Methodology

Our evidence standard

We prioritise peer-reviewed papers, formal proof artifacts, technical reports from the team responsible for the result, and direct responses from qualified experts. Press coverage and social posts are treated as leads, not final evidence.

A result is described at the scope in which it was tested. A matrix multiplication construction, an olympiad proof, a benchmark score, and a universal complexity-theory theorem are not interchangeable.

Evidence labels

<table class="impact-table">
  <thead>
    <tr><th scope="col">Label</th><th scope="col">What it means</th></tr>
  </thead>
  <tbody>
    <tr><th scope="row"><span class="status-pill status-peer">Peer-reviewed</span></th><td>Published after scholarly peer review.</td></tr>
    <tr><th scope="row"><span class="status-pill status-demonstrated">Demonstrated</span></th><td>A concrete result or public artifact exists within a stated setting.</td></tr>
    <tr><th scope="row"><span class="status-pill status-contested">Contested</span></th><td>Credible experts or evidence point in different directions.</td></tr>
    <tr><th scope="row"><span class="status-pill status-speculative">Speculative</span></th><td>A forecast or interpretation not yet supported by direct evidence.</td></tr>
  </tbody>
</table>

What we will not do

• Call a theorem established without a precise statement and a checkable proof.
• Treat a benchmark score as a formal proof of research ability.
• Turn a task-specific demonstration into a broad claim about replacing researchers.
• Fill missing evidence with confident wording, invented numbers, or unnamed experts.

Corrections and updates

Every substantive claim should carry a review date and a visible change history. Report a specific error through the corrections route. Substantive revisions are recorded in the public repository.