How the Pulse top 10 is selected

A worked example makes the ranking concrete. Imagine two stories detected on the same morning. The first is mentioned by a few hundred social posts, all repeating the same headline from a single outlet. The second starts with a handful of regional press articles in two languages, then picks up an official statement and a specialist forum thread within a few hours. The first story is loud but narrow; the second is smaller but accelerating across independent kinds of sources and languages. Pulse ranks the second higher — that pattern is what tends to precede mainstream coverage.

What the score measures

• Velocity — how fast attention around a topic is accelerating, measured against its own baseline rather than raw volume.
• Diversity — how many independent kinds of sources are talking about it, with institutional coverage weighed differently from social chatter.
• Recency — how fresh the signals are, with older mentions naturally fading out.
• Cohesion — whether the mentions genuinely talk about the same story, across languages.
• SEO opportunity — whether search demand is emerging faster than the competition for it.

The blend between these factors is calibrated per category and refined over time — breaking news and slow-burning scientific topics do not behave the same way, and the engine accounts for that.

The daily snapshot

Every day at 00:01 UTC, the 10 strongest emerging clusters — globally, then per region and per category — are snapshotted, given a one-line editorial description, and archived at a permanent dated URL. Snapshots are immutable: yesterday's top 10 never changes, which makes the archive a verifiable record of what was emerging and when.

Quality filters

Not every rising signal is news. Encyclopedic noise — such as disambiguation pages, species stubs or articles surfacing only because they were edited — is excluded from the public lists.

Honest limits

Trend dynamics vary by category and by country: breaking news moves in hours, policy and science move in days. The scoring blend is calibrated per category and refined continuously as new data comes in — a score is a decision aid, not an oracle, and the editorial judgement stays with the reader.

How AI is used across EarlyForge, how drafts are reviewed and how to report a correction: our editorial method.