ASSEMBLY-WP-0017-T06: Census private office construction (data centers) as investment indicator L5b; T05 unblocked

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Assistant: claude-code
Assistant-Model: opus
Assistant-Process: 14913@bnt-lap001
Assistant-Session: add7ef69-b901-4756-9a67-34cf1125d96f

checkpoints/IndicatorCatalog.md

@@ -236,3 +236,16 @@ channel, the strength of host access rules, and frontier access.
 
 CB2 is the observable for the access-rule strength a. At the 2026-10-13
 review it gave a ≈ 0 (at most 0.006 per year).
+
+## Investment indicator for the regime filter (ASSEMBLY-WP-0017-T06)
+
+| ID | Indicator | Jurisdiction | Source | Access | Cadence |
+|---|---|---|---|---|---|
+| L5b | Private construction put in place, Office (includes data centers), SAAR | US | Census Construction Spending, `https://www.census.gov/construction/c30/xlsx/release.xlsx`, Table 1 (second "Office" row = private) | file (fixed layout; last 5 months + year-ago month) | m |
+
+L5b is closer to organism investment than L5, which covers all
+information-processing equipment. The regime filter's reliability rests on
+an investment indicator (`research/ModelMatching.md` § 4.2). Hyperscaler
+capital expenditure (company filings) is a second candidate, but the 2026
+guidance figures conflict across sources and are not audited quarterly
+values.

simulations/gadm02/observed_series.json

@@ -252,6 +252,38 @@
      0.8519
     ]
    ]
+  },
+  "US_dc_construction": {
+   "indicator": "L5b",
+   "unit": "millions of USD, SAAR (private office construction, includes data centers)",
+   "source": "U.S. Census Bureau, Construction Spending release 2026-10-01 (https://www.census.gov/construction/c30/xlsx/release.xlsx), Table 1, private Office: Aug 2025, Apr-Aug 2026. Data centers are reported within Office and drive its growth (+29.8 % yoy while office proper is weak). Operator: proportional to gross organism investment; preferred over L5 for the regime filter",
+   "log_sigma": 0.05,
+   "points": [
+    [
+     0.0,
+     103300
+    ],
+    [
+     0.6667,
+     111220
+    ],
+    [
+     0.75,
+     115159
+    ],
+    [
+     0.8333,
+     121940
+    ],
+    [
+     0.9167,
+     128199
+    ],
+    [
+     1.0,
+     134119
+    ]
+   ]
   }
  }
 }

workplans/ASSEMBLY-WP-0017-reliable-flip-detection.md

@@ -96,13 +96,15 @@ years (`matching.py --synthetic`). Add the criteria as a slow test.
 
 ```task
 id: ASSEMBLY-WP-0017-T05
-status: wait
+status: todo
 priority: medium
 state_hub_task_id: "93477fd0-6d90-51dd-a41a-75a22715bf29"
 ```
 
-Waits for T06 (T04 passed, but the result rests on an AI-specific investment
-series). Then run matching at every full checkpoint (optionally at light ones), report the current regime and any flip candidates in the
+Unblocked 2026-10-13 (T04 passed, T06 provides L5b). First extend
+`main_observed` to the IMM filter with the observed series (adoption, L5b,
+discovery, sentiment once available), then run matching at every full
+checkpoint (optionally at light ones), report the current regime and any flip candidates in the
 checkpoint record, and coordinate the runner change with WP-0014 (governed
 execution profile).
 
@@ -110,11 +112,17 @@ execution profile).
 
 ```task
 id: ASSEMBLY-WP-0017-T06
-status: todo
+status: done
 priority: high
 state_hub_task_id: "97f58120-60b4-5583-82b2-a227031fd989"
 ```
 
+**Done 2026-10-13:** L5b, Census private Office construction (includes data
+centers; +29.8 % yoy), added to `observed_series.json` (Aug 2025, Apr–Aug
+2026) and to the indicator catalog. Hyperscaler capex was considered and set
+aside: the 2026 figures are guidance and conflict across sources. Noise for
+the filter: log sd 0.05, as tested.
+
 The filter's reliability rests on investment (without it, false flips are 5 of
 6). L5 covers all information-processing equipment. Find and wire a series
 closer to organism investment, for example hyperscaler capital expenditure