Principia — pipeline status

Hand-curated snapshot of running / queued / blocked / recently completed / coming soon. Updated when state changes.

As of: 2026-06-08

Registry census (current)

The database moved from "blocked on extraction" (the snapshot below the fold) to a live, queryable evidence registry. Current counts (npm run vision:scoreboard):

Entity	Count
Canonical entities (verified-provenance gate)	4,450 / 5,000 target
Constructs (canonical)	405
Effect sizes (meta-analytic, DOI-backed)	1,023
Synthesized Bayesian priors	772
Instruments	298
Canonical survey items	1,933
Citations (academic 97% verified)	5,578
Theoretical models	8
Domain families (taxonomy overlay)	25

SPEC §7 launch gates green except canonical-entity count (4,450 / 5,000) and queue clearing. Every prior traces to a real meta-analysis; nothing fabricated; curator-gated write path (no automation writes effect sizes directly).

New capabilities (2026-06)

• Methods API + research-methods wizards (PRN-101/102). /api/v1/methods/* — the evidence/rigor layer behind a study-designer and an analysis-wizard: test suggestion, severity-aware assumption checking, an assumption→remedy knowledge base cited to the canon, and execute-with-explanation. Principia owns the evidence + remedy KB; the toolbox owns the computation.
• Article deep-extraction (PRN-131). Open-access article → results-table effects + study-design facets + salient passages, automatically (OpenAlex/Unpaywall/PMC PDF resolution → structured extraction → curator-gated effect-size proposals). Reproduces hand-sourcing with table-level provenance; grew the registry by ~150 effects in one pass.
• Construct taxonomy overlay (PRN-134). A non-destructive secondary axis tagging the 405 constructs into 25 navigable domain families (multi-lens) — the "simpler structure" without flattening or merging the registry.
• Study-design classification + diagrams (PRN-095). ~half the corpus design-classified across an 8-facet taxonomy, rendered as X–O–R / research-flow diagrams on source dossiers.

Running

• Research automation loop (PRN-023). @people-analyst/literature + npm run loop:tick + npm run curator:digest. Consensus MCP configured for curator tasks. Vision + assignment queue: docs/specification/loop/research-ingestion-vision.md, PRN-024..035 in docs/AGENT-ASSIGNMENTS.md. Curator policy locked 2026-05-20: docs/specification/loop/curator-policy.md (digest hybrid, hold verified intake, rejected status, proposals-only automation).
• Engineering substrate. @measurement/core is live (v0.4.0) with the full canonical vocabulary — Construct, Instrument, Item, Measure, Model, EffectSize, Publication, Citation (with library_id + library_path), Evidence, DeploymentEvidence, CanonicalSurveyItem, ValidityEvidence, CulturalAdaptation, UsageRestriction, asset-id types, variable-code helpers. Principia consumes it via file: dep; src/types/index.ts re-exports the canonical primitives. npm run typecheck is clean.
• Engineering scaffold. ~/meta-factory/packages/measurement-ingest/ (v0.1.0) is scaffolded as the producer-side ingest layer. First adapter (constructs) exists as a typed stub; lift from scripts/extract-variables-to-taxonomy.ts is the next pass.

Queued

• First construct-family survey: engagement. Tier 1, position 1 in PROGRAM.md. Blocked on @measurement/core schema extraction (see Blocked below). Survey scope: UWES, Gallup Q12, MEI, JES, plus the Kahn-tradition qualitative measurement work. The methodology proof-of-method.
• Construct-family ordering re-confirmation. The roadmap in PROGRAM.md is the v1 sequence. Before the engagement survey kicks off, the ordering should be re-confirmed against any external pull from the People Analytics Platform's Calculus, AnyComp, or Reincarnation work. ~one work session.
• Initial database schema design (Postgres / Supabase). Once @measurement/core lands, a queryable schema is defined for Construct, Instrument, Item, Measure, Model, EffectSize, Publication. Optimized for "find all r values between construct X and construct Y where N>100 and quality_grade ≥ B" queries. Estimated one to two work sessions once unblocked.
• Verification-log surface. A small admin/contributor surface showing every extracted row, its verification status, and links to the source paragraphs the verifier used. Required before the engagement survey ships, because the survey's published novelty-verification log is a load-bearing part of the methodology.

Blocked

• Schema extraction. @measurement/core has not yet been extracted from CanonicAI's packages/core/. UNBLOCKED 2026-05-18. ASN-1013 landed (CanonicAI 64102e1); tight-coupling primitives followed 2026-05-19 (b355bdb); decisions schema + v0.3.0 (791d224); v0.4.0 with ValidityEvidence + CulturalAdaptation (this commit). All four jurisdictions typecheck clean against @measurement/core.
• Database build. Blocked on schema extraction. UNBLOCKED. Postgres / Drizzle schema can now be designed against the canonical types. PRIN-PRINCIPIA-05 in ~/meta-factory/docs/principia/handoff-principia-repo.md.
• Sibling-app subscriptions. UNBLOCKED. CanonicAI packages/core/ consumes via workspace symlink; principia + people-analytics-toolbox consume via file: deps; PA-site assignment PA-PRINCIPIA-01 ships next.

Recently completed

• 2026-05-19. @measurement/core v0.4.0 — added ValidityEvidence (convergent / divergent / predictive / concurrent / content / face / structural sub-types) and CulturalAdaptation (instrument × locale records) per SPEC §4 gaps #4 and #5. Closes the v1 entity surface.
• 2026-05-19. Six cross-doc decisions resolved (brand = "Principia" + peopleprincipia.com; HR-metrics: ingest both with entity_kind; flip-per-construct-family; external posture confirmed; closed UsageRestriction vocabulary applied to Citation). @measurement/core v0.3.0 shipped with the schema deltas.
• 2026-05-19. Tight-coupling follow-on — @measurement/core v0.1.1 added asset-id.ts (AssetDomain, AssetType, Asset, formatAssetId, parseAssetId) and moved variable-codes.ts from CanonicAI. Codifies Principia's tight coupling to CanonicAI's metadata coding in the type system. Principia's AGENTS.md gained a "Tight Coupling with CanonicAI" section.
• 2026-05-18. ASN-1013 landed — @measurement/core v0.1.0 extracted from CanonicAI's packages/core/. Seven schemas moved; new types (Publication, EffectSize, Measure, StudyQualityGrade, short canonical aliases); Citation gained library_id + library_path; SurveyItem gained reverse_coded + numeric_anchors + norms. Prior 2026-04-29 auto-flip with non-existent ship commit cd42b6a7b was a false positive; corrected in vela.
• 2026-05-18. v0.2.0 added Evidence, DeploymentEvidence, CanonicalSurveyItem per SPEC §5 deltas — citation-to-claim edges, production psychometrics back-flow, item-level dedup spine aligned 1:1 with Reincarnation's RID.
• 2026-04-30. Scaffold v1 — repo initialized at people-analyst/principia; AGENTS.md, README.md, OVERVIEW.md, methodology.md, PROGRAM.md, this file, bibliography seed, literature map seed all landed. Registered as the sixth product on peopleanalyst.com/research.
• 2026-04-30. Reconnaissance pass through CanonicAI and people-analytics-toolbox completed; existing measurement schemas inventoried; gap list identified.

Coming soon

The 30 / 60 / 90 horizon, contingent on schema-extraction unblocking:

• 30 days post-unblock. First construct-family survey (engagement) draft v1 in markdown — construct definition, instrument inventory (UWES + Q12 to start), reliability/validity evidence, novelty-verification pass for cited rows. Effect-size table populated for the 8–12 strongest predictor and outcome relations.
• 60 days post-unblock. Engagement survey v1 frozen; second tier-1 family (job satisfaction) draft v1 underway. Database schema designed and migrations applied. Verification-log surface live (basic table view).
• 90 days post-unblock. Three tier-1 families landed (engagement, job satisfaction, organizational commitment). Book introduction chapter draft. First external-reader pass on the methodology document — confirming the rubric, the verification log, and the construct-family ordering meet the standard the methodology section claims.

Process notes

• Updates land in the same commit as the work that triggers them. A construct-family survey ship updates this file in the same push.
• "Blocked" status is honest about what is blocked. If a queued item is blocked on Mike's decision rather than on engineering, the blocker line names that.
• This file is read by the peopleanalyst.com research surface as the canonical pipeline status. Replacing it with vapor-state will surface immediately as a contradiction with the construct-family roadmap in PROGRAM.md.