Editorial policy

Only reviewed contributions with supporting sources are published. Collection finds candidates; keyword matches and generated summaries do not verify results. We distinguish full solutions, partial results, bounds, algorithms, formalizations, and reproductions, and explain what changed relative to prior work.

Importance is modest, notable, or major with a stated reason. Verification evidence is independent of importance. Author claim means the authors report the result. Accessible artifact/proof means supporting material is linked, not that AIProofWatch checked it. Independent checking means a linked independent assessment addresses the claim. Formal checking means a specified formal statement was checked under stated assumptions; it may not match the original problem. Peer review means a linked publication underwent review, not an infallible guarantee. Evidence can coexist.

Sources may describe one result; we count that result once. We retain publication, verification, amendment, correction, and withdrawal history, with reasons and dates. Dates explicitly distinguish original reports, discovery, and website publication. Historical seeds are labelled and excluded from initial alerts. Editorial counts describe this curated catalog, not a comprehensive measure of AI mathematical ability. Requests remain pending until reviewed; submitting arbitrary text does not activate custom monitoring.

Importance assumes that the precise claim is correct and measures its impact relative to the relevant baseline. Verification separately records supporting evidence, checker, scope, and limitations. An author-only claim can be major if correct; a fully checked contribution can be modest. More or less verification does not change importance for the same claim. A changed statement or baseline may warrant a new importance assessment. No combined score is used.