Document Status Registry

This registry labels the major documentation surfaces so old roadmap material does not compete with the closed learning lab.

Status Labels

Status	Meaning
`active`	Current source for what to do next
`closed`	Current completion truth and fresh-experiment gate; no backlog is active
`evidence`	Results, attempts, artifacts, and validation history
`reference`	Useful background, not the active queue
`learning`	Owner curriculum and concept explanations
`parked`	Intentionally paused until it unblocks active factory work
`superseded`	Preserved for links/history, but replaced by a newer doc
`archive`	Historical material only

Registry

Doc / Folder	Status	Use It For	Do Not Use It For
`PROJECT_STATUS.md`	`closed`	completion candidate, release boundary, and fresh-experiment rule	old feature-by-feature backlog
`docs/README.md`	`closed`	golden path through the learning artifact and lab	detailed run evidence
`docs/NEXT.md`	`closed`	release receipt and fresh-experiment admission rule	an active task queue
`docs/factory/`	`reference`	factory contracts, reports, evals, packaging, enforcement	generic ML theory
`docs/techniques/`	`reference`	method-vs-recipe cards and closed technique lineages	active task selection
`docs/techniques/audit-inventory.md`	`evidence`	row-level treatment of `docs/audits/audit_2026.md` technique rows	run-level success/failure claims
`docs/attempt-ledger.md`	`evidence`	76 final experiment dispositions and lessons	full per-run logs or an active queue
`docs/audits/history-coverage-audit.md`	`evidence`	what historical work is normalized, classified, partial, or narrative-only	active task selection
`docs/audits/exactness-completion-audit.md`	`evidence`	proof that the docs exactness pass is complete and guarded	new roadmap scope
`docs/external-products-reviewed.md`	`evidence`	reviewed products, papers, and stolen techniques	exhaustive literature survey
`docs/factory/public-artifacts.md`	`evidence`	public artifact release state and blockers	internal-only scratch notes
`docs/learning-pipeline.md`	`learning`	ready owner learning sequence tied to repository labs	generic course catalog
`docs/learn/`	`learning`	curriculum and concept references	active project queue
`docs/prds/`	`reference`	acceptance criteria for a named/deferred lane	selecting the next active task
`docs/PLAN.md`	`reference`	historical feature inventory and shipped/skipped/TODO context	current source of truth
`docs/roadmap/`	`superseded`	old roadmap links and historical split	active roadmap
`docs/parked/`	`parked`	paused lanes and why they are paused	active tasks
`docs/archive/`	`archive`	historical snapshots	current claims
`docs/performance/`	`evidence`	kernel, memory, and throughput results and their method notes	active task selection
`docs/audits/`	`evidence`	coverage, feature, docs-quality and validation audits	current claims
`docs/techniques/` folder pages	`reference`	per-method concept cards (LoRA, MoE, quantization, …)	run-level results
`docs/integrations/`	`reference`	wiring posttrainllm to external tools and data sources	internal contracts
`docs/guides/`	`learning`	long-form model / training / study walkthroughs	current queue
`docs/sessions/`	`archive`	dated session records and handoffs	current claims

Conflict Rule

When docs disagree, trust the newest active/evidence source in this order:

PROJECT_STATUS.md
-> docs/NEXT.md
-> docs/README.md
-> docs/techniques/
-> docs/factory/
-> docs/attempt-ledger.md / run report / public artifact entry
-> docs/prds/PRIORITY.md
-> docs/PLAN.md
-> archived/session notes