EarlyForge

Key concepts

Glossary of EarlyForge terms — sites, briefs, trends, root events, scoring, connectors, citations, and webhooks.

Key concepts

This page explains the core concepts you'll encounter throughout EarlyForge.

Workspace

A workspace is your top-level space: it holds your team, sites, plan, and billing. When you sign up you get a private personal workspace (only you can see it). To collaborate, you create team/brand workspaces and invite members — one per client is a common pattern. See Manage your team → Workspaces.

Site

A site is an isolated publication inside a workspace. Each site has its own configuration, style guide, and connected publishing destination.

Think of it as a separate publication. If you manage a tech blog and a finance newsletter, you'd create two sites.

  • Each site has a unique site ID
  • Sites are isolated — articles, briefs, and settings don't leak between sites
  • You create sites yourself — a new account/workspace starts with none. A site is optional for a standalone article, and required for briefs and publishing.

Brief

A brief is a content plan that tells EarlyForge what to write about. It contains:

  • Topic — what the article should cover
  • Sources — URLs, documents, or a trend to use as input
  • Tone & style — formal, casual, analytical, and so on
  • Structure — desired length, heading structure, and SEO keywords
  • Schedule — optional schedule for automated generation

Briefs can be created manually through a wizard, or generated automatically from a detected trend.

Trend

A trend is a signal detected by EarlyForge from hundreds of sources. Each trend has:

  • Score (0–100) — a composite score from the scoring engine
  • Category — one of 28 categories (tech, finance, health, science, sports, crypto, and more)
  • Origin country — where the signal first surfaced
  • Language — one of 28 supported languages
  • Lifecycle stage — where the trend is in its life:
StageDescription
EmergingJust detected — low volume but accelerating
TrendingGrowing rapidly, increasing cross-platform presence
PeakingAt maximum velocity and highest score
SaturatingStill active, but growth has plateaued
DecliningLosing momentum, fewer new signals
ArchivedNo longer active, kept for reference

EarlyForge is built to surface trends in the emerging stage — before they peak — so you can publish ahead of the crowd.

Root event

A root event is an aggregation of related trends across languages and countries. It represents a single real-world story tracked globally.

For example, a major earthquake in Japan would generate trends in Japanese, English, French, Korean, and more. EarlyForge clusters these into one root event and tracks:

  • Geographic propagation — which countries are reporting, and when
  • Sub-clusters — the individual trends grouped under the root event
  • Propagation speed — the time between the first and fifth country reporting

Root events use semantic search and entity disambiguation to group trends accurately across languages.

Scoring engine

The scoring engine evaluates every trend using five factors:

FactorWhat it measures
VelocityHow fast the trend is accelerating
DiversityCross-platform presence — more platforms means a stronger signal
RecencyHow fresh the signal is
CohesionSemantic consistency across sources
SEO potentialSearch volume and keyword opportunity

All sources are weighted equally — none is treated as inherently more valuable than another. Differentiation comes from the scoring factors above.

Connector

A connector is a data-source integration that feeds trend detection. EarlyForge runs hundreds of connectors, organized by type:

TypeExamples
Search & socialSearch trends, Reddit, Hacker News, Bluesky, Mastodon
NewsNews aggregators and major news feeds
AcademicarXiv, PubMed, Semantic Scholar, OpenAlex
FinanceMarket data, central-bank, and economic indicators
GovernmentOpen-data portals and legislative feeds
Science & healthWHO, CDC, NASA, NOAA, weather and health feeds
CultureFilm, gaming, music, and books
ReferenceWikipedia, Wikidata, Internet Archive

Citations

Every generated article is factually grounded. Instead of writing from a model's internal knowledge alone, EarlyForge:

  1. Retrieves the most relevant passages from the indexed knowledge base
  2. Grounds the AI on those passages as context
  3. Generates an original article that cites specific sources

The result includes inline citations ([1], [2], ...) that link to the real source articles, so every claim is verifiable. See the Citations guide for details.

Webhook

A webhook is an HTTP callback that EarlyForge sends to your server when an event occurs (for example, an article is published). Webhooks are signed with HMAC-SHA256 so you can verify they genuinely came from EarlyForge.

See the Webhooks guide for setup instructions.

On this page