← Back

Glossary

Key concepts behind Verg

Verg Pattern

A detected cluster of assertions from unaffiliated sources sharing the same underlying thesis, filtered against citation graphs to discount amplification. Cross-community convergence is a stronger signal than virality, but unaffiliated does not guarantee fully causally independent. Classified as Solution, Problem, or Metaphor.

CI Score

Verg Index — composite of semantic similarity, independence score, and creator diversity. Weighted by stability duration and cross-platform verification. Range 0-1.

Independence Score

Measures cross-community presence and discounts direct citation/affiliation overlap. High (>0.7) = strong cross-community convergence. Low (<0.5) = potential echo chamber. Note: this catches surface-level dependence but not hidden common causes (shared catalysts, overlapping news cycles, pretraining contamination).

Token Bake

Measured token cost: sum of contributing source body_text length divided by ~4 (chars per token) for the raw side, JSON.stringify of every active pattern (excluding foreign-key indices) divided by ~4 for the curated side. Both numbers are computed from disk on every pipeline run and persisted to data/research/token_efficiency.jsonl. Reproducible, auditable, and never hardcoded.

Thought Leader

A tracked contributor whose content appears in verg patterns. Profiled via contribution attributes (Originality, Independence, Centrality, Source Depth) rather than reach or follower metrics.

Originality

Normalized independence_contribution signal (0-100). Measures how much a source generates novel vectors that appear in patterns independently of existing discussion threads.

Source Depth

Weighted score by content format: research papers (arxiv=1.0) and repositories (github=0.85) score higher than blog posts (rss=0.4) or social media (x=0.15), reflecting information density and verifiability.

Signal vs. Noise

Multi-factor filtering: citation graph pruning removes amplification chains, temporal analysis identifies origination vs. echo, and cross-platform verification ensures the signal isn't confined to a single information silo.

Accolades

Computed from signal data thresholds. Veteran: tenure >= 200 weeks. Cross-Platform: 3+ source types. High Originality: independence_contribution > 0.5. Published: arxiv presence. Active Builder: github + high content volume.

Verg Type

Solution verg = aligned approaches. Problem verg = shared obstacles. Metaphor verg = aligned cognitive frameworks. Classified during vector clustering based on assertion framing.