Hafeok · Aug 31, 2026 · Aug 30, 2026 · Aug 30, 2026 · Aug 30, 2026 · Aug 31, 2026
diff --git a/core/decisions/DDD-dec-34.yaml b/core/decisions/DDD-dec-34.yaml
@@ -0,0 +1,123 @@
+format: 1
+id: DDD-dec-34
+kind: decision
+statement: |
+  The upstream pin advances v5.9.0 -> v5.12.0 across three releases in one act, six pinned objects are re-instrumented and four unpinned ones are added, and the advance's firing is predicted in a committed file before graph/upstream.yaml is touched — because the node whose retirement forces this revision's largest repair, DDD-measure-06, fired nothing at all: it was never pinned.
+principal: Emil
+made: 2026-08-30, Paper A revision (GATE 3 ruling to proceed; the advance executed and verified at GATE 4)
+resolution: |
+  The advance, what it fired, and the finding that changes what gets pinned.
+
+  ## Why three releases at once
+
+  The pin was HELD at v5.9.0 by ruling, so canon could be repaired without dragging the projection
+  behind it. Three releases landed against that hold: v5.10.0 (the Phase 1a session — the discharge
+  partition and the measure boundary), v5.11.0 (retirement provenance and the validator) and
+  v5.12.0 (definitions that were never in the registry, and a collision the store won). The hold
+  was correct and it is discharged here, in one advance, by the paper it was protecting.
+
+  ## Predicted before, verified after
+
+  Stated in this order so the verification can disconfirm it, per DDD-dec-28 and DDD-dec-29, and
+  COMMITTED AHEAD OF THE OPERATION at
+  meta/sessions/2026-08-30-paper-a-revision/gate4-prediction.md, with the instrument that computed
+  it beside it (predict-pin-advance.py, which re-implements validate-core-order.py's own
+  pinned_content_digest rather than approximating it).
+
+    PREDICTED FIRING:
+      E12  0 — all 67 pinned ids still exist at v5.12.0
+      W5   exactly 1 — DDD-frame-15, projected -> retired
+      W6   exactly 6 — DDD-cost-09 (region); DDD-delivery-01 (statement); DDD-frame-02
+           (statement); DDD-frame-15 (statement and region); term:delivery (canonical_md);
+           term:residual-discretion (canonical_md)
+      W7   1, unchanged — term:maturation, declared against DDD-dec-21
+      and 61 of 67 pins unmoved.
+
+    OBSERVED, at ref v5.12.0 with the hashes NOT yet re-instrumented:
+
+      67 pins resolved against the pinned ref, 1 basis-loss, 6 content-drift, 1 shadowed id(s)
+
+  Every id and every hash was the one written down beforehand. **The prediction held in every
+  limb.** The two-stage method is the point: had the hashes been advanced in the same edit as the
+  ref, the drift count would have read 0 throughout, the pass and the skip would have produced
+  identical output, and there would have been nothing to check the prediction against.
+
+  After re-instrumentation: 71 pins resolved, 0 basis-loss, 0 content-drift, 1 shadowed id.
+
+  ## Reconciled against the migration seed's own prediction
+
+  meta/migration-plan-ground.md predicted three W6 for the next advance past both sessions —
+  term:delivery, DDD-cost-09, DDD-delivery-01. Those three appear exactly. The other three are the
+  Phase 1a session's repairs and are not ground-migration work, which is why the seed does not
+  list them. The seed also names seven W6 that fire when W1 executes IN CANON; none fired, because
+  this session carries W1 in the two manuscripts only and renames nothing upstream.
+
+  ## The finding: four pins added, and why it is a finding
+
+  DDD-measure-06 retires at v5.12.0 from established. It forces the largest prose repair in this
+  revision — the paper asserted its biconditional as **established** in three places, and the
+  companion measure note had already conceded both failure directions without canon being updated
+  to match, so this projection was stronger than the note sharing its graph.
+
+  **It fired nothing.** W5 only sees ids that are pinned, and DDD-measure-06 was not. Both existing
+  checkers passed all three lines, because none is a block quotation and none is an appendix row.
+  The gap was found by a sweep written at this session's GATE 1 for inline status assertions, now
+  shipped as papers/paper-a/check-status.py.
+
+  That is DDD-agent-01 applied to this repository's own apparatus: a basis cited without a pin is a
+  basis that can move silently. Four nodes are therefore pinned at this advance — DDD-measure-06,
+  its two successors DDD-measure-16 and DDD-measure-17, and DDD-frame-17, which supersedes
+  DDD-frame-15. Three of the four could not have been pinned earlier: they do not exist at v5.9.0
+  and pinning them there would have fired E12.
+
+  ## The other steps, and their results
+
+  - **The four failing quotations**, all repaired and verified: DDD-frame-15's statement (retired,
+    the four modes replaced by DDD-frame-17's three values), DDD-frame-02's (amended clause),
+    term:residual-discretion's (amended, the cryptographic-hash separation), and the compact form
+    in the conclusion — the cheapest, since the compact form survives verbatim in DDD-frame-17's
+    closing clause, so the citation moved and no prose did. check-quotations.py at v5.12.0:
+    **30 verbatim, 0 disclosed-partial, 0 failing.**
+  - **Appendix A regenerated wholesale** against the tag, never hand-edited, and re-read by the
+    independent script: **75 nodes rendered, 75 cited, 0 discrepancies**, up from 72 nodes and 13
+    discrepancies before regeneration. Idempotent across three runs, byte-compared.
+  - **check-status.py at v5.12.0: 93 inline status assertions, 0 wrong**, from 4 wrong before the
+    repair.
+  - **The declared ref** advanced in the manuscript's front matter, which is now the supplement's
+    S2.
+
+  ## What this act does not do
+
+  It does not merge anything; Emil ratifies. It does not file, amend or retire any claim in either
+  repository — the two retirements this revision carries were made upstream at v5.10 and are only
+  projected here. It does not rule on the accountability arity finding (S4 of the supplement),
+  which is routed as a successor item rather than ridden in on a pin advance.
+basis:
+  - DDD-dec-29
+  - DDD-dec-28
+  - DDD-dec-10
+  - DDD-agent-01
+  - DDD-frame-15
+  - DDD-frame-17
+  - DDD-measure-06
+  - DDD-measure-16
+notes: |
+  Numbering: decisions share one id space across the principle repository and this one. 33 is the
+  highest before this file (upstream, at v5.12.0), so this advance is 34.
+
+  Filed at the advance rather than staged [PROPOSED], because the tag it pins already existed:
+  v5.12.0 resolves to 81f6929, verified as the principle repository's default-branch head before
+  anything moved. The DDD-dec-28/29 near-miss — an instruction to bump arriving before the tag was
+  cut — does not apply here, and the file lands ratified rather than being tidied later.
+
+  The ordering constraint from DDD-dec-29 still binds and was followed: the ref written into
+  graph/upstream.yaml is also written into a published manuscript, so pinning a branch during
+  staging would put a branch sha into a paper. Tag first, always.
+
+  ## What the two-stage verification cost, and why it is worth it
+
+  Nothing but one extra checker run. The alternative — advance ref and hashes together — produces
+  a green run whether or not the prediction was right, which is presumed discharge in its exact
+  form: the artefact recording the pass identical to the artefact recording the skip. Three
+  advances have now been verified this way (DDD-dec-28, DDD-dec-29, this one) and the prediction
+  has held each time, which is an argument for keeping the method rather than for dropping it.
diff --git a/graph/upstream.yaml b/graph/upstream.yaml
@@ -69,7 +69,24 @@ upstream:
   # basis-loss, EXACTLY ONE content-drift — term:floor, daf43e07... -> 917f7e4d..., both hashes
   # as predicted — 0 status movement, the one standing W7 unchanged. The prediction held. No
   # pins added: v5.9.0 mints nothing.
-  ref: v5.9.0                       # tag, never a branch
+  # Advanced v5.9.0 -> v5.12.0 at the Paper A revision (2026-08-30), on Emil's GATE 3 ruling;
+  # the advance is DDD-dec-34 here. Three releases at once, because the pin was HELD at v5.9.0 by
+  # ruling so canon could move without dragging the projection — v5.10.0 (Phase 1a), v5.11.0
+  # (retirement provenance) and v5.12.0 (definitions and the store collision) all landed against
+  # a projection deliberately left behind. Predicted firing, stated before it ran and committed
+  # ahead of the operation at meta/sessions/2026-08-30-paper-a-revision/gate4-prediction.md:
+  # E12 0 - W5 exactly 1 (DDD-frame-15, projected -> retired) - W6 exactly 6 (DDD-cost-09 region;
+  # DDD-delivery-01 and DDD-frame-02 statement; DDD-frame-15 statement and region; term:delivery
+  # and term:residual-discretion canonical_md) - W7 1, unchanged. Verified at the bump, by
+  # advancing ref first and re-instrumenting the hashes second so the firing could be observed
+  # rather than assumed: 67 pins resolved, 1 basis-loss, 6 content-drift, 1 shadowed id, every id
+  # and every hash the one written down beforehand. The prediction held in every limb. After
+  # re-instrumentation: 71 pins, 0 basis-loss, 0 content-drift, 1 shadowed id.
+  # FOUR PINS ADDED, and the reason is a finding rather than housekeeping: DDD-measure-06 retires
+  # at v5.12.0 and fired NOTHING, because W5 only sees ids that are pinned. Paper A asserted it as
+  # **established** in three places and both of its checkers passed those lines. The gap was found
+  # by a sweep written for inline status assertions, now shipped as papers/paper-a/check-status.py.
+  ref: v5.12.0                      # tag, never a branch
   pins:
     # basis claims for DDD-dec-04 (the split decision itself rests on these)
     - id: DDD-frame-01
@@ -123,7 +140,7 @@ upstream:
     # the assurance-locus conversion core/13 §5's amortisation corollary rests on (GATE B carve)
     - id: DDD-cost-09
       status_at_pin: projected
-      content_hash: sha256:fecb317fb0b76a969b7f8f5eebdbfc94ea628faf75deb83142de7efb0dd4c013
+      content_hash: sha256:f6cc24f0f406dea4f43785106310669ed81a016202165c0d1c739df31b091ffe
     # the routing layer core/15 exercises and core/14's register consumes (Wave 2, 2026-08-10)
     - id: term:capability
       status_at_pin: settled
@@ -163,10 +180,10 @@ upstream:
     # 2026-08-15; staged with the v5.5.0 advance, DDD-dec-18)
     - id: term:delivery
       status_at_pin: draft
-      content_hash: sha256:1e7b2e4d5df5dad2869d966abe94f5d5677a4a2f4d3e0ff35356534f382ecac2
+      content_hash: sha256:c56909e53c93d1cfbf857df605df91607e75176ca146f07731386ef9d166e46e
     - id: DDD-delivery-01
       status_at_pin: projected
-      content_hash: sha256:67377ef2294de814afd95026f2e7ad0bdb768516696395d05f1ed1645aa6d804
+      content_hash: sha256:1340d483e45ad5db10acb754aa335453f719f01e8126c3ee0f028f8c5545a73d
     # Basis added by the freight session (2026-08-17): four upstream objects gained basedOn
     # edges from decisions filed here — DDD-dec-20 (the session convention), DDD-dec-22 (the
     # pin/shadow instrumentation) and DDD-dec-23 (ruling 16). Found by the GATE 6 basis-impact
@@ -242,7 +259,7 @@ upstream:
       content_hash: sha256:5c146793529821d24b0b45137f1d8e7893acd8c6392b3694d7e145078c9d5e3e
     - id: DDD-frame-02
       status_at_pin: projected
-      content_hash: sha256:f45492eedfaf5dc3e0a92776fb71b61639ac1941486816200b01894ba807f04a
+      content_hash: sha256:0002f60b446fc9d6a5561f21e61fb10472b3b209a416b4ed10216e37721fdbb3
     - id: DDD-frame-13
       status_at_pin: projected
       content_hash: sha256:4add58e0f47c0ff22bb752d6b78aef7f591a14b17d610cdb06677439e0b33276
@@ -274,8 +291,8 @@ upstream:
       status_at_pin: settled
       content_hash: sha256:b3d940d3167c8e9e18667a224a13d98bad17509844930b3be9e49834014aa19e
     - id: DDD-frame-15
-      status_at_pin: projected
-      content_hash: sha256:6d14509db3daaaf1773b6e6d24b6d6b742da2c7fa4f5f6bf18f2286a970df3f9
+      status_at_pin: retired
+      content_hash: sha256:2a65878dfdd95895ef91819fe4545ed125e542346eae2ed9b5e91e50e517d58b
     - id: DDD-frame-04
       status_at_pin: projected
       content_hash: sha256:eec0a4b49c4f996f68a8b83281e0990b829efd318964f803a16452fb4d32a0d3
@@ -304,4 +321,25 @@ upstream:
       content_hash: sha256:7714da9eb209a6839b862a36c0aecd48d411f6af18cab4b1acaa7a1dea032167
     - id: term:residual-discretion
       status_at_pin: draft
-      content_hash: sha256:77284b1467c8a100654279d5a3c86970c791badeb1d891e60f9618e13dc1541e
+      content_hash: sha256:9bbe31e3a02fd3cee70ad42d560268943cd136aa1c0d389ec081d7af0c579340
+    # [PROPOSED] Paper A revision (2026-08-30), added at the v5.12.0 advance. Four nodes the
+    # manuscript leans on that were NOT pinned, found by the Gate 1 status sweep rather than by
+    # the instrument: DDD-measure-06 retires at v5.12.0 and fired NOTHING, because W5 only sees
+    # ids that are pinned. The paper asserted it as **established** in three places and both
+    # existing checkers passed those lines. That is DDD-agent-01 about this repository's own
+    # apparatus — a basis cited without a pin is a basis that can move silently — and the four
+    # are pinned here so the next movement is instrumented rather than swept for. They could
+    # not be pinned before this advance: DDD-frame-17, DDD-measure-16 and DDD-measure-17 do not
+    # exist at v5.9.0 and pinning them there would have fired E12.
+    - id: DDD-frame-17
+      status_at_pin: projected
+      content_hash: sha256:bf8d2c699946fd388ee10af09a0e2441fde92d5d69c8e5e847a2e369ca23248d
+    - id: DDD-measure-06
+      status_at_pin: retired
+      content_hash: sha256:a7ae6e7ad5ccd0afd3bc35c5fd5ce061532b4f99f4d352464f9cc5b78ee01c39
+    - id: DDD-measure-16
+      status_at_pin: established
+      content_hash: sha256:2d763d1adf9e1c68c88d1f5c3e30550a5079e274d84e688c31f5f5bb51f1d94e
+    - id: DDD-measure-17
+      status_at_pin: projected
+      content_hash: sha256:164a18072145bc4a0cdb3c6e1ca8890fc64c1b36396c73be5ca8bd45c66114e5
diff --git a/meta/migration-plan-ground.md b/meta/migration-plan-ground.md
@@ -29,12 +29,22 @@ projection: **380 (67%) need no edit; 187 are bare and need re-expression.**
 
 | Wave | Content | Size | Pins |
 |---|---|---|---|
-| **W1** | S5 → *deployment distribution* | 87 mutable, 29 in merged papers | `term:verdict` unpinned |
+| **W1** | S5 → *deployment distribution* | 87 mutable, **31** in the two manuscripts (was recorded as 29) | `term:verdict` unpinned |
 | **W2** | 11 registry entries + 11 embeds + upstream `core/` bare prose + **`apparatus/`** | 11 + 11 + 65 | **7 fire W6** |
 | **W3** | projections, applications, papers, downstream `core/` | 122 bare | 0 |
 | **W4** | `product-cli` | 1,203 | ruled separately |
 | **never** | U (261, of which **142 identifiers**) + immutable (260) | 521 | — |
 
+**W1's counts, corrected again by the Paper A revision (2026-08-30).** Against
+`w0-full-v2.json`: S5 totals **97**, of which 9 are immutable, leaving **88 mutable** corpus-wide
+and **87** excluding `product-cli`. The two merged manuscripts hold **31** of those (measure note
+26, Paper A 5) — **35%, not 88%**, and the row above is corrected from 29. Of those 31, Emil's Gate 1
+rulings put **15 in scope for that session and 16 out of it**: generated or hand-carried node-table
+rows, passages quoting a live claim verbatim while canon still says `ground distribution`, one
+filename, and the three occurrences ruled ambiguous between S5 and S1. **On those rulings W1's
+remaining surface is 73 of 88, not zero** — the executed figure is recorded at that session's close
+and supersedes this line if it differs.
+
 **Corrected counts, after W0 and W0-bis:** S2 1,529 · S1 **567** · S3 369 · U 261 · S5 97 · S4 **22**.
 Canon is **S1-led** (182 S1 : 154 S3) and **canon's S4 is mechanically zero.**
 
@@ -106,3 +116,27 @@ Three mechanisms, all found by reading rows the instrument had already scored:
    per-rule precision, never one figure for a table. A windowed rule earns a sample of its own.**
 
 **This closes two mechanisms, not the possibility of a rare third.**
+
+### A third, found by the Paper A revision and added on Emil's ruling
+
+> **Classification data cites content, never position.**
+
+The W0 classification's line numbers for `papers/paper-a/paper-a.md` are stale, and not by error:
+the audit's extract was taken on 2026-08-24, and the item-4 session regenerated that paper's
+Appendix A the next day with a `kind` column, moving every appendix row down by eleven lines. Three
+of the five S5 rows in that file therefore point at the wrong lines — `1365`, `1371`, `1409` in the
+extract against `1376`, `1382`, `1420` at head.
+
+Nothing was lost, because a sense assignment travels with the occurrence rather than with the line.
+But **a migration executed by line number would have edited three wrong rows**, and the three wrong
+rows were all inside a generated appendix, where an edit forges agreement with the graph. The
+failure would have been silent in the only place the corpus cannot afford silence.
+
+The rule is cheap: an occurrence is addressed by its file plus a content anchor, and a line number
+is a convenience for a reader rather than an instruction to an instrument. `w1-enumerate.py`
+(`meta/sessions/2026-08-30-paper-a-revision/`) carries both numbers and reconciles neither silently
+— it fails loudly when a manuscript has moved under the enumeration, which is how this was found.
+
+**Corollary, and it is the wave-planning fact:** a generated table's rows are not the wave's to
+move at all. They carry canon's words and they move when canon moves. W1's own scope shrinks once
+that is applied — see the counts correction below.