Catherine Doyle (invented)

A note on how the manuscript was made: voice memos, journalist agent, librarian, review board, and a pre-publication candidate that is explicitly not a version. I also scanned the contents: 41 chapters in 10 parts, 75,000 words, and some chapters admitted to be placeholders.

Honest, but this is a production diary, not a cover letter. A commissioning editor wants to know what the book is and who buys it, and this front matter spends its words on process. Placeholders in a manuscript I am asked to price worry me.

Where the sources run out, the chapter says so, in a note addressed to Dinis. ✓ verbatim

Question: Is this a second edition or a new book? The note says it is still undecided.

Confusion: The cast of roles (board, advocate, cutter, checker) is explained to someone who has not yet been told what the book is about.

Next: Read on: the proposal is chapter 1, so I read it properly.

The pitch. Two readers, Gen Z entering work and Gen X subject-matter experts, joined as Gen ZX with a ZX Spectrum hook. The stated objective is that the reader starts a GenAI company by the end. The measure of success is companies created.

This is a real hook and a clear audience: the expert founder aged 40 to 55 is a buyer I recognise. The objective is bold, and bold objectives can be failed in public. The 'XZ backdoor' aside and the Spectrum nostalgia are charming but the book then promises company-formation content I suspect is not there.

Our job is to empower the reader to create a startup by the end of the book. ✓ verbatim

Question: Who verifies the 'fastest-growing one in a thousand' founder statistic, and is it a figure I could put on a cover?

Confusion: The same chapter says it is for two generations and then says millennials can read it too; the buyer is blurred.

Next: The proposal has a hook. Check the contents and the first part.

A two-line part introduction listing five chapters.

Nothing to judge. Five chapters before the argument starts is a lot of runway for a business reader.

Next: Read chapter 2 for the opening voice.

Wonder and optimism: 'you are lucky to be here', the list of technology waves the author lived through, the dot-com crash, and a rebuttal of AI fear. The author admits he could not read the Business Insider piece and cannot name the podcaster.

Warm, personal, a real voice. But several unsourced anecdotes and an admission of not having read the article he rebuts weaken it for a sceptical trade reader. A book that opens with 'lucky to be here' sells on mood, and mood dates fast.

I could not get past the paywall, so I cannot tell you what it argued. ✓ verbatim

Question: Why does a published book's opening rebut an article the author could not read?

Confusion: Notes to Dinis inside the chapter body make it hard to tell what is the book and what is the working draft.

Next: Skim the graphs chapter, then jump to the technical frame.

Graphs as networks of nodes and edges, the detective board, Wardley maps as graphs with positions, code as a layered graph. A claim about an sgit commit reaching nine of 72 commands.

Interesting if you already like graphs. For a business-book buyer it is a big conceptual ask this early, and it leans on the author's own tool as the proof.

What is crazy is to make decisions based on shadows of reality. ✓ verbatim

Question: Why does a founder need a chapter on code as a fractal graph before they have formed a company?

Confusion: Jumps from the detective board to AST and call trees without warning.

Next: Skip ahead to the frame, then sample what each part promises.

What generative AI is, in the author's working picture, followed by a list of corrections an AI agent made to him (vectors not graphs, the prompt does not change the model, hallucination has a second cause). The title note explains the Gen ZX name.

The corrections section is unusual and honest, but it also tells me the main explanation is a metaphor the author knows is wrong. For a non-technical reader the 'working picture' is clear; the corrections are for experts. The Lee Brimelow image is a rights problem I would have to clear.

I asked the agent that is helping me write this book to correct me where I was wrong. ✓ verbatim

Question: Which of these two pictures is the one the book relies on later, the metaphor or the correction?

Confusion: The chapter teaches a picture and then undoes it in the same chapter.

Next: Jump to the first part about jobs, then the centre of the book.

Hired for the workflow. Operator skills, hiring statistics from several surveys, the advice to send a tailored analysis with your CV.

Practical and sellable to the Gen Z half, packed with statistics from LinkedIn, Microsoft, PwC and Gartner, which will date within a year. Useful for a trade promotion if sourced to the letter.

An edge is not a job. ✓ verbatim

Question: Who checks 1,165 percent and 62 percent against the originals before print?

Next: Sample the working-with-agents part.

Sessions, briefs, role files, shared state, and rules that grow. Counts of commits and messages from the author's own repositories.

The substance is real and the 'every mistake added a rule' line is a good one. But the evidence is the author's own newsroom; to my reader it is a case study of one company.

Every mistake added a rule, and every new rule caused new mistakes. ✓ verbatim

Question: Which of this is advice for my reader's first company and which is the newsroom's internal practice?

Confusion: Sessions named with an @ prefix and the Librarian are introduced as if known.

Next: Now the centre of the book, Start the company.

Part introduction for the centre of the edition: ship, price, sell, fund, make the company real. Flagged as not legal or tax advice.

This is the part the whole objective stands on. Six chapters; I will sample two.

From experience, mine and the people I work with; not legal or tax advice. ✓ verbatim

Next: Read the pricing chapter, then the company chapter.

Four moves to a product: define, price, deliver at linear cost, offer free to people who know you. Free is not zero. Pay-after-value, the Stripe fee arithmetic, a reading meter at 10p an article.

Best chapter so far. Concrete numbers, a counter-example, an honest 'no results yet'. It is built on the author's own RiskMandate ladder, which makes it read as a case study of his company. The meter is a live experiment on his site, which I cannot sell as evidence.

A commercial product is one people take when the only cost is their own attention. ✓ verbatim

Question: What happens to the argument if the reading-meter review on 6 December 2026 fails?

Next: Read the company chapter, the one the objective depends on.

A chapter that says it is a placeholder. Two lines on Portugal and Lithuania, a section on Upwork, two notes to Dinis asking for memos.

This is the book's central promise and it is empty. The UK company formation, EU programmes and cloud credits are what a founder buys the book for. A publisher cannot take a manuscript where the headline objective is a stub.

This chapter is a placeholder, and it says so. ✓ verbatim

Question: When will the memos exist, and will the book wait for them?

Confusion: The title lists four jurisdictions and the text delivers essentially none.

Next: Check the sgit and privacy chapters, since the author's own products come in there.

Encrypted version control, the vault key, read keys, write-only lanes, key management, two corrections (constraints, GDPR). Quotes from the EDPB, ICO and the Court of Justice.

Well argued and carefully hedged, including 'None of this is legal advice'. But it is the author's product, beta, with a brief essay in the middle on why nobody buys it for the encryption. As a publisher I see a product chapter inside a business book, which has costs for trust and for shelf life.

that last part is a feature with a real cost, which I would rather say up front than have you find out ✓ verbatim

Question: Is a reader supposed to adopt sgit, or just understand the principle?

Confusion: Dense terms (append token, read key, vault of vaults) arrive faster than they are defined.

Next: Check the hands-on part, which should be what readers act on.

A folder per contact, five documents each, a six-line agent policy, a one-week exercise plan. Clear and usable.

The most usable chapter, and the kind of thing I can put on a back cover. The contents list a 'Your first workspace' chapter in Part 4 that looks like the same idea, so I suspect it is duplicated across Part 4 and Part 9.

Draft and never send; append and never rewrite; stay in the folder. ✓ verbatim

Question: Why is this workflow in the book twice?

Next: Check the form-your-company chapter, then the ending.

Another placeholder: what the chapter will cover, three facts, five decisions to make before registering.

Second stub for the same missing content. The five decisions are sensible but they are a checklist where a procedure was promised.

By the end of this chapter you will have a company: registered, with a name, a country and a bank account. ✓ verbatim

Question: How can a chapter promise a registered company when it says the steps are not written?

Next: Read the last chapter, then decide.

Two-line intro to the frame part.

Fine, nothing to say.

Next: Skim chapters 4 and 5.

Wardley maps introduced by drawing one first, the four stages, where do you fit, centre of gravity, jobs that never scaled.

Good teaching device, drawing before naming. But it is a third conceptual framework (graphs, maps, lenses) before any company is formed.

A map you cannot be wrong about is a diagram. ✓ verbatim

Question: Does a first-time founder need Wardley maps to follow the later chapters?

Confusion: The 'how to use this book' section tells readers to jump by centre of gravity, which conflicts with the instruction in chapter 2 not to skip.

Next: Skim chapter 5.

Steve Jobs dots, SG/Send as the author's own dot, multidisciplinary edge, maths teaching, learning, thinking with the AI.

Motivational, a generic trade-business chapter with some personal evidence. Could be shortened.

Your job now is to make dots worth connecting. ✓ verbatim

Confusion: Sections on maths teaching drift away from the founder theme.

Next: Skim chapter 7.

Lenses, roles, loops, tools, memory, and why long chats are compressed.

A good plain-language explanation of agents, which my reader needs. Chapter 32 then repeats the five-lens example, and chapter 33 repeats the loop.

Each of those roles is, in essence, almost a raw model plus a role description and a question. ✓ verbatim

Next: Skim chapter 8.

Anyone who can brief an agent is a developer; spreadsheets, errors, Portuguese as a programming language, what a briefing-only developer cannot do.

Good for the Gen X expert. The spreadsheet error statistics are a strong hook for a book cover.

Of 54 spreadsheets audited between 1997 and 2000, no fewer than 49 were found to contain significant errors ✓ verbatim

Question: The section on what the briefing-only developer cannot do is thin and flagged as such. Can it be answered?

Next: Skim chapter 9.

Explorer coding versus vibe coding, handover, villagers, bonus tests, no new features while you move.

The renaming is a good hook and probably the book's most shareable idea. The handover material has substance.

What people call vibe coding is exploration, and the word should say so. ✓ verbatim

Question: Does anyone outside the author's circle use 'explorer' and 'villager'?

Confusion: Wardley's explorer, villager and town planner terms are used without a quick refresher.

Next: Skim part 3 and chapters 11 and 12.

Intro to getting the job, or making it.

Good framing of the old chapter becoming a new one.

Next: Skim chapter 11.

Act as the job you want, the 2018 echo, do it in public, put your CV online.

Solid, easy to market to the Gen Z half. Quoted back to the 2018 book heavily.

the bottom line is the same: if you are looking for a job, you have time ✓ verbatim

Next: Skim chapter 12.

Publish constantly, the author's unfinished books on Leanpub, a book as a curated set of finished articles.

Disarming that the author lists his stalled books, but a publisher reading that note takes it seriously: who finishes this one?

a book is a curated, ordered set of finished articles, plus the editorial thread that connects them. ✓ verbatim

Question: What guarantees this book is finished, given the stalled ones?

Next: Skim chapter 13, then part 4.

The row becomes a folder: Record, Ledger, Messages, Notes, Summary, a six-line agent policy, five exercises.

This is the same workflow as chapter 30. Cut one or merge.

No developer, no new tool, no code. ✓ verbatim

Question: Which of the two is the canonical version?

Next: Skim part 4 and move to part 5.

Intro to working with agents.

Part 4 is only two chapters, one of them overlapping with chapter 30.

Next: Skim part 5 chapters.

Intro to direction and judgement.

Two chapters; good topic for a business book.

Next: Skim chapter 15.

Direction is the scarce skill, places of highest leverage, knowing when to stop, a budget per step, agents decide and people answer.

One of the strongest ideas in the book and well suited to the executive market.

Direction is the scarce thing. ✓ verbatim

Next: Skim chapter 16.

Ask for evidence, provenance, how to disagree rationally, reality corrects the graph, be the best-prepared person.

Useful but long and returns to graphs; chapter 35 repeats the preparation point.

Karl Popper's principle is short: a claim earns its status by being refutable. ✓ verbatim

Confusion: The 'privacy and Snapchat' passage is a side trip.

Next: Skim chapter 17.

The playbook that worked, why experts were always the better bet, where the opening is, where the company sits, there is a company in this.

The most direct statement of the audience. Discloses that two of the three expert founders are people the author helps; I would want that disclosed more loudly.

Question: Are the three expert founders real, named and willing to be quoted?

Next: Skim chapter 18.

Ship it, give it away then take it away, five days and ten customers, the loop is the moat.

Good argument and a good story, though the story is about the author's own product.

Not to plan, not to pitch, not to architect. Ship. ✓ verbatim

Question: Is the ten-customers story typical or exceptional?

Confusion: The text notes its own tension with 'vibe coded' versus explorer coding.

Next: Skim chapters 20 and 21.

Sell: answer the question in the room, sell value that lands the day they switch, sell a service before the product, what the buyer is about to do to you.

Credible and specific. Mentions a named consultant, which needs permission.

impressed by evidence and bored by claims ✓ verbatim

Next: Skim chapter 21.

Do not raise before you are profitable; who investors choose; sell rather than issue shares; what breaks.

A contrarian view with a flagged bias (the author's companies run on open source). Needs a legal read before print.

The worst possible time to raise money is before you are profitable, before you have the product, before you understand the fit. ✓ verbatim

Next: Skim part 7 and chapters 23 to 27.

Intro to building without a server.

Part 7 is where the author's products cluster.

Next: Skim chapter 23.

Serverless defined by what it costs when idle, with a pricing table dated 18 October 2026, cloud as Lego, files as the database.

The by-the-hour test is a good practical rule. The table will be out of date within a year.

Ship your company serverless. ✓ verbatim

Question: Will the table be refreshed for each printing?

Next: Skim chapter 25.

Mandate, reach and the gap for agents, the policy beside the prompt, who can pull the plug, secrets.

Good for corporate bulk buyers. It is also the pitch for RiskMandate.

A control bounds a grant only if it is enforced by something the grant does not include. ✓ verbatim

Question: Where does the advice end and the product begin?

Next: Skim chapter 26.

Open source: what it is and is not, the fork next door, not free, all of it or an evaluation version, sell the build not the code.

Useful and fair, but it reinforces the free-edition risk: the author's argument is that giving it away is the strategy.

It was never about contributions. ✓ verbatim

Question: Does the author's open-source stance cover this book's CC BY licence in the way that hurts print sales?

Next: Skim chapter 27.

A release that passed every check and never reached anybody; green does not mean live; tests, twins, and five synthetic people.

A strong story, but engineering-heavy for a founder reader.

Getting a working thing into a stranger's hands is still, in 2026, genuinely hard ✓ verbatim

Confusion: The 'five people who do not exist' section is unexplained at skim level.

Next: Skim part 8 and chapters 28 and 29.

Intro to the long game: graphs as memory and civic responsibility.

A short part at the back of the book that feels like an epilogue, but a heavy one.

Next: Skim chapter 28.

Backup your life, thinking in graphs, memory as context management, publish and the site becomes the memory.

Third graph chapter; this is the book's personal philosophy and risks going on.

You basically need to backup your life! ✓ verbatim

Next: Skim chapter 29.

Portugal, the OWASP keynote, EU regulation, state-paid code belongs to the commons, smaller models, sovereignty.

Interesting, and interesting for the Portugal and EU market, but it is a policy chapter in a startup book. A Portuguese edition may be the real home for it.

It only takes courage, focus and commitment ✓ verbatim

Question: Is this chapter for founders or for policymakers?

Next: Skim part 9 and its chapters.

Intro to the eleven hands-on workflows, dated 2026.

Good idea to date and renew them each year, which suits a series of annual editions.

Next: Skim chapters 31 to 36.

Draw a graph of something you know, a question not a subject, forty-odd lines in a text file.

Practical and the failure mode (picking a subject) is spot on.

if somebody tells you otherwise in the first month, they are solving a different problem from yours ✓ verbatim

Next: Skim chapter 32.

Five-lens exercise on an idea of your own.

Repeats the five-lens passage from chapter 7 almost word for word. Merge.

Confusion: I could not tell what this adds to chapter 7.

Next: Skim chapter 33.

Run an agent loop by hand with yourself as the harness.

Again restates chapter 7, but with an exercise. Good teaching if the earlier version is trimmed.

That is where the control lives. ✓ verbatim

Next: Skim chapter 34.

Storyboards as data drawn in the browser, checked quotes, the cast.

This is about how the website works. It reads as documentation for the author's own site, not a reader skill.

There is no image file to edit. ✓ verbatim

Question: Who is the reader for this recipe?

Confusion: I do not understand why a founder would build storyboards.

Next: Skim chapter 35.

Meeting briefs and follow-ups, a short chapter with the author's step by step still missing.

Useful idea, unfinished.

Now it is a brief and a feedback loop. ✓ verbatim

Next: Skim chapter 36.

Eight steps for a newsroom site and an inbox agent.

Another recipe tied to the author's site. Fine as an appendix.

Next: Skim chapter 38.

A table of startup credits and programmes as of 7 October 2026.

Commercially valuable to a founder, shelf life of months, and the unnamed vendors will be obsolete before the ink dries. Better as a web page.

Programmes change; check before applying. ✓ verbatim

Question: Who updates it, and when?

Next: Skim chapter 39.

Price it first, offer to people who know you, write a stop rule, count.

Repeats chapter 19's four moves, now as a workflow.

Confusion: The four moves appear in nearly identical words in chapter 19.

Next: Skim chapter 40.

A weekend recipe: static site, vault data, a model on the page, a payment link.

The most satisfying promise in the book, and it leans on the author's vault tool.

If it bills you by the hour when nobody is using it, it is not serverless. ✓ verbatim

Question: Can this be done without the sgit vault for readers who do not adopt it?

Next: Last, part 10.

Intro to the last part: how the book was made and the measure.

The process is the product, says the author; I disagree for a trade book.

Next: Done; give structural feedback.