Manual background checks mean hours of scattered searches, no record of which source backed which fact, and red flags you might miss. Case File retrieves the evidence, drafts the report, and stamps anything it can't confirm — so nothing slips through unverified.
An illustrative example, built on the same template real reports use. No evidence was actually retrieved for this subject — it exists only to show the shape of the output.
Every biography and career-history line is checked against the retrieved evidence. Unmatched claims are stamped UNVERIFIED, not deleted.
Findings run through a status taxonomy — allegation, investigation, formal charge, conviction, denial, acquittal, no sufficient evidence.
Dated events pulled from the corpus, laid out chronologically, each one still carrying its source citation.
Positive and negative drivers extracted separately, each one traceable back to the claim behind it.
Background context pulled from English or Indonesian Wikipedia — shown, but kept visually separate from the evidence-grounded analysis.
The report calls out where sources disagree or evidence runs thin, instead of papering over it.
Curators pull articles into the corpus from an Indonesian news aggregator. Admin-gated, so the corpus stays clean.
Enter a subject name. Case File retrieves matching articles and Wikipedia context, then sends the evidence set to the model.
Every claim is checked against its citation. Anything the system can't confirm is stamped UNVERIFIED and shown anyway — never hidden.
Enter a subject name and let the evidence do the talking.
Generate a report