Roadmap

• Repository bootstrap
• Constitution
• Specifications
• Knowledge model
• Repository generator
• Website
• Validation
• AI agents

Publishing

• Add a reproducible research-publisher build and GitHub Pages deployment workflow
• Normalize Markdown-formatted research and architecture documents to .md so the canonical state-constrained corpus is included in publisher discovery
• Confirm the first successful deployment from the main branch
• Assign unique canonical IDs and URLs to the excluded legacy split handbook and course-note documents

Research Validation Sequence

1. Long-horizon agent evaluation and verification
2. Context compilation, state, and memory
3. Permissions, security, and recovery
4. Agent orchestration and human escalation
5. Retrieval and knowledge-platform validation
6. Model routing and interface-agent evaluation

See research/AI-Research-Roadmap.md for decision gates and deliverables.

Evaluation Mission Status

• Cycle 001: charter, constructs, task taxonomy, audit protocol, and registries
• Cycle 002: integrity pilots for ET-004 and ET-014 (provisional pending independent review)
• Cycle 003: in-sample deterministic grader calibration and blind export
• Historical non-human experimental inventory, lineage, contradiction, and confidence synthesis
• Cycle 004a: independent review plus frozen-grader unseen-outcome challenge
• Cycle 004b: contained repeated baseline and exploratory variance estimate
• Cycle 005: context or system comparison

Detailed critical path: research/evaluation/12-research-and-engineering-roadmap.md.

Comparative evidence boundary and ranked experiments: research/analysis/non-human-experimental-comparative-analysis.md and research/analysis/non-human-next-experiments.md.

Containment, telemetry, and rollback now run in parallel with evaluation rather than waiting for a later security phase.

State-Constrained Architecture and Time Entry

• Separate the 01–12 research prompts, reports, and synthesis.
• Separate generic semantic-control architecture from Time Entry inputs.
• Group current Time Entry requirements, state model, and architecture by authority.
• Author and approve any still-needed numbered Time Entry contracts; none were present during repository organization.

Repository Fitness

• Backfill material pre-ROS work into validated historical attribution
• Remove case-only collisions for root README, current state, and roadmap
• Add a repeatable Git case-collision audit
• Add initial reference-integrity and stable-identifier fitness checks
• Measure metadata coverage and validate adopted relationship targets
• Define bounded canonical metadata adoption and calibrate reciprocal supersession checks on real edges

Research Frontier

• Enter the canonical research TODOs as proposed ROS backlog missions
• Establish the canonical repository frontier and health assessment
• RFR-001: obtain independent pilot integrity reviews
• RFR-002: run a contained repeated exploratory baseline
• RFR-003: validate graders on natural outcomes
• RFR-009 (Validation): coverage and relationship diagnostics calibrated; bounded metadata adoption and real supersession-edge validation remain

See research/frontier/FRONTIER-MASTER.md for the ranked backlog and dependency graph.