diff --git a/workplans/ASSEMBLY-WP-0011-cross-border-coupling.md b/workplans/ASSEMBLY-WP-0011-cross-border-coupling.md
new file mode 100644
index 0000000..1b99e02
--- /dev/null
+++ b/workplans/ASSEMBLY-WP-0011-cross-border-coupling.md
@@ -0,0 +1,118 @@
+---
+id: ASSEMBLY-WP-0011
+type: workplan
+title: "Cross-border coupling: jurisdiction-local and global scenarios"
+domain: communication
+repo: assembly-economics
+status: proposed
+flavor: implementation
+depends_on: [ASSEMBLY-WP-0013]
+owner: claude-code
+topic_slug: coulomb-social
+created: "2026-10-13"
+updated: "2026-10-13"
+---
+
+# Cross-border coupling: jurisdiction-local and global scenarios
+
+GADM is a closed economy. The jurisdiction archetypes (WP-0009-T04,
+`models/AdoptionBand.md` § 4) showed that this is the largest gap for both
+owner hypotheses ("Europe lags", "Argentina may thrive"). A jurisdiction
+with rights-first rules still runs on lineages, platforms and capital set
+abroad.
+
+Goal: two scenario modes on one model.
+
+- **Jurisdiction-local in a defined context:** one jurisdiction is simulated
+  in full; the rest of the world is an exogenous context scenario (given
+  paths for foreign lineages, prices, capital, sentiment).
+- **Global:** all jurisdictions are simulated together and exchange
+  cross-border flows at every time step.
+
+Starts only after WP-0013 gives a calibrated, reality-aligned starting
+point for each jurisdiction.
+
+## Cross-border channels: inventory and evidence
+
+```task
+id: ASSEMBLY-WP-0011-T01
+status: todo
+priority: high
+```
+
+List the channels through which jurisdictions affect each other, with
+evidence and a candidate indicator for each. Candidates: lineage import
+(models built under foreign rules: effective L, lin_open, m), platform and
+membrane reach (foreign operators' discovery and attention extraction),
+compute and capital flows (`aO` funded abroad), trade in organism output,
+talent migration (H, F), regulatory spillover ("Brussels effect"), and
+sentiment and media contagion (Σ). Choose the minimal set that can change
+the readings in § 4.3. Write `models/CrossBorder.md`.
+
+## Coupling structure in the model
+
+```task
+id: ASSEMBLY-WP-0011-T02
+status: todo
+priority: high
+```
+
+Design how jurisdictions share state: a coupling matrix per channel
+(weights from trade, platform market shares, lineage origin), and which
+stocks become partly shared (lineage control, discovery channel, organism
+capability frontier). Keep each jurisdiction's own levers. Decide the
+numerical scheme: synchronous step, all jurisdictions read the previous step
+of the others. The uncoupled case must reproduce the single-jurisdiction
+model exactly.
+
+## Multi-jurisdiction simulator
+
+```task
+id: ASSEMBLY-WP-0011-T03
+status: todo
+priority: high
+```
+
+Implement `simulate_world(jurisdictions, coupling, T)` on top of `gadm04`.
+Each jurisdiction has its own state, parameters and archetype. Tests:
+coupling 0 equals independent runs; symmetric coupling of identical
+jurisdictions equals one jurisdiction; conservation of shared quantities.
+
+## Context scenarios for jurisdiction-local runs
+
+```task
+id: ASSEMBLY-WP-0011-T04
+status: todo
+priority: medium
+```
+
+Define rest-of-world context scenarios (for example: "US-led concentrated
+frontier", "open-lineage world", "fragmented blocs") as exogenous paths of
+the coupled quantities, derived from global runs or set by hand. A local
+run is one jurisdiction × one context × one band anchor.
+
+## Global scenarios and hypothesis re-test
+
+```task
+id: ASSEMBLY-WP-0011-T05
+status: todo
+priority: high
+```
+
+Run EU, US, AR (and a J4 reference bloc) together across B1–B4, with and
+without the legitimacy loop. Re-test "Europe lags" and "Argentina may
+thrive" with lineage import. Compare local-in-context and global results:
+when does the context approximation fail? Update the break register.
+
+## Checkpoint indicators for the coupling
+
+```task
+id: ASSEMBLY-WP-0011-T06
+status: todo
+priority: medium
+```
+
+Add indicators for the coupling weights to the catalog (lineage origin of
+deployed models, cloud and platform market shares by jurisdiction, AI trade
+and investment flows), so that the matrix can be updated from checkpoints
+(WP-0010).
diff --git a/workplans/ASSEMBLY-WP-0012-model-matching-flipping.md b/workplans/ASSEMBLY-WP-0012-model-matching-flipping.md
new file mode 100644
index 0000000..5da3c87
--- /dev/null
+++ b/workplans/ASSEMBLY-WP-0012-model-matching-flipping.md
@@ -0,0 +1,109 @@
+---
+id: ASSEMBLY-WP-0012
+type: workplan
+title: "Model matching and regime flipping"
+domain: communication
+repo: assembly-economics
+status: proposed
+flavor: implementation
+depends_on: [ASSEMBLY-WP-0013, ASSEMBLY-WP-0010]
+owner: claude-code
+topic_slug: coulomb-social
+created: "2026-10-13"
+updated: "2026-10-13"
+---
+
+# Model matching and regime flipping
+
+Decide from gathered data which model (and which regime within a model)
+fits each jurisdiction best, and use the best-fitting one for predictions and
+simulations. When new measured or simulated data shows that another model or
+regime fits better, record a **flipping event**: a signal that the regime a
+jurisdiction or society operates under has changed significantly.
+
+Candidates are combinations of model version (GADM-0.2 … 0.4, later the
+coupled model of WP-0011), adoption anchor (B1–B4), archetype (J1–J4), and
+loop structure (legitimacy loop on/off, self-enhancement). Starts after
+WP-0013 provides the calibration machinery and a reality-aligned baseline.
+
+## Candidate set and fit criterion
+
+```task
+id: ASSEMBLY-WP-0012-T01
+status: todo
+priority: high
+```
+
+Define the candidate regimes and a fit criterion per jurisdiction: a
+likelihood or scaled-error score over the checkpoint indicators, with
+indicator noise and proxy uncertainty (catalog caveat: no indicator equals
+a model quantity). Penalize complexity (for example an information
+criterion), so that a richer model wins only when the data demands it.
+Write `research/ModelMatching.md`.
+
+## Matching engine
+
+```task
+id: ASSEMBLY-WP-0012-T02
+status: todo
+priority: high
+```
+
+Implement `matching.py`: for each jurisdiction, fit each candidate to the
+checkpoint series (using the WP-0013 calibration routine), score it, and
+produce a ranked table with weights (for example Akaike or Bayesian model
+weights). Output: `results/matching.md` and a machine-readable record per
+checkpoint.
+
+## Flipping events: definition and detection
+
+```task
+id: ASSEMBLY-WP-0012-T03
+status: todo
+priority: high
+```
+
+Define a flipping event: the best-fitting regime changes **and** the
+evidence for the change exceeds a threshold that persists over at least two
+checkpoints, so that noise does not flip. Distinguish measured flips (from
+observed data) and simulated flips (a run crosses into another regime's
+basin, for example the B4 boom-bust of BR-005). Record each event in a
+flip register (`research/FlipRegister.md`), linked to the break register.
+
+## Predictions from the matched model
+
+```task
+id: ASSEMBLY-WP-0012-T04
+status: todo
+priority: medium
+```
+
+Use the best-fitting regime, or a weighted ensemble, for the prediction
+register (`checkpoints/predictions.md`). Each prediction states which
+regime produced it. After a flip, new predictions use the new regime; old
+predictions remain and are scored as they were made.
+
+## Checkpoint integration
+
+```task
+id: ASSEMBLY-WP-0012-T05
+status: todo
+priority: medium
+```
+
+Run matching at every full checkpoint (and optionally at light ones), and
+report the current regime per jurisdiction and any flip candidates in the
+checkpoint record. Coordinate the runner change with WP-0010.
+
+## Test on synthetic data
+
+```task
+id: ASSEMBLY-WP-0012-T06
+status: todo
+priority: high
+```
+
+Before trusting real flips: generate synthetic indicator series from a known
+regime that switches at a known time, add realistic noise and gaps, and
+check that matching recovers the regime and detects the flip with an
+acceptable delay and false-alarm rate.
diff --git a/workplans/ASSEMBLY-WP-0013-calibration.md b/workplans/ASSEMBLY-WP-0013-calibration.md
new file mode 100644
index 0000000..8776495
--- /dev/null
+++ b/workplans/ASSEMBLY-WP-0013-calibration.md
@@ -0,0 +1,90 @@
+---
+id: ASSEMBLY-WP-0013
+type: workplan
+title: "Calibration of GADM to measured data"
+domain: communication
+repo: assembly-economics
+status: ready
+flavor: implementation
+depends_on: [ASSEMBLY-WP-0009, ASSEMBLY-WP-0010]
+owner: claude-code
+topic_slug: coulomb-social
+created: "2026-10-13"
+updated: "2026-10-13"
+---
+
+# Calibration of GADM to measured data
+
+Tune model parameters and initial states to the measured checkpoint data, so
+that scenarios start from a reasonably reality-aligned state per
+jurisdiction. WP-0011 (cross-border coupling) and WP-0012 (model matching
+and flipping) start only after this.
+
+Grounding in WP-0008 fixed some parameters from the literature. This
+workplan fits the remaining ones to observed levels and trends.
+
+## Calibration targets from checkpoint data
+
+```task
+id: ASSEMBLY-WP-0013-T01
+status: todo
+priority: high
+```
+
+Translate catalog indicators (`checkpoints/IndicatorCatalog.md`) into model
+targets per jurisdiction: an observation operator from the raw indicator to
+the normalized model quantity (for example AI firm-use share → K_Ω), with
+an uncertainty range that reflects proxy quality. Start with the stocks
+that have data: K_Ω (L1, L2), investment (L5), labour share (F1), attention
+(F4), domination (F2), sentiment (L6), ecology (E1). Write
+`models/Calibration.md` and `simulations/gadm02/calibration_targets.json`.
+
+## Calibration routine
+
+```task
+id: ASSEMBLY-WP-0013-T02
+status: todo
+priority: high
+```
+
+Implement `calibrate_data.py`: choose free parameters (those that GSA
+identifies as influential and grounding did not fix), fit them to the
+targets with a bounded, derivative-free optimizer (standard library only,
+for example Nelder–Mead or a coarse grid followed by local search), and
+report fit, residuals and identifiability (which parameters the data does
+not constrain).
+
+## Per-jurisdiction starting states
+
+```task
+id: ASSEMBLY-WP-0013-T03
+status: todo
+priority: high
+```
+
+Fit initial states and archetype levers for US, EU, DE and AR from the
+checkpoint 0 data. Where data is missing (AR adoption, sentiment), keep the
+prior and widen the uncertainty instead of guessing.
+
+## Re-run readings from the calibrated start
+
+```task
+id: ASSEMBLY-WP-0013-T04
+status: todo
+priority: medium
+```
+
+Re-run the band probe, legitimacy loop and jurisdiction archetypes from the
+calibrated starting states. Record which readings survive calibration and
+which change (break register).
+
+## Re-calibration at checkpoints
+
+```task
+id: ASSEMBLY-WP-0013-T05
+status: todo
+priority: low
+```
+
+Document how calibration is repeated when new checkpoint data arrives, and
+hand the routine to WP-0012 for model matching.