Skip to content

Working copy

Draft text, rebuilt from the manuscript on every change. Stubs and source passages are still here, and none of it is final. Back to the working copy contents

Drafted The chapter is written and still being revised.

Front matter

How to read this book

This book is a manifesto with a toolkit attached. Bots are starting to read construction tenders, and the book is about the standard they will be read against: what it is for, who gains from it and what it would take.

Each part answers one question. The prologue and Parts I and II give the argument in brief: what the problem is and how we should solve it. Part III asks why other tools and other contract types are not enough. Part IV is for the people who would use the standard, one chapter for each: buyers, bidders, trades, designers and consultants, vendors, legal departments and insurers. Read your own chapter first if you are short of time. Part V is how we get there, through what AI does badly, what it does well, the tools and the standard. Parts VI and VII look back two centuries and forward to where this ends up.

Where two true statements pull against each other, both are kept and the tension is named. Construction has more tensions to manage than problems to solve, and a book that resolved them all would be a book about some other industry.

The appendices hold what a reader can skip: the history of procurement as a deck of cards, a reading list, a glossary, the bibliography and a draft term sheet. A companion edition written for AI assistants carries the sources and asides this one leaves out.

Author's note on method

The project behind this book was run in public, and its decisions are numbered and kept. The chapters were drafted with AI assistants from my own notes and sentences, and every one has passed through my hands. Nothing here is legal or procurement advice.

Source passages (kept until Andre has read the draft)

Laid out by Claude on 2026-11-02. Every quoted passage is Andre's, copied unchanged from the file named under it; the source repository records these as his words, lightly edited from conversation. Notes in italics are Claude's. Delete this section as the front matter is written.

Epigraph candidates

The outline's second candidate, Amber Case's banana peel, in Andre's sentence. Credit the image to her (decision 75):

While tenders are likely to be resistant to earning Amber Case's Calm Tech designation (see the Protocol Institute's 2026 Symposium), the banana peel concept, that the outside should tell you what is going on inside, would still be a goal in designing the bots and the standards.

(positions §16, recorded 2026-10-10)

A lighter one:

If Tenders could become Calm Tech Certified, that would be a miracle for the good of mankind.

(positions §19, recorded 2026-10-11)

The Chic Sale line is not yet chosen; nothing on file quotes him.

For "How to read this book"

The rule for statements in tension, which the outline puts here. Chapter 2 (the divorce) also carries it as optional; use it in one place only:

I would need examples. There are likely many statements with unresolved tensions between them. Some of these could be resolved by a protocol to manage the specific tensions, but I don't have a rule for batch processing the issues. Perhaps the best approach would be to keep things in tension until a trade-off decision needs to be made. In the meantime, they should be labeled as Frontier, Monstrous, Kafka, or Tool-building tensions, based on my essay on the subject.

(positions §21, recorded 2026-10-11)

The four labels come from the Substack essay "Age of AI, the age of protocols" (andrecomeau.substack.com). Its definitions are not yet quoted anywhere in the book's sources; take them from the essay.

Possible chapter openers. The eight one-line tensions under the machinery figure are Claude's, tested on the bot council and PIBot, and wait on Andre's words (machinery-figure-chat-2026-10.md, section 2): "The lowest bid wins, mistakes and all." (the procurement system); "The owner's estimate against what bidders charge." (the estimator); "Who pays to find the drawings' errors?" (bot procurement); "If a robot builds the error, who answers?" (robots from the model). The others are on BIM, BIM for bots, bot admin and robotic construction. Since 2026-11-18 the third is called bot admin (Andre: "Bot Admin may be a more neutral term than Bot Monitoring. Workers will not find increased Taylorism to be to their liking"); the machinery figure follows since the same day, with Claude's new line "Is it built as modelled? Records keep up." for Andre to replace.

For "Author's note on method"

On how much generated material Andre reads, which explains why the book separates his sentences from the rest:

I will usually glance at the questions until I find one I want answered. You and the bot-council can generate text much faster than I can read, so it is unlikely I will do a deep dive into all the generated material.

(positions §17, recorded 2026-10-10)

The case study's own statement of method, which is Claude's wording and can be adapted: "The framing, the liability argument and the tensions are the author's own, offered directly and not drawn from a source." and "Nothing here is legal or procurement advice." (case-studies/andre-comeau/README.md, "Sources").