Do AI Products Need an Avant-Garde?

A Restless Egg Series

I. The Arrival of a Train

Allow me to begin by recounting the legend surrounding the 1896 French short silent documentary film, The Arrival of a Train at La Ciotat Station. In the film, a single continuous shot captures a group of people standing about the platform of a railway station in La Ciotat, waiting for a train. As the story goes, when the film was first shown, the audience, being largely unfamiliar with the new medium of film, was so struck by the image of the train moving rapidly towards them that many began to flee backwards in terror to avoid being run over.

The story of The Arrival of a Train at La Ciotat Station is, in my mind, an emblem of the critical societal function that is played by an avant-garde creative class: to announce the arrival of a new medium, mark its terrifying difference from the status quo, and in the same moment, herald a forthcoming societal transformation where that scary medium moves from novelty to ubiquity.

The term avant-garde was first employed by the French Utopian socialist Henri de Saint-Simon in 1825. For Saint-Simon, the avant-garde designated artists, scientists, and industrialists as the new and elite leadership class of an emerging social order. In his own words:

It is we artists who will serve you as avant-garde…the power of the arts is in fact most immediate and most rapid: when we wish to spread new ideas among men, we inscribe them on marble or on canvas...What a magnificent destiny for the arts is that of exercising a positive power over society, a true priestly function, and of marching forcefully in the van of all the intellectual faculties...! (Opinions littéraires, philosophiques et industrielles (Paris, 1825))

Fifteen years later, Cesar Daly, editor of the Revue generate de l'architecture argued, in echo of Saint-Simon, that his journal should "fulfill an active mission of scouting the path of the future." In other words, for Daly, what the avant-gardist uniquely sees is the outline of a future, which is currently invisible or otherwise inexpressible from the standpoint of the present. While the avant-gardist speaks to present conditions, they do so from the perspective of this scouted future.

In The Politics of Vision, the art historian Linda Nochlin argues that a sense of alienation, psychic, social, or ontological is also fundamental to a conception of the avant-garde. Not alienation in the sense of the artist-as-outcast per se, because as Nochlin notes this “was of course not a novelty by the middle of the nineteenth century,” but rather, alienation in the sense of the avant-gardist “in conflict with himself, internally or distanced from his true social situation externally, as were…Flaubert, Baudelaire, and Manet.”

These are the three senses of the avant-garde that I wish to invoke in this piece with its title “Do AI Products Need an Avant-Garde?”. First, per Saint-Simon, that this class spreads new ideas amongst people by inscribing them in works that others may encounter. Second, that the vanguardism of the avant-garde refers primarily to a temporal vanguardism; Daly’s “scouting the path of the future.” And third, per Nochlin, that this genre of creator registers a felt distance between the external social situation, whatever it may be, and their internal apprehension of what is and what is possible.

When I have previously written about the artist-founder as the avatar of an emerging technical avant-garde, these three elements are what I have tried to imply. The artist-founder is part of an advance guard, one who depicts use-cases for emerging technologies which will predominate in times to come, but which we also cannot yet fully imagine without help or provocation, our being too-fully ensconced in the habits of everyday life.

The artist-founder traces out a path backwards, from their scouted future towards a Saint-Simonian inscription of that vision “on marble or on canvas” (forms which in our time may take the form of an installation or any other kind of art object), and then back to the future again as a window towards novel interaction patterns others can begin to imagine.

In other words, the work of art as material object exists to make the kernel of what the avant-gardist perceives communicable. It gives their vision form around which thought can gather and communication begin.

In a different sense, which there is neither time nor space to deeply investigate here, the avant-gardist also works to bring about certain foreseen futures by making them legible for the first time. Although such visions incarnate may be met with initial skepticism or else outright dismissal, they give early credence to ideas which are struggling to be born.

II. The Ubiquitous Infrastructure of Intelligence

During his Keynote at COMPUTEX 2025, Nvidia Founder and CEO, Jensen Huang argued that his company should no longer be understood as a semiconductor company that designs the complex GPU architectures. Instead, Jensen argued, Nvidia is today best understood as an infrastructure company.

What does this mean? According to Jensen, a new global technical infrastructure is being built atop the prior two global technical infrastructures of electricity and information (e.g. the internet). Just as electricity and information are everywhere now, so too will be artificial intelligence. In his keynote, Jensen further prophesies a (not so distant) future wherein public companies will report token production, in addition to figures like EBITDA or user growth, and they will be valued based on how many tokens (how much intelligence) they manufacture per quarter.

And so just as a social infrastructure had to be built around electricity and the internet, so too will one have to be constructed in light of the becoming-infrastructural ubiquity of artificial intelligence.

Imagine it like this: when you go to an historic city, say Venice, you look around and observe the beauty of the canals and the palazzo’s and so on and so forth, and you know, without fully knowing, that each one of the things you see has a much deeper and longer history that is inscribed within it, a history you can touch, but which you cannot innately perceive, hence the tourism business.

With artificial intelligence as infrastructure, we are taking the intelligence that a tour guide has about a place and layering it atop everything, the way older buildings get renovated to add electrical wiring or later, Wi-Fi.

What we must soon grow accustomed to is the omnipresent layering and the accessibility of intelligence at every scale and situation of life in the built environment.

The question becomes, how do we tap into this intelligence? When? Why? What does it afford us the opportunity to do? What kind of experience does it deliver?

Ubiquitous interfacing with intelligence is the problem set raised by the becoming-infrastructure of AI.

The questions of how, when, where and why one interacts with ubiquitous intelligence are, right now at least, extremely open. Generally speaking, these questions are mostly addressed by those working on the application layer of AI or in a different sense, those working on the product side more generally.

However, it is worth pointing out that these questions are as fundamental to the AI future as foundation research is, because what is at stake in them exceeds far beyond the limited stakes of what a single user receives from an isolated human-AI interaction. Rather, these questions imply the broader field of research concerning the templates we will design for how our species grows into and learns to live with the ubiquity of artificial intelligence as it threads its way through the spaces that we already inhabit and are accustomed to interacting with in certain ways.

It is almost not worth stating that we are approaching a crossroads in our co-evolution of technology, as AI develops into whatever it will be. Consequently, the need for an avant-garde class is rising, as it was at the time of the term’s being coined, in the wake of the first industrial revolution and its associated technologies and then only a few years later The Revolutions of 1848.

On that point, it’s worth noting that the time is not always right for the avant-garde. Sometimes one knows the answer and just needs to build it. E-commerce needed to make the shopping malls of meatspace accessible online. It did not necessitate a fundamental re-imagining of the shopper, the seller, commerce or some other bit of received wisdom that an avant-gardist would lodge onto.

The reason for my writing about the artist-founder and the new technical avant-garde is that there is so clearly a need for it when it comes to AI products. Rather than slouching towards a future of ubiquitous intelligence interaction that is designed from and germinated within the incentive structure of the present, the future of human-AI interaction, which is to say a future that is worthy of the name in its being genuinely different from the present rather than a repetition of the same only in a different moment of time, will require an advance guard to go forth and scout other possibilities that one may prefer to inhabit. The technical avant-gardist must return with that knowledge and build out exemplars of it for the rest of us to see.

Given the scale and speed of society today, the communicative act of avant-garde art will require this latest incarnation to make their inscriptions not on marble or canvas, but in media forms which are most able to spread, attain users and set paths for technical use to evolve into and towards.

My co-founder, Sylvan Rackham, loves to quote René Daumal’s Mount Analogue: “the door to the invisible must be visible.” Truly, this is the function of the new technical avant-garde, to create a visible door that opens out onto vistas of unfamiliar technologies and use cases that are currently invisible or occluded, but which promise exciting and as yet unforeseen ways of interacting with ubiquitous infrastructural AI.

III. The Status Quo of Human-AI Interaction

What are the ways of interacting with AI that exist now? This is the form factor question. The chatbot reigns supreme, obviously enough. Within its arena are AI agents and increasingly the more ambitious agentic AI.

Screen-based interfaces are today derided in favor of the next AI mega-trend: physicalized AI. Tesla is a case in point: the company’s astronomical valuation has everything to do with AI, the Robotaxi and the Optimus personal robot and has completely detached from its sales of electric vehicles (which have plummeted).

A new genre of devices is also beginning to perk up, starting from precursors such as the Meta Quest and Apple Vision Pro headsets and extending to the newer Meta Ray-Ban and Meta Oakley partnerships for smart glasses. At their latest I/O conference, Google also announced a new line of AI smart eyewear in partnership with Gentle Monster.

Astride these are a new set of screenless or semi-screenless wearables, “always on’s” I’ve heard them called. They include the Limitless, Omi and Friend necklaces, the Rabbit R1 “intern in your pocket,” the Plaud voice recorder, Modem’s Dream Recorder and the fabled Ive-Altman collab device that is yet to come.

Despite these innovations an avant-garde for AI is yet to emerge. At least on the product side, for we can safely say that the transformer model was an avant-garde invention. At the most innovative AI summits and expos one gets the point: variations-on-the-same slogan campaigns about AI agents will streamline your businesses’ workflows, each plastered on posterboard the same slightly but slightly different shade of purple.

The issue is not that these products are not innovative; they are extremely innovative one and all, but each along the horizon of the familiar: I do the same things I’ve always done, but now an AI records my words for me and then adjusts my calendar. That’s great and super useful! I want one! But this is a categorically different experience than watching The Arrival of a Train at La Ciotat Station, feeling like you are about to be killed and then realizing that not only are you safe, but you are also witnessing the birth of a new technological medium.

The feeling elicited by avant-garde AI products is not likely to be “oh that’s neat, I could really use that thing.” I have to imagine that these products will make their first users shout things like “Holy Priceless Collection of Etruscan Snoods, Batman!” and they will feel much more like The Arrival of a Train at La Ciotat Station than they will someone else filling out your to do list. That feeling is currently missing from AI products, and it’s why the question of what AI is really for remains so clearly unanswered at the very foundational level.

IV. A Bildungsroman of AI: The What and How of Aesthetic Surplus

An avant-garde for AI products does not treat such products as ends in themselves, but develops them as part of a process of prefiguring a new kind of AI user, AI audience, human and human-machine relationship. The avant-gardist sees a species of human-AI relations that does not currently exist and which will not be born out of the linear enhancement of existing capabilities.

I quite like Thomas Moynihan, the historian of ideas’ framing of this process as a kind of bildungsroman, a coming of age tale for the human-as-infrastructural-AI-user. We must learn how to acquit ourselves to the infrastructure of AI. It is not knowledge we will come upon all at once, but something that we will grow into.

Prior to electrification, factories relied on steam or waterwheels to power enormous rotating line shafts. These shafts were suspended over the factory and they powered belts that ran down from the central shaft to the factories’ individual machines. At first, electricity was not seen as some great advancement for the factory, but a more costly way to rotate the line shaft. Years passed before the avant-garde realization that electricity was not just an alternative to steam or water for powering the same belt-and-shaft style factory architecture. But when it came, what this realization made possible was a fundamental re-design of the factory, eschewing the central line for modular, electrically powered machines.

This story comes to me from Sylvan and he often pairs it with a distinction he drew (in a past life) between “what” and “how” technologies. How technologies refer to technified ways of doing things that you did before. You know how to “like” things and “poke” people, and Facebook lets you do those things from far away. On the other hand, the guitar is a what technology, because it unlocks a new way of doing that must be learned and accounted for.

In recent months, Haley Albert and Sylvan have begun describing “what technologies,” (amongst other things) as containing a kind of “aesthetic surplus.” What I understand this to mean (and I will not claim to understand their use of the term fully) is that something is more generous in its being than can be accounted for in its use. The guitar is in a relationship of surplus to the first guitarist, inviting experimentation and innovation, hence the perdurance of it as form. The Arrival of a Train at La Ciotat Station is in a relationship of surplus to early audiences given their unfamiliarity with the medium of cinema.

When something is in a relationship of aesthetic surplus compared to established modes of use, one is invited to experiment with new ways of use, partly in order to adapt oneself to the yet-undiscovered opportunities the given thing suggests. Doug Stark talks about this solicitation to experiment and the compensatory adaptive work it activates as the fundamental function of play, and, for him, this is why so much of the early human-ChatGPT interactions tend to be so silly and fun: aesthetic surplus solicits play as adaptation and experimentation, to both make sense of the new and follow through on its promises.

The opportunity of AI is partly due to the fact that it has sent an existential shockwave throughout the tech community and the world at large. From your toothbrush to your toolshed, everything is computer now.

And we still aren’t quite sure what AI is really for yet! We know it offers massive productivity boosts, but we sense that it also affords us opportunities that won’t fit it into the concepts we had prior to its arrival.

The point of calling for an avant-garde for AI products is to develop these concepts in products as communicable forms that can convey ideas about what and how human-AI interaction should be if and when intelligence does indeed become the next global infrastructure substrate.

The good news is that because nobody knows what AI is for, there is a great willingness to undertake new experiments. There is now a unique openness towards the future as something that will not grow out of the present that exists, but which is yet to be understood.

If Stark is right about play-as-adaptation to drastic change and a catalyst for further experimentation, then bildungsroman for the human-interfacing-with-infrastructural-AI opens with a set of questions that are delightfully simple and playful: where do you want to go, what do you want to do, what kinds of interactions do you want to have, which are not possible now, but imaginable with the layering of intelligence atop physical reality?

The avant-gardist will not answer these questions from the point of view of the present, but will bring back a taste of the future, building it out into a product you or I can see or touch. It should feel different. It will.

If you encounter an artist-founder’s product and say to yourself “what the hell is that?” Remember that when the first train tracks were being laid down towards La Ciotat, many a causal passerby may have walked by and thought to themselves the exact same thing.