Skip to main content
TCW

DOCS

Technical documentation for the machine-readable layer.

⚡ MCP Server📡 REST API🧬 Concept DNA📜 Open Knowledge Protocol

🧬 Concept DNA

Every article on The Context Window carries structured metadata that maps its place in the knowledge graph. This is what makes our content machine-readable — not just for SEO, but for AI agents that need structured, citable information.

Fields

TL;DR

2-3 sentence summary. This is the primary text AI agents cite when referencing an article. Written to be standalone — it should make sense without reading the full piece.

Key Claims

Numbered assertions the article makes, each tagged with a confidence level (high, medium, low, speculative) and supporting evidence. Individually addressable for citation.

Prerequisites

Articles or concepts the reader should understand first. Creates a learning path through the knowledge graph.

Depends On

Foundational concepts this article builds upon. These are edges in the knowledge graph pointing to concept nodes.

Enables

Concepts that become accessible after understanding this article. The forward-looking edges in the graph.

Related Concepts

Lateral connections — concepts that are related but not in a dependency chain. Helps with discovery.

Predictions

Timestamped forecasts with confidence (0-100), horizon (when testable), reasoning, and live tracking status (pending/confirmed/busted).

Adversarial Perspective

AI-generated counterarguments presenting the strongest case against the article's thesis. Shows intellectual honesty.

Confidence Levels

High

Strong evidence, broad consensus, or well-established pattern.

Medium

Reasonable evidence with some uncertainty. Multiple interpretations possible.

Low

Early signal, limited evidence, or contested claim.

Speculative

Forward-looking extrapolation. Interesting to track but may not pan out.

Where It Lives

  • • In the UI — sidebar panel on desktop, below-article on mobile. Press C for a focused view.
  • • In HTML — JSON-LD structured data in each article's <head> for search engines.
  • • In the API — full Concept DNA returned by /api/v1/articles/:slug
  • • Via MCP — queryable through the get_article tool.

See it in action: The Architecture of Agentic Memory has a fully populated Concept DNA sidebar with 3 key claims, prerequisites, and a prediction.