Certify Build N US Populace (populace-us-2024-buildn-sparse-rmloss100-c3e378a) as the certified default

[MaxGhenis]

Certify the Build N US Populace release as the certified default, replacing Build M:

• New: populace-us-2024-buildn-sparse-rmloss100-c3e378a-20260722T010408Z (published 2026-07-22, immutable tag on policyengine/populace-us; latest.json deliberately untouched)
• Outgoing: populace-us-2024-buildm-sparse-rmloss100-98bf731-20260717T075101Z
• Model: policyengine-us==1.764.6 (unchanged; compatibility_basis: built_with_model_package)

Build N sparse vs Build M on the release scorecard: final loss 0.028242 vs 0.02666, within-10% 89.49% vs 90.21%, 5,672 vs 5,667 compiled targets. The headline pair is not surface-comparable: Build N repaired the target surface itself — populace#486 (capital-gain-distributions level repair) and populace#488 (cross-vintage aging chain completion) corrected 518 target values (state capital gains +$298.5B, taxable interest +$21.6B), so N is scored against a strictly harder, truthful surface than M was.

Gates and receipts, all green on the published release manifest:

• All 8 build gates passed (base_population_scale, immigration_composition, ecps_parity, hours_worked_signal, snap_take_up_signal, eligibility_inputs_signal, pregnancy_signal, snap_discretionary_exemption_signal). The Build-M-era SSI reconcile-loop gates no longer exist: populace#477 deleted the loop and replaced it with the one-shot Bernoulli law gate, which reports bernoulli_law_violation_count = 0.
• Export-mass parity: 0 failures on both legs (base_frame_vs_reference, export_vs_base_frame).
• Reform-coverage smoke: 39/39 probes green, including keogh_distribution_neutralization.
• QRF tail concentration: 14-entry evidenced exclusion register (sha-recorded, fails loud if dormant), owners populace#481/LRUCache.add silently keeps the old value for an existing key #487/Remove policyengine-us-data from the certified bundle package scaffold #432/Notify consumer repos after PyPI release #451.
• Publish contract: green with medical_expense_deduction_total at the adjudicated 0.25 broad-fit bound (populace#490; published row −20.57%). The durable fix (shared critical-target register + boosted loss weight) is staged on populace branch loss-contract-alignment; the 0.15 bound restores once a boosted run holds it.
• CD surface ships off per the populace#449 verdict (CD-on halved within-10%); the parity generator carries a fail-closed CD_SURFACE_REGIME = "off" constant.

Runbook: docs/engineering/runbooks/build-m-us-populace-certification.md (same flow; #470 common case — model package unchanged, so pyproject and household snapshots are untouched). The merge of the certification PR is the default flip and stays with the maintainer.