B3c: one @2 freeze rule, warm-start combined total, fit acceptance check (companion of autofit_inference#7)

REFERENCE.md

@@ -44,6 +44,11 @@ setup nor sums stages. Diagnostics carry `status` and `values`. Samples carry
 availability and access limitations; safe output paths are hints, not verified
 public archive links. Large samples remain in project storage.
 
+Warm starts (additive, v1 and v2): a record may carry `warm_start` (provider
+provenance), `provider_wall_s` and `consumer_total_s`. Its `total_s` is then the
+producer's combined attempt cost, with a `total_s` definition saying so; the board
+shows that total and names the provider's share. The reader still sums nothing.
+
 Comparisons declare id, records and protocol. The reader refuses changed scientific
 target/dataset/model/pipeline/stage/seed identities and incompatible clock definitions.
 Unknown controls are not proof of parity. It displays refusals without ratios or

campaigns.yaml

@@ -92,13 +92,14 @@ campaigns:
 - id: gaussian-x3-search-benchmark
   title: PyAutoFit search benchmark on gaussian_x3 (autofit_inference)
   status: blocked
-  updated: '2026-10-08'
-  next: Run the 297 deferred wave-1 runs (and the JAX blend reference seed 1) on RAL partition ral on a PyAutoFit
-    revision after A2, then freeze gaussian_x3@2 before any wave-2 row.
+  updated: '2026-10-14'
+  next: Freeze gaussian_x3@2 from the existing 17-row wave-1 calibration before any scored run (protocol §7;
+    the pilot is not restarted); then run scored wave 2, which absorbs the 297 deferred runs, and the JAX blend
+    reference seed 1 on RAL partition ral on a PyAutoFit revision frozen after A2 and A4.
   recent_progress: Wave-1 pilot stopped at 223 of 520 expected runs by the B3 wrap-up ruling (2026-10-08);
     pilot ranks nothing; catalogue, calibration record and inference-summary@1 published.
-  blockers: Phase A2 (PyAutoFit#1679, gradients under JAX) must merge first; NSS waits on A3b; no compute
-    submission is authorized by this ledger.
+  blockers: Scored wave 2 waits for phases A2 (PyAutoFit#1679, gradients under JAX) and A4; NSS waits on A3b;
+    no compute submission is authorized by this ledger.
   job_status: unknown — live scheduler not queried; the local pilot was stopped and this ledger records no
     RAL submission
   evidence: https://github.com/PyAutoLabs/autofit_inference/issues/4
@@ -170,7 +171,7 @@ tasks:
   status: blocked
   priority: normal
   path: tasks/gaussian_x3_search_wave1.md
-  updated: '2026-10-08'
-  next: Pilot partial (223 of 520 runs); run the deferred cells on RAL partition ral after A2, then freeze
-    gaussian_x3@2.
+  updated: '2026-10-14'
+  next: Pilot stopped at 223 of 520 runs and is not restarted; freeze gaussian_x3@2 from its 17-row calibration
+    before any scored run; the deferred cells move to scored wave 2 (after A2 and A4).
   evidence: https://github.com/PyAutoLabs/autofit_inference/issues/4

insight/board.py

@@ -104,6 +104,19 @@ def md(v):
     return e(v).replace("|", "\\|").replace("\n", " ")
 
 
+def total_value(r):
+    """The record's ``total_s`` as the producer stated it. When the producer declares a
+    warm-start provider (additive ``warm_start`` / ``provider_wall_s`` fields) the total
+    is its combined attempt cost and the provider's share is named; the reader never
+    sums clocks itself."""
+    total = r["timings"]["total_s"]
+    provider = r.get("provider_wall_s")
+    if total is None or type(provider) not in (int, float):
+        return total
+    name = (r.get("warm_start") or {}).get("provider") or "unnamed"
+    return f"{total} (incl. {name} warm-start provider {provider})"
+
+
 def _dump(v):
     return json.dumps(v, sort_keys=True, ensure_ascii=False)
 
@@ -174,7 +187,7 @@ def _detail(s, now=None):
                     e(r["execution"]["status"]),
                     e(r["scientific"]["convergence"]) + " / " + e(r["scientific"]["acceptance"]),
                     e(r["timings"]["sampling_s"]),
-                    e(r["timings"]["total_s"]),
+                    e(total_value(r)),
                     e(r["diagnostics"]["status"]),
                     e(r["samples"]["availability"]),
                     paths,
@@ -359,7 +372,7 @@ def render_markdown(views, now=None, campaign_data=None):
                 r["scientific"]["convergence"],
                 r["scientific"]["acceptance"],
                 r["timings"]["sampling_s"],
-                r["timings"]["total_s"],
+                total_value(r),
                 r["archived"],
             ]
             out.append(

receipts/fit.json

@@ -3,15 +3,15 @@
   "repo": "autofit_inference",
   "github": "PyAutoLabs/autofit_inference",
   "branch": "main",
-  "resolved_commit": "900648b5c66514167dd79a4859aed400fcc9551e",
+  "resolved_commit": "6df69441bba42917bcb000f4170299361029dacf",
   "summary_path": "dashboard/summary.json",
   "source": "remote",
-  "fetched_at": "2026-10-08T19:54:10Z",
+  "fetched_at": "2026-10-14T09:49:29Z",
   "outcome": "ok",
   "schema": "inference-summary",
   "version": 1,
   "record_count": 223,
   "errors": [],
   "cached_from": null,
-  "refreshed_at": "2026-10-14T09:06:30Z"
+  "refreshed_at": "2026-10-14T09:49:29Z"
 }