[MaxGhenis]

Brings the federal and Alaska to Hawaii entries in source_conflicts/candidates.yaml up to date with the code on main, and adds candidates for conflicts the merged code's own comments record. One of three PRs splitting this refresh by state; the follow-up promised in #10349.

Why. #10287 wrote these candidates while their implementing PRs were still open, so each said code: [] or "Code not yet on main". A PR each one names has since merged: #8861, #9072, #9162, #9213, #9243, #9467, #9611, #9623, #9624, #9625, #9636, #9643, #9661, #9696, #9709, #9858, #9878, #9944, #9967, #9968, #10022, #10060, #10068, #10087, #10089, #10097, #10135, #10137, #10176, #10194, #10272, #10279, #10299.

Since the first review. This head addresses the GPT-6.1 Sol review:

• Merged main after Mark nine source conflicts now in the shared register #10441 moved nine of the file's entries to the shared register. Where a block this PR edited moved, the block is deleted rather than kept.
• Re-scanned main for entries whose PR has merged since the first round and refreshed those: the Arkansas casualty-loss pair (Apply separate-filer casualty limits in Arkansas and statutory loss limits in Puerto Rico #9944), the four Georgia LIHEAP entries (Add Georgia LIHEAP regular heating assistance #9968) and the dependent standard deduction worksheet entry (Use W-2 box 1 wages in the dependent standard deduction worksheet and Medicaid filing test #10272). Two of them flip follows; see below.
• Added the 110%-floor candidates that main records for the District (Apply the federal 110% poverty floor to state LIHEAP programs (DC, MA, MS, NJ, OR) #10279) and Riverside County (Apply the federal 110% poverty floor to state LIHEAP programs (DC, MA, MS, NJ, OR) #10279).
• Hawaii disabled exemptions: the note now says a nondisabled spouse's disabled exemption is zero.
• ACTC housing-deduction bar: the note now excepts 2021, when the credit is fully refundable, and lists fully_refundable.yaml and foreign_earned_income_exclusion_bar_applies.yaml.
• Filled the Georgia manual and plan URLs from the code's own references.
• Refreshed the Alaska and Hawaii ABAWD timing entries and the FY2026 large-household cap entry: Follow the SNAP statute on ABAWD timing in Alaska and Hawaii and the FY2026 large-household cap #10299 merged, so their notes no longer call the change in flight and say what the merged code does.

follows flips. Each changed because the merged code does something different from what the candidate recorded:

• us-ar:casualty-loss-status-four-spouse-floors: conflicting-source to statute. Changed by Apply separate-filer casualty limits in Arkansas and statutory loss limits in Puerto Rico #9944: PE-US now follows the statute. For spouses filing separately on the same return, each spouse's loss is reduced by the $100 exclusion and then by 10% of that spouse's Arkansas AGI; only joint-return spouses count as one individual, and the input records one casualty per person.
• us:dependent-deduction-worksheet-earned-income: unknown to gap-fill. PE-US follows the worksheet: earned income is W-2 box 1 wages (after elective deferrals), earned disability benefits and Schedule 1 business and farm income, less the deductible part of self-employment tax, summed for the head and spouse and floored at zero; the Medicaid filing test and Oregon's and Hawaii's dependent deductions read the same measure.

Refreshed candidates (38). Each now lists the tracked files that decide it under pe_us.code. Its note says which source that code follows, as read on main, and quotes nothing the code does not show. pe_us.follows stays unknown wherever the readings begin "No statute supplied".

Candidate	Code	What the code follows
us-ak:snap-abawd-implementation-date	hr1_in_effect.yaml, is_snap_abawd_hr1_in_effect.py	PE-US follows the statute: since #10299 Alaska's HR1 toggle is true from July 4, 2025, and the parameter's comment says section 10102 has no later effective date, that USDA applies the amended exceptions from enactment, and that no authority for a statewide delay before Alaska's November 1, 2025 good-faith exemption was found (statute)
us-al:irrevocable-trust-loss-deduction	gross_income_sources.yaml, al_agi.py	PE-US follows the statute and current rule: estate_income counts signed in Alabama gross income, so a beneficiary's loss passes through (statute)
us-al:liheap-insurance-payments	unearned.yaml, al_liheap_countable_income.py	The unearned income list cites manual sections 5.5-5.5.2 and has no insurance or annuity input of its own; it counts pension_income, disability_benefits and workers_compensation (unknown)
us-al:liheap-lump-sums-and-refunds	al_liheap_countable_income.py, unearned.yaml	PE-US follows the manual: gambling winnings count net of losses, and the unearned income list has no refund, rebate or other lump-sum input (unknown)
us-al:medicaid-parent-dependent-child-age	age_limit.yaml	PE-US follows the administrative code: Alabama's age limit is 19 from 2014 with no student condition, and the file's comment records the S25 conflict (unknown)
us-ar:casualty-loss-aggregate-agi-floor	income_floor.yaml, ar_casualty_loss_deduction_joint.py, ar_casualty_loss_deduction_indiv.py	PE-US follows the statute and form: one 10% floor applies to a return's combined losses after the exclusion, the couple's total on a joint return and each spouse's own losses when spouses file separately on one return (statute)
us-ar:casualty-loss-status-four-spouse-floors	ar_casualty_loss_deduction_indiv.py, exclusion.yaml, income_floor.yaml	Changed by #9944: PE-US now follows the statute. For spouses filing separately on the same return, each spouse's loss is reduced by the $100 exclusion and then by 10% of that spouse's Arkansas AGI; only joint-return spouses count as one individual, and the input records one casualty per person (statute)
us-ar:gambling-loss-deduction-before-2015	ar_gambling_loss_deduction.py	PE-US follows DFA's guidance: ar_gambling_loss_deduction has no 2015 start date, and its 2014 test deducts losses up to winnings (conflicting-source)
us-ar:income-tax-above-table-formula	threshold.yaml, ar_income_tax_helpers.py	PE-US follows the bracket card's formula at and above the table threshold, and the threshold's comment records the conflict with the booklet note (gap-fill)
us-ca:care-fera-housing-subsidies	income_sources.yaml, ca_cpuc_countable_income.py	PE-US follows D.14-08-030 and PG&E's stated practice: housing assistance leaves the countable income list from 2015, and the list's comment rests FERA coverage on PG&E's June 5, 2026 reply brief, not on the decision (unknown)
us-ca:itemized-limit-casualty-exclusion	excluded_deductions.yaml, ca_casualty_loss_deduction.py, ca_itemized_deductions.py	PE-US follows the statute: California's own casualty deduction is excluded from the high-AGI limitation (statute)
us-ca:santa-clara-ga-vehicle-boundary	ca_scc_general_assistance_countable_vehicle_value.py, vehicle_value_limit.yaml	PE-US follows the handbook: a vehicle worth exactly $4,650 is exempt (unknown)
us-ca:ssp-nmohc-couple-standard	out_of_home_care.yaml	PE-US follows the statute: the 2026 allowance is $1,153.31, from a $3,252.14 couple standard, twice the $1,626.07 individual rate, and the file's comment lists the sources on each side (statute)
us-co:oap-initial-application-income-test	co_oap.py, co_oap_own_countable_income.py	PE-US follows the regulation: each month it compares total countable income after the 3.533 disregards with the grant standard, and has no separate gross-income test at initial application (unknown)
us-dc:eitc-no-tin-child-schedule-before-2023	qualifying_child_tin_required.yaml, dc_eitc_has_qualifying_child.py	PE-US applies the childless schedule in every year from 2015 when no qualifying child has an SSN or ITIN (unknown)
us-de:529-subtraction-exact-agi-limit	de_529_plan_subtraction.py, de_529_plan_subtraction_joint.py, agi_limit.yaml	PE-US follows the statute: a filer whose federal AGI equals the limit keeps the subtraction, and the limit's comment reads the statute as barring only AGI "greater than" the limit (statute)
us-de:elderly-disabled-exclusion-age-sixty	de_elderly_or_disabled_income_exclusion_joint.py, de_elderly_or_disabled_income_exclusion_eligible_person.py, age_threshold.yaml	PE-US follows the worksheet: a filer or spouse aged exactly 60 qualifies (age at least age_threshold), on joint returns and on other returns (conflicting-source)
us-ga:liheap-income-floor-vs-smi-limit	ga_liheap_eligible.py, fpg_floor.yaml, fpg_year_lag.yaml, ga_liheap_income_limit.py	PE-US follows the statute: income below 110% of the prior year's poverty guideline, rounded to cents, qualifies even above the 60% SMI limit. The comment says the limit and the printed matrices fall below that floor from size 16 in FY2025 and FY2026 and that Georgia's plans assure the floor (statute)
us-ga:liheap-lawful-noncitizen-household-members	ga_liheap_household_size.py, ga_liheap_eligible.py	PE-US follows the eligibility sections: lawful immigrants count in household size and undocumented members do not, and the comment records page 57's bar on all-noncitizen households and the glossary's exclusion from household size (unknown)
us-ga:liheap-plan-manual-income-inclusions	person.yaml, ga_liheap_income.py	PE-US follows the manual: the adult income list counts gambling winnings, cash gifts (financial_assistance), disability benefits and workers' compensation, and its comments record that the FY2025 and FY2026 plans leave those categories unchecked (unknown)
us-ga:liheap-subsidized-renter-energy-burden	ga_liheap_eligible.py, _ga_liheap_housing.py	PE-US follows the manual: every household needs an energy burden, which a subsidized renter with heat in rent shows through a heating bill or a direct electric bill, and the comment records that FY2026 plan item 2.3 answers No to differing policies for subsidized renters (unknown)
us-hi:joint-return-claimable-spouse-exemption	exemptions_count.py, hi_regular_exemptions.py	PE-US follows the per-individual statute: exemptions_count leaves out only the filer who can be claimed, so the other spouse keeps a regular exemption, and hi_regular_exemptions adds the age exemption only for filers who cannot be claimed (statute)
us-hi:joint-return-one-disabled-spouse-exemptions	hi_disabled_exemptions.py, hi_exemptions.py	PE-US follows the regulation and instructions: each spouse takes the larger of their disabled exemption ($7,000 if disabled, otherwise zero) and their own regular and age exemptions, so the nondisabled spouse keeps theirs (unknown)
us-hi:snap-abawd-implementation-date	hr1_in_effect.yaml, is_snap_abawd_hr1_in_effect.py, hi.yaml	PE-US follows the statute: since #10299 Hawaii's HR1 toggle is true from July 4, 2025, and the parameter's comment records DHS's announced November 1, 2025 start and says no good-faith exemption covering July through October 2025 was found (statute)
us:additional-child-credit-housing-deduction-bar	section_911.py, refundable_ctc_barred_by_section_911_exclusion.py, fully_refundable.yaml, foreign_earned_income_exclusion_bar_applies.yaml	PE-US follows the instructions: elects_section_911_exclusion treats a filer who claims only the section 911(c)(4) housing deduction as electing section 911, so refundable_ctc_barred_by_section_911_exclusion denies the additional child tax credit from 2015 except in 2021, when the credit is fully refundable and the bar does not apply; the helper's docstring says this case rests on the IRS instructions, not the statute (conflicting-source)
us:dependent-deduction-worksheet-earned-income	standard_deduction_worksheet_earned_income_person.py, standard_deduction_worksheet_earned_income.py, earned_income_sources.yaml, basic_standard_deduction.py	PE-US follows the worksheet: earned income is W-2 box 1 wages (after elective deferrals), earned disability benefits and Schedule 1 business and farm income, less the deductible part of self-employment tax, summed for the head and spouse and floored at zero; the Medicaid filing test and Oregon's and Hawaii's dependent deductions read the same measure (gap-fill)
us:education-benefits-joint-claimed-spouse	is_barred_from_education_credits_by_dependency.py, student_loan_interest_ald_eligible.py	PE-US follows the Form 1040 instructions: a claimed head or spouse bars the whole joint return from the education credits and the student loan interest deduction (conflicting-source)
us:eitc-trust-royalties-investment-income	eitc_relevant_investment_income.py, passive_estate_income.py	PE-US follows the worksheet: the passive basket counts passive_estate_income, Schedule E Part III column (d), and leaves out the box 5 royalties, annuities and income in respect of a decedent that go in nonpassive column (f) (conflicting-source)
us:filing-chart-a-blindness-and-equality	tax_unit_is_required_to_file.py, applies.yaml	PE-US follows the statute for 2018 to 2025: a return is required only when gross income exceeds the full standard deduction, blindness included (statute)
us:filing-claimable-married-separate	tax_unit_is_required_to_file.py	PE-US follows the statute: a married individual filing separately, claimable or not, must file with any positive gross income, since the separate threshold is the exemption amount, zero while it is suspended (statute)
us:filing-claimable-spouse-unearned-income-additions	tax_unit_is_required_to_file.py	PE-US follows Chart B and Publication 501 Table 2: the unearned income threshold for a filer who can be claimed is the section 63(c)(5)(A) amount plus their own additional standard deduction for age or blindness, on joint returns too, and the comment records that section 6012(f)(2)(D) names the 63(c)(5)(A) amount alone (conflicting-source)
us:hsa-limit-indexing-windows	uprating_extensions.py, self_only.yaml, family.yaml	PE-US follows IRS practice: projections after the published limits use 12-month windows ending March 31, except the 2016 C-CPI-U-to-CPI ratio, which keeps the window ending August 31, and the docstring says these windows reproduce every published 2019-2027 limit (conflicting-source)
us:separate-filer-zero-income-threshold	tax_unit_is_required_to_file.py	PE-US requires a separate filer to file only with positive gross income, matching neither the literal statute, which reaches a separate filer with none, nor the IRS's $5 relief (unknown)
us:snap-bbce-income-chart-rounding	tanf_non_cash_gross_income_limit.py, meets_tanf_non_cash_gross_income_test.py	PE-US uses the exact percentage of the poverty guideline, not any state chart's rounding, and the comment notes that where a chart rounds up, a household between the two limits is ineligible here though the chart admits it (unknown)
us:snap-conditional-entrant-waiting-period	subject_immigration_statuses.yaml, meets_snap_qualified_alien_waiting_period.py	PE-US follows the regulation: conditional entrants are on the five-year clock from April 2003, the list's comment records the FNS memo's conflicting label, and battered aliens have no immigration status value (unknown)
us:snap-large-household-cap-effective-date	max_allotment.yaml, snap_max_allotment.py	PE-US follows the statute: since #10299 snap_max_allotment applies the cap from October 1, 2025, citing 7 U.S.C. 2012(u)(3)(C)'s annual computation and FNS's September 4, 2025 memo, and the cap's comment says the FY2026 table prints no cap (statute)
us:snap-minimum-allotment-rounding-base	snap_min_allotment.py, published_adjustment.yaml	PE-US rounds 8% of the rounded-down one-person maximum, as the regulation does, and adds a published adjustment where USDA's figure, computed from the unrounded cost, differs: $1 in the 48 states and DC for FY2027 only (unknown)
us:snap-waiting-period-age-based-ssi	is_snap_alien_waiting_period_exempt.py, is_usda_disabled.py, disabled_programs.yaml	PE-US exempts anyone receiving SSI on any basis: is_snap_alien_waiting_period_exempt reads is_usda_disabled, which counts any SSI receipt among the benefits in gov.usda.disabled_programs (conflicting-source)

New candidates (2). The merged code records each conflict in its comments; none had a candidate or a register entry.

• us-ca:riverside-liheap-income-floor-vs-smi-limit (statute). Can Riverside County's LIHEAP income limit exclude a household of 20 or more whose income is below the federal 110% poverty-guideline floor? PE-US applies the floor to the prior year's guideline, rounded to cents, with the statute's strict "less than"; its own limit is 60% of SMI, which the comment says falls below the floor from size 20 in 2025 and size 21 in 2026 and 2027 (ca_riv_liheap_eligible.py).
• us-dc:liheap-income-floor-vs-smi-limit (statute). Can the District's 60% SMI limit exclude a household of 34 to 39 whose income is below the federal 110% poverty-guideline floor? PE-US applies the floor to the prior year's guideline, rounded to cents, with the statute's strict "less than" (dc_liheap_eligible.py).

follows left unknown beyond the "No statute supplied" convention. These readings name no statute or leave the statutory question open, so statute, conflicting-source and gap-fill would each assert a legal reading the code does not make: us-dc:eitc-no-tin-child-schedule-before-2023 (the readings leave the statutory question open); us-hi:joint-return-one-disabled-spouse-exemptions (the readings leave the statutory question open); us:separate-filer-zero-income-threshold (the readings leave the statutory question open); us:snap-bbce-income-chart-rounding (no source is a statute); us:snap-conditional-entrant-waiting-period (no source is a statute); us:snap-minimum-allotment-rounding-base (the readings leave the statutory question open).

Verification.

• policyengine_us/tests/code_health/test_source_conflicts.py: 103 passed at this head. I ran a copy outside the package, so pytest doesn't import policyengine_us.
• The file is main's plus this PR's entries: every other entry is byte-identical. Each changed entry was re-dumped with the PyYAML settings the file was written with (width 110).
• Ids stay sorted and unique, and every code path is a file git tracks.
• New floor candidates: each figure was recomputed from the code's formula and parameters on main, and each plan page was read.
• The three PRs touch disjoint entries; merged together they apply cleanly. After all three, no candidate says "Code not yet on main" or has code: [] while naming only merged PRs.

Not in this PR. The nine Florida and South Dakota LIHEAP candidates landed in #10349. Candidates whose pull requests are still open keep code: []. The other two parts of this refresh are #10351 and #10352.

axiom: n/a: source-conflict backlog metadata only; no model change

🤖 Generated with Claude Code

[MaxGhenis]

Hand-off to the US hub, at head 1660833.

• What it does: this PR points candidates whose PRs have merged at their code on main and rewrites their notes. It also adds candidates for conflicts that merged code comments record. It edits only source_conflicts/candidates.yaml, plus a changelog fragment. It already contains main at 41d218c, so the twelve candidates that Mark nine source conflicts now in the shared register #10441 moved to the register stay deleted.
• Review:
  • GPT-6.1 Sol's round-2 delta review approved this PR at 7b496b2.
  • An Opus 5.5 delta review approved it at 59ef2f7, with minor wording findings. The Codex lanes were closed.
  • 1660833 applies that review's wording verbatim.
  • A second Opus 5.5 check of that commit alone also approved. It confirmed that the commit changes only the fields those findings name, and that each new claim holds on main: the Hawaii waiver in waived_counties/hi.yaml, and the 2025-07-04 dates in the Alaska and Hawaii hr1_in_effect.yaml parameters.
• Checks:
  • Locally, the guard test (test_source_conflicts.py) passed all 103 tests at this head. It also passed on the file that all three PRs produce merged together.
  • CI is still running at this head. The PR is mergeable. Merging is blocked only by CI and the required approving review.
• Merge order: no dependencies. This PR, Point the IN-MT source-conflict candidates at their code on main #10351 and Point the NM-WY source-conflict candidates at their code on main #10352 merge with one another in any order without conflicts. I checked this pairwise and all three together with git merge-tree.
• Follow-ups owed: none from this PR. The SSP, MD CCS and NY HEAP candidates are in Add source-conflict candidates for SSP, Maryland CCS and NY HEAP #10444.

[MaxGhenis]

US + core hub merge audit, us#10350 at 1660833e4bcded3d7d9ce918126b9e3e73f417d3 (squash): Point the federal and AK-HI source-conflict candidates at their code on main.

• Review: GPT-6.1 Sol round-2 delta APPROVE at 7b496b2 (~/reviews/source-conflict-candidates-2026-10-21/review_r2.md); then, with the Codex lanes closed, Opus 5.5 delta reviews of the main merge at 59ef2f7 (review_r3_opus.md, minor wording) and of the wording-fix commit 1660833 (review_r4_opus_delta.md, APPROVE; checked each new claim on main).
• Gates checked live at merge:
  • gh pr checks exits 0;
  • the latest Pull request run at the head concluded success;
  • MERGEABLE;
  • not a draft;
  • no reviewer's latest review requests changes;
  • the semantic-overlap guard;
  • head pinned.
• Impact: None: the PR edits only source_conflicts/candidates.yaml (the conflict backlog, read by the guard test, not the model) and a changelog fragment (hub read the file list).
• Order: No dependencies; Point the federal and AK-HI source-conflict candidates at their code on main #10350, Point the IN-MT source-conflict candidates at their code on main #10351 and Point the NM-WY source-conflict candidates at their code on main #10352 merge pairwise and together without conflicts (owner's git merge-tree check).
• Axiom: n/a: source-conflict backlog metadata only; no model change