# 2 · The first book, and the pivot that saved the second

*After this chapter you will know how a book gets written in three days and why that was
the easy part, and you will recognise the single most valuable thing an author can do
with an agent: stop it, out loud, in writing, before it answers a question nobody asked.*

---

## Three days to a book

The repository's first release, v0.1.0, is dated 21 August 2026 at 14:47 UTC. Its note
is short and tells you what kind of project this is going to be:

> First cut. The pipeline before the prose: validate → tag → deploy on every push to
> dev, with a seven-check pre-release gate and automatic minor tagging verified against
> the commit subject. Then the site: the front page and its three altitude doors …

The pipeline came before the prose. Before there was a single chapter, there was a
validator with seven checks, a machine that tags releases, and a deploy step that will
not run if validation fails. That ordering is not an accident and it is not fastidiousness.
It is the reason the next eighty-seven releases were possible.

Two hours and fifty-six minutes later, v0.2.0:

> The book. A new /book/ section — the site's content as a book, Meaning Through
> Connectivity: sixteen chapters in six parts, in three reading modes — chapter pages
> with a persistent table of contents, the whole book in one page, and a 79-page PDF
> printed from that page.

![The first edition at v0.2.0, roughly three hours after the repository's first commit.](figures/15__v0.2.0__the-first-book.png)

*Figure 2. The book section at tag `v0.2.0`, 21 August 2026.*

By the end of that first day the book had a print PDF in a standard technical-book
interior format, a second screen-styled PDF for tablets, centred title pages and
properly typeset tables, each fix arriving as its own release because the founder was
reading the PDF on a phone and reporting what looked wrong.

By the end of the second day it had a cover generated as SVG from the book's own
vocabulary, a review workflow, two reviews from the founder normalised into a typed
register, a decisions register, and an altitude ladder: the same argument rendered at
five levels of compression, from one paragraph to full prose.

![The altitude ladder at v0.3.15: one book, five altitudes, built as an experiment rather than described in a plan.](figures/16__v0.3.15__the-altitude-ladder.png)

*Figure 3. The altitude ladder at tag `v0.3.15`, 22 August 2026.*

By the end of the third day the chapters had moved to markdown, twenty-one source
documents had been carried into the repository whole, concept graphs had been computed
per document, and the first edition was complete enough to freeze.

Sixteen chapters in six parts, 20,838 words of page markdown, thirty-four releases to
the freeze at v0.3.26, three days.

That is the part that looks impressive and is actually the easy part.

## The plan for the second book, and the memo that stopped it

On 23 August the repository shipped v0.3.27, a large planning document:

> The plan to write the book again, from the top down. Two founder voice memos on 23
> August cancelled the refactor that was planned and asked for something larger: enough
> structural evidence, tooling and workflow now exist to start again and write the book
> the way it was meant to be written …

The plan was ten files. It audited every unit of the first edition and gave each one a
carry, lift, rewrite or drop verdict. It proposed a spine of five statements. It defined
five altitudes and what each would hold. It was, by any normal standard, a good plan,
and the founder said so.

Then, reading section 5, he recorded a voice memo that cancelled most of it. It is
published verbatim as brief 20, and this is the part that matters:

> the brief is amazing, right? Like it's really cool. I'm really good. I'm I'm on I'm on
> section five, but one of the things I'm already noticing is that we are already
> defining the answer before we know what questions we're answering, and and this is very
> obvious in section two, where you say what exists today, because this is already making
> a number of judgement calls, but we don't know what the shape of the book is

And a little later, the sentence that reordered the whole project:

> I feel, if you think about it, that we cannot know what each of the levels, level one,
> level two, level three, will work will look like until we have done this. So, so I
> think this is like a whodunit, right? Like we need to come up with the plot. We need to
> come up with what is the sort of the whodunit mystery machine. Like what is the
> sequence of events? What is the narrative? What is the story that we want to tell on
> the book?

The instruction that came out of it inverts the normal construction order for a book.
Not spine first with evidence attached to it, but all the reference material first,
decomposed into concepts, facts, claims and evidence, and only then the plot lines drawn
through it. The brief's own one-paragraph summary of itself puts it this way:

> the construction order inverts: not spine first and evidence attached to it, but **all
> the reference material first**, decomposed into concepts, facts, claims and evidence,
> and only then the plot lines drawn through it. A book is a journey with pace,
> punchlines and characters, and the question the author is really answering is a
> whodunit

## Why this is the most valuable move in the book

Four things are worth pulling out of that moment, because each one generalises.

**He stopped it at section 5, not at the end.** He was reading the plan, noticed the
plan was answering questions that had not been asked yet, and said so immediately rather
than finishing the document and writing a considered review. Speed of objection matters
more than polish of objection when the build cycle is measured in hours. A vague
complaint delivered inside the hour beats a precise one delivered tomorrow, because the
agent will have built tomorrow's version by then.

**He named where the fault was.** "Section two, where you say what exists today, because
this is already making a number of judgement calls." That is not "I do not like the
plan". It is a location and a mechanism. An agent can act on a location and a mechanism.
It cannot act on a feeling.

**He praised the part that was right.** "The most amazing things we've done when we
mapped the book and we created the concepts maps is this section where you talk about
the the book's graphs hold six families of nodes." The six node families survived the
cancellation and became the backbone of everything built afterwards. Cancelling a plan
while naming the part to keep is the difference between a redirect and a restart.

**The memo went into the repository unedited.** Including "I'm really good", which is
otter.ai mishearing something, and "the the" doubled four times. Nobody cleaned it up.
Chapter 3 is about why.

The consequence of brief 20 is the entire second half of this book. Because the
construction order inverted, the fifty releases that followed it went into building the
machinery for decomposing documents into graphs, rather than into writing chapters. The
book you would expect to have been written in that time does not exist. What exists
instead is a working surface for writing it, and a lot more confidence about what it
should say.

## Freezing the first edition

Before the new order could start, the old one had to be put out of reach. v0.4.0, on 23
August:

> The first edition moves to /v1/ and freezes. The second edition begins, empty. Phase
> 0 of the dev pack, and a move rather than the copy the plan originally proposed: the
> founder's decision, and the better one. A copy leaves two live trees and no rule about
> which is authoritative, so the first thing that happens is somebody edits the wrong
> one. A move leaves exactly one, and the v1/ prefix makes an edition's boundary visible
> in every path.

That parenthesis is worth a paragraph of its own. The plan proposed copying the first
edition to `/v1/` and continuing to work in the root. The founder said move it instead.
The reason is that a copy leaves two live trees and no rule about which one is
authoritative, and within a week nobody remembers which files they are supposed to be
editing.

The move cost something immediately: every URL of the first edition changed. The project
handled that by generating redirect stubs at all the old addresses, and then four
releases later, at v0.4.7, deleted all 108 of them:

> The redirect stubs are retired: 108 pages deleted, and the root is now exactly the
> model. Founder call: the stubs at the pre-move addresses added complexity to the
> repository structure, and there are no external users of this content who would hit
> the moved paths, so the insurance was costing more than it covered.

Two decisions, four days apart, both of which trade a small amount of user-facing
politeness for a large amount of structural clarity. Both were the founder's call, both
are recorded with their reasoning in the release table, and the second one supersedes a
technique the project had already written into its own methods register, where it now
sits marked "superseded at v0.4.7" rather than deleted.

## The empty second edition

What v0.4.0 created was a directory called `/v2/` with nothing in it. Not a draft. Not
an outline. A frozen first edition on one side, an empty space on the other, and a
memo saying build the universe first and find the plot afterwards.

Five days later that empty space held: an extraction pipeline, a document reader that
is genuinely an instrument, a core graph that decomposes a document to the individual
word, an identity ledger, a deterministic transformer engine with twelve operators in
their own folders, a file explorer, thirty-five registered methods, and nineteen briefs.

It still holds no chapters of the second book. That is not a failure of the method. It
is what the method said would happen, in the founder's own words, on day three.

---

**Where the live estate shows this.** The frozen first edition is at `/v1/`, its book at
`/v1/book/` and in one page at `/v1/book/single.html`. The memo that inverted the order
is at `/v2/memos/20-founder-memo-the-universe-first.html`, with its raw markdown at
`v2/briefs/20__founder-memo__the-universe-first.md`. The plan it reshaped is the ten-file
dev pack at `/v2/dev-pack/`.