[derek73]

Part of #762. docs/design/decisions/suffix-acronym-collisions.md and docs/design/decisions/vocabulary-collisions.md each get a ## Standing block, one present-tense line per decision in force, each pointing at its log entry, plus a ## Log heading. Both keys come off tests/v2/decisions_without_standing.json. These two files go next because #752 (vocabulary phase 2) rewrites the ambiguous marking they record, and it should start from blocks that say what is in force. A one-line head note also lands in decisions/C8 (below).

suffix-acronym-collisions

• Bundle line. The file's title dates a bundle of undated bullets (2026-09-07, #342/#454). So, per the README, the Log opens with one dated line naming the bundle, with the title's words as its verdict.
• mai and esq, which carry their date only in prose. I gave each its own dated index line: 2026-09-08 #316/#489 — ESQ LEFT SUFFIX_ACRONYMS: … and 2026-10-16 #687 — MAI LEFT SUFFIX_ACRONYMS: …. The alternative was to point them at the bundle line, but a pointer carries the decision's date, and 2026-09-07 is the wrong date for both. The added lines follow the bundle convention's own reasoning: each repeats the date its bullet's text already gives, adds no decision and edits no entry. Rewriting the bullets to open with their dates would be rewording an entry's opening, which none of the four edit kinds allows. The bundle line says the two are indexed separately. The README does not name this case yet. If you want it written down, it fits as one clause after the bundle sentence.
• Parts reversed elsewhere, marked under edit kind 1. Each is a parenthesized "superseded" note at the head of the entry it sits in, and the bundle line lists them:
  • Accepted cost: Ahmad Jayadi, CHA now reads suffix, by decisions/S9's 2026-10-01 Should John Smith, XYZ read XYZ as a credential by default, even though John Smith XYZ doesn't? #564.
  • ba: Smith, BA now reads suffix, by decisions/S6's 2026-09-14 (Derek).
  • Roman-numeral accident: retired.
  • Parking lot: closed; its own 2026-09-15 head also notes that John Smith, XYZ moved under Should John Smith, XYZ read XYZ as a credential by default, even though John Smith XYZ doesn't? #564.
  • Not decided here: se and om are now marked, by the 2026-10-21 Mark 33 credential readings ambiguous where a surname reading is substantial #796 entry.
  • esq: the E.S.Q. measurement is overtaken by the dotted shape.
  • Should meng and lac be ambiguous suffix acronyms? wang meng has had no family name since 2.0 and renders Wang MENG on master #540's head: 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 reversed its acceptance of Wang M.Eng. and its run-ending costs.
• Elsewhere: lines replace Standing bullets for decisions recorded under other keys:
  • the case lean (S6)
  • 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 (S7's "A CREDENTIAL RUN KEEPS ITS AMBIGUOUS MEMBERS")
  • the two shape switches (S9)
  • the roman-numeral retirement (S3)
  • sa's dual, which rests on decisions/C8's PR fix(assign,vocab): a one-word family comma reads its credential run as suffixes (#296, #325) #428 review entry. C8's own Should parse("Smith, Jr.") put "Jr." in title rather than suffix? #296/Smith, Ph. D. Jr. puts the credential in first (regression from 1.4.0) #325 head now notes that its "ms and sa into the ambiguous set" was reversed there (edit kind 1, found in review).
• Not standing: the Donald Mc loses its family to a post-nominal reading — the comma-less twin of #380 #454 bullet's "the fix(suffix-routing) ledger rule stays" (the ledgers were retired in Replace the differential ledgers with a committed snapshot of corpus readings #664), and the rules/P6 repair, which was carried out.
• Open: Reports say "read as a credential" for a period-marked title after a comma #793 and Phase 2: derive ambiguity from clashing readings and extend the gate to every reading kind #752.
• Edit kind 2: the Mark 33 credential readings ambiguous where a surname reading is substantial #796 entry's "Recipe as in the "Luo Si" and "Min Chi" lose their family name: Si and Chi read as credentials #791 entry above" is corrected to "below", with a note. This error predates the branch.

vocabulary-collisions

This is a log, not a reference list: it has dated entries and decisions in undated bullets. The keystone's undated bullets (C-i, C-ii, the do demonstration, Applications, the C-ii per-word bullet) were written with the 2026-08-16 keystone entry. git blame on the pre-split decisions.md shows them in f7c18a4, 77de1e6 and 1f5e0b8. They gain a bundle line like the one above, so their pointers land on a line that names them rather than on the base-rate verdict.

Changes in the entries:

• C-i's bullet gains a head note. For suffix acronyms, "iff borne" is superseded by suffix-acronym-collisions' relative-frequency criterion: being borne only opens the question.
• The C-ii per-word bullet gains an edit-kind-2 correction. It said do and mc inherit vd's answer. do was already in the ambiguous suffix half, so the precedence decides nothing for it (rules/P6). mc does inherit, and Donald Mc loses its family to a post-nominal reading — the comma-less twin of #380 #454 accepted that.
• The 2026-08-17 call to re-judge lo/el goes under Not standing:. Mc Donald and Ste Marie read the particle as the given name #360 judged lo ambiguous and left el open, and neither was added to the vocabulary.
• The Open: #348 line moves into the block.

Review

• Two rounds of the design-docs-reviewer: one on the first commit, and one on the first fix commit, as AGENTS.md asks.
• Round 1 found 10 issues. The main ones:
  • a line that ignored S6's case lean
  • C-i stated as an unqualified iff
  • the do inheritance
  • 30 of the 33 Mark 33 credential readings ambiguous where a surname reading is substantial #796 words meeting the threshold, not 33
  • four bullets that belong on Elsewhere: lines
• Round 2 found 9 issues in the first fix's own text. Most were the same scope error repeated in a rewritten line: a case-scoped claim about the ba/meng/si class that failed at the bare and comma sites (Anna BA, SMITH, BA, John Smith, Meng). The final line names measured examples and leaves the readings to S6, S4 and C4 rather than restating them by case.
• Both rounds' findings are fixed. The second fix commit is small and had no third review round. I checked every example it cites by parsing it on the tree.

How it was checked

• Every word a line names was looked up with tools/vocab/query.py, and --issue 796's 33 were checked against Lexicon.default().
• Every example a line cites was parsed on the tree.
• The whole folder was grepped for each issue and example. The reversals were in S9, S6, S7, S3, C8 and P6.
• A script resolved each pointer and printed the entry it lands on, so each one names the entry intended, not just some entry.
• Negative control: deleting the mai index line, the bundle line, or a twin's distinguishing words makes standing_findings fail.
• The full suite passes on 3.11 and 3.15. The 3.15 run was at the first commit; later commits change only docs, and the suite was re-run on 3.11.

Stale, left as is

• The old Open: #348 handle said the census "needs a given-name frequency corpus this repo does not have". tools/name_usage.py now has SSA counts and nam_dict given-name bands, so Should a word that is both a title and a given name report an ambiguity? (Baron Davis → title 'Baron', no given name) #348's stated blocker may be out of date. The block's handle no longer repeats it.
• Mark 33 credential readings ambiguous where a surname reading is substantial #796 entry: "ra (also a title: John Smith Ra. reads title Ra. with no flag)" describes the marked option. Unmarked, the tree reads suffix Ra.. The sentence is true but reads like a description of today's parse. The same goes for its edd sentence, which the Standing line rewords as a counterfactual.
• The 2026-08-17 C-i correction was committed at 23:23 PDT on 2026-08-16 (2026-08-17 UTC). I left the date alone.
• In the mai and esq index lines, the Log is not in date order: index lines first, then the bundle's bullets. That is the file's existing shape.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.05%. Comparing base (c54bfa6) to head (ec47d4f).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #802   +/-   ##
=======================================
  Coverage   99.05%   99.05%           
=======================================
  Files          47       47           
  Lines        4642     4642           
=======================================
  Hits         4598     4598           
  Misses         44       44           

☔ 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.