Skip to content

docs(C1): a capitals-settled run opened by a class word reports - #570

Merged
derek73 merged 5 commits into
masterfrom
docs/c1-silence-sentence
Oct 1, 2026
Merged

derek73 merged 5 commits into
masterfrom
docs/c1-silence-sentence

Conversation

@derek73

@derek73 derek73 commented Oct 1, 2026 •

Copy link
Copy Markdown
Owner

Follow-up from #569's review. rules.md#C1 said a comma credential run whose every ambiguous-class word is in capitals "reads whole in silence". That's only true when another credential opens the run.

Input Reads Reports
John Smith, PhD MA suffix PhD MA nothing
Smith, PhD MA suffix PhD MA nothing
Doe, MA PhD suffix MA PhD suffix-or-name on MA
John Smith, MA MA suffix MA MA suffix-or-name on MA

A class word opening the part is the first word after the comma, and C1 reports that word's decision either way. You confirmed the report is wanted, since nobody repeats MA at the end of their name on purpose, so this PR corrects the statement and leaves the behavior alone.

Changes

  • rules.md#C1: the sentence now names the exception, and there's a new example line "Doe, MA PhD" → ambiguities=("suffix-or-name",). That name was already in the contract corpus, so the gate and ledgers don't move.

  • decisions.md: a dated C1 bullet with the measurement and a recompute recipe. Over every 2–3-word all-caps run behind John Smith, and Smith, , all 900 runs opening with a class word report. 554 of the 560 opening with another credential are silent, and the other six are Why does John Smith, PhD DO DO read given 'PhD' and family 'DO DO John Smith'? #562 particle pairs.

  • release_log.rst: the Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544 bullet repeated the old claim. It now says an all-capitals run is silent when another credential opens it, and an opening acronym reports on itself (found in review).

  • rules.md#S2: its first-slot precedence sentence ("the count decides there first") now points to C1's exception. A run of two or more words, settled on its capitals, keeps the family comma and reports only an opening class word. John Smith, MA MA reports on the first MA alone, while one word (John Smith, MA) still flips by the count. This is prose only; the case row comma_run_the_capitals_already_settle_is_not_flipped already pins the behavior.

  • rules.md#S2 and the SUFFIX_OR_NAME docstring: both listed "the segments beyond" a second comma as a slot that reports. No suffix-or-name report lands there: Smith, John, MA and John Smith, Jr., MA are silent. Both docs now say a part past the second comma is no reporting slot and leave the rest to C2, because that part can still raise comma-structure or unbalanced-delimiter. S2 gains C2 in its interacts: list.

Decided in review (Derek, 2026-10-01), no change:

Review: the design-docs review ran all eleven axes on the first commit, and again on the three that follow. The second round found two errors in that round's own wording, both fixed in the last commit:

The new S2 sentence held over about 98k parses, and the past-the-second-comma claim held over about 731k.

Checks: no parse moves. The full suite passes (10,613), and the differential gate exits 0 at all five baselines.

🤖 Generated with Claude Code

rules.md#C1 said a run whose every class word is written in capitals
"reads whole in silence". That holds only behind another credential:
a class word opening the part is the first word after the comma, whose
decision C1 reports either way, so 'Doe, MA PhD' and 'John Smith, MA
MA' report suffix-or-name on that word. The behavior is kept (Derek,
2026-10-01) and the statement corrected, with 'Doe, MA PhD', already a
contract-tier corpus name, as its example. No parse moves.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@derek73 derek73 added this to the 2.4 milestone Oct 1, 2026
@derek73 derek73 added the docs Documentation fixes and updates label Oct 1, 2026
@derek73 derek73 self-assigned this Oct 1, 2026
@codecov

codecov Bot commented Oct 1, 2026 •

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.87%. Comparing base (b39c370) to head (b34a34e).
⚠️ Report is 8 commits behind head on master.

Additional details and impacted files
@@           Coverage Diff           @@
##           master     #570   +/-   ##
=======================================
  Coverage   98.87%   98.87%           
=======================================
  Files          45       45           
  Lines        3990     3990           
=======================================
  Hits         3945     3945           
  Misses         45       45           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

derek73 and others added 4 commits October 1, 2026 12:11
… recipe

The #544 release bullet still said a capitals-settled run reports
nothing; it now says so only when another credential opens it, and an
opening acronym reports on itself. The decisions bullet's grid is drawn
with repetition, and 14 of its reporting runs are #562 particle pairs
whose flip reports over the whole part.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
S2 said the count decides the first slot after a family comma before
case. That holds for one word ('John Smith, MA' flips by the count)
but not for a run C1's shortcut settles on its capitals, which keeps
the family comma and reports only an opening class word ('John Smith,
MA MA' reports on the first 'MA' alone). Prose only, so the next
reader does not rediscover it (Derek, 2026-10-01).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
S2's list of slots where the ambiguous credential class reports ended
with 'and the segments beyond it', and SUFFIX_OR_NAME's docstring said
the same. No suffix-or-name report lands past the second comma: 'Smith,
John, MA' and 'John Smith, Jr., MA' are silent, and such a part reports
only C2's structural flag. Both now say so and point at C2; S2 gains
C2 in interacts:. The silence is kept (Derek, 2026-10-01).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Second review round: 'reports only COMMA_STRUCTURE' was false -- an
unclosed delimiter in a tail adds unbalanced-delimiter ('John Smith,
Jr., (Bob') and a maiden clause standing in one is read as one. S2 and
the SUFFIX_OR_NAME docstring now say only that the tail is no reporting
slot and leave the rest to C2. The decisions bullet's history is
corrected too: at #530's merge a tail could still report suffix-or-name
('John Smith, Jr., PhD Do Ma'), silenced by the 2026-09-28 #544 bullet.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@derek73
derek73 merged commit fc682e3 into master Oct 1, 2026
9 checks passed
@derek73
derek73 deleted the docs/c1-silence-sentence branch October 1, 2026 19:58
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

docs Documentation fixes and updates

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant