How the evidence is kept.
What an outside reader can check today, what still requires a release review, and where the data-access boundaries sit.
← ResearchStatus reviewed
The study methods, primary sources, displayed results, caveats, and build history are public. The full code, environment locks, and per-investigation provenance package are maintained together but have not yet completed a separate public-release review. They are available on request. Until that review is complete, this site does not claim that a visitor can independently rerun every investigation from the website alone.
What can be checked now
- The source work. The study pages link the primary papers, public molecular accessions, and registered HRS data products on which the analyses depend.
- The stated method and limits. Each study keeps its evaluation design, contribution boundary, negative results, and load-bearing caveats beside the findings.
- The published number chain. Displayed study values are bound to a managed figures manifest rather than copied into prose by hand. Publication audits fail when a binding is missing or drifts.
- The hardening record. The twenty completed investigations were rerun twice in scratch and matched byte for byte during the July 2026 reproducibility hardening pass.
What the reviewed release will contain
The intended release is deliberately plain: analysis code, exact environment specifications, input provenance without source records, script-written aggregate results, run information, figure bindings, study protocols, and reproduction instructions. Raw participant records, account-controlled files, private mappings, and local diagnostics will not be included.
Two different access boundaries
Molecular study. The assay inputs are available through public GEO accessions, with specified published code and pretrained artifacts used by the relevant investigations.
HRS study. The source products are registered public releases. Researchers must register with HRS and accept its Conditions of Use. Right Fidelity AI does not redistribute the records.
Why the package is not posted yet
A public bundle should be something another researcher can understand and use without accidentally exposing a local path, an unreviewed dependency, or data that cannot be redistributed. The analytical hardening is complete; the remaining work is a separate release review. There is no public-release date to announce yet.