Divergent Compute.AI Economic Think Tank

First Principles

How AI actually works.

An open, visual, rigorous textbook of the whole AI stack — from what a token is to the math of attention to the economics of the build-out. Every chapter is layered, so the same page is a two-minute answer for a newcomer and a deep dive for an expert. And because this is Divergent Compute, every concept connects to what it costs and why the build-out exists — the bridge nobody else draws.

01 The answer02 Intuition03 Mechanics04 The math05 The code06 The economics07 Sources
Open & citable (CC-BY)·Versioned·Built from the primary sources·Interactive where it teaches
Search

Complete. All six parts, 34 chapters — the whole book is live, from what a token is to the tool landscape you build with, by way of the transformer, the inference stack and the warehouse-sized cluster, the craft of building with a model, and where the mechanics meet the money — scaling laws, the economics of a token, and the agentic reliability wall. Every chapter: seven layers from intuition to the math to runnable code, an interactive you can drive, and the bridge to why it drives a $500B build-out. Sourced, versioned, and forkable — CC-BY.

⇩ Download the whole book  ·  PDF, 253 pages