. Emerson's "the only thing worth naming to do that" is the latter. Moves to info. `puffery-words`' bare `nestled` turned out to be the literal verb (a head nestled against a shoulder) as often as the puffery sense in the corpus. Narrowed to require a following in/among/between, the shape the cliché actually takes -- severity unchanged, this one was a pattern bug, not a confidence question. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KKRi7mTZKtDeo47FnYMP3t
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.
Summary
Other follow-up from #6: 21 of 27 rules were still
warning, most of them shipped before theinfotier meant anything — beforeno-x-no-y-fragandnot-x-but-ysplit off the ambiguous half of their patterns,warningwas just "not that one." Now that the split has real weight, the rest of the catalog deserved the same scrutiny those two got.What changed
Probed all 21 still-
warningrules against a 1.06M-word corpus of the same five authors used to justify prior rationale (Austen, Melville, Madison, Thoreau, Emerson).19 held up clean — zero hits, or hits that were genuinely the described cadence occurring rarely in careful prose (
exactly-the,rich-tapestry), the same shape asno-x-no-yandem-dash-overuse, which already went through this in earlier PRs and stay untouched here.Two didn't:
is-real-and-notrequires nothing about what follows the conjunction, so "is real, and/but/not" fires on any predicate-adjective sentence regardless of what comes next. Emerson's "my debt to my senses is real and constant" isn't a both-sidesing move, just two adjectives — indistinguishable from the tell on surface form alone. Moves toinfo.worth-namingcollapses the AI meta-signpost with the plain sense of a thing worth mentioning. Emerson's "the only thing worth naming to do that" is the latter. Moves toinfo.Along the way, found a real bug rather than a confidence question:
puffery-words' barenestledwas matching the literal verb (a head nestled against a shoulder, a kitten nestled into a blanket) as often as the puffery sense. Narrowed to require a following in/among/between, the shape the travel-brochure cliché actually takes. Severity unchanged.Testing
rspec— 229 examples, 0 failures. Every corpus quote is public domain and named in a comment perdocs/SPEC.md's provenance rule; no corpus itself is committed.🤖 Generated with Claude Code
https://claude.ai/code/session_01KKRi7mTZKtDeo47FnYMP3t