What does inverse clickbait mean?

A high clickbait score lowers the ranking — "inverse" because the formula treats less clickbait as better. A headline scoring 100 on clickbait would lose roughly 10 ranking points; one scoring 0 would lose none.

Can the editor override the ranking?

Yes, before send. Edits are logged in the audit trail. Most days no overrides are needed; on busy days one or two items are reordered to fit the brief's section flow.

Are these scores ever shown per headline?

Not yet on the public site. The scores drive ranking and are used to seed the Bullshit Index. A per-headline disclosure is on the roadmap for a future widget release.

Why six-hour urgency half-life and not twelve or twenty-four?

Empirical. With three brief drops per day, six hours roughly aligns urgency decay to the gap between briefs. A twelve-hour half-life would let prior-day news dominate the next morning; a three-hour half-life would crowd out anything that broke before lunch.

How SignalPop scores 18,000 headlines a month

Eighteen thousand headlines a month is roughly six hundred a day. SignalPop does not publish all of them; it ranks them and surfaces a few dozen. The ranking depends on three numbers computed at ingest time, plus a fourth that updates as the day progresses.

Criticality is a 1-to-10 score that asks one question: how consequential is this story for a reader who is not a specialist in the subject. A Fed rate decision is a 9 in money; a labor-market release is a 7 to 8; a CEO press release is a 4. The score is assigned by the same lightweight model that does headline rewriting, but it is constrained by a per-category rubric (rate decisions are 9; earnings beats are 6; analyst calls are 4) so the model cannot drift toward overrating drama.

The clickbait score is a regex-and-vocabulary pass — interrogative headlines, hyperbolic verbs, ALL-CAPS, ellipses, and the words 'shocking,' 'devastating,' 'blasted,' 'slammed,' and a list of similar markers. A headline scoring high on clickbait is rewritten to remove the markers before display; the original is kept in the database for audit. The rewritten version is what appears in the brief and on category pages.

The corroboration count is the number of distinct sources ingested in the last forty-eight hours that report the same underlying story. We cluster headlines using a combination of cosine similarity over title embeddings and shared named-entity overlap. Items with corroboration counts of two or more get a substantial ranking boost; single-source items with low criticality get pushed below the fold.

The fourth number, urgency, decays over time. A story that broke at 6 a.m. is more urgent at 7 a.m. than at noon, regardless of criticality. The decay is exponential with a six-hour half-life. The morning brief therefore tends to feature the prior-evening news that has had time to corroborate; the night brief features the day's late-breaking developments.

The brief assembler takes the top items by a weighted combination of these four numbers — criticality 40 percent, corroboration 30 percent, urgency 20 percent, inverse clickbait 10 percent — and presents them in section order. There is no editorial thumb on the scale at this stage. The human editor reviews the result before send and can override individual placements, but the default ranking is mechanical and reproducible.

FAQ