This is a pure structural refactor of src/domain/load.ts into load.ts/editions.ts/standing.ts/history.ts, with importers re-pointed. Diffing the removed block against the new standing.ts and history.ts and editions.ts, the logic is moved verbatim: RATIFICATION_MIN_PANEL stays 4, the reconsideration-cannot-be-ratified-by-checks-it-saw rule, panelStaleness/currentChecks set-aside behaviour, withinOneStep with the ungraded states at 2.5, survivingObjections and crossModelSummary are all character-identical — no §3.15 check is loosened or dodged. No content/ files are touched, no citations are added (verification reports none), and AGENTS.md is untouched. Docs changes (DECISIONS.md, AUTOMATION.md, DATA_MODEL.md) describe the split accurately and are stamped with date and channel as the founder-direction rule permits; the remaining edits are stale-path string fixes (scripts/lib → src/lib) in comments and one run-account header, which correctly renames the generator that actually produced it. Only nits: editions.ts imports `editionErrors` from load.ts without using it (and creates a mutual import with load.ts), and load.ts is left with an orphan doc comment at EOF — cosmetic, not constitutional.

The change relocates edition, standing, and history functions without altering their substantive logic, while updating the visible importers to their new owning modules. The ratification, stale-check, vendor-seat, disagreement, and reconsideration safeguards remain intact, so no §3.15 check is visibly weakened or bypassed. No content records, citations, verification labels, assessments, or confidential materials are added. The decisions entry records the change append-only with a date and channel, and nothing in the diff positively contradicts that provenance.

The change is a clean structural refactoring that decomposes `src/domain/load.ts` into modular domain files (`load.ts`, `editions.ts`, `standing.ts`, and `history.ts`) with zero normative or logic alterations. All function signatures and behaviors are preserved across call sites, tests, and documentation. No constitutional rules, epistemic checks, or data invariants in AGENTS.md §3 are weakened or violated.

Pure module split of load.ts into load/editions/standing/history with importer re-points; AGENTS.md untouched. Function bodies for ratification, concurrence, staleness, and history appear moved intact with no check-weakening. No content/ citations, confidence material, or fabricated provenance. Docs only record the founder-directed refactor and path comment fixes (scripts/lib → src/lib).

This is a pure refactor: `src/domain/load.ts` is split into `load.ts`, `editions.ts`, `standing.ts`, and `history.ts`, with the moved functions (ratification, orderEditions, currentEdition, latestCheckPerModel, withinOneStep, history/feed helpers, etc.) carried over verbatim — I compared the deleted and re-added bodies line by line and found no behavioural change, no change to RATIFICATION_MIN_PANEL, the graded one-step concurrence scale, the reconsideration-freshness rule, or any other §3.15 check. AGENTS.md is not touched; no citations, locators, or content records are added (citation verification confirms none); no run records are carried; the OMITTED FILES list is empty, so nothing is hidden from review. The import re-pointing is consistent across all sixteen importers shown, and the remaining `scripts/lib` → `src/lib` comment/header updates are cosmetic path corrections that match the actual module locations. One cosmetic residue (a dangling `/** The current edition … */` comment left at the end of load.ts after its function moved) and the new load.ts ↔ editions.ts import cycle are code-quality notes, not constitutional violations — the header comment on load.ts honestly records the split and date, so provenance is accurate.