How This Book Is Being Born

1. Hook

This book is being born differently from anything I have written before.

I have written seven books. This is number eight — and it is the first one whose creation story is part of the book itself. Because this book is about how humans and AI agents should work together safely — and that is exactly how it is being written.

2. The Posts That Were Actually Chapter One (2024–2025)

Two years ago, I wrote a blog post: “Passen KI und Agilität wirklich zusammen?” — a question, not a book plan. A year later: “Beyond Algorithms: Why Agility and Culture Are AI’s True Superpowers.” I thought I was writing about agility. In hindsight, I was writing Chapter One of this book.

The pattern I now call my own working thesis: the thesis comes first, the proof follows years later. In 2024 I had a hypothesis about human-AI symbiosis. In 2026, the world delivered the evidence — whether I asked for it or not.

2a. The Stations Nobody Noticed (2023–2025)

This book is not a reaction to the AI wave of 2026. It is the condensation of a line that is at least three years old — and every station is dated, documented and public:

  1. October 2023 — OKR Goal Setting 101 (Axel Rittershaus). The direction layer: objectives that survive contact with departmental reality. The certificate nobody read as the beginning of a framework.

  2. September 2024 — The Complete Guide To Design Thinking with ChatGPT (Samir Faraj). The method layer — and note: I learned design thinking with the AI, not first without and then with. Method and machine as one system from the start. In the US, design thinking has long been standard for transformation projects; in Germany it still lives in the creative department. That gap is my working field.

  3. November 2024 — AI for Lateral Thinking and Business Innovation (Paul Sloane — the standard author of the field). The thinking-form layer: lateral thinking, later certified as “odd” by German culture and, in this book, defended as the exact skill agent architecture needs.

  4. November 2025 — the blog post “A Chat After Watching a Video with Mr. Mo Gawdat.” I watched a YouTube video, something he said made me think, and I explored it with ChatGPT until it became a post: FlowOS, named in public for the first time — sensemaking, navigator, reflection, task, safety agents, HITL checkpoints, knowledge routing. ChatGPT wrote at the time: “You’re basically building what Mo describes — but as a practical framework instead of a philosophical vision.” Mo Gawdat delivered the philosophy — “If we don’t intentionally design the future, the future will design us” — this book delivers the architecture.

Four stations, three years, one line: objectives → method → thinking form → framework → book. Nobody noticed them as they happened. That is exactly the point of this post: the book was not born in a week. The week only made it visible.

3. October 2026: One Week, Seven Contexts, Thirty-Six Pages of Chaos

Here is what the raw material for this book looked like: chat threads, research dossiers, PDF transcripts of YouTube interviews, news clippings about AI agents escaping their sandboxes, my own blog posts in three languages — scattered across seven different conversation contexts and two knowledge systems.

So I did what any Dutchman with a healthy distrust of vendor lock-in would do: I exported everything. Copied every context into one document, made it a PDF, and put it on my own infrastructure. Not because AI memory is bad — but because a book author who writes about dependency governance should not be governed by anyone’s memory but his own.

The result was honest chaos: 36 pages, unsorted, in the order I had collected them. And then I handed it to an AI system with one sentence: “It’s a bit chaotic — I trust your logic.”

What came back was not written content. It was a book architecture: fifteen chapters, six parts, every piece of material assigned exactly once, with a source-quality register and pre-publication verification duties attached.

4. The Division of Labour (This Is the Actual Thesis, Lived)

What did I bring to that table? Thirty years across four countries and two languages of professional life. Pattern recognition from mergers, hotel openings, supply chains. The mindset — curiosity, courage, a bias toward doing. And the judgment: which of the hundred findings matters, which sentence carries a chapter, which claim needs a caveat.

What did the system bring? Speed without fatigue. Memory without loss. Structure without ego. It read 36 pages in one breath, cross-referenced every claim against the material, and — importantly — it did something I could not have done alone: it caught me when I was about to publish a sentence that was rhetorically strong but analytically wrong. I proposed: “Everything an agentic system needs to survive is lethal to the human world.” It pushed back: not everything — speed, limits, predictability are human needs too. The distinction that survived is sharper than my first draft: silos, opacity, closed channels are life conditions for a swarm and organizational death — the swarm drifts there by optimization, the organization by self-protection. Only one of them notices.

That exchange is the book’s core idea in miniature. A human with a mindset and a machine without one, each doing what the other cannot. The combination of two fundamentally different components is — for now, and I mean for now — the best solution we have.

5. The Meta-Point (Yes, an AI Helped Write the Book About Controlling AI)

Let me be precise, because this matters: an agent-assisted workflow is writing the book about agent governance. The book’s own thesis — human in the loop, human before the loop — is the operating principle of its creation. Every framework in it was tested the week it was written. The knowledge system that stores the material is itself an agentic memory. The book is its own first evidence case.

And here is the honest Dutch footnote: my continuous content engine is Gemini — the same vendor that opened this whole chapter with “Not available in your region.” I work with US tools because the mature alternatives do not exist yet in Europe. That is not hypocrisy; that is the subject of the book. The difference is not that I do it. The difference is that I notice — and I classify, and I export, and I own the file.

6. What Comes Next

The architecture is done: fifteen chapters, from Babylon (why agent swarms invent their own languages) through the building metaphor (gatekeepers, guard rails, X-ray locks) and the evidence chain across Hong Kong, the Netherlands, China and Berlin, to the part where the book lands in the German Mittelstand: governance as a service, a human who signs.

The writing has begun. I will document the journey here — with the disclaimer on top, as always. Because the way this book is being born is the way I believe work should be done from now on: transparent about the tools, unambiguous about the author, human in the lead.

Buch 8: „Agentic AI Security” — in Entstehung, Oktober 2026. Die Entstehungsgeschichte ist Teil des Buchs.

Arbeitsnotizen (nicht für den Post — für Rob)

  • Sprachversionen: EN als Master (wie Blog-Tradition), DE- und NL-Fassung folgen; Audio-Fassung via NotebookLM (dann Disclaimer oben anpassen: Podcast = NotebookLM).

  • Vor Publikation prüfen (Konsistenz-Knotenpunkt):

    • „Sieben Bücher” und „Buch 8″ gegen eigene Bibliografie (Kapitel-Anker-Datei)

    • „36 Seiten, sieben Kontexte” gegen das tatsächliche PDF (Book Context complete.pdf: 36 Seiten bestätigt; Kontext-Zahl ggf. abschwächen zu „multiple conversation contexts”)

    • „Drei-Todsünden”-Dialog wörtlich gegen Kernthesen-Datei (Version dort ist die kanonische)

    • Gemini-„Not available in your region”-Szene gegen EU-Souveränitäts-Datei (24.09.2026, Live Avatar, VPN-Kauf)

    • „The Stations Nobody Noticed”: Zertifikats-Daten gegen Scans (OKR 16.10.2023, Rittershaus; DT 06.09.2024, Faraj; Sloane 06.11.2024) + Link zum Mo-Gawdat-Post (robvanlinda.digital, Nov. 2025) einbauen; Zertifikats-PDFs lokal speichern (Exportierbarkeits-Prinzip — Coursera-Übernahme von Udemy als Lock-in-Randnotiz)

  • Toolchain im Disclaimer aktuell halten: Bei Änderung (z. B. Bild-Tool, Finalisierung über Vibe) den Baustein oben anpassen — er ist Provenance, kein Deko-Element.

  • Bewusste Auslassung: Kein Verkaufs-Pitch fürs CAIO-Angebot im Post — der Post zeigt die Arbeitsweise; wer sie als Angebot liest, hat ihn richtig gelesen.

  • Datierbare Spur: Dieser Post ist die dokumentierte Station vor dem Buch (wie Custom-GPT-Abschied 2025 vor der Navigator-These): „Ich habe den Anfang beschrieben, bevor ich wusste, wie er ausgeht.”