Jammy2211 · Oct 15, 2026 · Oct 15, 2026
diff --git a/campaigns.yaml b/campaigns.yaml
@@ -44,11 +44,8 @@ campaigns:
 - id: interferometer
   title: Interferometer likelihood and streaming
   status: active
-  updated: '2026-10-13'
-  next: 'No new committed measurements. Decision matrix remains filled; streaming 1e8 CPU/A100 rows and the
-    #332 memo after-measurement still need authorization. #411 merged 2026-10-13: the historical numba kill
-    gate is INCONCLUSIVE (four rounds). Curvature preload needs the fixed-mapper science decision; FFT levers
-    remain unstarted.'
+  updated: '2026-10-15'
+  next: 'Three PyAutoArray tasks filed 2026-10-15 from the Aris 512×512 profiling (PyAutoPulse#62): #626 sparse data vector memory (active, --auto), #627 AdaptSplit log-det backend parity and #628 numba route-gate cost model (queued behind the PyAutoArray claim). Streaming 1e8 CPU/A100 rows and the #332 memo after-measurement still need authorization; curvature preload needs the fixed-mapper science decision; FFT levers remain unstarted.'
   evidence: https://github.com/PyAutoLabs/autolens_profiling/blob/4db4d988e520a82219152743748c3188d485a4ba/wiki/campaigns/interferometer_likelihood.md
 - id: linear-solver
   title: Linear-solver accuracy and cost
@@ -303,6 +300,16 @@ tasks:
     in apply_sparse_operator (memory).'
   issue: https://github.com/PyAutoLabs/autolens_profiling/issues/368
   evidence: https://github.com/PyAutoLabs/autolens_profiling/blob/719459fefb40c7e5be157f652818dbf923830557/wiki/campaigns/interferometer_streaming.md
+- id: interferometer_large_mask_aris
+  campaign: interferometer
+  title: 'Interferometer large-mask likelihood (Aris 512×512): sparse data-vector memory, AdaptSplit log-det backend parity, numba route-gate cost model'
+  status: active
+  priority: high
+  path: tasks/interferometer_large_mask_aris.md
+  updated: '2026-10-15'
+  next: 'PyAutoArray#626 active under --auto (worktree interferometer-sparse-data-vector; witness 7.2 GB -> <= ~3.5 GB at 5,746 vertices); #627 and #628 queued in Mind planned.md behind the PyAutoArray claim, in that order. Record each before/after witness table in the task file at close-out.'
+  issue: https://github.com/PyAutoLabs/PyAutoPulse/issues/62
+  evidence: https://github.com/PyAutoLabs/PyAutoArray/issues/626
 - id: interferometer_w_tilde_fft_size_levers
   campaign: interferometer
   title: 'Interferometer W~ curvature matrix is FFT-bound on the mask extent: pruned padded FFT and real-space

diff --git a/dashboard.html b/dashboard.html
diff --git a/dashboard.md b/dashboard.md
@@ -1,6 +1,6 @@
 # PyAutoPulse — profiling dashboard
 
-<!-- pulse-campaigns:b95c0bf384d49e4f8bd1f183ef71eae3e05fa2e4892d3ba474b05d4261563690 -->
+<!-- pulse-campaigns:a53f6bbfc21f511541a477cdac96aaf7b677a7e058726ca4f839a61947a26133 -->
 ## Profiling Check In
 
 <details><summary>Full check-in prompt</summary>
@@ -94,7 +94,7 @@ Reviewed 2026-10-13. No campaign wiki/results change or new matching implementat
 
 <details><summary>Interferometer likelihood and streaming — active</summary>
 
-Reviewed 2026-10-13. No new committed measurements. Decision matrix remains filled; streaming 1e8 CPU/A100 rows and the #332 memo after-measurement still need authorization. #411 merged 2026-10-13: the historical numba kill gate is INCONCLUSIVE (four rounds). Curvature preload needs the fixed-mapper science decision; FFT levers remain unstarted.
+Reviewed 2026-10-15. Three PyAutoArray tasks filed 2026-10-15 from the Aris 512×512 profiling (PyAutoPulse#62): #626 sparse data vector memory (active, --auto), #627 AdaptSplit log-det backend parity and #628 numba route-gate cost model (queued behind the PyAutoArray claim). Streaming 1e8 CPU/A100 rows and the #332 memo after-measurement still need authorization; curvature preload needs the fixed-mapper science decision; FFT levers remain unstarted.
 
 [Campaign evidence](https://github.com/PyAutoLabs/autolens_profiling/blob/4db4d988e520a82219152743748c3188d485a4ba/wiki/campaigns/interferometer_likelihood.md)
 
@@ -103,6 +103,7 @@ Reviewed 2026-10-13. No new committed measurements. Decision matrix remains fill
 - [Interferometer fixed-mapper searches: reuse the W~ curvature matrix across likelihood calls via `preloads.curvature_matrix`](https://github.com/PyAutoLabs/PyAutoPulse/blob/main/tasks/interferometer_fixed_mapper_curvature_preload.md) — ready: Phase 1 Mind draft (2026-10-04) still unstarted; run it through start_dev. Human science call first - no production SLaM stage holds the mapper fixed today.
 - [fnnls warm-start memo: stop it slowing scattered evaluation streams (interferometer CPU, autolens_profiling#332)](https://github.com/PyAutoLabs/PyAutoPulse/blob/main/tasks/interferometer_nnls_memo_scattered_stream_guard.md) — active: Library half shipped: PyAutoArray#615 (closes #613) merged 2026-10-06, in 2026.10.7.1. Local solver-only witness (n=576): iid memo on/off 1.49x → 1.17x; walk 0.18x unchanged. Owed: re-run the autolens_profiling#332 ALMA Delaunay 2.17x after-measurement (needs compute authorization).
 - [Campaign: interferometer streaming (array-free) vs in-memory — memory and time scaling to 2e8 visibilities](https://github.com/PyAutoLabs/PyAutoPulse/blob/main/tasks/interferometer_streaming_scaling.md) — active: Phase 1 CPU cells + results/streaming_scaling rows merged (autolens_profiling#375, 2026-10-04); wiki/campaigns/interferometer_streaming.md. Optional: 1e8 CPU row and A100 rows (compute authorization). Library candidate unfiled: chunk transformer.image_from in apply_sparse_operator (memory).
+- [Interferometer large-mask likelihood (Aris 512×512): sparse data-vector memory, AdaptSplit log-det backend parity, numba route-gate cost model](https://github.com/PyAutoLabs/PyAutoPulse/blob/main/tasks/interferometer_large_mask_aris.md) — active: PyAutoArray#626 active under --auto (worktree interferometer-sparse-data-vector; witness 7.2 GB -&gt; &lt;= ~3.5 GB at 5,746 vertices); #627 and #628 queued in Mind planned.md behind the PyAutoArray claim, in that order. Record each before/after witness table in the task file at close-out.
 - [Interferometer W~ curvature matrix is FFT-bound on the mask extent: pruned padded FFT and real-space pixel scale on the A100](https://github.com/PyAutoLabs/PyAutoPulse/blob/main/tasks/interferometer_w_tilde_fft_size_levers.md) — ready: Unstarted (no PR/issue/branch/results, 2026-10-04). Select one bounded step when prioritised.
 
 </details>

diff --git a/tasks/interferometer_large_mask_aris.md b/tasks/interferometer_large_mask_aris.md
@@ -0,0 +1,44 @@
+# Interferometer large-mask likelihood — memory and routing (Aris 512×512)
+
+Campaign: interferometer
+Task: interferometer_large_mask_aris
+Reviewed: 2026-10-15
+Status: active
+
+Filed 2026-10-15 from the Aris interferometer profiling session (Delaunay + `AdaptSplit`,
+`use_jax=False` numba route and JAX jit, real-space masks 125×125 at 5,496 masked pixels and
+512×512 at 91,507 masked pixels, 5.5M visibilities, 682×682 W~ extent). The human asked for the
+three resulting PyAutoArray tasks to be logged and monitored as one Pulse speed-up campaign.
+Mind owns the repository claims and PR lifecycle (issues below); this task records the
+campaign intent, the baseline numbers the tasks must beat, and each check-in.
+
+## Members (Mind lifecycle, PyAutoArray; one claim at a time, in this order)
+
+| # | Issue | Pass condition (Witness) | Mind state 2026-10-15 |
+|---|-------|--------------------------|-----------------------|
+| 1 | [PyAutoArray#626](https://github.com/PyAutoLabs/PyAutoArray/issues/626) sparse data vector | 512×512, 5,746 vertices, numba route: peak RSS 7.2 GB → ≤ ~3.5 GB, logL unchanged to 1e-8 | active, `--auto`, Heart RED development override |
+| 2 | [PyAutoArray#627](https://github.com/PyAutoLabs/PyAutoArray/issues/627) AdaptSplit log-det parity | numba vs JAX `log_det_regularization_matrix_term` agree < 1e-3 (now 2443.22 vs 2561.98), regression test | queued behind #626 (planned.md) |
+| 3 | [PyAutoArray#628](https://github.com/PyAutoLabs/PyAutoArray/issues/628) numba route gate cost model | both masks pick the faster of direct_conv vs FFT at every tested vertex count | queued behind #627; DelaunayNN #624/#625 merged so no factory conflict |
+
+## Baseline measurements (s/call, 1 thread, laptop; PROGRESS4 steps 8.3–8.4)
+
+512×512 mask, 91,507 masked pixels, regular-step Delaunay meshes:
+
+| vertices (nnz/col) | NumPy FFT route | numba direct_conv | JAX jit (cores) | peak RSS |
+|---|---|---|---|---|
+| 1,459 (188) | 28.3 | 38.4 | 18.3 (1.59) | 4.13 GB FFT / 3.22 DC / 3.69 JAX |
+| 2,577 (106) | 44.1 | 34.2 | 32.2 (1.75) | 4.34 / 3.41 / 5.02 |
+| 5,746 (48) | 97.3 | 41.2 | not run (8.95 GB XLA compile temp) | 7.14 / 7.17 / — |
+
+- Default gate 60 nnz/col mis-routes 2,577 vertices to FFT (1.29× slower); measured crossover ~140–150 nnz/col on this mask vs ~60 (Delaunay) / ~77 (rectangular) on 125×125 (autolens_profiling#226).
+- Memory culprit: `sparse.py:104` `data_vector = mapping_matrix.T @ dirty_image` materialises the dense (91,507 × S) matrix: 1.07 / 1.9 / 4.2 GB at 1,459 / 2,577 / 5,746 vertices, route-independent.
+- Backend parity: chi², curvature, data vector, curvature log-det identical to 1e-16 relative; only the AdaptSplit regularization log-det differs (125×125: 2443.22 numba vs 2561.98 JAX; 512×512 @1,459: 2576.59 vs 2574.73). ~1 in 4 draws non-PD (numba LinAlgError vs JAX -inf).
+
+These are laptop timings under load and qualify no baseline; the campaign's after-measurements
+re-run the same cached setups and are recorded here per task. Caches live in a session
+scratchpad (temporary); the numbers above are the durable record.
+
+## Check-in 2026-10-15
+
+- Filed: #626 active (worktree `interferometer-sparse-data-vector`), #627 and #628 queued in Mind planned.md. Implementation worker running on #626; read-only root-cause investigation running for #627.
+- Next: after #626 PR-open and human `/prm`, start #627 (`/start_library`), then #628. Record each task's before/after witness table here on close-out.