About — author, funding, and methodology

The Competency Restoration Observatory tracks how well five states meet their own legal deadlines for psychiatric competency-restoration services — the process of treating criminal defendants found mentally incompetent to stand trial so they can be evaluated fairly. Washington, Oregon, and Colorado operate under federal court orders; California under a state-court mandate; Texas had no binding deadline until a July 2026 federal ruling, now under appeal. Each state publishes its own compliance data, in its own format, on its own schedule — this site re-derives a common, comparable series from those public source documents by code, republished with full provenance rather than summarized by hand.

This is descriptive measurement, not advocacy: the site reports what the public record says, attributes no intent to any agency, and draws no causal conclusions about why a number moved. Every chart is a view on the underlying dataset, not the other way around — see the Data page.

Sudhanshu Rai, Independent Researcher.

This project is conducted in an individual capacity, unaffiliated with any court, state agency, advocacy organization, or party to litigation in any of the five states covered. It is not conducted on behalf of, or reviewed by, any employer.

This project receives no funding of any kind — no grant, institutional support, advocacy-organization backing, or sponsorship. It is self-funded and free to use. There is nothing further to disclose because there is no funding relationship to disclose.

Every figure on every state page traces back through the same pipeline:

1. Source documents. Court-monitor reports, special-master filings, legislative budget briefings, and state agency waiting-list reports — all public, all cited by filename and a SHA-256 fingerprint (first 12 hex characters) on the Data page, so a copy can be checked against exactly the file this site actually read.
2. Extraction by code, not by hand. Each state has its own extraction script (tables and, where the source is narrative rather than tabular, dated prose) that runs directly against the source PDFs and writes a structured dataset. No number on this site is typed in by hand.
3. A verification gate. Every extracted dataset is checked against a set of hand-confirmed anchor values before anything ships — a change that breaks a known-correct number fails the build rather than silently publishing a wrong one.
4. Per-row source citation. Every exported row cites the exact PDF page it came from, and where the value is genuinely table-derived, the exact table — not just the source document as a whole. Values drawn from narrative prose are never given a false table citation; this was a real bug caught and fixed during this site's own build (see the Data page's own disclosures for which datasets are which).

Not everything is independently re-derivable from raw PDFs within this pipeline yet. Where a series is a verified data copy rather than something this specific codebase can regenerate today, that is disclosed directly on the page and dataset it affects — never silently treated as equivalent to a fully re-executable series. This site never computes or displays a cross-state average: the five states measure genuinely different things, on different legal clocks, and an average across them would manufacture a comparability the underlying data doesn't support — see Compare.

If a number here doesn't match the source document, or you know of a more recent report this site hasn't picked up yet, that's a genuinely useful correction to send in — use the contact link below. Every correction to an already-published value or source citation is logged, dated, and explained on the Changelog page — a rebuild that changes a number with no matching log entry fails this project's own verification gate, so a correction can never ship silently.