Contents

Graph Engineering Course — Docs

This directory is the whole course. Everything readable on GitHub and everything rendered by the companion website is generated from these same markdown files — there is exactly one copy of the content, not two that can drift apart.

If you haven't been here before, don't start on this page. Visit the router first and answer a couple of quick questions about where you're coming from; it will point you at the right entry point below.

Two ways through the material

The core path (roughly two hours). Steps 1 through 13, plus Projects 1 through 4. This gets you from "I have one loop and one memory file" to "I can build a fact graph, feed a worker a bounded slice of it, and check a claim against real edges instead of trusting how confident it sounds." Most readers stop here on a first pass and that's a complete, useful stopping point.

The second read. Steps 14 through 17, Projects 5 through 8, the full pattern library, and the advanced tier. This is where governance, scale, and the judgment calls about when not to build a graph live. Come back for it once the core path has had time to sink in, or once you're actually running more than one loop against a shared graph and need the harder material.

Skill tracks

Track	Level	Walking in, you already know…	Walking out, you can…	Where it lives
G1 · Foundations	New to graphs	The Loop Engineering vocabulary — heartbeat, spine, maker/checker	Point to the exact moment a lone memory file quits working once a second worker starts touching it, and tell a trail of attempts apart from a store of checked claims	Prerequisites + Foundations + Part 1
G2 · Practitioner	Comfortable with the two-graph split	That work-history and facts are different graphs with different jobs	Push a real claim through extraction, resolution, and provenance, then hand a worker a bounded subgraph and check its output against real edges	Parts 2–4 + Projects 2–6
G3 · Engineer	Can stand up a single working graph	How one graph is built and read	Connect several loops into a governance graph, and name — and patch — each of the four ways a lone loop quietly goes wrong	Part 5 + Projects 7–8 + pattern library
G4 · Ultra-Pro	Has shipped a graph already	How to run a graph as production infrastructure	Recognize when a graph is the wrong tool for a job, scale one across a bigger system, and write a new pattern for someone else to reuse	Parts 6–7 + advanced tier + certification

Before Part 1

Section	What it's for
`00-start-here/`	A short router — 2–3 questions that point you at the right starting page
`01-prerequisites/`	Confirms you have Loop Engineering and Harness Engineering, with recap primers if you want a refresher
`02-foundations/`	Vocabulary and mental models this whole course leans on: the glossary, mental models, the two-graph split at an intro level, and comprehension debt

The 17-step roadmap

Links below point at where Day 2 of this build will place each step's page. Until then they're placeholders for the shape of the course, not live pages — that's expected at this stage of the build.

Part 1 — The Memory Problem

Part 2 — The DAG of Work

Part 3 — The Graph of Facts

Part 4 — Working From the Graph

Part 5 — The Graph of Loops

Part 6 — One Graph, End to End

Part 7 — Staying Grounded

Reference material

Section	What it's for
`methods/`	The build-a-graph method, the pattern picker, and the pre-build decision framework
`operating/`	Anti-patterns, failure modes, safety notes, and observability guidance
`advanced/`	The Ultra-Pro (G4) tier: scale, federation, and org-level governance
`projects/`	The eight hands-on projects, from a first hand-drawn graph to the two-loop capstone
`appendix/cheatsheets/`	Quick-reference sheets per tool
`assessments/`	The final exam, capstone rubric, and Graph Ready certification

Patterns and starter kits

Runnable material lives outside docs/ at the repo root: patterns/ for the pattern catalog and starters/ for clone-and-run kits. Full attribution for every idea this course draws on is in resources/sources.md.