Software Engineering handoff

Objective

Bootstrap Software Engineering as a greenfield Repository Operating System pilot.

Current state

• ROS 1.0.0 greenfield profile installed on 2026-07-29.
• Project charter is a draft.
• No first vertical slice, evidence record, hypothesis, or experiment has been accepted for implementation.
• MS-AICODING-2026-A1C4 completed the first bounded research cycle.
• RP-AICODING-2026-B7E2 is in review with five evidence records, supported hypothesis HY-AICODING-2026-7C2D, candidate theory TH-AICODING-2026-44D0, and proposed experiment EX-AICODING-2026-2D77.
• RP-HANDOFF-2026-A8E1 provisionally selects a handoff reconstruction evaluator as the first slice; EX-HANDOFF-2026-0D7A is proposed.
• MS-BOUNDARY-2026-C21E completed twelve bounded body-of-work maps. RP-BOUNDARY-2026-5A7C and DF-BOUNDARY-2026-3F72 reject a distinct discipline at review stage; an experimental integration profile remains unvalidated.
• MS-MEASURES-2026-4C2A is complete. RP-MEASURES-2026-71B4 rejects reconstruction time, document quality, and clarification count as sufficient primary outcomes. It specifies safe continuation success plus independent harm and lifecycle-cost gates.
• research/frontier/PORTFOLIO-QUEUE.md is the ranked graph view. Safeguards is complete in synthesis, as is the artifact baseline.
• JR-SAFEGUARDS-2026-7F14 defines the minimal-data, sandbox, consent, accessibility, retention, and stop boundaries. It does not authorize participant research.
• DF-BASELINE-2026-48C0 selects a competent conventional status note as the artifact control. MS-ROSATTRIB-2026-31D9 preserves whole-ROS attribution as a separate future question.
• MS-PREREG-2026-6B9F is active. Its internal protocol draft is complete through estimands and analysis, but independent preflight and authorization gates remain.
• The operating system is under evaluation.

Validation

Last run on 2026-07-30:

./ros registry build
./ros registry check
./ros validate

Registry build updated six generated registries; registry check and validation passed. git diff --check also passed.

Unresolved questions

1. Can safe continuation and critical harm be scored reliably on held-out cases?
2. Can disabled participants use the complete study path equivalently?
3. Which ethics/human-subject rules and retention requirements apply to the actual sponsor, jurisdiction, participants, and intended publication?
4. Is confirmatory recruitment feasible? A simple 50%-to-60% scenario needs about 388 recipients per arm before clustering or attrition.
5. What evidence could attribute value to ROS beyond the handoff artifact?

Next action

Supply the independent and accountable roles required by EX-HANDOFF-2026-6E42: study owner, data steward, qualified ethics reviewer, statistical reviewer, two rubric reviewers, and direct accessibility participants. Freeze preregistration only after those gates pass. Independently review all four completed REPs and three Decision Records. Do not run EX-AICODING-2026-2D77 until a partner, telemetry review, and preregistration exist.