Jammy2211 · Oct 23, 2026 · Oct 23, 2026
diff --git a/dashboard.html b/dashboard.html
@@ -334,7 +334,7 @@
 - Open autofit_inference: https://github.com/PyAutoLabs/autofit_inference
 - Open euclid_detailed: https://github.com/PyAutoLabs/euclid_detailed
 - Open cowls: https://github.com/PyAutoLabs/cowls
-- Open cowls_ansofi: https://github.com/PyAutoLabs/cowls_ansofi</textarea></details><div class="orchestration-freshness" data-freshness="grey"><details data-freshness-stamp data-refreshed-at="2026-10-23T14:36:18Z"><summary><span data-freshness-label>Last updated 2026-10-23 14:36:18 UTC</span></summary><time datetime="2026-10-23T14:36:18Z">2026-10-23 14:36:18 UTC</time></details><a data-refresh-link href="https://github.com/PyAutoLabs/PyAutoCortex/actions/workflows/dashboard_refresh.yml" title="Open the dashboard refresh controls">↻ Update</a></div></div><p class="orchestration-status" role="status" aria-live="polite"></p></section>
+- Open cowls_ansofi: https://github.com/PyAutoLabs/cowls_ansofi</textarea></details><div class="orchestration-freshness" data-freshness="grey"><details data-freshness-stamp data-refreshed-at="2026-10-23T15:27:28Z"><summary><span data-freshness-label>Last updated 2026-10-23 15:27:28 UTC</span></summary><time datetime="2026-10-23T15:27:28Z">2026-10-23 15:27:28 UTC</time></details><a data-refresh-link href="https://github.com/PyAutoLabs/PyAutoCortex/actions/workflows/dashboard_refresh.yml" title="Open the dashboard refresh controls">↻ Update</a></div></div><p class="orchestration-status" role="status" aria-live="polite"></p></section>
 <nav class="board-nav" aria-label="Board sections"><a class="board-nav-card" href="#summary"><span class="board-nav-count">16</span><span class="board-nav-label">Running</span></a><a class="board-nav-card" href="#summary"><span class="board-nav-count">47</span><span class="board-nav-label">Open</span></a><a class="board-nav-card" href="#projects"><span class="board-nav-count">12</span><span class="board-nav-label">Projects</span></a><a class="board-nav-card" href="#orchestration-cortex"><span class="board-nav-label">Check in</span></a></nav>
 
 <p class="muted mdsrc"><a href="https://github.com/PyAutoLabs/PyAutoCortex/blob/main/dashboard.md" class="board-source-link" title="Markdown version" aria-label="Markdown version"><svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.6" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true" focusable="false"><path d="M14 3H6a1 1 0 0 0-1 1v16a1 1 0 0 0 1 1h12a1 1 0 0 0 1-1V8Z"/><path d="M14 3v5h5M8 12h8M8 16h6"/></svg></a></p>
@@ -517,13 +517,13 @@ <h4>Current position</h4>
 </div>
 <h4>Runs (0)</h4>
 <p class="muted">nothing on the cluster</p>
-<details class="sub"><summary>Recent log — 5 of 8 entries</summary>
+<details class="sub"><summary>Recent log — 5 of 9 entries</summary>
 <ul class="log">
+<li>2026-10-23 — <i>note</i> — <details class="full"><summary><span class="clip">second wave-1 finding closed (PyAutoFit#1713, PR #1714 merged, autofit_workspace_test#116): the `invalid value encountered in add` warnings on seeds 25 and 172 are a readout-only defect — under the mode projection an accepted truncated-nor…</span><span class="less">show less</span></summary><p>second wave-1 finding closed (PyAutoFit#1713, PR #1714 merged, autofit_workspace_test#116): the `invalid value encountered in add` warnings on seeds 25 and 172 are a readout-only defect — under the mode projection an accepted truncated-normal scatter message sits ~19 sigma into its tail, its log_partition underflows to −inf (true log Z ≈ −187.9), the sigma evidence goes +inf and the EP log_evidence NaN from sweep 4; posteriors bit-identical to the ens_n5 sidecars. The library now names the term once per fit and returns an explicit NaN. The cure (tail-stable log_partition via log_ndtr) is message maths, filed in the Mind as draft/bug/autofit/truncated_normal_log_partition_tail_underflow.md for the human.</p></details></li>
 <li>2026-10-23 — <i>note</i> — <details class="full"><summary><span class="clip">criterion 6 amended (PyAutoFit#1711, PR #1712): the 19/400 zero-SUCCESS seed-legs of wave 1 split into 13 NO_CHANGE-only prior factors (zero-iteration Laplace convergence at the start point, projection accepted, closed-form match a 5e-5 /…</span><span class="less">show less</span></summary><p>criterion 6 amended (PyAutoFit#1711, PR #1712): the 19/400 zero-SUCCESS seed-legs of wave 1 split into 13 NO_CHANGE-only prior factors (zero-iteration Laplace convergence at the start point, projection accepted, closed-form match a 5e-5 / b 2e-5 on seed 140 leg A — a converged update, not stale; library reporting aligned, no EP maths changed) and 6 BAD_PROJECTION-only hierarchical factors (the #1405 scatter state, cured by projection="moments" #1656; seeds 29 and 71 re-run under moments give HierarchicalFactor SUCCESS 28 and 23). Criterion 6 now counts SUCCESS or NO_CHANGE as a completed update; the matching ensemble_seed.py any_factor_never_succeeded change is the human's to apply in the science checkout. Side finding: seed 29 leg B under moments still misses the sigma cell (a 0.225 / b 0.416) while sigma stays inside [q05, q95].</p></details></li>
 <li>2026-10-14 — <i>note</i> — <details class="full"><summary><span class="clip">PyAutoFit#1656 (merged 2026-09-30, released 2026.10.7.1) ports the minimal-EP referee's moment matching as LaplaceOptimiser(projection="moments"); on analytic_gaussian.py leg B autofit EP sigma 6.3401 ± 2.4609 vs closed form 6.5667 ± 2.883…</span><span class="less">show less</span></summary><p>PyAutoFit#1656 (merged 2026-09-30, released 2026.10.7.1) ports the minimal-EP referee's moment matching as LaplaceOptimiser(projection="moments"); on analytic_gaussian.py leg B autofit EP sigma 6.3401 ± 2.4609 vs closed form 6.5667 ± 2.8832 (mode path 9.3722, BAD_PROJECTION=19). A wave-2 ensemble on the moments path needs a private PyAutoFit checkout on RAL: the shared mirror is 404b3e5f7 and lacks it. The criterion-2 mu-threshold diagnostic is still unrun.</p></details></li>
 <li>2026-10-14 — <i>note</i> — output/ vaulted to z_vault/ on laptop and RAL (mv, nothing deleted), anticipating reruns; output/ left empty, results/ untouched; z_vault/ gitignored locally</li>
 <li>2026-09-30 — <i>note</i> — <details class="full"><summary><span class="clip">astra (Codex gpt-6-astra) on criterion 2, verbatim: 'Neither is established as "wrong" by this result alone. I favour inadequate calibration of the threshold, with moderate confidence, but would not yet certify the control.' Decisive diagn…</span><span class="less">show less</span></summary><p>astra (Codex gpt-6-astra) on criterion 2, verbatim: 'Neither is established as "wrong" by this result alone. I favour inadequate calibration of the threshold, with moderate confidence, but would not yet certify the control.' Decisive diagnostic: independently reconstruct the same EP factorisation at the fixed point on a passing, an edge and the worst mu seed and compare tilted vs q moments; audit the per-site sigma&gt;0 clip (analytic_ep_minimal.py:333). Opus reviewer concurs independently (a_mu = 0.041*(E_ref[mu]-m0)/std_ref over 200 seeds, corr 0.98; calibration seeds 0-4 had |a_mu| 0.002-0.020 vs 95th pct 0.077). Full answer: PyAutoCortex#50 comment https://github.com/PyAutoLabs/PyAutoCortex/issues/50#issuecomment-5908448194</p></details></li>
-<li>2026-09-10 — <i>result</i> — accepted ensemble_parity (R-20260910-04): ok accept and intake the things to address. we can do another run down the line so also achieve the results in the analytic_gaussian project for future comparison</li>
 </ul>
 <p class="muted"><a href="https://github.com/PyAutoLabs/PyAutoCortex/blob/main/projects/analytic_gaussian.md">full log</a></p>
 <p class="muted">active · ral · <a href="https://github.com/PyAutoLabs/PyAutoCortex/issues/34">PyAutoCortex#34</a></p>

diff --git a/dashboard.md b/dashboard.md
@@ -321,13 +321,13 @@ Use the cortex skill. — resume analytic_gaussian: read PyAutoCortex projects/a
 
 _nothing on the cluster_
 
-<details><summary>Recent log — 5 of 8 entries</summary>
+<details><summary>Recent log — 5 of 9 entries</summary>
 
+- 2026-10-23 — *note* — second wave-1 finding closed (PyAutoFit#1713, PR #1714 merged, autofit_workspace_test#116): the `invalid value encountered in add` warnings on seeds 25 and 172 are a readout-only defect — under the mode projection an accepted truncated-normal scatter message sits ~19 sigma into its tail, its log_partition underflows to −inf (true log Z ≈ −187.9), the sigma evidence goes +inf and the EP log_evidence NaN from sweep 4; posteriors bit-identical to the ens_n5 sidecars. The library now names the term once per fit and returns an explicit NaN. The cure (tail-stable log_partition via log_ndtr) is message maths, filed in the Mind as draft/bug/autofit/truncated_normal_log_partition_tail_underflow.md for the human.
 - 2026-10-23 — *note* — criterion 6 amended (PyAutoFit#1711, PR #1712): the 19/400 zero-SUCCESS seed-legs of wave 1 split into 13 NO_CHANGE-only prior factors (zero-iteration Laplace convergence at the start point, projection accepted, closed-form match a 5e-5 / b 2e-5 on seed 140 leg A — a converged update, not stale; library reporting aligned, no EP maths changed) and 6 BAD_PROJECTION-only hierarchical factors (the #1405 scatter state, cured by projection="moments" #1656; seeds 29 and 71 re-run under moments give HierarchicalFactor SUCCESS 28 and 23). Criterion 6 now counts SUCCESS or NO_CHANGE as a completed update; the matching ensemble_seed.py any_factor_never_succeeded change is the human's to apply in the science checkout. Side finding: seed 29 leg B under moments still misses the sigma cell (a 0.225 / b 0.416) while sigma stays inside [q05, q95].
 - 2026-10-14 — *note* — PyAutoFit#1656 (merged 2026-09-30, released 2026.10.7.1) ports the minimal-EP referee's moment matching as LaplaceOptimiser(projection="moments"); on analytic_gaussian.py leg B autofit EP sigma 6.3401 ± 2.4609 vs closed form 6.5667 ± 2.8832 (mode path 9.3722, BAD_PROJECTION=19). A wave-2 ensemble on the moments path needs a private PyAutoFit checkout on RAL: the shared mirror is 404b3e5f7 and lacks it. The criterion-2 mu-threshold diagnostic is still unrun.
 - 2026-10-14 — *note* — output/ vaulted to z_vault/ on laptop and RAL (mv, nothing deleted), anticipating reruns; output/ left empty, results/ untouched; z_vault/ gitignored locally
 - 2026-09-30 — *note* — astra (Codex gpt-6-astra) on criterion 2, verbatim: 'Neither is established as "wrong" by this result alone. I favour inadequate calibration of the threshold, with moderate confidence, but would not yet certify the control.' Decisive diagnostic: independently reconstruct the same EP factorisation at the fixed point on a passing, an edge and the worst mu seed and compare tilted vs q moments; audit the per-site sigma>0 clip (analytic_ep_minimal.py:333). Opus reviewer concurs independently (a_mu = 0.041*(E_ref[mu]-m0)/std_ref over 200 seeds, corr 0.98; calibration seeds 0-4 had |a_mu| 0.002-0.020 vs 95th pct 0.077). Full answer: PyAutoCortex#50 comment https://github.com/PyAutoLabs/PyAutoCortex/issues/50#issuecomment-5908448194
-- 2026-09-10 — *result* — accepted ensemble_parity (R-20260910-04): ok accept and intake the things to address. we can do another run down the line so also achieve the results in the analytic_gaussian project for future comparison
 
 [full log](https://github.com/PyAutoLabs/PyAutoCortex/blob/main/projects/analytic_gaussian.md)
 

diff --git a/projects/analytic_gaussian.md b/projects/analytic_gaussian.md
@@ -11,6 +11,7 @@ Wave 1 (342413, 200 seeds at N=5) is accepted as the baseline: autofit EP is exa
 
 ## Log
 
+- 2026-10-23 — note — second wave-1 finding closed (PyAutoFit#1713, PR #1714 merged, autofit_workspace_test#116): the `invalid value encountered in add` warnings on seeds 25 and 172 are a readout-only defect — under the mode projection an accepted truncated-normal scatter message sits ~19 sigma into its tail, its log_partition underflows to −inf (true log Z ≈ −187.9), the sigma evidence goes +inf and the EP log_evidence NaN from sweep 4; posteriors bit-identical to the ens_n5 sidecars. The library now names the term once per fit and returns an explicit NaN. The cure (tail-stable log_partition via log_ndtr) is message maths, filed in the Mind as draft/bug/autofit/truncated_normal_log_partition_tail_underflow.md for the human.
 - 2026-10-23 — note — criterion 6 amended (PyAutoFit#1711, PR #1712): the 19/400 zero-SUCCESS seed-legs of wave 1 split into 13 NO_CHANGE-only prior factors (zero-iteration Laplace convergence at the start point, projection accepted, closed-form match a 5e-5 / b 2e-5 on seed 140 leg A — a converged update, not stale; library reporting aligned, no EP maths changed) and 6 BAD_PROJECTION-only hierarchical factors (the #1405 scatter state, cured by projection="moments" #1656; seeds 29 and 71 re-run under moments give HierarchicalFactor SUCCESS 28 and 23). Criterion 6 now counts SUCCESS or NO_CHANGE as a completed update; the matching ensemble_seed.py any_factor_never_succeeded change is the human's to apply in the science checkout. Side finding: seed 29 leg B under moments still misses the sigma cell (a 0.225 / b 0.416) while sigma stays inside [q05, q95].
 - 2026-10-14 — note — PyAutoFit#1656 (merged 2026-09-30, released 2026.10.7.1) ports the minimal-EP referee's moment matching as LaplaceOptimiser(projection="moments"); on analytic_gaussian.py leg B autofit EP sigma 6.3401 ± 2.4609 vs closed form 6.5667 ± 2.8832 (mode path 9.3722, BAD_PROJECTION=19). A wave-2 ensemble on the moments path needs a private PyAutoFit checkout on RAL: the shared mirror is 404b3e5f7 and lacks it. The criterion-2 mu-threshold diagnostic is still unrun.
 - 2026-10-14 — note — output/ vaulted to z_vault/ on laptop and RAL (mv, nothing deleted), anticipating reruns; output/ left empty, results/ untouched; z_vault/ gitignored locally

diff --git a/state.json b/state.json
@@ -4,7 +4,7 @@
   "repo": "PyAutoCortex",
   "status": "yellow",
   "headline": "YELLOW \u2014 12 active \u00b7 16 running",
-  "updated": "2026-10-23T14:36:18Z",
+  "updated": "2026-10-23T15:27:28Z",
   "pages_url": "https://pyautolabs.github.io/PyAutoCortex/",
   "items": [
     {