cortex: migration — 21 phases, 12 rulings, the two 2026-08-31 science batches, four epics; Migrated-from key; decisions 53–56 (PyAutoMind#383)

REFERENCE.md

@@ -82,6 +82,7 @@ sections are fixed: `## Question`, `## Witness`, `## Where to look`, `## Runs`,
 | `Review-minutes:` | integer | a seed, not a measurement |
 | `Epic:` | slug | shared with the Mind — the join key across the two dashboards |
 | `Filed:` | `YYYY-MM-DD` | |
+| `Migrated-from:` | source path or ledger anchor | the Mind prompt, review file or project-ledger entry this phase was transcribed from (phase 4 of the birth epic) |
 
 Gate refs are matched by `GATE_REF_RE`, copied verbatim from
 `PyAutoMind/scripts/lifecycle.py`:
@@ -313,6 +314,7 @@ The human's words, verbatim.
 | `Reviewed-at:` | timestamp | |
 | `Review-minutes-actual:` | integer | |
 | `Follow-ups:` | comma-separated GitHub refs | `GATE_REF_RE`; **the issue is created before `rule` runs** |
+| `Migrated-from:` | source path or ledger anchor | the review file or project-ledger entry a backfilled ruling was transcribed from |
 
 Body: `## Ruling` — the human's words verbatim; `## Evidence` — pointers.
 
@@ -375,13 +377,19 @@ names a member of it.
 - reviewed-at: <free text>
 - review-minutes-actual: <integer or (not given)>
 
+### Follow-ups accepted
+- <accepted follow-up, one per line>
+
 ## <slug> — <HEALTH>
 - decision: accept|rerun|drop|leave-to-finish|(none)
 - ruled: yes|no
 
 <note verbatim, or (no note)>
 ```
 
+`Follow-ups accepted` is optional and is `###`, not `##`, because `check`
+requires every `##` to name a member of the slot's batch record.
+
 `<HEALTH>` ∈ `HEALTHY | SUSPECT | FAILED | RUNNING`. `(none)` with
 `ruled: no` is a member the human left untouched — listed, never silently
 dropped. `ruled: yes` requires a verb; a verb with `ruled: no` is a leaning the

batches/2026-08-31-am.md

@@ -0,0 +1,54 @@
+# Batch 2026-08-31 am
+- dispatched: 2026-08-29T23:44Z (retroactive — the members were hand-dispatched before the batch framework shipped)
+- review-at: 2026-08-31T13:00Z (stated at review, not at dispatch)
+- shift: overnight 29th→31st, pre-framework
+- lane: local-dev
+- review-minutes-planned: 121 (packet estimate; no dispatch-time plan existed)
+- migrated-from: PyAutoMind/batches/2026-08-31-am.md
+- members:
+  - failed_submissions_342008_10: phases/inference_programme/failed_submissions_342008_10.md — 342008, 342009, 342010 — 1 — ready
+  - smc_probe_342018: phases/inference_programme/smc_probe_342018.md — 342018 — 25 — ready
+  - knn_rerun_342016_7: phases/inference_programme/knn_rerun_342016_7.md — 342016 — 15 — ready
+  - refs_5_6_342016_56: phases/inference_programme/refs_5_6_342016_56.md — 342016 — 15 — ready
+  - slogdet_ab_342017: phases/inference_programme/slogdet_ab_342017.md — 342017 — 12 — ready
+  - delaunay_adapt_split_pl_sersic_0: phases/subhalo_validation/delaunay_adapt_split_pl_sersic_0.md — 342027 — 10 — awaiting-ruling
+  - delaunay_adapt_split_pl_eff_0: phases/subhalo_validation/delaunay_adapt_split_pl_eff_0.md — 342027 — 8 — running
+  - delaunay_adapt_split_pl_eff_1_outer: phases/subhalo_validation/delaunay_adapt_split_pl_eff_1_outer.md — 342027 — 8 — running
+- collected: 2026-08-31T14:46Z
+- reviewed-at: 2026-08-31T13:34 (local)
+- delivered: 7/7 runs produced results (2 of them failed on their own evidence: the 342008-10 submissions and the SMC probe)
+- packet: batches/packets/2026-08-31-am.html
+- review: batches/reviews/2026-08-31-am.md
+- review-minutes-actual: excluded — workshop session (the framework was built and polished mid-review); the first calibration point is the next batch
+- notes: |
+    Transcribed into the Cortex grammar on 2026-09-01 (birth epic phase 4). The Mind
+    original stays intact at PyAutoMind/batches/2026-08-31-am.md with a `- moved-to:`
+    line — a verbatim history has to live somewhere live.
+
+    Two shape differences from the Mind record, both forced by the Cortex grammar:
+    the Mind's single member `subhalo-pl_eff-342027_12` covered two lenses on one
+    array, and decision 56 gives each lens its own phase, so it appears here as two
+    members (`delaunay_adapt_split_pl_eff_0`, `delaunay_adapt_split_pl_eff_1_outer`)
+    carrying the Mind member's 8 review-minutes each. And the Mind's eighth member
+    `euclid-phase2-pr46` was a **dev** retrospective (PR #46, `structure-ok`); the dev
+    surface has its own vocabulary and its own record, so it is named here and not
+    transcribed as a science member.
+
+    Retroactive record: every member was hand-dispatched on 2026-08-29 night, before
+    the batch framework shipped, and the Mind file was written at collection. The
+    review session doubled as the framework workshop (review surface v2, laptop mirror,
+    flexible slot, packet archive all built mid-review), which is why the minutes
+    figure is excluded from calibration.
+
+    The four profiling members were REJECTED wholesale, with no ledger writes: the
+    human rewound the JAX profiling epic to Phase 1 the same day (autolens_profiling#200).
+    Do not certify refs 5/6/7, do not write the slogdet verdict, do not ingest SMC
+    rates — the rewound epic takes precedence over this packet's proposed follow-ups.
+    In the Cortex those four rejections are R-20260831-02..05, filed as `drop`
+    (`reject` is the dev surface's word).
+
+    Packet defect found in use and fixed same-day: the submit modal could not close on
+    Pages (the fragment relied on the artifact viewer's [hidden] reset); packets are
+    standalone documents now. `batches/packets/2026-08-31-am.html` here is the archive
+    of record from now on; the Mind's copy stays at its Pages URL so the localStorage
+    the human's browser holds against it keeps working.

batches/2026-08-31-pm.md

@@ -0,0 +1,59 @@
+# Batch 2026-08-31 pm
+- dispatched: 2026-08-31T18:40Z
+- review-at: 2026-08-31T21:00Z ("5pm, 2.5 hours from now" — stated by the human at dispatch)
+- shift: afternoon (first timed batch under the review workflow)
+- lane: local-dev
+- review-minutes-planned: 74
+- migrated-from: PyAutoMind/batches/2026-08-31-pm.md
+- members:
+  - inference_refs_v1_redo: phases/inference_programme/inference_refs_v1_redo.md — 342091 — 20 — awaiting-ruling
+  - mge_pos_ref_reuse: phases/inference_programme/mge_pos_ref_reuse.md — 340210 — 5 — awaiting-ruling
+  - mge_fp64_retro_baseline: phases/inference_programme/mge_fp64_retro_baseline.md — 339070 — 4 — awaiting-ruling
+  - delaunay_fp64_retro_baseline: phases/inference_programme/delaunay_fp64_retro_baseline.md — 339071 — 4 — ready
+  - delaunay_adapt_split_pl_sersic_0: phases/subhalo_validation/delaunay_adapt_split_pl_sersic_0.md — 342027 — 10 — accepted
+  - delaunay_adapt_split_pl_eff_0: phases/subhalo_validation/delaunay_adapt_split_pl_eff_0.md — 342027 — 8 — awaiting-ruling
+  - delaunay_adapt_split_pl_eff_1_outer: phases/subhalo_validation/delaunay_adapt_split_pl_eff_1_outer.md — 342027, 342093 — 8 — awaiting-ruling
+  - rectangular_adapt_pl_sersic_0: phases/subhalo_validation/rectangular_adapt_pl_sersic_0.md — 342094, 342095 — 15 — awaiting-ruling
+- refreshed: 2026-08-31T22:53Z — inference_refs_v1_redo pulled (first: 2 of 10 tasks ended clean, 1 running, 7 queued behind the single A100)
+- refreshed: 2026-08-31T22:57Z — the subhalo members pulled (second: 342093 landed; 342094 landed, 342095 released and running; both pl_eff chains still inside subhalo[2] with ~3 h to wall)
+- refreshed: 2026-09-01T13:01Z — both laptop legs pulled (third: the 342091 wave complete, 10/10 delivered clean; both pl_eff witnesses landed inside the 2-day wall)
+- refreshed: 2026-09-01T17:54Z — rectangular_adapt_pl_sersic_0 witness landed (evidence_increase +0.2538, the project's first grid-derived witness)
+- collected: 2026-08-31T20:57Z
+- reviewed-at: 2026-09-01T15:13 (transcribed)
+- delivered: 8/8 science members produced results; 3 ruled this slot, 5 carried
+- packet: batches/packets/2026-08-31-pm.html
+- review: batches/reviews/2026-08-31-pm.md
+- review-minutes-actual: (not given)
+- notes: |
+    Transcribed into the Cortex grammar on 2026-09-01 (birth epic phase 4). The Mind
+    original stays intact at PyAutoMind/batches/2026-08-31-pm.md with a `- moved-to:`
+    line; this record holds only the **science** half of that slot.
+
+    The Mind slot was mixed: nine cloud dev members (autofit-resampling-info,
+    autofit-prodigy-49, autofit-multistart-iterations, autonerves-colab-silence,
+    release-smoke-env, organ-board-github-link, memory-queue-filing-gate,
+    numba-vs-jax-sparse, euclid-3a-prose) and two laptop science members. The dev
+    members have their own vocabulary (merge | tweak | reject | defer) and their own
+    record; they are named here and not transcribed. `review-minutes-planned:` above
+    is the science half of the Mind's 117 + 13.
+
+    The two laptop members expand into the eight phases above: the Mind's
+    `jax-inference-phase1-refs` carried the 342091 wave plus three legacy-reuse ruling
+    members added to the packet by that leg (mge-pos-ref-reuse, mge-fp64-retro-baseline,
+    delaunay-fp64-retro-baseline), and the Mind's `subhalo-follow-up-wave` covered four
+    lens/pipeline runs that decision 56 gives a phase each.
+
+    Both laptop members were read-and-record only: no RAL job was submitted, cancelled
+    or modified, no pipeline or library code was edited, and no ruling was written in
+    the session — the rulings are this file's review, filed 2026-09-01T15:13Z.
+
+    CARRY-FORWARD, adopted at the human's direction during this slot: long-running
+    members do not hold a review. The slot reviews everything ready; in-flight members
+    are recorded carried and enter the next batch's packet automatically. Five members
+    are carried out of this slot — the refs wave (its mesh rows are now spent evidence
+    under R-20260901-03 and must be re-checked against the positions-penalty rule before
+    certification) and the four subhalo phases.
+
+    The parked second `rectangular_adapt` submission stays PARKED: its gate (a clean
+    pl_eff witness) opened when both pl_eff witnesses landed, but which lens goes next
+    is a human decision that follows the rulings on the two pl_eff phases.