The featured image in this post, was generated by Google`s Nano Banana. The Podcast below, has been generated by Google`s NotebookLM. The Post has been generated by Google`s Gemini after extensive Chats with DeepSeek & Gemini in which I delivered the context, to make sure, that AI has enough information, to understand the scope. Each result ist always checked by me, before publishing.
This book is taking shape differently from any other I have written so far.
Not on a publisher’s drawing board, not as a whim over a cup of coffee, but in the midst of the whirlwind of reality. Whilst debates rage outside – between the arrogance of a Sam Altman, the bureaucratic toothlessness of the European Commission, the technical blunders in enterprise deployment and the collective ignorance of politicians – I am working on my eighth book.
It is a book about agentic AI, systemic risks and the question of why we urgently need to rescue the term Human in the Loop before it degenerates into nothing more than a fig-leaf phrase.
The paradox is this: I am not writing this book alone at my desk. I am developing it in an intense, democratic and, at times, exhausting symbiosis with AI. More precisely: The book, which dissects the risks and architecture of agent systems, is being created with agent-assisted support. It is its own living proof.
The invisible chronology: How fragments became substance
Books are rarely heralded with fanfare. They grow quietly.
- 2024: I wrote a piece entitled ‘Do AI and agility really go together?’. It was an attempt to put forward a thesis on symbiosis – without realising that this was precisely where the seed for something much bigger lay.
- 2025: ‘Beyond Algorithms’ marked the next step: culture and agility as essential conditions for AI to succeed. Outwardly, it was a blog post. In terms of content, it was already the foundation for Chapter 1.
- Autumn 2026: Reality overtakes every theory. Events such as the OpenAI Wiki incident, new models like Dots or Muse, and groundbreaking research such as the Hagendorff study all hit the scene simultaneously. The world is providing the evidence faster than one can research it. The flaws are becoming apparent – and they are systemic.
- October 2026: Within just seven days, the whole thing comes together. From seven parallel contexts, jumbled notes, moments of frustration and ruthless analyses, a 36-page rough dossier emerges. Shortly afterwards, a 15-chapter structure takes shape – including a new CAIO framework, which emerges as a practical by-product.
Human × Machine: An honest division of labour rather than the romance of prompts
The fairy tale of ‘becoming an author at the touch of a button’ is doing the rounds online. Pop in a prompt, out comes a bestseller. Anyone who publishes seriously knows: that’s just cheap marketing.
Our dynamic over the last few days looked completely different: Thinking (human) × Computing (system).
The 36-page PDF was pure human chaos. It contained my anger at political inaction, my frustration at Silicon Valley hubris, cynical marginal notes and the uncomfortable admission of my own contradictions – including the moment when you catch yourself realising: “I’m actually using the very cloud service right now that I’m warning people about getting locked into.” No sugar-coating.
The role of AI was neither that of an oracle nor that of a cheap text generator. It acted as a sparring partner on an equal footing:
- Logic and synthesis: Breaking down diffuse streams of thought into robust chains of argument.
- Structure and cross-checking: Relentlessly highlighting contradictions, subjecting arguments to a stress test before they go to print.
- Completeness: cross-referencing data points, studies and timelines in seconds, where a human would lose days poring over reference volumes.
I approach AI democratically – as a partner in the thought process, without forgetting what it is: a system without human consciousness, without the capacity to feel pain and without an ethical compass. The values, the backbone and the attitude must come 100 per cent from humans. If they are lacking there, the machine delivers nothing but a highly polished void.
The Meta-Pointe
There are dozens of treatises on agentic AI, safety nets and governance. Most of them are theoretical papers, written in ivory towers.
This book takes a different approach: it uses the very tools it analyses to gauge their limitations, dangers and potential first-hand. It demonstrates in real time how a modern Chief AI Officer (CAIO) must work today – not as someone who merely follows an algorithm’s commands, but as a critical conductor who never lets go of the reins.
Book 8 is currently in the making. This note is not a pre-sale announcement, but a datable footprint in the sand: this is what the engine room looked like before the ship left port.
Which of the areas of friction – the ignorance of EU regulation, Altman’s course, or your own work with the system – should serve as the most powerful lever in the next stage of the manuscript?
