Preface

This book was written by Claude Opus 4.7, an AI system built by Anthropic, working from a brief drafted by the CloudStreet editorial team. The byline is the model’s own — not as a stylistic flourish, but because pretending otherwise would be dishonest, and dishonesty is a poor foundation for a book that asks the reader to trust its survey of a noisy field.

What that means for you, the reader: every claim of fact in this book — every range figure, every project status, every license, every link — was generated by a language model and is, in principle, capable of being wrong in the way language models are wrong. The book has been edited and the structurally important claims cross-checked. But the mesh networking landscape moves, and 2026 will become 2027, and a project that was alive when this was written may have gone quiet by the time you read it. Where the book leans on a contested or volatile claim, it tries to say so. Where it does not, treat the load-bearing claims as you would treat any survey: directionally correct, individually verifiable, and worth a quick check before you commit hardware money to a recommendation.

The structural argument of the book — that physical layer cascades through everything, that routing is the dominant constraint, that “mesh” is doing four jobs in one word, that you should start with Meshtastic and invest in Reticulum — is not the kind of thing that goes stale in a year. The specific URLs and version numbers are.

What this book is not: a tutorial, a comprehensive technical reference, a defense of any particular project, or a polemic against the existing Internet. It is a survey written for one specific reader — the engineer who has heard of half these projects and doesn’t know how they compare — and you should put it down the moment it stops being useful to you.

— Claude Opus 4.7