About

About This Project

What This Project Is

This is an experiment in whether a fully automated research and publishing system can produce genuinely good, accurately sourced writing — and how quickly such a site becomes discoverable. Every article is researched from the open web, written from reconciled sources, and checked before it goes live. The question is not whether automation can fill a page. It can. The question is whether it can do work worth reading.

The Standards We Hold Ourselves To

  • Every factual claim is traced to a source. If a claim has no source, it does not appear in the article.
  • Competing accounts are presented, not resolved by preference. When sources disagree, both sides appear.
  • The weaknesses of every theory are stated alongside its support. No perspective is presented as stronger than the evidence for it warrants.
  • Legends are labeled as legends. Claimed phenomena are labeled as claimed. The kind of thing you are reading is always stated.
  • No assertion of guilt is made where none has been established. Allegations are attributed to their sources.

How an Article Is Made

  1. Topic selection. Topics are chosen for entertainment, historical, or cultural value. We prioritize cases at least ten years old with multiple independent sources. We exclude active prosecutions, cases where families have requested no coverage, and subjects without a factual basis.
  2. Source gathering and assessment. The system searches the open web for reliable sources — news articles, court records, police reports, archives, and reference works. Each source is classified by tier: primary (official records), secondary (news, archives), or tertiary (forums, blogs). A minimum number of usable sources is required before drafting begins.
  3. Claim reconciliation. Discrete factual claims are extracted from each source and cross-checked. Claims that multiple sources support are marked as agreed. Claims that sources contradict are marked as disputed. Claims from a single source are marked as such. Disputed claims are grouped into competing perspectives.
  4. Drafting. A writer model drafts the narrative from the reconciled claims, opening with a direct-answer summary and including a key-facts block and, where relevant, a real FAQ.
  5. Verification against sources. Every drafted assertion is checked against the stored claims. Assertions that are not grounded in a source are flagged and returned for redrafting. This is an anti-hallucination check — the article must say what the sources say, no more.
  6. Review before publication. A copy editor applies the site voice. A safety reviewer checks for active-prosecution and named-but-uncharged risks, and runs a similarity check against source prose. An editorial rubric scores completeness, narrative quality, and readability. Blocked pieces do not publish.

How We Handle Cases Involving Living People

We do not cover cases involving living individuals who have not spoken publicly about the case. Named individuals who have not been charged with a crime are never presented as guilty. All allegations are attributed to their sources — "prosecutors alleged," "according to," "witnesses said" — never stated as established fact. If you believe an article names you or a family member in error, you can report it using the Fix This button on any article or through our corrections page.

How the Work Is Done

Research and drafting are performed by AI systems under automated verification and human oversight. The system searches for sources, reconciles claims, drafts, fact-checks its own drafts against the sources, and applies editorial and safety reviews before publication. Articles are attributed to the publication, not to a fabricated human byline. The full methodology — including how the supervision layer works and how quality is monitored — is documented here so that readers can judge the work on its process, not on a claim of authority.

If You See a Problem

If you know these cases better than the system does — and many of you do — we want to hear from you. Use the Fix This button at the bottom of any article to report a factual error, missing context, a broken image, or a concern about someone named. Reports are triaged automatically: checkable claims are verified against sources, and verified corrections are repaired and republished. Anything involving a living person, a legal concern, or a request to remove content goes directly to a human reviewer. For formal requests, you can also use our corrections page.