I remember reading “Zen and the Art of Motorcycle Maintenance” on a hot summer vacation in Palermo. We stayed in an AirBnB across from the Palazzo Butera and, having trouble connecting with the flow of the city, I retreated into the books I brought. Pirsig’s autobiographical, philosophical novel gripped me in particular, and it prompted me to, years later, mention the book in a talk.
And then, recently, it gripped me again. I thought of LLM writing and why it felt so uncanny despite its apparent simulation of good writing. It’s hard to pin down. Yes, it deals in tropes, but so does most writing. So does my writing. What, then, made it different?
I think Pirsig gives the only answer I’ve found epistemologically satisfying so far: Quality. He defines Quality as the aether around subject and object from which both are abstracted, the unseen force that shapes the observer’s relation with the observed. It’s there in all human endeavors: art and writing of course, but also in science (the beauty of Euler’s identity eiπ + 1 = 0 is apparent but difficult to account for).
Machine-generated writing apparently lacks such Quality, and the connection between it and its readers feels thin, at least if its critics (I among them) are to be believed. But because Quality cannot be reduced to material properties, it’s hard to pin down.
There is another author who thought about this, and who makes this explicit. In Christopher Alexander’s “Timeless Way of Building” (a philosophical treatise on architecture, of all things), he calls it the “quality without a name”. Spaces either have it or don’t, and we can only relate to it by enumerating its aspects.
I suppose if I were to grade some machine-generated prose by Alexander’s later “properties of living structure”, it would fall flat in some of them. But so would most prose. So would mine. And so the quality remains nameless.
But I do believe in its existence, because I believe in life having mystery, anachronistic as that might be. Perhaps trying to pin it down would already be to misunderstand it.
///
In January of 2020, I read two books side by side that accidentally matched quite well: Foucault’s “Order of Discourse” (in the German “Die Ordnung des Diskurses”) and Baudrillard’s “Simulacra and Simulation”. My reading notes from then suggest that I quite enjoyed Foucault, and had trouble with Baudrillard’s highfalutin French philosophy prose (I called it “a little bit like poetry on crack”).
Reflecting on the two in 2026, I picked an uncannily good foundation for the years to come. Foucault reminded me to watch who controls the discourse and what their motives are, Baudrillard reminded me that simulacra are mirages and when you take them too seriously, you get sucked in and the mirage gets willed into existence.
Because this is the topic du jour, let’s talk about AI.
In Baudrillard’s pithily titled “The Gulf War Did Not Take Place”, he does not assert that bombs were not dropped and people did not die. He observes that most people seemed to think that they had a good idea of what was going on on the ground because the media supplied them with a steady stream of videos, commentary, and infographics. The simulacrum, then, was the war as constructed in the viewer’s head, which might or might not map onto reality: the narrative became its own event, with its own coherence and effects.
In the same vein, when I talk to an LLM, I map my own thinking processes onto the silicon. The output looks the same. In that sense, the LLM’s consciousness is simulated (maybe even a simulacrum, depending on your view of consciousness as a concept). We are watching the war reporter’s account of cognition, and our boots are not on the ground.
This gets worse when we realize that those who speak as though the simulacrum were the relevant reality get amplified in the discourse (back to Foucault). It serves the narrative of the day, and that of the important stakeholders: investors, politicians, the market at large. If our peers believe the simulation, its hold on us gets stronger (peer pressure is an odd thing). Note that none of the people served by this discourse need to believe it. They need a discourse in which treating the GPU as a thinker appears reasonable, inevitable, and economically valuable.
There is a “fun” little addendum here that is worth discussing: as the prevalence of the simulacrum grows, it feeds back into reality. As we read more machine-generated writing, our voices change. It also becomes a conceptual scheme through which the real is understood: human thought gets mapped onto context windows, token generation, and pattern completion.
And because this was all flowery and conceptual, let me end on a concrete note: the depth of the human experience is immeasurable. We are in the process of flattening it, because the invisible hand of the market compels us to. I, like many other people, believe myself to be rational, and rationality means to use a tool that is supremely useful. But it also means to think as I am using it, not just about the immediate but also the downstream effects, and not just for me, but for everyone else. It’s the intellectual equivalent of not littering, because the cognitive Pacific garbage patch is big enough as it is (just remember that, much as with littering, the biggest polluters are institutions, not individuals).
///
I’ve been thinking about Camus’ Myth of Sisyphus a bit. Specifically about its take on the absurd and its culmination in this poignant last sentence.
“Il faut imaginer Sisyphe heureux/One must imagine Sisyphus happy”
I always thought I misunderstood it, somehow. That it was more profound than about labour itself having value. Indeed, he’s also talking about accepting his fate is his final victory over the gods, all that jazz. But then I researched it again, and I realized the line before strips this all away again.
“The struggle itself toward the heights is enough to fill a man’s heart”
And so we are back at labour. I spend most of my programming work on things that are purposefully pointless. Martial and cringe as it may be, I always imagined myself as the soldier taking their weapon apart and putting it back together, over and over. I can do it blind. I can do it with one hand. I can do it in my mind. It’s fun, and it’s rewarding.
But I realize not all people function that way, especially in relation to their labour, and so accepting the futility of it all is painful. And that’s why Camus feels painful: “One must imagine Sisyphus happy” is just “love the process”, but with the pain and sorrow that this acceptance brings laid bare. It’s the whole shark, teeth and all, not just the fin swimming around in the distance.
And this is where, perhaps most ridiculously of all, I’m reminded of a recent interview with Peter Steinberger of OpenClaw fame. In conversation with Lex Fridman, when questioned about why OpenClaw succeeded against its competitors, he said something (accidentally?) metaphysically profound.
“It’s hard to compete against someone who’s just there to have fun“
This is the full chain from deep metaphysics to motivational poster to grindset capitalism, and to everyone’s surprise, they all say fundamentally the same thing. Once the process is the endgame, failure doesn’t lose its bark, but it loses its bite. I’ve written so many terrible programs, I’ve failed at so many of my projects. But I would not change a thing. I accidentally got sucked into a thing I truly love doing, and you cannot take that away from me.