Repository navigation
Conversation
|
I pushed commit 56c59623 and opened #11019 as a draft against The two partner files change 24 expected numeric literals and 14 grant/date comment blocks. Both inverse overlays recover the original files exactly. All 20 case names and input objects, all 41 output paths, order and 0.1 margins remain unchanged. Bulletin 25-15 and Bulletin 26-16 supply the four three-month increments: $21, $15, $18 and $6. Existing annual literals supply conditional API anchors; native HEAD values only corroborate the arithmetic. The complete diagnostic collected all 4,962 native observations across 237 cases and 2,481 assertion leaves per phase. BASE matched every old assertion. HEAD matched 2,457 and changed only the 24 proposed fields; 17 other outputs in the two files and 2,440 selected off-file pairs stayed unchanged. All 144 affected case-month checks kept their prior branches and confined changes to fourth-quarter cash. The 386 unselected cases remain unmeasured by this diagnostic. Before commit, the one bounded verification passed all 25 existing cases, including both entire partner files and the five grant cases/94 new source controls. Pytest took 9.32 seconds; native wall time took 62.79 seconds, with 3,133,718,528 bytes peak RSS. This commit matches those tested source bytes. Required The first GitHub workflow at this head, 43803162385, completed with a skipped conclusion. All seven check contexts skipped, so that run supplies no passing CI evidence. The PR remains a draft. At 2026-11-16 04:29 UTC, Subfleet queued FULL1 partner-review job The old above/below wage names retain the inherited 0.33 software boundary; all four wages pass the source Part A-only exact-third calculation. These updates maintain that software contract and do not qualify complete legal awards or legal cliffs. #10971 retains the Part A/Part B, receipt/history, EID, election, floor and precision questions; #10639 retains its resource work. Axiom’s existing C-110 block-2 module remains the grant-component target. The source payload supplies the 44 printed cells, additional-person rule and 94 September/October monthly component companions. The 24 legacy annual cash totals do not supply legal companion goldens. I filed the grant-only source issue with |
|
I pushed commit 56c59623 and opened #11019 as a draft against The two partner files change 24 expected numeric literals and 14 grant/date comment blocks. Both inverse overlays recover the original files exactly. All 20 case names and input objects, all 41 output paths, order and 0.1 margins remain unchanged. Bulletin 25-15 and Bulletin 26-16 supply the four three-month increments: $21, $15, $18 and $6. Existing annual literals supply conditional API anchors; native HEAD values only corroborate the arithmetic. The complete diagnostic collected all 4,962 native observations across 237 cases and 2,481 assertion leaves per phase. BASE matched every old assertion. HEAD matched 2,457 and changed only the 24 proposed fields; 17 other outputs in the two files and 2,440 selected off-file pairs stayed unchanged. All 144 affected case-month checks kept their prior branches and confined changes to fourth-quarter cash. The 386 unselected cases remain unmeasured by this diagnostic. Before commit, the one bounded verification passed all 25 existing cases, including both entire partner files and the five grant cases/94 new source controls. Pytest took 9.32 seconds; native wall time took 62.79 seconds, with 3,133,718,528 bytes peak RSS. This commit matches those tested source bytes. Required The first GitHub workflow at this head, 43803162385, completed with a skipped conclusion. All seven check contexts skipped, so that run supplies no passing CI evidence. At 2026-11-16 05:51 UTC, I cancelled FULL1 job The old above/below wage names retain the inherited 0.33 software boundary; all four wages pass the source Part A-only exact-third calculation. These updates maintain that software contract and do not qualify complete legal awards or legal cliffs. #10971 retains the Part A/Part B, receipt/history, EID, election, floor and precision questions; #10639 retains its resource work. Axiom’s existing C-110 block-2 module remains the grant-component target. The source payload supplies the 44 printed cells, additional-person rule and 94 September/October monthly component companions. The 24 legacy annual cash totals do not supply legal companion goldens. I filed the grant-only source issue with |
Texas HHSC raises TANF maximum grants from October 1, 2026, in Bulletin 26-16. This change adds all 44 applicable printed grant amounts and the $89 additional-person amount, preserves earlier entries, and adds September/October component tests. The one-person/two-parent slot remains inapplicable.
The two partner files also update 24 annual expected values and 14 comment blocks for the fourth-quarter grant change. Case names, inputs, margins, order and every other byte remain unchanged. The annual expectations retain each existing fixture’s software-contract premises: old literal + three months × (new printed grant − old printed grant). Bulletin 25-15 supplies the old grants through September 2026. Both bulletins leave budgetary and recognizable-needs amounts unchanged.
Every changed partner field follows below. Paths share
policyengine_us/tests/policy/baseline/partners/analytics_coverage/edge_cases/.Source groups give the three-month increments: WITH_SECOND_3 uses 418→425, giving $21; WITHOUT_SECOND_2 uses 331→336, giving $15; WITHOUT_SECOND_3 uses 382→388, giving $18; NON_CARETAKER_1 uses 130→132, giving $6. January–September retain the old grants; October–December use the new grants.
The old above/below wage names retain the inherited 0.33 software cutoff. All four wages pass the source Part A-only exact-third calculation, so those names do not establish legal cliffs. These updates preserve the existing software boundary purpose and do not qualify complete legal awards. #10971 retains the broader Part A/Part B, receipt/history, EID, election, floor and precision questions; #10639 retains its resource changes. This change alters no eligibility, deduction, history or precision formula.
Validation:
make formatand its Ruff check passed. Ruff left all 8,441 files unchanged; byte checks preserved all 22,545 source files across formatting. No test rerun followed unchanged formatting.At 2026-11-16 05:51 UTC, I cancelled FULL1 job
20261115-232903-history-tx2026-56c59623-full1before any attempt started. I queued replacement20261116-005113-history-tx2026-56c59623-full1-canonicalfor this head to correct the review cache namespace. The replacement continues the first full review round, and its submission receipt records zero attempts and no accepted result. At 2026-11-16 05:53 UTC, GitHub reported workflow 43806342989 in progress at this head. That reviewer must recompute the 24 changed values from primary grant sources and the preserved contract premises, inspect all 41 outputs and the boundary purposes, and confirm the intended scope. The local tests and author source checks do not supply that independent review.Axiom still needs the monthly component/effective-date update in the existing
us-tx/policies/hhs/texas-works-handbook/c-110-tanf/block-2.yamlmodule and adjacent companion tests. The grant-only source issue supplies the 44 printed cells, size-16 extension controls and 94 September/October component expectations. The 24 conditional annual partner totals do not supply legal Axiom companion goldens. Authenticated bulletin citation versions and signed encoding remain pending; no billed encoder run has authorization.axiom: TheAxiomFoundation/rulespec-us#2288 queued