Methodology

The pipeline

Loreon follows a consistent path from raw input to actionable output. Each stage is source-agnostic, so new providers can be added without changing the downstream model.

• Ingest — collect signals from web sources, community discussion, developer activity, and market data.
• Normalize — extract clean text and entities, then map everything to a shared schema for narratives, founders, and projects.
• Correlate — cluster related signals and link them to the ecosystems and people they involve.
• Score — compute an attention score and momentum for each entity. See Attention Score.
• Surface — rank and present results across Discovery, Narratives, Founders, Projects, Ecosystems, and Markets.

Design principles

• Early over loud. The goal is to surface signal before it becomes obvious, not to echo what is already trending.
• Source diversity. No single source dominates; signals are corroborated across categories.
• Transparency. Every surfaced item carries its provenance so you can judge why it appeared.

Update cadence

Signals are designed to refresh continuously. Attention scores reflect rolling windows so that momentum, not just absolute volume, drives ranking.