How I accomplished this work

Before I touch a tool, I think through the trajectory and its consequences, step by step, over weeks. I work with paper and pen. I make sketches, mind maps, and doodles. In parallel I record voice notes, my thoughts on each single point, the for and the against, and create transcripts from them.

Then I give different AI systems my transcripts with the instruction to order them and then go through and discuss them point by point. I used Claude and Grok, initially also ChatGPT, but I dropped that one because it produced too much chatter and regularly distracted me from my work through engagement questions, almost like a salesperson.

The decisive point: I brought the large theses and considerations and set them against the system, the way one sets an argument against a sharp conversation partner.

This requires a particular persistence, because I quickly learned: I could not trust the AI's answers. On the one hand it tried to soften the consequences of arguments, on the other hand to reject my considerations and theses as unscientific and unprovable.

I did not let up, because my arguments seemed valid and logical to me. I searched for studies and evidence, brought new arguments, and contradicted many times, often over ten times, until the AI finally had to concede the validity or the provability.

The machine accelerated the research and sharpened the text. The responsibility for what is true and what counts lies with me.

What was special about it was that it enabled considerations that I would otherwise have had to conduct in discourse with people competent in those fields. Much more intensely, and without human pause or interruption.

Instead of five feedback loops with a human who gets tired after a few hours or wants to talk about another topic, and where the next meeting is often only possible many days later, no limits were set for me here. For three months I could question, research, sharpen, and try to falsify my theses for sixteen to eighteen hours a day. Instead of three years I completed this work in three months.

Beyond that I used AI systems as instruments: for source research, for checking evidence, for formulating prose, and for translation between German, English, and Spanish.

I print the results on paper and go through them once more. I make handwritten notes, improvements, and additions on these sheets, and highlight parts with colors and markers. Then I cut the sheets into individual pieces and arrange them on the floor in a different structure, to glue them onto an A2 sheet and hang them on the wall for reflection.

After that I usually had new insights, and noticed connections I could not see before.