Writing is how I think. The document it leaves behind is a by-product. For thirty years I have written extensively for work, and for twenty of those years I commuted between the US, Europe and Asia, analysing markets and trends and turning what I found into sales and marketing approaches. Thousands of those documents survive, much of it very precisely documented. It was written so that I could work out what I actually thought.
Writing does that for me in a way reading never has. I forget most reports and analyses within days of reading them. Stories stay, but the sober material I work with rarely arrives as a story. So I attended writing seminars over the years and practised writing meditations, aiming to sit longer inside a subject before I judged it. The method worked on the same premise every time: I do not know what I think about something until I have written a full paragraph about it, crossed most of it out and found the one sentence that survived.
That state is often called flow, something close to ease. What I feel while writing is closer to strain: holding a context and a chain of logic in place while watching for the next distraction that will knock it loose. A tab switched to check something, a document searched for, a number verified, a meal, a drink, any of these can cost hours, because the thought was not finished when the interruption arrived and it will not reassemble the same way twice. I even put off going to the bathroom until the very last moment, because the thought disappears faster than it forms.
I write in German, because it is my mother tongue and the sentences come easier there. Afterwards I translate, first with tools like DeepL and increasingly with AI like Claude, and I revise those translations less each year. I do not use editors or writing tools. The tool that changed things sits beside the writing and answers questions while the sentence is still open.
For most of my career the harder problem was retrieval. I wanted to write down what thousands of documents from those twenty years of travel had taught me, because every wrong decision in that career traced back to one large why, and I could feel that why without being able to state it. Technology made the data fast to reach, and for a while that fastness worked against me: I went looking for a figure or a precedent, found the search itself absorbing and did not come back. The context was gone and the paragraph had gone cold.
AI closed the distance between a question that arises mid-sentence and the answer to it. Research that used to take days is now a matter of phrasing the prompt correctly. I run a knowledge base alongside the writing, built from years of my own documents and texts, and it hands me a number or a name in the middle of a sentence instead of after a detour that breaks the sentence. Many of the essays I have written would not exist without that speed, because the old version of me would have gone looking and not returned. Along the way I found a use for ontology, a field I had always worked intuitively without developing it systematically; the tool gave the intuition a structure.
It would be easy to let a model write the essay outright. I do not, because that would stop me learning. The paragraph I write and mostly discard is where I find out what I think; a model asked to produce the finished version skips the discarding, and with it the reason I write.
Who this is for
The public argument over who wrote a text, a person or a model, strikes me as the wrong question. The question that matters is who the text is for. My answer has been consistent from the start: I write for myself first, for my own understanding. The response from other readers came later, as something I had not planned for. Essays became the form that serves both jobs at once, mine while I write them and someone else’s once they are finished.
Timothy Shoup of the Copenhagen Institute for Futures Studies forecasts synthetic content at 99 to 99.9 percent of the internet by 2030. The forecast is a projection, a marker of the scale of the question. The share measurable today is far lower. If a machine can produce most of what gets published, the person who typed the words stops being the interesting fact about a text.
What decides authorship, in my case, is who decided that this fact mattered more than that one, and who stands behind the sentence once it is published. Every essay I have written was made the same way: I wrote, and the tool handed me back my own accumulated experience while I did it.