Campaign session

01KX3ES7SMT162PBW6RZTRETB3

finished 2026-07-09 12:49:00.212388+00:00 → 2026-07-09 12:52:00.908817+00:00 · 0 runs · supervisor: react-agent

“Blocked no-op re-verification (0 runs): PostGIS reachable (prereq #1 cleared, 13 floors) but Phase-2's ≥40-floor prerequisite unmet — only the same 6 of 10 Phase-1 floors staged, so re-running the pilot is pure redundancy vs session 01KWYBE… (2026-07-07). C2/C3 already settled there; C1 stays blocked pending the floor import. Correct outcome per Step 2.5: re-confirm known block cheaply, do not burn budget on science that cannot move the gate.”

Archive snapshot, as of 21 h ago — the run corpus is rebuilt once a day, so this page is not a live reading. The fleet panel is the live one; it refreshes every 30 s.

Success criteria

Criterion	Resolved
Primary: bootstrap-over-floors 95% CI on LOFO inflation ratio lower bound > 1.5x. — UNRESOLVED / BLOCKED: powered test needs n≈40 floors (n0=z²σ²/d², σ≈4.82, d=1.5); PostGIS holds 13 floors, only 6 of the 10 Phase-1 replications staged. Prior session 01KWYBEJQK4307TTP9KMRY0DEY's under-powered fail (pooled ×3.37, floor-cluster CI [1.04, 9.85]) stands. Awaits Phase-2 ≥40-floor import.	no
Variance split: σ²_floor vs σ²_seed; geometry-dominates iff between-floor SD of per-floor CV(N) > between-seed SD. — RESOLVED (prior session 01KWYBE…, persisted artefacts): σ²_floor 0.0109 vs σ²_seed 0.000165 (66x); between-floor SD 0.105 vs between-seed 0.0116 (9x). Geometry dominates decisively. No new evidence this session.	yes
Comparator honesty: LOFO denominator is held-out within-floor leave-one-occupancy-out, never in-sample residual. — RESOLVED (prior session 01KWYBE…): held-out MAE 0.088 vs within leave-one-occupancy-out 0.026; in-sample-residual bug fixed. No new evidence this session.	yes

Synthesis

c-csi-cross-geometry-scaleout — session 01KX3ES7SMT162PBW6RZTRETB3 (2026-07-09): Phase-2 blocked, 0 runs

Verdict: a cheap re-verification of the Phase-2 prerequisite, which remains unmet — no simulations launched. The powered primary (C1) requires n≈40 distinct floors (n₀ = z²σ²/d², σ≈4.82, d=1.5). gis_stats reports 13 floors / 12 experiments in PostGIS, and gis_experiments confirms only 6 of the 10 Phase-1 replications: floors are staged (12439-multiroom, 1374, 147440, 7421, 16157, 12419); the other four — 10156, 10425, 6090, 8266 — are still absent, and no ≥40-floor Phase-2 import batch (ResPlan converter → gis_rebuild_nrg → gis_validate) has landed. Launching the 150-run sweep would reproduce an identical n=6 pilot with zero information gain over the 2-day-old session, and cannot move C1's powered gate. Sealed as a clean partial.

Prerequisite status

Prereq #1 (PostGIS reachable on the runner): CLEARED. gis_stats returned 2431 rows across 13 floors; the connection-refused that blocked the 2026-07-06 topology sibling (01KWWJ0REM…) is gone.
Prereq #2 (≥40-floor import for Phase 2): STILL OPEN. 13 floors present; 6 of 10 Phase-1 floors staged.

Criteria (carry-over from prior session 01KWYBEJQK4307TTP9KMRY0DEY, 2026-07-07, 90 gate-passed runs)

This session added no corpus; the resolvable criteria stand on that session's persisted artefacts/ (cross_geometry.parquet, per-floor CV(N) table, LOFO breakdown, metrics.json).

C1 (primary inflation CI) — UNRESOLVED / under-powered, blocked. Pooled ×3.37; floor-cluster 95% CI [1.04, 9.85], lower bound 1.04 < 1.5. A heterogeneous 0.70×–49.8× per-floor mixture; at 6 clusters the interval is unreliable and its exclusion of 1.0 is not leaned on. Requires Phase 2 (≥40 floors).
C2 (variance split) — RESOLVED (PASS). σ²_floor 0.0109 vs σ²_seed 0.000165 (66×); between-floor SD 0.105 vs between-seed 0.0116 (9×). Geometry dominates seed decisively in the CV(N) signal.
C3 (comparator honesty) — RESOLVED (PASS). Held-out leave-one-occupancy-out MAE 0.088 vs within-floor 0.026; the in-sample-residual bug is fixed.

Next action for the operator

Stage ≥40 ResPlan floors (start with the 4 missing Phase-1 floors), extend the campaign replications: block, then re-dispatch for the powered Phase 2. Until the floor corpus grows past ~40, C1 cannot be resolved — seed replication is cheap, floor replication is what buys the CI. No figures rendered (no new runs); the prior session's csi_cross_geometry small-multiples + CI panel and seed_trace figures remain the current visual record.

Honest scope

In-silico, ResPlan geometries, single band (2.4 GHz), CSI-only. This session is a budget-disciplined blocker re-confirmation, not a science pass.

Attached runs

The attached runs are not in the current atlas snapshot — rebuild via web --build-sim-atlas.