Beyond the Dead Internet: How AI Writing Becomes Philosophical Dialogue
This article was co-written with Claude AI for translation, reformulation, and structure. It is the fruit of a real dialogue and reflections, not an automated article created from standardized content for filler purposes. I wish everyone a good read.
Context: The Risk of Global Absurdity on a Path to Dead Internet Theory
Dead Internet theory basically says the Internet will be filled with auto-generated content so much that people won’t come to the Internet anymore. One can then wonder, as an amateur blog writer, what’s the meaning of continuing all of this?
Balancing Authenticity with Efficiency
LLMs definitely bring efficiency in writing (I use Claude Code with visualization in Obsidian). I will not go through all the things it can do; I want to take more time to write about how, as a human, I can thrive with this capability at hand while not losing myself.
I do not exactly believe in Dead Internet Theory as such. I believe in the fiercest competition for authenticity and substance because the form can be polished by AI. I believe that originality and depth of ideas will be a competitive advantage when everyone will have access to easy-to-write, easy-to-publish shallow and soulless clickbait factories. At the age of brainrot emergence (2025), I believe the challenge will be finding subtlety, finesse, mind-boggling and mind-opening content, stopping the escalation of dumbness demonstration or emotional shock engineering (hopefully).
Moreover, I don’t believe people will enjoy reading, in the long term, mixed-up gibberish resulting from poor intention (meaning from prompts without authenticity or very short and information-poor prompts, for example), even if well formatted. Readers can sense when content lacks genuine thought behind it.
Therefore, the strong point is to find balance between authenticity and efficiency, not being diluted in the most probable answer, but making the most from the incredible LLMs machine. That’s why I’m sharing here feedback on my writing process, in hope for critical feedback, reuse, and maybe self-understanding through the writing process of this article (written meta-cognition).
Also one point: Authentic does not mean utterly unique. I fairly acknowledge that people on this earth may have the same ideas as I do (most probably more brilliant and insightful). Still, authenticity is not only from your point of view as a reader, but also authenticity to myself. I want to be true to myself through this writing and expression process, and this research can make me create more original content.
First, I Ask Myself Why I Write
What’s the deep motivation behind writing? It could be expressing myself, defending an idea, sharing what I consider important and useful to the world, preparing future context for future ideas, writing for the pleasure of writing…
Knowing (or at least having a short glance) helps activate emergence.
Second, I Write Big Ideas in Emergence
Ideas in disordered manners as they occur.
I see ideas like “neural activations, like a network of light bulbs attached to neurons getting lit together for a short moment”: It’s not linear, and it’s much more natural to have it disordered. This mirrors what we call hypergraph knowledge—where multiple concepts can connect simultaneously to multiple other concepts, unlike traditional linear structures where each idea links to only the next one in sequence. Think of it as the difference between a chain (linear) and a web (hypergraph) where any point can connect to any other point directly.
Third, Choosing an Outline
Difficult part because choosing an outline, a plan, a linear path turns multidimensional hypergraph information into ordered content, which is easier to read (because prioritization of content to explore is quite obvious) but loses the connectivity in non-adjacent points (for example, if there is a link between sections 3 and 5, between sections 2, 5, 7, and between 1 and 9, it is not represented).
This dimensional reduction is like projecting a vector in a lesser-dimension hyperplane; we lose information. In more Platonic terms, imagine you have a three-dimensional object, and you project its shadow onto a wall—the line that forms the shadow is your outline. You inevitably lose the depth, the volume, the full richness of the original object.
Yet there is a part of choice, a part of existentialism in this step. It feels like I am not only a recorder, but a decision maker. This is where I feel I really count as an individual; otherwise, I just consider my ideas as a natural occurrence of connection (like in Foundation: it has to happen, no matter who does that).
I can choose the outline partially (like for this article, and I also ask AI for propositions of outline).
Fourth, A Mix of Writing Myself and AI Reformulation
I usually don’t ask AI for ideas, mostly for reformulating my flawed “Frenchy” English into elegant paragraphs. Then I bounce back on Obsidian and read, add some more, then come back to Claude Code.
My requests to Claude Code alternate between:
- Writing instructions “Please rewrite (a specific section)…”
- Questions for reference “Find references about study about (topic)…”
- Targeted feedback “What’s your opinion on (a property) of (a specific section)?”
- Questions of curiosity “What is the state of the art and what are the last discoveries about…”
I think of writing as the fruit of a dialog, a dialog between me and the training data through Claude, and the article being the minutes of our meeting.
Fifth, Last Clean-Up
After that, I ask Claude to clean up writing, references, and various titles to make them consistent, always by asking “propose me” instead of “change”. The element of choice, however cumbersome, is essential for me to pull out what I want (usually the internal benefits) from this writing exercise.
A Brief Digression: Why Mix Process with Philosophy?
You might wonder why I’ve woven together practical steps (motivation, emergence, outline, dialog, cleanup) with conceptual explorations (hypergraphs, existentialism, dimensional reduction). This isn’t accidental—it reflects how my mind actually works during AI-assisted writing.
When I write, I’m simultaneously operating on multiple levels: the immediate task (getting words on paper), the structural challenge (organizing ideas), and the deeper questions (what does this choice mean? how do ideas really connect?). The philosophical concepts aren’t decorative additions; they’re the cognitive frameworks that emerge naturally when you pay attention to your own thinking process.
This is perhaps what makes AI-augmented writing different from traditional writing advice. We’re not just optimizing a human process—we’re discovering what happens when human cognition meets artificial intelligence. The concepts bubble up because the experience itself is conceptually rich.
The 3 Lessons for AI-Augmented Writing
1. Climb Up the Scale
This is what I would recommend for starting AI-assisted blog writers:
- Write down your ideas in a disordered note-taking tool to centralize your data
- Write very short articles without AI
- Start practicing writing then AI review
- Try the dialog process
2. Reuse the Context
- Use your notes to let AI figure out your style of “coucher les idées” (laying down ideas), your style of writing, your values. It will make suggestions consistent with the rest
- Drafts and disordered notes are gold for that
3. Ask Yourself About
- What’s (deep) inside your life experience that is not already an easy-to-find answer (at least to you)
- Who (if any) are you writing for
- What tools fit you most for dialog, notes, etc.
Conclusion: Opening Up a Path for Individual-Training Data Emergence
It would be so easy to talk again about AI and human combination. I wrote a lot about this concept and my site is full of that. What I would like to push forward is rather another angle: my data + training data emergence.
Interestingly, this AI writing exercise naturally leads me to philosophical questioning. Somewhere, philosophy is a practice of concepts and ideas, quite close to deliberate metacognition liberated from the lone individual—which is also the work of writing. The reference to external concepts, the verification of coherent combinations, is also part of the writing work (at least mine).
If I look at myself, like a beginning philosopher who ignores himself and doesn’t know the extent of the concepts he could question or reuse, I write by making the same choices on a smaller scale. If I look at the “Claude + me” system, then we’re having a philosophical dialogue between an ignorant person (me) and his questions, and a knowledgeable one (Claude) and his accumulated erudition. This is how we obtain a unique mix.
The AI writing process creates a combination of my data (as it appears in my mind as I write) and the training data of the LLMs, leading to a new mix and terrain to explore. I hope you can enjoy that as much as I do.