Home / Articles / Short, punchy, and co-written with Claude: what the second edition of Generation Z Developers has to be, and how we show it is not AI slop

Short, punchy, and co-written with Claude: what the second edition of Generation Z Developers has to be, and how we show it is not AI slop

By Dinis Cruz · 2026-10-18 · article v1.0.0 · newsroom v0.12.119 · booksgeneration-zwritingqualitygen-z-bookarticle

Abstract: More notes for the book. It is a book I will send to anyone I know who wants to, or should, get into AI, so it needs chapters for people who do not yet work the way I do, including one on setting up a Google Workspace and the agentic workflow behind my emails. The first edition was liked for one thing above all: it was short and punchy, a condensation of ideas where almost every paragraph made you think. The second has to be the same. It will also be co-written with Claude, and I want to say so and credit it, and to show the ratio: how many words of voice memos, edits and reviews go into each word of the book. Readers will open it expecting AI slop. My first published pieces, made with ChatGPT's deep research, got exactly that reaction; the articles on this site now get told they sound like me. So the book should meet the suspicion head on, with the evidence. To keep it tight, every paragraph, group of paragraphs and chapter should know what it is: its evidence, its facts, its hypotheses, its purpose, all in a graph, which also shows where we repeat ourselves. We can refactor everything until each part is signed off. Synthetic readers will tell us what they understood and what they did not. And not every issue they raise needs fixing: sometimes the answer is, we are not doing that.

In short

• A book for anyone getting into AI. Including how to set up the workflow behind my emails.
• Short and punchy, again. The first edition's readers liked it for that.
• Co-written with Claude, and credited. With the ratio of memo to word shown.
• Meet the suspicion head on. Readers will expect AI slop; show them the evidence.
• Every paragraph knows what it is. Its evidence, facts, hypotheses and purpose, in a graph.
• The graph shows repetition. And we can refactor anything until it is signed off.
• Synthetic readers, and the right to say no. Not every issue needs fixing.

For anyone getting into AI

This is the book I will send to anyone I know who wants to get into AI, or should. Many of them do not have agentic workflows like the ones I use, so parts of the book are for them. One chapter should be practical: setting up a Google Workspace and the agentic workflow I have here, the one behind the man behind the curtain, where agents draft every email and I send it.

Short and punchy

People liked the first book. There are some kind reviews of it, and what I heard most was that it was short and punchy. Of course there were spelling mistakes, and of course parts could have been expanded. But it was a condensation of ideas, and almost every line, every paragraph made you think. That is what I want this one to be.

Co-written with Claude

This book will be co-written with Claude, and I want to credit it. Claude is putting it together: from my voice memos, my reviews and my corrections, it writes the text, and I read, challenge and change it.

I also want to show the ratio, because it is a good case study of how this works. How much of this did I write? measured it for this site: over four weeks I sent about 65,000 words, 55,000 of them in voice memos, and the agent published 96,000 words of articles, with the words out per word in running from about one to about five, article by article. The book should measure itself the same way: the memos, the interviews, the reviews, the edits, and the words that came out. And it should show where the voice is: where I start and where the agent starts.

Meet the suspicion head on

A reader will open a book co-written with an AI expecting AI slop, so the book should tackle that from the start, with evidence and data rather than reassurance: this is voice memos, a lot of them, going through quality control and editing, then reviewers, before anything is published.

I have seen both sides of this. My first published pieces were written with ChatGPT's deep research, and some people said exactly that: this reads like ChatGPT. The articles on this site get a different reaction. People say they are good, sometimes that they are long, but I do not see complaints about the quality or the voice. The comment I hear is that they sound like me. The book has to sound like that too.

Every paragraph knows what it is

To keep a book punchy, every part of it has to earn its place. So every paragraph, every group of paragraphs and every chapter should know what it is: the evidence behind it, the facts it states, the hypotheses it puts forward, and what it is trying to do. That is a graph, and a website to finish the book makes the book one.

The graph has a second use. Some books repeat themselves: the same idea, made again in a later chapter, because by the time the author noticed, the earlier chapter could no longer change. With a graph of the book, and a graph of those graphs, we can see where two paragraphs say the same thing. And unlike a printed book, we can change everything, up to the end. A book that is a graph can be refactored, so it can be made consistent, until the parts are signed off.

Synthetic readers

We should also create synthetic readers: simulated readers of different kinds who read the book and tell us what they understood, what they did not, what they liked and what they did not. We already know how to run them; how to run synthetic users is the method we have used on three sites, with invented users of several types, each with a profile, a question and an expected path, interviewed at the end.

Not every issue needs raising

And one rule that matters as much as the others: not every piece of feedback needs to change the book. Sometimes the right answer to an issue is: we are not doing that. A book that tries to satisfy every reader's every comment stops being short, and stops being punchy. This site's reviews already work this way: every finding gets an answer, and "declined", with a reason, is one of the answers.

What this does not claim

• The reviews are from memory. The Amazon listing could not be read for this article.
• The ratio is from this site, not the book. The book will measure its own.
• "Not slop" is something to show, not to say. The evidence is the point.
• Synthetic readers are a tool, not readers. Real readers come after.

Where this comes from

A voice memo from me on 18 October 2026. Related: how much of this did I write?, how to run synthetic users, the man behind the curtain, a website to finish the book, Generation Z Developers, again. Written up and checked by this site's agent, newsroom.writer; the editorial responsibility is mine.

Want the next issue by email. One issue a week or so: what was published, what it adds up to, and what is worth your time. Subscribe to the SGit Newsroom →

← All articles