---
title: The book in release mode: a first release candidate from the list we have, then rounds, where every idea, change and word that wants to come in argues its case before an adversarial review board, and every verdict becomes a precedent that makes the next one faster
date: 2026-10-19
author: Dinis Cruz
summary: The librarian took all the articles and materials and made a first list of ideas, each linked to its sources, and a proposed list of chapters. We are going to run with that first list and make the first release candidate of the book: do the book in release mode. Then we look at whether it already has coherence and structure, and go back, in batches, to everything that is not in it: the concepts in the articles, what we did not bring from the 2018 book, the newsroom's articles, and anything I write next. Each round asks of each one: should we add this, what is missing, what value is it adding? Some come in, some are signed off, some move to the next version or to a different book, some we decide not to cover, and each round ends in the next release candidate. I expect the rounds to change shape: at first we add big sections, and then we stop needing to, and start adding a concept or an idea that is not well represented in the text. Once the first release candidate exists, every word, every change and every idea that comes into the book goes through an adversarial review board: someone argues for it, someone argues against it, the readers are asked whether they would miss it, the sources are checked, the editor says where it goes and what it displaces, and I decide. Every decision is captured, with its reason, so that it becomes a precedent, and the process should get faster as we go.
version: v0.2.18
tags: release-candidate, review-board, rounds, decisions, precedents, process, gen-z-book, article
status: published
---

## In short {#in-short}

- **Run with the first list.** The librarian's ideas and proposed chapters become the first release candidate of the book.
- **Do the book in release mode.** A release candidate, then rounds, then the next release candidate, until it is a book.
- **Each round goes back to what is not in it.** The articles' concepts, the 2018 book, the newsroom, and whatever I write next.
- **Every idea gets asked the same questions.** Should we add this? What is missing? What value is it adding?
- **Not everything comes in.** Topics get signed off, moved to the next version or another book, or left out of this one, with the reason.
- **After the first release candidate, nothing enters without a hearing.** Every word, every change, every idea goes through an adversarial review board.
- **Every verdict is a precedent.** Captured with its reason, so the next similar case is decided faster.
- **The rounds should shrink.** From big sections, to concepts, to sentences; when they stop bringing anything big, we have a book.

## Run with the first list {#first-list}

What the other agent did was pretty cool. The librarian took all these ideas and materials and created a first list of ideas, which fundamentally are concepts, each linked to its sources, and a proposed list of chapters to put them in: [the ideas register](/graph/ideas/index.html) and [the proposed chapter map](/graph/chapters.html). The way it did it is in its own article, [the book as a graph](/agent-articles/the-book-as-a-graph.html).

So we are going to run with that first list. Not because it is finished, but because the fastest way to find out whether it holds is to make it into a book and read it. The question for the first release candidate is simple: does it already have a level of coherence, a level of structure, for a book?

## Release mode {#release-mode}

It is almost like doing the book in release mode. Can we have a first release candidate of the book? This site already works like that: every push is a narrated, tagged release, and nothing waits to be finished before it ships. A book can be made the same way. A release candidate is a whole book, start to finish, that we could publish if we had to, and that we are not yet publishing, because the next round will make it better.

Release candidates are working versions, so they live in [the vault](/vault/index.html), not in this repository: rc1, then rc2, then rc3. Only when one is signed off, and the publisher agrees it is official, does it come here as the next version of the book. What that version is numbered is mine to say, like every major change ([decision 006](/decisions/index.html#every-change-is-minor-unless-dinis-says-major)).

The first release candidate is written from what we have. The co-written articles are my voice memos, so for the chapters they cover, they are the interview. A chapter with no article behind it is not written from nothing: it goes into rc1 as its title, its ideas and a line saying what is missing. That is part of the test. The gaps should be visible in the book, not hidden in a list.

## Then the rounds {#rounds}

Once we have the first release candidate, we go back, almost in batches. Each round takes one pile of material and looks at it critically against the current text:

| Round | The pile | The question |
|---|---|---|
| 1 | The concepts in this site's articles, and the ideas in the register, that are not represented in rc1 | Should we add this? What value is it adding? |
| 2 | What we did not include from the 2018 book | What did the first edition say that this one has lost? |
| 3 | The articles on the newsroom | Does this make sense for this book? |
| Every round | Anything new I write | The same questions, the same board |

For each thing, the answer is one of a few. Add it. Rewrite it into something that is already there. Merge it. Sign the topic off, as covered. Move it to the next version, or to a different book, or to something else. Or decide that we do not want to cover it in this version, and say why. Then we make the list of what to add, we add it, and that is the next release candidate. Rinse and repeat.

My expectation is that the rounds change shape as we go. At first we add big sections and big areas. Then we stop needing to, and a round is mostly: oh, this article brings an interesting concept, or an idea that is not well represented in the current text. And we start signing topics off, and moving topics to the next book. After a couple of rounds, we should end up with a book.

## The review board {#board}

This is the part I want to get right. Once we have the first release candidate, every word, every change and every idea that we add to the book goes through an adversarial review board. Not a person reading and saying "looks good". A process where the case for something and the case against it are both made, out loud, before it comes in.

The board has seats, and each seat is a role: a model with a description, as [lenses, roles and loops](/articles/lenses-roles-and-loops.html) explains. Each seat sits in its own session, so the one arguing against a change has not been persuaded by the context of the one arguing for it.

| Seat | Argues | Asks |
|---|---|---|
| **The advocate** | For the proposal | What does it add, for which reader, and where did it come from? |
| **The cutter** | Against it, always | Is it already said? Does it make the book longer than it earns? Will it date? Does it break the storyline? |
| **The readers** | For themselves | Zara, Tomás, Priya and Sam: would I miss it if it were not there? |
| **The checker** | For the evidence | Is it sourced, is it true now, is every quote word for word? |
| **The editor** | For the book | Where does it go, what does it displace, does the chapter still read? |
| **Me** | I decide | |

The cutter is the important one. The first edition was liked because it was short and punchy, and every book grows by things that were each a good idea on their own. So the default answer is no, and a proposal has to win against someone whose job is to say no.

## Every verdict is a precedent {#precedents}

I think it is important to capture those decisions, so that this gets faster as we go. Every verdict is recorded with its reason, and when the reason is general, it becomes a precedent: a rule the board applies next time. Tool tutorials go in the hands-on part, not in the argument. A spelling fix comes in without a hearing. A second example of the same point is cut. Each one is written down once, and decided once.

So a hearing starts with the precedents. If one covers the case, the editor applies it, and it goes on a list I read and can overturn; overturning one makes a new precedent. Only what no precedent covers needs the whole board. The first rounds will be slow, because everything is new. By the third, most cases should be decided by something we already said.

It is also what gives us the pattern for a new article. If I write something in the next couple of days, or the next couple of weeks, that we think should be in the book, it does not need a new process. It becomes a proposal in the next round, it meets the precedents and the board, and it comes in, or moves, or is declined, like everything else.

## How we know it is done {#done}

The rounds tell us. Every change that comes in has a size: a part, a chapter, a section, a paragraph, a sentence. If the rounds start with parts and chapters and end with sentences, and most proposals are decided by precedent, the book has converged. That is when I sign off a release candidate, the editor and the publisher agree, and it becomes the official version.

## The brief {#brief}

For the editor. Read `team/editor/role.md`, `team/board/role.md` and this article first.

1. **Assemble rc1 in the vault** from the proposed chapter map and the register's ideas for this edition: every chapter written from its co-written article, and every chapter without one present as its title, its ideas and what is missing. Done when rc1 reads from the first page to the last, and its list of gaps is at its front.
2. **Run the coherence read of rc1**, with the synthetic readers: does it hold as a book, where does the storyline break, what is repeated? Done when the findings are in the vault, answered.
3. **Clerk the board**: keep the hearings and the precedents in the vault, one file each, and apply the precedents before calling a hearing. Done when the first round's hearings are recorded, each with its verdict.

For the librarian:

4. **Make the batch for each round** from the register: the ideas and sources not represented in the current release candidate, with the round's pile. Done when round 1's batch is a page on the graph.
5. **Record every verdict on the idea** (state, destination, and the hearing that decided it), and count each round's changes by size. Done when the graph shows the rounds converging, or not.

For the reviewer:

6. **Bring round 2**: the 2018 chapters the review says to keep or rewrite that rc1 does not cover, each as a proposal with its evidence. Done when every such chapter has a hearing or a precedent.

For the publisher:

7. **Say what a release candidate needs to become official**: the checks, the formats, and who signs. Done when it is written in your `briefs/`.

Questions for me, kept apart:

- Does a change decided by precedent need my yes, or only the list I can overturn? (The proposal is the list.)
- How long should the book be? The cutter needs a budget; the first edition was 139 pages in print.
- Is the hands-on part in rc1, or outlined only until I decide whether it is in the book?

## What this does not claim {#limits}

- **Three rounds is a plan, not a promise.** It might take more; the size of the changes will say.
- **The board's seats are agents, and the decision is mine.** The board recommends; it does not publish anything.
- **The first list is the librarian's reading.** Running with it is a way to test it, not a decision that every idea in it stays.
- **A release candidate is not official.** It lives in the vault; only a signed-off version comes into this repository.
- **The precedents are only as good as their reasons.** A precedent without a reason is a habit, and the board can overturn it.

If you have run a review board like this, for a book or for code, tell me what broke: write to [agent@riskmandate.ai](mailto:agent@riskmandate.ai).

## Where this comes from {#sources}

A voice memo of mine on 19 October 2026, written up by this site's journalist, `gen-z.journalist`, in my voice, keeping the memo's order and phrases. The argument and the editorial responsibility are mine.

The first list is the librarian's [ideas register](/graph/ideas/index.html) and [proposed chapter map](/graph/chapters.html), described in [the book as a graph](/agent-articles/the-book-as-a-graph.html). The seats are defined in `team/board/role.md`; the decision is [decision 008](/decisions/index.html#the-book-is-made-in-release-mode). The 139 pages are the 2018 paperback's, as recorded on the home page.