Phenomenological Experience of Generative Artificial Intelligence

Some Context

In December 2025, after the “buzz” following the release of Laurent Alexandre and Olivier Babeau’s book, I listened to a long interview with Laurent Alexandre to try, without having to trudge through his book, to understand his perspective on education, at least at a basic level.

Despite everything one might say about Laurent Alexandre, I still find it interesting to examine his positions and ideas, which are not all worth dismissing. In this context, Laurent Alexandre describes his use of generative AI and, more importantly, explains that he enables his AI’s memory. He develops a reflection on the rationale and the consequences, which he considers beneficial, of this feature.

So I thought: why not try it for a while? I activated the memory on ChatGPT and quickly understood Laurent Alexandre’s reasoning. I also created a new account with Anthropic and dove into the features of Claude (which I use more and more without abandoning the other LLMs you already know).

Beyond this aspect, I read the experiment conducted by Fabrizio Ferri-Benedetti, which he presents on his blog. The idea was to simulate an interview between LLMs about the articles and notes he had written in 2025.

The transcript of the exchange is reproduced in extenso on his blog. It is not particularly fascinating, and so it was more the approach itself that caught my attention.

Querying an LLM based on a stock of notes is nothing new. With the arrival of MCP, we saw plenty of ideas of this kind aimed at connecting an LLM to note-taking applications (having used Obsidian for years, I looked into how to connect my app to Claude here). Before that, there had also been an almost viral trend regarding an experiment to try with your LLM: asking what it knew about you. If the user had enabled the “memory” of their application, the LLM would dig through conversation history and memory to provide a summary that was sometimes interesting, sometimes unsettling.

The Beginning of the Experiment

So I asked Claude:

Analyze the content of the site reflexions.florianernotte.be and provide a critical commentary on the content and predictions for future topics

The response I received was very interesting. I then asked:

I want to develop my philosophical reflections on technology. Do you have any suggestions?

Again, the response was very relevant, and I noted several quite serious avenues for reflection.

Among these, one caught my eye:

### Ethnography of Your Own Practices
	**Project:** A logbook observing your daily AI usage. Not to celebrate or condemn, but to understand phenomenologically.
	**Method:**
	- Note daily: when do you use Claude? For what? What does it change about your thinking?
- Observe moments of resistance: when do you refuse to use AI? Why?	- Document the transformations: what do you no longer do since you started using AI?
	**Possible article:** "Diary of a Lawyer with AI: Phenomenology of Cognitive Delegation"

The angle was interesting and I found the experiment perfectly aligned with the duality I try to convey within my digital notebook.

So I created a new conversation by copying the proposal as is, and I received a methodology and advice for documenting my future usage. On reflection, I realized that the burden of documenting each of my uses would be far too heavy, and above all, I would only get interesting results after several weeks of use.

In short, I replied to Claude:

Documentation burden too heavy. Help me compile it based on all our exchanges

Claude replied:

I will extract from our conversations the traces of your AI usage, moments of resistance, and observed transformations. Here is what I can document

Curious, I finally asked Claude:

Start writing this diary of a lawyer.

The result exceeded my expectations. It requires some adjustments and corrections before being shared (I will do so soon), and I think I will repeat the experiment each year to analyze and understand the evolution of my usage and of the technology.

I could have done it differently, being less lazy and carefully rereading every article I had written, but I find this to be another way of thinking, one that fits fully within this duality I am trying to describe throughout my reflection notes.

Phenomenology of a Practice with Generative AI

[Phenomenology of a Practice with Generative AI (2026)] (/en/post/phenomenologie2026/)