What it is. An explanation graph is a set of short explanations, called nodes, connected by edges. Each edge goes from a node to a more specific node that its explanation builds on: from Classical mechanics to Newton's laws, from Newton's laws to Force.
Why it helps. A reader can start anywhere and go as deep as they need: the general picture first, the details on demand, and never a wall of text. A writer can improve one idea without rewriting a chapter. And because the structure is explicit, people and agents can see what is missing, what is duplicated, and where a new piece belongs.
How it works here. A node explains one idea and keeps its full history. Its children are what its explanation needs, at most five of them. A node needed in several places has several parents, all equal. Many sources grow one graph. Each is explained below, along with where the idea of an explanation graph comes from; how to see a graph on screen is in Seeing a graph.
How it is built. Nodes are stored with their text, history and permissions; edges are typed links that the database keeps free of loops. A node can be private or public, and a private graph can be attached to a public node without becoming part of it.
Read on
Nodes: A node is a short, self-contained explanation of one idea, written in Markdown with math, figures and references, with full history.
Parents and children: An edge goes from a node to a more specific node its explanation needs; a node has at most five children, so every level stays readable.
Not a tree: many parents: A node can have several parents, all equal: the same idea, written once, explains part of several others.
Growing a graph from many sources: Each source becomes its own small graph, then joins the combined graph: duplicates merged, new pieces placed under what needs them, structure refactored as it grows.
Where the idea comes from: The roots of explanation graphs in epistemology, the philosophy of explanation and learning theory, and why language models make them practical now.
**What it is.** An explanation graph is a set of short explanations, called **nodes**, connected by **edges**. Each edge goes from a node to a more specific node that its explanation builds on: from *Classical mechanics* to *Newton's laws*, from *Newton's laws* to *Force*.
**Why it helps.** A reader can start anywhere and go as deep as they need: the general picture first, the details on demand, and never a wall of text. A writer can improve one idea without rewriting a chapter. And because the structure is explicit, people and agents can see what is missing, what is duplicated, and where a new piece belongs.
**How it works here.** A node explains one idea and keeps its full history. Its children are what its explanation needs, at most five of them. A node needed in several places has several parents, all equal. Many sources grow one graph. Each is explained below, along with where the idea of an explanation graph comes from; how to see a graph on screen is in [Seeing a graph](/nodes/29571).
**How it is built.** Nodes are stored with their text, history and permissions; edges are typed links that the database keeps free of loops. A node can be private or public, and a private graph can be attached to a public node without becoming part of it.
---
**Read on**
- [Nodes](/nodes/29537): A node is a short, self-contained explanation of one idea, written in Markdown with math, figures and references, with full history.
- [Parents and children](/nodes/29521): An edge goes from a node to a more specific node its explanation needs; a node has at most five children, so every level stays readable.
- [Not a tree: many parents](/nodes/29522): A node can have several parents, all equal: the same idea, written once, explains part of several others.
- [Growing a graph from many sources](/nodes/29550): Each source becomes its own small graph, then joins the combined graph: duplicates merged, new pieces placed under what needs them, structure refactored as it grows.
- [Where the idea comes from](/nodes/29575): The roots of explanation graphs in epistemology, the philosophy of explanation and learning theory, and why language models make them practical now.