Sources and release scope

Every public claim is tied to a source, vintage, and release.

Worker measurement, occupation composition, establishment staffing, and firm-side AI use enter the observatory under explicit source and interpretation boundaries.

Source register

Four sources support distinct parts of the evidence.

Primary worker measurement

RPS / GenAI Adoption Tracker

RPS supplies worker adoption, recurrence, assisted work hours, and reported time savings. The authorized aggregate release path uses registered FRED series to generate bounded attributed public views plus derived publication artifacts.

Industry composition

Current Population Survey

Official Q2 2025 and Q2 2026 Basic Monthly public-use files supply worker-share and actual-main-job-hour occupation weights. Composition results are descriptive; design-based uncertainty for custom pooled composition vectors remains unsupported.

Establishment robustness

OEWS · 2025-05

Official staffing data provide an independent establishment-side check on industry occupation composition. Population differences from CPS remain explicit.

Firm-side triangulation

BTOS

BTOS supplies current employer-business AI use by sector. The industry explorer uses its relationship with worker RPS adoption as cross-construct descriptive triangulation.

Provenance

Publication detail stays inspectable.

RPS provenance and public scope

The core worker source is the Real-Time Population Survey Generative AI Adoption Tracker by Bick, Blandin, and Deming, surfaced through FRED. The canonical registry contains 131 work-focused series records: five national constructs, 60 industry records, and 66 occupation records.

The source registry records owner permission for published aggregate project use and derived aggregate analysis. Release 1 publishes seven complete national-history quarters through Q2 2026 plus the latest complete industry and occupation A/H/S cross-sections. Historical subgroup panels, bulk download, and generic source-query access remain outside the current public scope. Private source-input bytes are excluded from public release artifacts.

FRED RPS release page · Peer-reviewed RPS paper

Public, private, and validation boundaries

Public release: Contains source metadata, formulas, crosswalks, bounded attributed RPS views, longitudinal diagnostics, CPS/OEWS composition evidence, BTOS triangulation, methodology, and the technical essay.
Private release-candidate workspace: Holds authorized RPS source inputs used to validate the registered source history and regenerate release artifacts. Source-input files are hash-bound to each candidate and excluded from public review and release bundles.
Composition evidence: Contains Q2 2025 and Q2 2026 CPS composition artifacts, occupation-adjusted industry-context residual diagnostics, and 2025-05 OEWS robustness artifacts.
Outside current claims: Full historical RPS subgroup distribution, custom CPS design covariance, causal firm effects, and measured labor-productivity effects require additional evidence or publication rights.

Each promoted release passes source and data contracts, governance checks, type validation, optimized production builds, route smoke tests, and rendered browser accessibility checks. The release process binds those checks to the exact candidate and preserves prior release history.

Human screen-reader traversal, physical mobile-device testing, and field Core Web Vitals remain outside current validation evidence. The public deployment is audited against the exact promoted release commit.

Read data & methods →