About this site

What it is

A knowledge base covering 35 medical conditions. Each has a set of topic pages, a bibliography of every paper cited, a registry of clinical guidelines worldwide, a sheet of published statistics with provenance, and a patient-experience layer. Each also maintains an explicit list of open research questions — the gaps and contradictions its own literature exposes — and a what's new page showing how recent the evidence behind it is.

How it was built, honestly

Pages are written and audited by AI agents under one binding rule: no citation may be written that was not retrieved from a live PubMed or ClinicalTrials.gov query in the same working session. Nothing is cited from model memory. Where a claim cannot be verified, it is removed or marked unverified rather than asserted.

A second, independent pass then audits each condition: it re-fetches every cited record, checks that the paper exists, that its authorship and year match, and that the claim attached to it is actually supported by the source; it corrects numbers, softens overstatements, re-runs searches behind any claim that something does not exist, and promotes a page only if it comes back clean. That pass regularly finds real errors — a trial recorded as pending that had in fact reported a negative result, a statistic attributed to the wrong population, an unsupported claim about a paper's authorship — and it has refused to promote pages it could not stand behind until they were repaired.

Its limits — please read these

• Machine-written and machine-audited. Human clinicians have not reviewed these pages. Where a page has been reviewed by a person, its curation log will say so.
• Audit status varies. Pages that have not passed the audit carry a warning banner; the table on the home page shows how many pages of each condition are audited.
• Currency varies. Each page states when it was last curated, and each condition's what's new page shows how recent its underlying evidence is.
• Coverage is uneven. Some conditions are considerably deeper than others.
• Recency is not quality. An older paper is often the definitive one.
• Bibliographic data comes from PubMed/NCBI and ClinicalTrials.gov. Abstracts and full texts remain the copyright of their publishers; this site links to sources rather than reproducing them.

Corrections

If something here is wrong — a misread finding, a mischaracterised paper, a citation that does not support its claim — please report it. Corrections are applied to the source repository and recorded in the affected condition's curation log.

Last built 2026-09-03.