Certify US Populace default: Build J 57k (rmloss100)#470
Merged
Conversation
Certifies populace-us-2024-buildj-sparse-rmloss100-75d5add-20260710T094201Z as the US default, replacing Build I. latest.json already flipped via the tracked publish. Build J re-runs the full gated stack (SSI asset stage + coverage gate + reform smoke) so an SSI resource-limit reform no longer scores $0: the ssi_asset_limit_10k_20k probe scores +$5,153,370,118. Generated by scripts/certify_data_release.py (bundle manifest + US TRO sidecar rebind + changelog fragment); pe-us stays 1.764.6, so pyproject and the model-surface snapshot are unchanged. Test release-id constants updated. UK TRO left for the post-merge generate --include-tros step. Refs PolicyEngine/populace#368, PolicyEngine/populace#359. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
MaxGhenis
marked this pull request as ready for review
July 10, 2026 11:43
This was referenced Jul 10, 2026
MaxGhenis
added a commit
that referenced
this pull request
Jul 12, 2026
) Register a staged, non-default Populace US artifact (populace_us_2024_acs_local, published in its own immutable Build L release) so it can be loaded by name without ever becoming the certified default. - Add a sibling dataset_overlays.{country} manifest section merged into the resolvable dataset registry by get_release_manifest. Overlays are strictly additive (may not shadow the certified default or any certified dataset) and survive re-certification, because certify_data_release only rewrites data_releases. - Register dataset_overlays.us.populace_us_2024_acs_local pinned to its Build L revision + sha256; rebind the US TRO sidecar to the new manifest. - Document loading the local-area dataset and filtering it for state/CD breakdowns (docs/microsim.md), and the overlay mechanism (docs/release-bundles.md). - Add a fill-in-the-id Build M US Populace certification runbook replaying #470 (docs/engineering/runbooks/), including verifying overlay survival. Fixes #471 Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
MaxGhenis
added a commit
that referenced
this pull request
Jul 17, 2026
Fixes #474. BUILD_M_RELEASE_ID populace-us-2024-buildm-sparse-rmloss100-98bf731- 20260717T075101Z replaces populace-us-2024-buildj-sparse-rmloss100- 75d5add-20260710T094201Z; policyengine-us stays 1.764.6 (built_with_model_package), so pyproject and household snapshots are untouched - the #470 common case, per the Build M runbook. Build M sparse vs Build J: final loss 0.02666 vs 0.03084, within-10% 90.21% vs 89.00%, 5,667 vs 5,510 compiled targets, all 14 release gates passed. The UK TRO sidecar's two sha fields update because it binds the same (changed) bundle manifest; not a degraded regeneration. Verified: bundle.py check green; local-area overlay preserved; manifest+certify suites 70 passed; models+US regions 57 passed against the published tag; overlay selection 7 passed; format+lint clean. Merging this PR is the default-flip decision and stays with the maintainer. Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This was referenced Jul 22, 2026
MaxGhenis
added a commit
that referenced
this pull request
Jul 22, 2026
Fixes #480. BUILD_N_RELEASE_ID populace-us-2024-buildn-sparse-rmloss100-c3e378a- 20260722T010408Z replaces populace-us-2024-buildm-sparse-rmloss100- 98bf731-20260717T075101Z; policyengine-us stays 1.764.6 (built_with_model_package), so pyproject and household snapshots are untouched - the #470 common case, per the Build M runbook. Build N sparse vs Build M: final loss 0.028242 vs 0.02666, within-10% 89.49% vs 90.21%, 5,672 vs 5,667 compiled targets. The pair is not surface-comparable: populace#486/#488 repaired 518 target values (state capital gains +$298.5B, taxable interest +$21.6B), so N holds 89.49% against a strictly harder, truthful surface. All 8 build gates passed; export-mass parity 0 failures both legs; reform-coverage smoke 39/39 including the keogh probe; the SSI one-shot Bernoulli law gate reports 0 violations (populace#477 deleted the reconcile loop). Publish contract green with medical_expense_deduction_total at the adjudicated 0.25 bound (populace#490; published row -20.57%). The UK TRO sidecar's two sha fields update because it binds the same (changed) bundle manifest; not a degraded regeneration. Merging this PR is the default-flip decision and stays with the maintainer. Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
MaxGhenis
added a commit
that referenced
this pull request
Jul 23, 2026
BUILD_O_RELEASE_ID populace-us-2024-buildo-sparse-rmloss100-22bd902- 20260722T232627Z replaces populace-us-2024-buildn-sparse-rmloss100- c3e378a-20260722T010408Z; policyengine-us stays 1.764.6 (built_with_model_package), so pyproject and household snapshots are untouched - the #470 common case. Build O attempt 1 vs Build N: final loss 0.028131 vs 0.028242, within-10% 89.56% vs 89.49%, 5,672 targets on the same truthful surface (directly comparable, unlike N-vs-M). First release to pass every gate on its first attempt - the populace#491 register alignment at work. Input coverage 163/163 now includes the populace#500 adult-care and SE-health stages: CDCC adult-care reforms score real dollars instead of a silent $0. Known boundaries stated in full in the populace#368 verdict, led by the inherited populace#507 SSI aged-band defect (do not cite for SSI baselines); attempt 2 with the populace#508 fix supersedes this default when it lands. Strict-dominance rationale in the verdict. The UK TRO sidecar's sha fields update because it binds the same (changed) bundle manifest; not a degraded regeneration. Merging this PR is the default-flip decision and stays with the maintainer. Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Certifies the Build J 57k sparse artifact as the US default, replacing
populace-us-2024-buildi-sparse-rmloss100-6e8e929-20260709T034135Z(July 9).Release:
populace-us-2024-buildj-sparse-rmloss100-75d5add-20260710T094201Z(latest.json already flipped via the tracked publish)Dataset sha256:
9d1e460d…· 57,240 households · pe-us 1.764.6 · generated byscripts/certify_data_release.pyBuild J re-runs the full gated stack on populace
main(the #369 coverage gate + reform-coverage smoke, #373 SCF asset stage, #384 TANF fix, #385 release-native loading, #387 CMS substitution register) with every gate enforced and no bypass flags. It exists to close populace#368: an SSI resource-limit reform must no longer silently score $0.Why promote (promotion scorecard — populace#368 verdict, all figures verbatim from artifacts)
The SSI probe — the reason populace#368 exists — is green at +$5,153,370,118 at 2024 (baseline SSI $55,192,291,256 → reform $60,345,661,375, $1B floor enforced). The three SSI countable-resource asset columns (
bank_account_assets,stock_assets,bond_assets) ship on the H5 with signal and are hard requirements in the coverage manifest (no exclusion), exactly as the issue demanded.Registry: Build J compiles 5,510 specs (
091ace02f962) vs Build I's 5,514 (d71c59514e3a). The delta is the #387 RI CMS Medicaid substitution (+1 injected spec: RI anchored at the ledger-verified November 2024 count, cannot-rot on backfill) and 5 new TANF thin-state zero-support exclusions (−5:fl/ia/mo/ne/njbasic assistance, which lost support once #384 made TANF take-up seeding live). Base pool data is identical to Build F/H/I; selection is exactly Build I's rmloss100 57,240 (identity join: 0 unmapped, 0 ambiguous).Caveats carried forward
Refs populace#368, populace#359.
🤖 Generated with Claude Code