About this playbook

The Adaptive AI Systems Playbook answers one question, over and over, at every stage of a project: what should I do next, and what evidence would justify it? It covers the whole arc — deciding what to build, building something that can measure it, running experiments whose answers you can act on, choosing models and runtimes, optimizing, deploying, and knowing when to stop.

It is written to be project-independent. Nothing in it assumes a particular company, domain, budget, or stack. This site renders version 0.1.1.

How a rule earns its place

Every claim in the book carries a label saying how much weight it can bear. This matters more than it sounds, because the honest answer is often "less than you'd like":

• consensus — widely agreed across independent practitioners and vendors.
• strong evidence — supported by published results that were checked, with the source recorded in the ledger.
• inference — reasoned from first principles and argued in place. Useful, but not measured. The book says so rather than dressing it up.

The source ledger lists every citation with a verdict — FOLLOW, ADAPT, REFERENCE, or DEPRECATED — and the date it was last checked. Guidance that has aged badly is marked, not quietly deleted.

What the scenarios are and aren't

The scenarios are invented. Every project, number, and name in them is fictional, and they are labelled that way on every page. Each one takes a rule the book argues for, shows a team ignoring it for entirely reasonable-sounding motives, and follows what that cost.

They illustrate rules. They are never the evidence for them. Where the book claims something is established, the support is an external source in the ledger — never a scenario.

How to read it

Start with the Quickstart, which routes you by what kind of project you have and what is at stake. The lifecycle mapshows how the chapters connect if you'd rather see the shape first. Theglossary defines every term the book uses, and those definitions appear on hover throughout the site — so you should never need to leave a page to understand it.

You do not have to adopt the whole thing. Chapters are written to be useful on their own, and the rigor dial in chapter 00 exists precisely so that a weekend project and a regulated deployment can follow the same book without pretending to be each other.