PolicyEngine · juaristi22 · Sep 10, 2026 · Sep 10, 2026 · Sep 10, 2026 · Sep 10, 2026
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
@@ -3,8 +3,8 @@ name: Tests
 on:
   push:
     branches: [main]
+  # Stacked pull requests need the same acceptance checks as main-bound work.
   pull_request:
-    branches: [main]
 
 # Cancel outdated PR runs; give each main push its own independent group.
 concurrency:
@@ -243,6 +243,7 @@ jobs:
           packages/microcosm-build/tests/test_spec_seed_identity_owned_temp.py
           packages/microcosm-build/tests/test_spec_seed_identity_system_metadata.py
           packages/microcosm-build/tests/test_spec_seed_identity_per_code_context.py
+          packages/microcosm-build/tests/test_spec_seed_identity_parameter_assets.py
       - name: Derive candidate spec and seed identities (diagnostic only)
         id: spec_seed_diagnostics
         timeout-minutes: 15

diff --git a/changelog.d/uk-full-build.changed.md b/changelog.d/uk-full-build.changed.md
@@ -0,0 +1,7 @@
+Consolidate UK source construction, geographic cloning, target selection,
+calibration, exact-count sizing and acceptance evidence into one executable
+graph. Calibrate all applicable geographies by default, with country-only as
+an explicit filter and independent pool/output sizing. Retire obsolete H5
+migration stages and independent calibration entrypoints. Restore bound stage,
+gate and diagnostic artifacts on replay, validate exported bytes, and report
+unsigned certification readiness from the full graph.
diff --git a/changelog.d/uk-full-graph-contracts.added.md b/changelog.d/uk-full-graph-contracts.added.md
@@ -0,0 +1,4 @@
+Add explicit graph weight normalization, immutable Frame evidence projections,
+ordered target-selection receipts and portable calibration problems/results.
+Register UK dense calibration, informed size search, exact-count draws and
+refits as separate resumable nodes over the original pool.
diff --git a/docs/evidence/spec-engine/us-f0-coverage.json b/docs/evidence/spec-engine/us-f0-coverage.json
@@ -1656,13 +1656,13 @@
           "compiler_ir.node_slices"
         ],
         "expected": {
-          "map_sha256": "87ba50531d9fa6683096ecb39a31655b331ab8acff3a5876c2f62b33562a0885",
-          "protocol_sha256": "fd3e4b06f11be4e8c13ea19fef9469ab95cbbe3e2dcfce351e860dd3e00709e4"
+          "map_sha256": "c15bff65f27e51a7b4554b1e3edb43609c6d4fa34ccca613d7a6776947d2fd79",
+          "protocol_sha256": "a775cccdc4ffe64c79bf78be1a0a3c48ad274e3cc5a5aa49f907c320a48827b7"
         },
         "failures": [],
         "observed": {
-          "map_sha256": "87ba50531d9fa6683096ecb39a31655b331ab8acff3a5876c2f62b33562a0885",
-          "protocol_sha256": "fd3e4b06f11be4e8c13ea19fef9469ab95cbbe3e2dcfce351e860dd3e00709e4"
+          "map_sha256": "c15bff65f27e51a7b4554b1e3edb43609c6d4fa34ccca613d7a6776947d2fd79",
+          "protocol_sha256": "a775cccdc4ffe64c79bf78be1a0a3c48ad274e3cc5a5aa49f907c320a48827b7"
         },
         "status": "covered"
       },
@@ -1677,7 +1677,7 @@
           "compiler_ir.seed_stream_map"
         ],
         "expected": {
-          "implementation_sha256": "fd3e4b06f11be4e8c13ea19fef9469ab95cbbe3e2dcfce351e860dd3e00709e4",
+          "implementation_sha256": "a775cccdc4ffe64c79bf78be1a0a3c48ad274e3cc5a5aa49f907c320a48827b7",
           "protocol": "legacy-v1",
           "streams": [
             "build_model",
@@ -1698,7 +1698,7 @@
         },
         "failures": [],
         "observed": {
-          "implementation_sha256": "fd3e4b06f11be4e8c13ea19fef9469ab95cbbe3e2dcfce351e860dd3e00709e4",
+          "implementation_sha256": "a775cccdc4ffe64c79bf78be1a0a3c48ad274e3cc5a5aa49f907c320a48827b7",
           "protocol": "legacy-v1",
           "streams": [
             "build_model",
@@ -2599,7 +2599,7 @@
       "country": "us",
       "schema_id": "country_spec",
       "schema_version": 1,
-      "spec_sha256": "35a02b6b19c921faba1407d441e0b9d9623c496e2cd5b711be014def281a95c6"
+      "spec_sha256": "813fcb2bceef0673fe396ae67fe147947c897e44161e5f4e0d8d7c669132c0ed"
     }
   },
   "report_schema_version": 3,
@@ -2609,7 +2609,7 @@
     "country": "us",
     "schema_id": "country_spec",
     "schema_version": 1,
-    "spec_sha256": "35a02b6b19c921faba1407d441e0b9d9623c496e2cd5b711be014def281a95c6"
+    "spec_sha256": "813fcb2bceef0673fe396ae67fe147947c897e44161e5f4e0d8d7c669132c0ed"
   },
   "status": "pass"
 }
diff --git a/docs/graph-acceptance.md b/docs/graph-acceptance.md
@@ -296,6 +296,26 @@ Adding a normative field with a default changes the canonical projection
 of every node that carries it, so node keys moved with amendments 11 and
 13's sibling field `entrants`; no released artifact pins a graph key yet.
 
+### UK full-build registration extensions
+
+The UK registration adds two explicit contracts to the frozen declarations:
+
+- `WeightUpdate(entity, kind, reason, mass)` declares normalization without
+  changing the weight kind. It requires `conserve` or `declared` mass and an
+  ordered entity-ID receipt; it preserves design-weight ancestry.
+  `WeightTransition` retains its existing forward-only kind check.
+- `KernelContext.frame_metadata` and `frame_mass_log` expose immutable
+  population metadata and legacy mass records alongside declared table
+  slices. `frame_column_order` gives the original order of projected columns
+  only. The mutation guard covers these fields. A consumer that needs a
+  completed legacy ledger must depend on a structural population checkpoint
+  or explicit predecessor evidence; incidental execution order does not
+  establish that dependency. Graph mass accounting remains executor-authored.
+
+`test_weight_update.py` covers aligned identity, kind checks, mass validation,
+design ancestry, immutable context and replay. These extensions do not grant
+kernels access to undeclared population columns.
+
 ## Ownership
 
 Max's ruling (2026-09-01): the agents build all of it. The "implementer ≠

diff --git a/docs/uk-chronicle-feed-repin.md b/docs/uk-chronicle-feed-repin.md
@@ -1,9 +1,10 @@
 # Re-pin the UK Chronicle consumer feed
 
 The national and local target surfaces have independently reviewed Chronicle
-artifact pins. National calibration reads `uk/national_chronicle_feed.json`;
-local calibration and validation retain their existing pins. A national
-update does not authorize changes to local census membership or values.
+artifact pins. The full build reads `uk/national_chronicle_feed.json` for
+national targets and the separate local declarations for local targets and
+validation. Updating one pin does not authorize changes to the other surface's
+membership or values.
 
 Rebuild the complete UK bundle and consumer artifact in
 `PolicyEngine/chronicle` at the declared commit. Keep the resulting
@@ -28,9 +29,11 @@ parity receipts affected by that update with
 local regeneration test accepts either the default `.codex-work` files or a
 `CHRONICLE_UK_LOCAL_FACTS` override and skips only when neither is present.
 
-The national calibration runner refuses a feed whose facts or manifest digest
-differs from its committed pin. `--allow-unpinned-feed` is an explicit diagnostic override and
-is recorded in the run manifest; it is not a re-pin procedure.
+The canonical `tools/build_uk_full.py` command refuses a feed whose facts or
+manifest digest differs from either reviewed declaration. Omitting explicit
+CLI hashes uses the committed pins. A country-only target selector retains
+those source checks. There is no `--allow-unpinned-feed` bypass in the full
+build; changes require the reviewed re-pin procedure above.
 
 After the `ec7169b` re-pin, census household targets use the same Chronicle
 compile path as every other bound UK local family. The OA ladder now supplies

diff --git a/docs/uk-dataset-size-plan-355.md b/docs/uk-dataset-size-plan-355.md
@@ -76,7 +76,7 @@ Use the inputs and environment from the existing
 Pass the same pinned source arguments to the existing driver and add:
 
 ```bash
-uv run python tools/build_uk_rowwise_candidate.py \
+uv run python tools/build_uk_full.py \
   --input-h5 "$UK_SPINE_H5" --input-sha256 "$UK_SPINE_SHA256" \
   --ladder "$UK_LADDER_NPZ" --ladder-sha256 "$UK_LADDER_SHA256" \
   --ledger-facts "$UK_LEDGER_FACTS" \

diff --git a/docs/uk-dense-release-assembly-runbook-762.md b/docs/uk-dense-release-assembly-runbook-762.md
@@ -1,147 +1,7 @@
-# UK dense line release assembly runbook (#762)
+# UK dense release assembly
 
-The dense line `microcosm-uk-2024-25-dense` is the spine cloned K=15 times
-through the OA geography ladder and calibrated to the national and local
-target surfaces in one solve (`tools/build_uk_rowwise_candidate.py`). It ships
-on the **inspect lane only**: a constant release id, an immutable per-cut tag,
-`dataset_role: non_default_local_area`, an empty `default_datasets` map, and
-`--no-latest` at publication, so it can never displace the default artifact.
-It is registered as `("uk", 2025, "dense")` in the private repo
-`policyengine/populace-uk-private`. Publication is a separate human step.
+Use the [UK full-build graph runbook](uk-full-build-graph.md). The dense reference is the output of the same full build when `--dataset-households` is omitted; exact-count builds use its informed search, selection and refit branch.
 
-The R16/R17 release verdicts recorded in the historical receipts used the
-previous gate policy. They do not satisfy the current contract: four quality
-gates now block release, and the incumbent-surface evaluation is mandatory.
-No historical run was re-signed or recalibrated by the PR #870 review fixes.
+The standard target scope includes all applicable geographies. Strict release-candidate checks retain the maintained source pins, solve doctrine, full sample, single engine block and applicable holdouts. Exporting a candidate does not authorize publication or establish native or matched-size certification.
 
-## Prerequisites
-
-- The four pins the run stood on (`spine`, `ladder`, `facts`, `manifest`) and
-  the signing key in `MICROCOSM_UK_TERMINAL_GATE_SIGNING_KEY` (base64, 32 bytes).
-- The Ledger consumer artifact, the spine H5 with its sidecar, the OA ladder.
-- The incumbent extraction (`tools/extract_uk_local_incumbent_surface.py`) for
-  the head-to-head score.
-
-## 1. Pre-flight the environment
-
-```bash
-uv run --no-sync python tools/preflight_uk_local_release_candidate.py --env \
-  --pins <pins.txt> --spine-h5 <spine.h5> --ladder build/uk/uk_oa_ladder_2021.npz \
-  --ledger-facts <chronicle-uk-artifact-dir>
-```
-
-Fails closed and by name on a missing or malformed key, a missing pin, a
-digest mismatch, or doctrine constants that are not the ruled ones.
-
-## 2. Run the release candidate
-
-```bash
-uv run --no-sync python tools/build_uk_rowwise_candidate.py --release-candidate \
-  --input-h5 <spine.h5> --input-sha256 <spine> \
-  --ladder build/uk/uk_oa_ladder_2021.npz --ladder-sha256 <ladder> \
-  --ledger-facts <chronicle-uk-artifact-dir> --ledger-facts-sha256 <facts> \
-  --ledger-manifest-sha256 <manifest> --seed 42 \
-  --logbook-prev-row-digest <previous row> --out <candidate-dir>
-```
-
-`--release-candidate` pins the doctrine (bound 10, `grain_equal`, K=15, 1500
-epochs), resolves the engine in a single block, and runs the rotated holdout.
-Expect about 3.5 hours and 10 GB at K=15.
-
-## 3. Pre-flight the finished run, then score it
-
-```bash
-uv run --no-sync python tools/preflight_uk_local_release_candidate.py --candidate-dir <candidate-dir>
-uv run --no-sync python tools/score_uk_local_candidate.py ... --output-json <candidate-dir>/score_vs_incumbent.json
-```
-
-The pre-flight checks the manifest and the signed gate report for everything
-the contract will demand: release posture attested, shippable, every
-release-blocking gate passed, single-block engine, the doctrine values, the
-A15 census household uprating and A17 tenure application, the measure
-exclusions and their windows, the holdout, the
-Logbook row, the artifact digest.
-
-Measure the full pinned incumbent surface before assembly:
-
-```bash
-uv run --no-sync python tools/evaluate_uk_incumbent_surface.py \
-  --candidate-h5 <candidate-dir>/microcosm_uk_2025_local.h5 \
-  --candidate-manifest <candidate-dir>/rowwise_candidate_manifest.json \
-  --ledger-facts <chronicle-uk-artifact-dir> --ledger-facts-sha256 <facts> \
-  --ledger-manifest-sha256 <manifest> --engine-blocks 1 \
-  --incumbent-manifest <incumbent-dir>/incumbent_local_surface_manifest.json \
-  --incumbent-metrics-csv <incumbent-dir>/household_metrics.csv \
-  --incumbent-weights-csv <incumbent-dir>/wide_weights.csv \
-  --out-json <candidate-dir>/incumbent_surface_evaluation.json \
-  --out-md <candidate-dir>/incumbent_surface_evaluation.md
-```
-
-Use the actual metrics and weights filenames from the extraction manifest.
-The evaluator remains diagnostic: missing optional incumbent inputs and poor
-fit produce a failed assessment, not permission to publish. Only one engine
-block is accepted. Assembly requires the complete authenticated evaluation,
-including finite candidate measurements on every national and local row and
-finite realized incumbent estimates on every local row. National comparisons
-use the pinned incumbent targets; they do not claim realized incumbent fit.
-Signed deferrals stay in this evaluation. A missing or unmeasurable row blocks
-release until its measurement is supplied.
-
-The same existing absolute quality limits apply to this surface: every row
-within 25%, and at least half each family's rows within 25% when the family
-has at least five rows. The within-10% family share remains diagnostic.
-The candidate's fitted score uses uniform rows on its active local surface.
-Its holdout uses the separately recorded weighting rule over held local
-grains. Their shared cap does not make the losses directly comparable; no
-ranking of fitted versus holdout losses is reported.
-
-## 4. Assemble the release directory
-
-```bash
-uv run --no-sync python tools/assemble_uk_dense_release_dir.py \
-  --candidate-dir <candidate-dir> --spine-h5 <spine.h5> \
-  --incumbent-manifest <incumbent-dir>/incumbent_local_surface_manifest.json \
-  --out-dir releases
-```
-
-Assembly verifies the hash join (every manifest output against its bytes, the
-spine against its pin, the gate report against the Logbook build id), re-runs
-the candidate pre-flight, mints the cut tag
-`microcosm-uk-2024-25-dense-<YYYYMMDDTHHMMSSZ>-<uuid8>` from the run's attempt
-id, clones the H5 beside itself as `microcosm_uk_2025_dense.h5`, stages
-`build_manifest.json`, `release_manifest.json`, `calibration_diagnostics.json`,
-`gate_summary.json`, `uk_source_coverage.json`, the signed `uk_local_gates.json`,
-`score_vs_incumbent.json`, `incumbent_surface_evaluation.json`, the original
-`rowwise_candidate_manifest.json`, `source_calibration_diagnostics.json`,
-`incumbent_manifest.json`, and `sha256sums.txt`, validates the directory with
-`microcosm.data.contract.validate_release_dir`, and only then renames it into
-`releases/microcosm-uk-2024-25-dense/`. Re-assembling requires removing the
-previous directory first. The JSON summary prints the publication command.
-
-Assembly and every later directory validation require measured clean code
-(`code.git_dirty` exactly `false`) and full measure-exclusion provenance.
-Approval and expiry dates must be valid ISO dates and in force on the current
-validation date; expiry-day validation is allowed, the following day is not.
-The upload path invokes this validator again before uploading bytes. Separate support
-and binding adjudications keep their own policies. The gate thresholds and
-existing approvals have not been widened or renewed.
-
-## 5. Publish for inspection (human step)
-
-Run the printed command. Its shape is:
-
-```bash
-uv run python -m microcosm.data.publish_cli releases/microcosm-uk-2024-25-dense \
-  --repo-id policyengine/populace-uk-private --artifact-root <candidate-dir> \
-  --no-latest --tag-name microcosm-uk-2024-25-dense-<timestamp>-<uuid8>
-```
-
-`--no-latest` is mandatory and enforced: publication refuses to move
-`latest.json` for a non-default role. The artifact is reachable by its tag and
-by the registry key `("uk", 2025, "dense")` only.
-
-## Promotion is a separate change
-
-Making the dense line (or a sparse successor via the L0 penalty, #762 I10) a
-default dataset is a registry and contract change with its own review; nothing
-in this runbook promotes anything.
+Historical dense receipts use their recorded gate policy. They do not satisfy the current full-build contract merely by being re-signed or repackaged. Current graph certification requires its identified gate, holdout, export/readback and incumbent-comparison evidence; exact-count promotion also needs a matched-size comparison.