How I Write the Articles Published Here

I want to be clear about how I write the texts published on Le secrétaire de Fernand.

I do not write my articles in the classic way, sitting at a keyboard and drafting a first version from beginning to end. My method is different. It goes through voice, through conversation, through iteration, and yes, through the use of tools like ChatGPT and Claude.

I would rather explain this openly, because to me, transparency matters just as much as the text itself.

My starting point: speaking before writing#

When a subject stays with me, I often begin by talking about it out loud on my phone.

I record voice messages in ChatGPT or Claude. I develop the idea orally, the way I would in a conversation: I explore an intuition, clarify a nuance, go back, try one angle, abandon another. I am not trying, at that stage, to produce a literary form. I am first trying to bring the thought out.

This is an important step for me, because speaking lets me go further than spontaneous writing. When I try to write directly, I sometimes simplify too quickly. I cut, I summarize, I flatten. Speaking, on the other hand, lets me unfold an idea more fully, with its hesitations, its turns, its clarifications, and sometimes even its contradictions. And very often, that is closer to what I really think.

Then: using AI as a critical interlocutor#

Once that raw material is there, I do not simply ask the tool to “write an article.”

Most of the time, I first ask something like: what do you think of this?

At that point, what I want is not flattery or a polished reformulation. I want criticism. I want to be told:

• this part of your argument needs more nuance;
• here, you may be going in the wrong direction;
• here, you are missing an important angle;
• here, a key distinction is missing;
• here, this point should be developed or clarified.

In other words, I use ChatGPT or Claude as interlocutors capable of giving me objections, tensions, blind spots, or simply leads for deeper reflection.

That helps me work on the substance before the form. It helps me refine what I want to say. It helps me avoid, as much as possible, formulations that come too quickly or ideas that have not been thought through enough.

A text that emerges through several back-and-forth exchanges#

The process is rarely linear.

I often go back to voice after a first response. I clarify. I correct. I reformulate my intention. I expand a point I had only sketched out. Then I ask for critical feedback again. Then I go back once more.

So there are several back-and-forth exchanges between my speech, the transcription, the critique produced by the tool, and my own return to the reflection.

This is not a button I press to “generate an article.” It is more like a working conversation, sometimes a long one, that allows me to clarify my thinking before giving it a publishable form.

Then comes the writing#

At a certain point, when the substance feels mature enough, I ask ChatGPT or Claude to write an article based on everything that has been developed.

That first text is not the end of the process. It is a structured draft. A shaped version. A proposal.

Then I read it carefully.

I check that the article truly respects the thread of the conversation, the logic of the reasoning, the chosen angle, and above all the spirit of what I wanted to say. I may then change the structure, move passages around, remove sentences, rewrite certain paragraphs, simplify, clarify, or, on the contrary, give the whole thing a little more breadth.

It also happens that I let the text rest. Some articles mature before publication. I come back to them later with fresher eyes.

Why I prefer this method#

The main reason is simple: speaking allows me to stay closer to my own thinking.

When I speak, I often go faster and further than when I write directly. I produce more material, more nuance, more links between ideas. I censor myself less early on. I let the real movement of the reflection show up more fully.

That gives the tool more material to work with, but above all material that is closer to what I am genuinely trying to express.

In a way, this method paradoxically allows me to be more present in the text, not less. Not because every sentence came straight from my hand, but because the intellectual material, the initial intuition, the nuances, the choices, and the final validation genuinely come from me.

What this means concretely for you as a reader#

The articles published here are therefore neither texts “written on their own by AI,” nor texts written in a fully traditional way.

They come out of a hybrid process.

I provide the subject, the intuition, the point of view, the core material, the nuances, the corrections, and the choices. AI comes in as a tool for transcription, contradiction, exploration, structuring, and assisted writing.

I think it matters to say this clearly, so that the reader knows what they are reading.

I am not trying to maintain the illusion of purely solitary writing if that is not the reality. But I do not want to understate the work of reflection, selection, judgment, and revision that comes before each publication either.

What I take responsibility for#

I fully stand by this way of working.

First, because it genuinely helps me think better and write better. Then because it corresponds to a very real evolution in our tools of expression. And finally because, to me, what matters is not preserving a romantic image of writing, but being honest about how a text is produced.

What matters to me, in the end, is that the published article remains faithful to what I want to say, that it has been worked on seriously, and that I can sign it without reservation.

If a text is published here, it is because, after this process, I consider it mine in its intention, in its direction, and in its responsibility.

In summary#

My writing method today relies on four steps:

1. I speak about the subject out loud;
2. I use ChatGPT or Claude to get criticism and deepen the reflection;
3. I then ask them to draft a text based on that material;
4. I reread, revise, restructure, sometimes let it mature, and then publish.

I would rather make this method explicit than leave it in the shadows.

Because writing is changing. Because tools are changing. And because trust deserves to be built in clarity.