What you can do

Status labels are plain English. Available means demonstrated today. Limited prototype means a pinned experimental workflow with disclosed assumptions. Planned is roadmap-only. Not a valid use is outside project claims even if someone could run code that looks similar.

Research task What works today Technical assistance needed Current limitations Verification reference
Study Medicare-like file structure, keys, and joins Available via deterministic fixture workflows Install/CLI help, or a colleague who can export tables Teaching fixtures, not a full CMS extract Scenarios & validation; catalog CLI
Run basic code and cohort-definition exercises Available with local setup Python/uv or exported tables for R/Stata Fixtures are small and structural Getting started; first run
Export ordinary fixture-based CSV/Parquet bundles Available through developer tooling Someone to run export and share files Not a calibrated research MVP download Export and release
Generate a 2021 FFS diabetes–any-inpatient MVP world Limited prototype (Python API / Phase 5 scripts) Developer familiar with the synthesis/worlds APIs Selected moments only; independent evaluation still Phase 6 Worked example; calibrated worlds; reports/phase5/
Compare multiple worlds under alternative assumptions Limited prototype Same as above Same targets can yield different free-structure outcomes Phase 5 family summary (e.g. 404 vs 572 inpatient claims at n=6,000)
Choose any disease, year, and outcome Planned (not a calibrated product today) — MVP is pinned to one context ROADMAP
Download calibrated datasets in one click from the website Planned (optional follow-on; not available now) — No hosted calibrated CSV product This matrix; documentation-maintenance non-goals
Make population, causal, clinical, or policy claims Not a valid use — Synthetic outputs are not Medicare findings Limitations; interpreting results

How to read “adjusted toward published statistics”

For the MVP, three selected public quantities are used as calibration targets (with mismatched source years and uneven evidence strength). Matching those moments improves contextual plausibility for methods work. It does not mean the synthetic joint distribution equals Medicare, that subgroups are validated, or that Phase 6 research-utility evaluation is complete.