# Verg

> Verg is a protocol for convergence intelligence. It detects when independent minds arrive at the same conclusion without coordinating, verifies the independence through social graph analysis, and exposes the resulting signal to humans and AI agents. Built for the paradigm shift into agent-native media.

Verg ingests from 485+ sources (arXiv, GitHub, YouTube, podcasts, RSS), runs a daily pipeline that detects convergence patterns with 99.8% token compression, and scores its own predictions (currently 80% accuracy). Every claim traces back to primary sources. Contribution is measured by what you produce, not by followers. Low-quality signals are classified (noise, frontier, archived), not pruned — preserving potential paradigm-shifting ideas that haven't converged yet.

Verg is an attempt to meet mind to mind with the parts of the paradigm.

## Core

- [Whitepaper](https://verg.dev/whitepaper.md): Protocol specification — convergence thesis, token bake economics, independence verification, frontier classification (living document)
- [API documentation](https://verg.dev/docs): REST endpoints, query parameters, response shapes, integration examples
- [Glossary](https://verg.dev/glossary): CI score, independence score, token bake, convergence type definitions
- [Dashboard](https://verg.dev): Live convergence patterns, thought leaders, emerging signals with full source provenance (requires auth for browsing)

## API

- [OpenAPI 3.0 spec](https://verg.dev/api/openapi.json): Machine-readable tool definitions for agent integration
- [GET /api/patterns](https://verg.dev/api/patterns): Convergence patterns with source provenance, CI scores, independence verification — query ?limit=N&min_ci=0.3
- [GET /api/leaders](https://verg.dev/api/leaders): Thought leaders ranked by contribution profile — query ?limit=N&type=architect&min_score=0.2
- [POST /api/feedback](https://verg.dev/api/feedback): Submit pattern or source quality feedback

## Key Facts

- Convergence detected by independence verification, not volume or trend metrics
- Independence verified via PageRank + Louvain community detection on the citation graph
- Token bake: typical pattern distills 144,000 raw tokens into ~280 curated tokens (99.8% compression)
- Prediction scorecard: 80% accuracy on graded predictions, published and auditable
- Leader scoring dimensions: originality, independence, centrality, source depth (never follower count)
- MCP server available with 5 tools: verg_patterns, verg_leaders, verg_emerging, verg_search, verg_predictions
- Public API: no auth, rate-limited to 60 requests per minute per IP
- Daily pipeline: runs autonomously at 4am, self-grades its predictions, prunes noise, classifies frontier ideas
- Built by @lazerhawk5000 — open methodology, open scorecard, no VC funding, no growth hacking

## Optional

- [Transmissions blog](https://verg.dev/blog): Rants and raw thinking from the frontier
- [Meeting the Paradigm](https://verg.dev/blog/meeting-the-paradigm.md): First post — raw thesis on convergence, agentic internet, information absorption
- [Full context bundle](https://verg.dev/llms-full.txt): Complete llms.txt + whitepaper + blog bundled for deep ingestion
- [Support](https://verg.dev/tip): Stripe tip mechanism for supporting the protocol
- [X @lazerhawk5000](https://x.com/lazerhawk5000): Author on X