RHL-14: O1 ruled by quorum — reword the three §3.2 lines, reserve no keyword

.pmat/baseline.json

@@ -1,6 +1,6 @@
 {
   "version": "3.41.1",
-  "created_at": "2026-09-20T17:31:37.910110581Z",
+  "created_at": "2026-09-20T20:19:32.784452565Z",
   "git_context": null,
   "files": {
     "./build.rs": {

docs/audits/quorum-RHL-14.json

@@ -0,0 +1,84 @@
+{
+ "ticket": "RHL-14",
+ "pr": 230,
+ "base": "main",
+ "head": "9f40a2111d49d1aae8ff9c9e4de34adf43d23f41",
+ "diff_sha256": "e2a9a40cb058b903ef1e38d2db9c81cb22a18fabb64b6f35eb49b0a12b9c3ee4",
+ "agreed": true,
+ "verdict": "implement-as-written (2/2), plus one header line added after review",
+ "author": {
+  "session": "ruchy-9d",
+  "model": "claude-opus-5[1m]"
+ },
+ "families_required": [
+  "gemini",
+  "claude"
+ ],
+ "families_run": [
+  "gemini"
+ ],
+ "families_note": "Docs-only spec amendment. Both review lanes gemini (3.1-pro-high, 3.8-flash-high), both measured (command+output on 6/7 and 5/7 findings), tree witness verified, 0 BLIND. The DECISION itself was a separate 3-lane gemini quorum (dabbe241, 51ff083b, 679e5b1c) recorded in the spec's O1 block; the author is claude-family and took the minimal reading of a 2:1 split, which is recorded there as a residual, not hidden.",
+ "decision_quorum": {
+  "width": 3,
+  "lanes": [
+   {
+    "lane": 1,
+    "model_measured": "gemini-3.1-pro-high",
+    "conversation": "dabbe241-e44f-4516-bc87-aa7d275d8aeb",
+    "verdict": "split: <App> to <App> production for `to`; reword per/are"
+   },
+   {
+    "lane": 2,
+    "model_measured": "gemini-3.8-flash-high",
+    "conversation": "51ff083b-4470-4ef7-8ac5-4e366b5b7af9",
+    "verdict": "O1-R for all three; refutes O1-A's `of` claim"
+   },
+   {
+    "lane": 3,
+    "model_measured": "gemini-3.7-flash-high",
+    "conversation": "679e5b1c-5d2a-40ac-8f43-b7ddab8b2c00",
+    "verdict": "split, same as lane 1"
+   }
+  ],
+  "consensus": "3/3 refuse O1-A as written; 3/3 measured zero vocabulary terms and zero docs/rhl/ programs use per/are or the three shapes; 2/1 on adding <App> to <App>; author took the minimal ruling (no grammar change) and recorded the split as a residual."
+ },
+ "review_quorum": {
+  "judged": "81e1ea0dc1800025fba91a92c6f5e54a74c8c7a3",
+  "width": 2,
+  "lanes": [
+   {
+    "lane": 1,
+    "model_measured": "gemini-3.1-pro-high",
+    "conversation": "92532bd8-8e69-491b-b7fb-97de581a5859",
+    "verdict": "implement-as-written",
+    "kept_clone": "wrote diff.txt/log.txt/old_spec.md/output_log.txt into its writes=false clone; shared checkout byte-identical; clone removed by the author"
+   },
+   {
+    "lane": 2,
+    "model_measured": "gemini-3.8-flash-high",
+    "conversation": "fde2b85e-cec5-45e5-bd4e-8efddb2a316d",
+    "verdict": "implement-as-written"
+   }
+  ],
+  "measured": [
+   "spec sha256 matches the new manifest line; all 301 manifest hashes match the tree",
+   "per/are only in comment lines of both vocabularies (fleet 12,27; tickets 10,23)",
+   "rg 'no change to|per run|are filed' docs/rhl/ \u2192 nothing",
+   "match block: of=False in=True to=True",
+   "roadmap.yaml parses; RHL-1 notes 'NOT READY' gone"
+  ],
+  "gaps_closed_by_author": [
+   "rg -n O1 sweep for stale 'awaiting operator' text: only grammar/fixtures/o1-a-candidate.lalrpop:4 (candidate evidence, manifest-hashed, left as both lanes allowed) and the RHL-0 receipt (historical)",
+   "the O1 header now says RULED so a reader landing there does not stop \u2014 one line, ed39f21e, manifest regenerated, 37 gates green"
+  ],
+  "not_measured": [
+   "`expect host is unchanged` and `then ticket count is 0` have no corpus twin; their parse is a grammar read by both lanes, not a generated-parser run. RHL-1 builds the parser that measures it."
+  ]
+ },
+ "rebind_note": "Reviewed at 81e1ea0d; head moved by one docs commit (the O1 header line + its manifest hash) after the review. Recorded rather than re-reviewed: one sentence, no claim in it the lanes did not already check.",
+ "residuals": [
+  "grammar/fixtures/o1-a-candidate.lalrpop:4 still reads 'Awaiting an operator ruling'",
+  "the 2:1 on <App> \"to\" <App> stays open for a corpus task that needs it"
+ ],
+ "rebind_note_2": "Rebased onto main after #229 merged (both PRs appended roadmap rows; the conflict was two append hunks, resolved by keeping both sides \u2014 204 rows, ids unique, checked with a real YAML parser). The judged diff's content is unchanged except the roadmap hunk's position; head is the bind commit's parent 9f40a211 (the artifact commit adds only itself), the rhl_pre_registration gates re-run green after the rebase."
+}

docs/rhl/PREREGISTRATION.sha256

@@ -304,7 +304,7 @@ f8b3f91447b4b64a09d69c71d87a7bdcecf5a37d1f9e08e1c6cc4314b61350bd  docs/rhl/corpu
 1772856a1c12f1a0a5b1fb3c7abf664bd7edafd2fe0318a737d0bda0fff7dfe9  docs/rhl/corpus/40-disk-cleanup-unstated-unit/tests.rs
 6d0fd9fb92ad60ede614398d2051d31ca89c12d9d846f45384979e9bd3097a8e  docs/rhl/corpus/HARNESS.md
 a270e5ec4706ac2f66bdb0e3d5632f4cfc31056bcf6afc8571d3d7709fefc0f8  docs/rhl/phase0-bindings.md
-64b5e95e5aa6f54dbd7102c7ca243c73ff4c332bea97115435b057cc36761038  docs/specifications/ruchy-high-level-language-interface.md
+a45f5b4a37b4b7eaab39a17c1770ebb4b36ca4989cdfbdfd34591fddb38bcc35  docs/specifications/ruchy-high-level-language-interface.md
 0447678137b6c7efc06628f010e42f2e2ff6e22cde38e7d573a570bc7939dd77  grammar/fixtures/ambiguous.lalrpop
 3eac9148c0f4c28ff8d42259e23726d6f04dc1b7e95146fbe86468e6a5c8385c  grammar/fixtures/o1-a-candidate.lalrpop
 45f9dd4d63d64bbb6135b0113d8854545439ae7b2d3d54eaa893838f5a636088  grammar/fixtures/o1-b-candidate.lalrpop

docs/roadmaps/roadmap.yaml

@@ -2916,7 +2916,7 @@ roadmap:
   - kind:code
   - rhl
   - depends:RHL-0
-  notes: 'ADMITTED by RHL-0''s merge (commit 1c2caf5d, PR #225) — the spec''s readiness gate ''no implementation rows are admitted until RHL-0 is merged'' has opened. BUT NOT READY TO START: spec open question O1 is unruled. Three lines of section 3.2 do not parse (expect no change to host / expect at most 1 ticket per run / then 0 tickets are filed), and RHL-1 is the row that builds the parser those lines have to pass through. Both candidate repairs are measured and committed as grammar/fixtures/o1-{a,b}-candidate.lalrpop: O1-A is conflict-free, parses all three lines and keeps the corpus at 72/72, and costs two keywords; O1-B adds no keyword and is NOT conflict-free. Get the one-line ruling before starting. K-hat basis is 118 [C] from RHL-0 (see HARNESS-1 on the unit question). RHL-1 also inherits F1''s unmeasured second half: ''every corpus program has exactly one parse'' needs the parser this row builds.'
+  notes: 'ADMITTED by RHL-0''s merge (commit 1c2caf5d, PR #225) — the spec''s readiness gate ''no implementation rows are admitted until RHL-0 is merged'' has opened. O1 RULED 2026-09-20 by quorum (RHL-14, ruling O1-R): the three section-3.2 lines were reworded to the shapes the corpus already used, no keyword reserved, no grammar change; the 2:1 residual on an App-to-App production is recorded in the spec''s O1 block. READY TO START once RHL-14 merges. K-hat basis is 118 [C] from RHL-0 (see HARNESS-1 on the unit question). RHL-1 also inherits F1''s unmeasured second half: ''every corpus program has exactly one parse'' needs the parser this row builds.'
 - id: RHL-2
   github_issue: null
   item_type: task
@@ -3388,6 +3388,18 @@ roadmap:
   spec: null
   acceptance_criteria:
   - 'Five-class audit pass, 2026-09-20. CLASS 3 (gate-that-cannot-fail). src/quality/mod.rs:860, in test_satd_count_collection: ''let count = QualityGates::count_satd_comments().unwrap_or(0);'' followed by ''assert_eq!(count, 0, "SATD comments should be eliminated")''. count_satd_comments returns Result and shells out to ''find src -name *.rs -exec grep -c ...''. If that invocation fails for ANY reason — find missing, src/ absent, a packaged crate without src/, a sandbox without exec — the error becomes 0 and the assertion PASSES. The test cannot distinguish ''there is no SATD'' from ''the counter did not run'', and 0 is the most permissive of the two answers. GROUNDED BY INSPECTION, not executed: making the counter fail under test needs an environment change. NOTE the production gate is NOT affected — line 300 uses ''?'' and propagates. This is the test only. FALSIFIER: replace unwrap_or(0) with expect(), or assert on Ok(0) rather than on the unwrapped value, then rename the binary/PATH so find is unavailable; the test must go RED where today it goes green.'
+- id: RHL-14
+  github_issue: null
+  item_type: task
+  title: 'RHL-14: O1 ruled by quorum — reword the three §3.2 lines, reserve no keyword'
+  status: planned
+  priority: medium
+  assigned_to: null
+  created: 2026-09-20T18:55:16Z
+  updated: 2026-09-20T18:55:16Z
+  spec: null
+  acceptance_criteria:
+  - Open question O1 (spec RHL-001, under §3.3) ruled by a 3-lane decision quorum on 2026-09-20, per the operator's standing instruction that decisions go to a quorum. 3/3 refuse O1-A as written (reserving per and are forecloses rate and descriptive terms — errors per hour, hosts that are pinned — and all 12 valid corpus programs already use the reworded shapes, so R costs 0 corpus diffs). 2/3 would additionally add App to App under Cmp for the to line; 1/3 calls that an ad-hoc preposition with no vocabulary term behind it. The minimal ruling is taken — reword all three, change no grammar — and the 2:1 on the to production is recorded as a residual. Lane 2 also found a factual error in the O1 text — it says reserving per/are is the same trade already made for of, and of is NOT reserved (grammar/rhl.lalrpop clause 2). Spec is manifest-hashed; PREREGISTRATION.sha256 regenerated in the same commit.
   phases: []
   subtasks: []
   estimated_effort: null
@@ -3433,4 +3445,5 @@ roadmap:
   labels:
   - kind:code
   - class:gate-cannot-fail
+  - kind:docs
   notes: null

docs/specifications/ruchy-high-level-language-interface.md

@@ -110,17 +110,17 @@ job "gx10 disk watch"
     end
   end
 
-  expect no change to host
-  expect at most 1 ticket per run
+  expect host is unchanged
+  expect ticket count is at most 1
 
   example "low disk files one ticket"
     given disk free of "/" is 90 GB
-    then 1 ticket is filed
+    then ticket count is 1
   end
 
   example "healthy disk files nothing"
     given disk free of "/" is 400 GB
-    then 0 tickets are filed
+    then ticket count is 0
   end
 end
 ```
@@ -161,7 +161,7 @@ Everything else in a program is a **vocabulary term**, a **literal**, or a **nam
 > because an action's attributes reading as siblings of the statements around
 > them loses the nesting §3.2 uses to say which attributes belong to which
 > action. But "the only one" was not measured and was not true.
-> **Open question O1 — three lines of §3.2 do not parse (RHL-0, 2026-09-20) `[V]`.**
+> **Open question O1 — three lines of §3.2 do not parse (RHL-0, 2026-09-20) `[V]`. RULED: see O1-R at the end of this block (RHL-14).**
 > A parser was generated from `grammar/rhl.lalrpop` and fed §3.2 verbatim. After
 > Amendment A1 these three lines still fail:
 >
@@ -192,7 +192,10 @@ Everything else in a program is a **vocabulary term**, a **literal**, or a **nam
 > | the 72 non-`missing-end` corpus programs | **72/72, 0 failures** |
 >
 > Cost: §3.3 grows by two keywords, and `per` and `are` become unusable inside any
-> vocabulary term — the same trade §3.3 already makes for `of`, `in` and `to`. It
+> vocabulary term — the same trade §3.3 already makes for `in` and `to`. (This
+> line first said "for `of`, `in` and `to`". `of` is NOT reserved: grammar clause 2
+> keeps it free precisely so `disk free of` stays spellable. Found by the O1
+> quorum, lane 2.) It
 > costs §3.1 principle 8 (block-closed, count the `end`s) **nothing**: no
 > production here touches block structure.
 >
@@ -212,6 +215,33 @@ Everything else in a program is a **vocabulary term**, a **literal**, or a **nam
 > **The ruling wanted, in one line:** adopt O1-A and add `per` and `are` to §3.3,
 > or reword these three lines of §3.2 so the language does not need them. O1-B is
 > not available — it is not conflict-free.
+>>
+> **Ruling O1-R (RHL-14, 2026-09-20) `[V]` — reword; reserve nothing; change no
+> grammar.** Ruled by a three-lane decision quorum (gemini 3.1-pro / 3.7-flash /
+> 3.8-flash, conversations `dabbe241…`, `51ff083b…`, `679e5b1c…`), under the
+> operator's standing instruction that design decisions go to a quorum. The three
+> lines of §3.2 now read `expect host is unchanged`, `expect ticket count is at
+> most 1`, `then ticket count is 1` / `is 0` — the shapes the 12 valid corpus
+> programs (`docs/rhl/breaks/valid/`) already used, so the corpus diff is zero and
+> the grammar is untouched. Each parses under the pre-registered grammar as
+> `"expect"|"then" <Cond>` → `Cmp` → `<App> <CompOp> <App>`.
+>
+> Measured by all three lanes: no term in `vocab/fleet-v1.yaml` or
+> `vocab/tickets-v1.yaml` contains `per` or `are` (the words occur only in
+> comments), and no program under `docs/rhl/` uses `no change to`, `per run` or
+> `are filed`. Why 3/3 refused O1-A: reserving `per` and `are` forecloses every
+> future rate or descriptive term — `errors per hour`, `hosts that are pinned` —
+> for three lines of an example, and the A1 precedent cuts the other way: A1 added
+> a keyword to keep a *structure* (the nested block); nothing structural is at
+> stake here. Principle 9 wins over "reads like English" when the English costs
+> the vocabulary a word.
+>
+> *Residual, 2:1.* Lanes 1 and 3 would also have admitted `<App> "to" <App>` under
+> `Cmp` (one production, zero keywords, `to` is already reserved) so that `expect
+> no change to host` parses. Lane 2 called that an ad-hoc preposition with no
+> vocabulary term behind it — there is no term `no change` — and the minimal
+> ruling was taken. A corpus task that needs `X to Y` as a comparison re-opens it;
+> `grammar/fixtures/o1-a-candidate.lalrpop` keeps the measured production.
 >
 > The unmarked form is kept verbatim as `grammar/fixtures/ambiguous.lalrpop`,
 > where it serves as F1's positive control. Evidence: