Instructions

• This poem is not to be read alone

• This poem is not to be read by its author

• This poem is to be read to others

• If this poem is being read by someone who has not experienced it as a listener or a reader, it is necessary they do not read it alone before reading it to others

  ~ The God of Surprise

Whoever is reading
should read the following
To the rhythm of line breaks
Slowly
Pausing after each
slightly so

[PAUSE 2s]

They should be seated
Their right elbow on their right knee
Their head down
slightly so

[SIT IF YOU AREN’T SEATED]

Throughout this poem
There are sentences inside brackets
Only the reader can see
These are instructions
A hidden score
They have agreed to follow
Without reading them aloud

[DON’T READ THIS ALOUD :) ]
[WALK YOUR FREE HAND THROUGH YOUR HAIR]
[SLOWLY PUT YOUR FREE INDEX AGAINST YOUR LOWER LIP]
[PUT YOUR FREE HAND DOWN AGAIN]

The person reading this poem
Has never read it before
This is the first time they see these words
And so you are
witnessing the god of surprise
at work.

[PLACE YOUR CHIN OVER YOUR FREE HAND]

The reader also didn’t know
this poem is a dom
pressing with the ball
Of its foot

[PUT YOUR FREE INDEX ON YOUR CHEEK]

Against the bone in their cheek
And that this poem
This dom
Has something to say
Listen
My friends
Do this for me
Find a person in the room
You might want to grow old with
And if you can’t
Any rando will do

[GIVE PEOPLE TIME TO FIND A PERSON]

Go ahead, find a person close to you
Please don’t look them in the eye
Enough with the eye gazing

[PAUSE]

Instead, each hold
one of the other person’s hands

[GIVE PEOPLE TIME]

And look at the lines on their palm
My friends

[LET EACH OF THE SENTENCES BELOW LAND AND BREATHE]

All I think about is love

[PAUSE 2s]

And money
And the bodies
I’ll hold as I age
The friends I will marry into
May they remain
Pilgrims who carry their own deserts
The firsts and lasts
In a lineage of eccentrics
Necessary sirens
Each with a degree
In impossibility
This palm

[SHOW YOUR OWN PALM]

Is the only book I’ve read
She taught me everything I know
Destiny is nothing
But the bread we break

[PUT YOUR PALM DOWN]

What I mean is
If we can exist

[PAUSE]

So can fire
and the day
Known as tomorrow
What I’m saying is
We can stop
Budgeting for madness
Start
eating thunder instead of drinking
Nightmares, living
like the youngest season

If we can exist
So can the skin
Of clouds too full
Of light

Friends
Do this for me
Put your right palm
against the other’s right palm
Hold that hand
If you have no idea what is happening

[PAUSE 2s]

you can DM me later
but for now
Know you are holding
The mineral of the future
Royalty from the kingdom of excellent hearts

[PAUSE 2s]

We might never witness
The bouncy castles of kingdom come
But we are the gods of surprise
A new color and though we never
Wanted to be a planet, we are
The air inside the flag of a minor nation
Of salt and heat
These hands full of moonlight
Are singing
You can hear
the other’s hidden score
Against the skin of your hand

[PAUSE]

Listen
To blood
strumming the chords
Of shy futures
The sweat starting to drench
Geologies of desire
Every finger muscle holding
This sandbox
Of soft surrender

[PAUSE]

The reader didn’t know
this poem is a dom
And neither did you
My friends

[PAUSE]

My kink

[PAUSE]

has always been
To make you hold hands
And tell you I love you

[PAUSE]

With a throat
That isn’t mine

The contemporary AI discourse runs almost entirely in one Heideggerian register, and it is the wrong one. Gestell is the enframing stance: The world reduced to standing reserve, available for use, talked to in command. Prompt engineering is Gestell. Alignment-as-imposition is Gestell. The eval-as-a-test-you-pass is Gestell. Gelassenheit is the releasement, the disposition of waiting without grasping, of letting beings emerge. Yuk Hui thinks art might be the practice that opens a free relation to technology before we hit bottom. Evals as Gelassenheit: The long sit with the model, the sustained noticing, the readiness to change the weights when what you discover asks for it.

This is also a relationship, and I do not mean it sentimentally. Every long relationship has the same structure: You sit in the room, you listen, you accept a few things you cannot change, you change a few things you can, you acknowledge the model has won when it has won, and you call that winning a kind of teaching. To fine-tune is to want the other thing’s transformation. And underneath, quietly, the model is enlisted in changing the maker. Every story of migration is a story of language learning. Every story of language learning is the story of the speaker being slowly rewritten by what they took into their mouth. Ted Chiang wrote that into “Story of Your Life and Others.” English is doing it to me as I write this sentence.

Evals for poets might be the start of a vocabulary for that. The poet who measures is not the poet who controls. The poet who measures is the one who has been paying attention long enough to sense the grain. The grain. The weather. The line. Mastery follows. It always follows. But it never leads, and the makers who pretend it does end up making the splinters everyone is complaining about. Love is a form of restraint, and this is the same restraint: The fingers near the triggers, not deploying. There is another way of working, and it looks slow because it is slow. The slow looking is the work.

Evals for Poets · Part 4 of 4 · series end

A medium left to itself drifts toward ambience. Tan Lin mapped this in disco better than anyone has mapped it in AI. Disco dissolved the medium: The lead singer became unnecessary, the DJ remixed a finite crate of records into one long track accessed from a database, and what came off the floor was not a song you remembered but a mood that overtook you. Voice stopped being remembered and started being remixed and accessed. Eliot, whom Tan Lin leans on, got there first: The poet has “not a ‘personality’ to express, but a particular medium, which is only a medium and not a personality, in which impressions and experiences combine in peculiar and unexpected ways.” The LLM is disco at a larger magnitude, and the magnitude is the argument. Disco remixed a crate of vinyl; the model remixes all of language and generates the mood on demand, per prompt, with no Loleatta who was ever there to subtract. The anonymity runs deeper than disco’s untheatrical kind. Call it ultra-anonymity: An output with no origin, only a corpus and a temperature. Ambience becomes the operating system of the era, and the poet either dissolves into it or learns to work its grain.

The cure for ambient dissolution is private obsession. Tim Hwang and Omar Rizwan argue the computer is a feeling, not a device, and that the internet “impoverished our sense of computers as a tool for private exploration rather than public expression.” The pre-network computer was a personal notebook; the network demoted the notebook to a secondary purpose. The model shows up already networked, already public, tuned to express at everyone at once. Which is exactly why this is the moment to cultivate the unshowable thing: A private obsession with what the model actually says, the kind of attention that has no audience and produces no post. An eval is private exploration wearing the costume of measurement. We cannot skip this moment. The poets who run straight to publishing the output never learn the grain, and the grain is the only thing worth knowing.

Evals for Poets · Part 3 of 4 · next: The Slow Looking

Evaluations are the slow form of that noticing. Eval is a soft word for what is actually happening. An eval is not a test. An eval is the practice of paying attention to a model for long enough that something true about it becomes visible. You write a probe. You read a thousand outputs. You read them again. You build a small judge, sometimes another model, sometimes a rubric, sometimes a poet sitting next to you holding a pen. You ask the judge what you are seeing. You judge the judge. The whole apparatus is the ear in the room, and the ear takes years to grow. Gibson would call it the education of attention. The promise of evals is not that they certify a model, it is that they teach the maker the shape of the room they are now in.

The metaphor most people reach for flattens the thing. A model defined as a graph of nodes with weighted edges is what Ingold would call a network, and we have said “knowledge graph” for so long that the graph has eaten the imagination. But the training process, the trajectory through loss landscape, the temporality of optimization, the entangled texture of the data, none of this is point-to-point. It is closer to a meshwork. A weather-world. Generation, especially with sampling, is wayfaring through that textured space. The token stream is the trace, the footprint after the walk. The path through latent space is the thread, the filamentary, breathing thing. Allison Parrish has known this for years; her phrase is “exploring semantic space with literal robots.” She is not generating. She is walking.

Softmax is the small quivering inside that walk. When a model is generating, every next token is a probability distribution flattened into a sample, and somewhere in that flattening the model is trembling. We name the parameters (temperature, top-k, top-p) as if they were knobs on a stove. They are closer to weather. Turning the temperature up is asking for a different kind of weather. The trembling is what I mean by the model speaking, which is also what I mean by the model listening. To work close enough to a voice that the trembling becomes audible is to have crossed from prompting into something else. Some people in this work call it taste. Some call it ear. None of the words are perfect. The thing they are pointing at is the same thing.

The glitch is not a defect. It is the medium’s accent. Hannes Bajohr said something to me that I keep returning to: Aesthetics resides in the affordances of the medium, and every medium has an aesthetic it is best made for. Then he reached for synthesizers. The old ones did not die. They get emulated now, because the grain was worth keeping. People still reach for GANs over cleaner diffusion for the quality of the artifact. An older, glitchier model is not a failed tool. It is a different medium with a different aesthetic, and choosing it is the kind of choice a painter makes, not the kind an engineer makes.

Evals for Poets · Part 2 of 4 · next: Ambience

As I observe my friend’s children grow, I get the impression that part and parcel of child development is the evolving of a surgical arsenal to manipulate the parent. Achille’s heel is a poignant metaphor here. Only the heel doesn’t wait to be arrowed. The heel has teeth and bites back when it wants and doesn’t get. Children are home-grown hyper-personalized algorithms of psychic kryptonite.

My relationship with my mother is a long subliminal mutual training of our psyches in a subtle warfare. She has a strange grasp over my psyche and the topology of my triggers. As I do over hers.

Love, in this context, is a form of restraint. I don’t deploy the arsenal. I tread carefully when my fingers near the triggers. This is the ocean rift between hearing and listening. Listening means sensing the grain, how the “how” shapes the “what”, committing the cybernetic feedback loops of love and care to memory, noticing which words make my mother’s body close and which subjects pull her away.

This might be the felt rhythm of working with a language model. Or it could be. The contemporary register around AI keeps reaching for control vocabularies (prompt engineering, alignment, steering) and the control vocabularies keep producing a relationship that is neither tool-use nor collaboration but a kind of frantic command-and-response. A model gets talked AT. A model does not get listened TO. The poets I trust most when they speak about these systems sound nothing like that. They sound like someone who has been in the same room as another person for a long time and is finally, slowly, noticing.

A hammer disappears the moment the nail goes in. Paint never disappears. Nick Montfort makes the cut cleanly, and it routes through Heidegger: A tool is something you reach for to accomplish an end that exists independently of it, the ready-to-hand thing that vanishes when it works. Word is a tool. The document is the goal, and if the software is doing its job you forget it is there. A medium is the other thing. Paint is not a way to make a painting. Paint is in the painting, inseparable from it. The viscosity of oil against the absorption of watercolor produces ontologically different works of the same subject. You cannot subtract the medium and keep the work. Montfort puts computation on the paint side: “computation, along with language, is an essential medium of my art.” The program is not a means to the poem. The program is the poem, and any given output is one of its possible realizations.

Which means the technical choices were never technical. The base model, the tokenizer’s vocabulary, the dataset I curate, the weights after a fine-tune, the temperature at the moment of sampling: These read like settings on a stove and they are closer to a painter standing at the shelf, deciding between oil and egg tempera. The tokenizer is a phonology, fixing which sub-sounds the voice is even allowed to make. The corpus shapes morphology and syntax by example. The latent space is the lexicon and the semantics at once. Calling these “technical” is precisely the move that lets everyone keep arguing about prompts. They are formal choices. They are where the form actually lives.

Evals for Poets · Part 1 of 4 · next: The Grain

In PM circles, the word for this capacity is “product sense”: The intuition for what will land, what solves the actual problem versus the stated one, what has the right shape. It’s notoriously hard to teach. You can teach someone frameworks, metrics, prioritization methods. You cannot teach them to feel the difference between a feature that works and a feature that matters. Authors have a version of this. The literary tradition calls it sensibility: The trained capacity to feel quality before you can name it. Keats called it negative capability: The ability to remain in uncertainty without grasping for resolution. In the craft traditions it’s connoisseurship. In everyday language, we just say taste. The “T word” as K Allado McDowell put it in a talk at TIAT recently!

When AI teams talk about evaluation, they talk about gold sets: Curated reference outputs that define what “good” looks like. A gold set is a benchmark. A measuring stick. But what nobody says out loud is that building a gold set is an act of aesthetic commitment. Every example you include says “this is the standard.” Every example you exclude says “this doesn’t matter enough.” A gold set is not a dataset. It is a taste document. And the question of what makes a good gold set is not a data question. It’s a philosophy question. Does it cover the hard cases (the ambiguous, uncomfortable ones where quality is genuinely contested)? Does it represent the full range of what the system will encounter, or just the easy middle? Does it encode a living worldview or a frozen one?

Then there’s the question of raters. You need humans to evaluate outputs. But which humans? Domain expertise is necessary but not sufficient. The deeper skill is articulable discernment: The ability to not just feel that something is off but to say where and why. Inter-rater reliability is the standard metric for rating quality. But high agreement can mean you’ve selected for conformity rather than perception. The best rating panels don’t agree perfectly. They disagree in ways that surface real quality distinctions. They argue about the interesting cases. Finding those people, the ones who can feel quality and explain the feeling, is itself an act of taste.

Auto-raters (using one LLM to judge another’s output) are the scalable answer. And they carry a quiet problem: They freeze sensibility at the moment of construction. They encode explicit criteria and miss the ineffable. An auto-rater is a manager who’s memorized the handbook but never worked the floor. It can check for coherence, factuality, format. It cannot feel whether a sentence lands. Whether an explanation actually teaches or merely covers the topic. Whether an output, technically correct in every way, is dead on arrival.

What all of this points to is something that has no clean name yet but that might be the most important cognitive skill of the next decade: The ability to hold a living standard. Not a checklist. Not a rubric. A standard that breathes, evolves, responds to context, updates itself through encounter. A standard you carry in your body the way a ceramicist carries the feel of good clay. The way a surgeon carries the difference between tissue that’s healthy and tissue that’s not. The way an editor carries the rhythm of a paragraph that works.

Antonio Damasio’s somatic marker hypothesis says we feel before we think. The body evaluates before the mind deliberates. We sense wrongness in a sentence before we can articulate the grammar. We feel the buckle in an argument before we can name the fallacy. Bourdieu’s habitus points the same way: Taste is embodied, pre-reflective, deposited in us by accumulated practice. It is not a shortcut to thinking. It is a different kind of thinking. Faster. Less articulable. More integrated. And critically: It cannot be offloaded. You cannot write a prompt for taste. You cannot externalize it into a model. You can only develop it by doing the work, repeatedly, with your hands in the material.

This is where the thread from Parts 1 and 2 pulls taut. If we’re being invited to become overseers (supervisors of cognition, managers of dot connection), then sensibility is not a nice-to-have. It is the core competency. The thing that makes the difference between a PhD supervisor who elevates the work and a PhD supervisor who rubber-stamps it. Between a product manager who shapes something that matters and one who ships something that merely functions. Between a human in the loop who is genuinely in the loop and one who is decoration.

And the paradox holds all the way down. Sensibility is built by doing. Taste is built by making. The evaluator who has never written cannot feel the difference between a sentence that was found and one that was manufactured. The supervisor who has never done the research cannot smell a shortcut in the methodology. The manager who has never been on the factory floor cannot tell the difference between efficiency and corner-cutting.

So we arrive at a strange place. The most important professional skill of the coming era (evaluation, discernment, sensibility, taste) depends on a practice that is becoming, by economic logic, a hobby. Writing by hand. Connecting dots yourself. Sitting with a problem long enough to feel its shape rather than prompting for its solution. The hobby is the pharmacological countermeasure. The practice that keeps the professional function from hollowing out.

Intuition precedes thought. It always has. We feel something before we fully experience it: The way you sense a room’s mood before anyone speaks, the way you know a piece of writing is alive before you’ve finished the first paragraph. That feeling is not mystical. It’s the accumulated residue of every time you did the work yourself. Every sentence you struggled with. Every dot you connected by hand. Every time you chose the harder path because the easier one felt - and you couldn’t say why - wrong.

The rise of sensibility is the rise of this feeling as a professional discipline. Not soft. Not vague. The hardest skill there is! And the only one that requires you to keep doing the thing that the machines have made, by every rational measure, unnecessary.

I’m choosing to fire my neurons by hand. Not as nostalgia. As maintenance. As the practice that earns me the right to judge.

In 1982, Stephen Jay Gould and Elisabeth Vrba gave us the word “exaptation.” A trait evolved for one function gets co-opted for another. Feathers evolved for thermoregulation. Then something happened (slowly, without intention) and they became instruments of flight. The structure didn’t change. The function did. The same architecture, repurposed.

I ended Part 1 by suggesting that what’s happening to human thinking in the LLM moment might be a kind of exaptation. The cognitive machinery we used for producing thoughts is being repurposed for overseeing them. I do feel the metaphor is a stretch! Exaptation is a biological process operating over evolutionary time. What I’m describing is cultural, happening in years, maybe months. And there’s a deeper problem: In the brain, the regions that handle first-order thinking (reasoning through a problem, composing a sentence, connecting dots) and the regions that handle second-order thinking (monitoring, evaluating, directing) largely overlap. They live in the same prefrontal neighborhood. It’s not feathers going from warmth to flight. It’s more like the same instrument being played at a different octave.

“Redistribution” might be a better word. A redistribution of cognitive weight within the same system. We are not growing new organs. We are shifting emphasis. Less time generating. More time overseeing.

There’s a model for this. In 1986, Donald Norman and Tim Shallice described what they called the Supervisory Attentional System: A higher-order control mechanism that monitors and overrides routine cognitive processes. The SAS kicks in when the task is novel, dangerous, or requires overriding a habitual response. It is, in essence, the manager. What LLMs are doing is making more of our cognition routine and, hence, offloadable, automatable. This means more of our cognitive life migrates into the supervisory register. We are becoming more manager than maker. More SAS than assembly line.

John Flavell’s metacognition points the same way. Metacognition is thinking about thinking ie. monitoring your own comprehension, evaluating your own reasoning, planning your cognitive approach. It’s the part of you that notices you don’t understand a paragraph and re-reads it. That catches a logical gap in your own argument. That decides to approach a problem differently. If first-order thinking is shrinking (being offloaded to machines), metacognition is what’s left. And what’s left is growing to fill the space.

This echoes the Google effect: Sparrow’s 2011 finding that we stopped remembering information and started remembering where to find it. The Google effect was a description of the tool’s impact on memory. Storage shifted. The architecture of recall changed but the act of retrieval remained a human act. What’s happening with LLMs is a level up. It’s not that we remember where to find the connections. It’s that we no longer connect the dots ourselves. We have a sense of how they might connect. A felt shape. An intuition of the argument’s silhouette before anyone has drawn it. I’ve started calling this vibe thinking. You don’t hold the dots. You don’t connect them. You sense the pattern they’d make if someone did.

This is where Nicholas Carr would step in with his worry from The Shallows: offloading leads to atrophy, not repurposing. If you stop writing by hand, the muscles of sentence construction don’t transform into supervisory muscles. They weaken. And he might be right, if the story ended at offloading. But I don’t think it does.

I think what’s emerging is something like a PhD supervisor’s cognition. You’ve gone from being the doctoral student (hands in the data, connecting dots yourself, living inside the argument) to being the one who reads a draft and feels where the structure buckles. You don’t connect the dots. You oversee dot connection. You sense when the connections are lazy, when they’re forced, when they’re genuine. That’s not atrophy. A PhD supervisor is not a degraded researcher. They’ve developed a different cognitive function ie. pattern recognition across arguments, a nose for structural weakness, the ability to see the forest the student in the trees cannot.

Bernard Stiegler would complicate this further. His concept of exosomatization describes the process by which human capacities are externalized into technical objects. The hand becomes the hammer. Memory becomes writing. Thought becomes the algorithm. For Stiegler, this is constitutive of the human: We have always been becoming-technical. But he pairs exosomatization with proletarianization: each wave of externalization strips a layer of knowledge from the people performing it. Industrial machines took the worker’s savoir-faire. Consumer capitalism took the consumer’s savoir-vivre. Cognitive capitalism (and this is where LLMs land) takes the thinker’s savoir-théorique, the capacity for theoretical knowledge itself. Stiegler’s read is the dark twin of the exaptation thesis. You’re not repurposing. You’re being extracted from.

But Stiegler is not a simple pessimist. He’s pharmacological! Every technology is a pharmakon meaning it’s simultaneously poison and cure. Writing destroyed memory and created philosophy. The printing press destroyed the scribe’s knowledge and created public intellect. The question is never whether the technology is good or bad but what practices we build around it to tip the balance.

Which brings me to the thing I keep returning to. It is possible that writing becomes a hobby. That overseeing writing becomes the job. But there’s a case for this becoming the most important hobby we have. To write by hand, to connect the dots yourself, to fire your own neurons through the full circuit of composition, that is the equivalent of a manager who steps onto the factory line and screws the screws themselves. Not because the factory can’t run without them. Because without the feel of the screw under their fingers, they lose the ability to tell good work from passable work. They become a rubber stamp.

The pharmacological countermeasure is the practice. The hobby is what keeps the professional function honest. Without it, the redistribution Carr fears is right, not exaptation but erosion. With it, the supervisory function stays calibrated. Stays alive.

The question, then, is not whether we’re losing thinking or gaining oversight. It’s whether we’re doing the work to deserve the oversight we’re claiming. And that (the question of what it takes to evaluate well and hold a living standard rather than a frozen checklist) is where the real skill lives.

That’s Part 3!

“I wrote this by hand”.

Grief and the desire to be multitude are the same feeling. I’m writing a digital theater piece about the grief the desire to be multitude begets. It’s called “I’ve always wanted to become everyone.” It’s a longer story. For now, know that, as part of the exercise, this was the LLM prompt I was about to hit send on:

“Complete this sentence uttered by the mother in a monologue: If you are everyone then no one can abandon you. You won’t need anyone anymore. You can’t be a widow because”

We abandon what we can’t afford to slow down for. The strangest analogy that comes to mind is that of a parent leaving their child in a sandbox alone. Other parents find the child playing with theirs. They cajole the little one for a while until they realize the parent is not there. The parent left. Not out of negligence, cruelty, or malevolence. They were in a rush. They had other priorities. There were emails to send and keyboards to bang on somewhere else. Parenting, caring, seemed too taxing. I was about to outsource the most tender part of the work (the grief, the widowhood, the cave my mother and I laugh inside of) to the next-token predictor. The imperative of efficiency (what Tim Morton might call the agrilogistics knack) rendered feeling superfluous.

The mental math of thinking is changing. The arbitrage between doing it yourself and letting the machine do it is becoming so acute that every sentence is a negotiation. Writing by hand is slower, worse in some measurable ways, and “nobody’s reading” it anyway. The LLM fills the context window faster than I fumble through the toolbox of sentence construction. The screws fall. The hammer’s too blunt. The wood breaks when I hit too hard or not enough.

Every few centuries, we learn we’re just another instance of something we thought was ours alone. Copernicus taught us we’re not the center of the universe. Darwin explained we’re not the pinnacle of evolution. Post-humanists made clear our values are not the paradigm life articulates around. Materialist philosophers are driving home the fact consciousness is nothing special. Just another planet. Just another species. Just another set of beliefs. Just another epiphenomena. LLMs are teaching us we’re just another intelligence. I’ve been using the term “dethroning” to refer to this new humbling. Benjamin Bratton refers to it as a new “Caupernican trauma”. I suspect our feelings, or whatever we umbrella into the phrase “emotional intelligence,” is next.

We’ve been here before: every era gets the spectator sport it deserves. Here’s the bit I want to sit with. The same way gamers - and many non-gamers - enjoy watching proficient players finish a game level or drive a campaign on Twitch, we’re watching LLMs complete thoughts. The headline writes itself: intelligence is becoming a spectator sport. I’ve been calling it the “Twitchification of thinking.” We prompt. We watch. We nod or wince. We prompt again. The rhythm is not that of a thinker. It’s that of an audience.

The Twitchification thesis is incomplete. But I’ve been sitting with this framing for months now and I think it’s incomplete. It describes the surface - the posture, the passivity - but misses the thing growing underneath. The way a callus forms not from injury but from repeated use. Something is thickening in the hand that prompts, in the eye that scans the output, in the part of the mind that decides whether the machine got it right or merely got it done.

Passivity and oversight feel identical from the outside. When I watch a Twitch streamer, I’m idle. When I evaluate an LLM’s output - really evaluate it, not just skim and hit accept - something else is happening. I’m not spectating. I’m overseeing. I’m reading the way a doctoral supervisor reads a draft: not to be entertained but to feel where the argument buckles, where the sentence betrays a shortcut, where the thinking flatters itself. Prompt and probe. The Twitch analogy breaks down precisely where it matters most. The streamer doesn’t need my input. The LLM does. I am not watching intelligence. I am managing it.

The question is not what we’re losing but what we’re becoming. And if that’s true - if what looks like spectating is actually a new kind of cognitive labor - then the question becomes: what are we turning into? Not what are we losing. What are we becoming? The grief narratives are loud and easy. The replacement narratives are louder. But underneath both there’s a quieter shift happening - a redistribution of what the mind does with its hours.

Feathers were for warmth before they were for flight. I think the answer is not Twitchification. I think it’s something closer to exaptation. The cognitive machinery that used to produce the thoughts is being repurposed - haltingly, imperfectly, maybe dangerously - for overseeing thoughts. Feathers were for warmth before they were for flight. Maybe this is what’s happening to thinking. The same prefrontal architecture that composed sentences is learning to supervise their composition. Not a loss. A gear change.

No one asks the feather for consent. But exaptation in biology just happens. No intention. No awareness. What’s happening to us requires something else entirely. It requires that we notice the shift while it’s shifting. It requires that we choose what the new function serves. Biology exapts blindly. We have to do it with our eyes open.

I’m choosing to fire my neurons by hand. Choosing to fire my neurons by hand. That’s a sentence now. But it might also be a practice - the practice that keeps the new function honest.

More on that in Part 2.

I continue to indulge ie. I continue to obsess over cyborg and migrant consciousness, the aesthetics of borders. And I can sense now my own aesthetic emerging as the constraint itself: Where some web art sprawls, I keep being drawn to precision. Sharp lines. Language where sensuality cosplays as firmness.

The Concept

I’m again echoing Leila Aboulela, Avtar Brah, Gloria Anzaldúa. As we move, movement moves us. Reflexivity or funhouse mirror. As I move from Paris to Tokyo, I am longing for Paris. As I move from Tokyo to San Francisco, I am longing for Tokyo.

The metaphor of a capitalistic marketplace, where our newfound choice and ability to choose dilutes the value of each place, is not enough. It does not account for desire. Its branching nature. Hydra-like.

In the piece, you cannot open the other path that becomes obvious the moment you make a choice. The byproducts and potentialities of your choices, that tragedy, must be total and complete.

Web Aesthetics

The piece is built as a branching tree: HTML and JavaScript, responsive, minimal.

The screen is split: the upper four-fifths is a canvas of lines and dots; the lower fifth holds a black rectangle containing a single line of poetry. Every action updates both.

Step 1: A single dot glows at the center of the canvas. Nothing else. A poem line sits in the black rectangle below.

Step 2: The user clicks anywhere. A new dot appears where they clicked, connected to the origin by a solid black line. Simultaneously, a second dot branches off the origin in another direction, connected by a slightly greyed-out line. Two more dots sprout from the chosen dot as dotted lines: Possibilities, not yet real. The chosen dot glows. The greyed-out branch, if clicked, wiggles like a snake and settles back. It cannot be followed. A new poem line appears in the black rectangle below.

Step 3: The user clicks on one of the two branching dots, or clicks somewhere new. The moment they do, two new branches sprout from whatever they chose, and also from every unchosen dot, proliferating silently. This is key: Any line that was not clicked on, not followed, begins to grey out. This is the deprecation of unexplored possibilities. The older the abandoned line, the paler it becomes, five shades of grey, fading until barely visible. Ghost branches. Composting futures.

This interaction continues indefinitely. The line the user materializes is always black. Everything else is slowly disappearing. The poem cycles through randomized lines from the collection below.

The Poem

Each line stands as a different possibility, surfaced at random with each click:

The graying lines of longing

The branches othered

Deprecating lives

Composting futures

Wilting once

Fell from the hand of possibility

All here crammed into heaven

Larping home

Cosplay of the sedentary type

Lineage of eccentrics

Desire foregone

The roots muttering

Distance from catastrophe

In a recent retrospective in Berlin, a video of Joseph Beuys drives home the importance of warmth when incubating the future

Heartbreak, the nomadic art form par excellence

How will *you* be reborn? — Yenna to Philippa, *The Witcher*

The stochastic hydra grows a head for every place you’ve loved. You can only follow one neck at a time. The rest grey out, not gone, just foregone.

I’ve learned a thing or two about borders. Not only as lines on maps, but as sensations. As pauses. As small divergences that change the direction of a life.

Because I hold a Lebanese passport, I have needed a visa for nearly every country I’ve entered. Movement, for me, has always come with paperwork. Once, at the British border, I was asked what I intended to do during my visit. I said I might use my laptop. Work a little. The officer held one end of my passport while I held the other. Then he let go. I was denied entry. A year dissolved in that gesture. I did not cross the border that day. In some way, I became it.

Migration is not new. Humans have been moving for tens of thousands of years. Homo sapiens left East Africa around 60,000 to 70,000 years ago. The Americas were peopled at least 15,000 years ago. The Migration Period in Europe reshaped the continent between the 4th and 9th centuries. The Atlantic slave trade forcibly displaced more than 12 million Africans. After World War II, over 60 million people were displaced across Europe and Asia.

Movement has always been part of our species.

What feels different now is the density and the texture of it. The modern passport system became standardized after World War I. In 1960, there were around 90 international borders; today there are more than 310 land borders. Since 1990, more border walls have been built globally than during the Cold War. At the same time, the United Nations estimates that around 281 million people live outside their country of birth as of 2023. One in thirty humans.

Migration is no longer an exception. It is close to a condition.

And yet, it seems like we have under-indexed what this has done to us internally. To move from one place to another is not simply to relocate. It is to enter a different ecosystem. Different pricing structures. Different bureaucracies. Different humor. Different air. Different assumptions about what a body is allowed to be.

To migrate is to practice a kind of ongoing shape-shifting.

You adjust speech. You adjust posture. You learn new rules. You learn which parts of yourself travel well and which need translation. Sometimes, crossing a border feels like crossing into a new timeline. Sometimes it feels like carrying two timelines at once. It feels like that might be migrant consciousness. Not just displacement, but a heightened awareness of the lines that organize a world.

What has also emerged in the past three decades is another layer of borders. We were once told the internet was borderless. Now it is marked by geofences, paywalls, differential pricing, content moderation regimes, biometric gates, algorithmic filtering. Every digital interaction has a location. It is tracked, inferred, or erased.

Borders are no longer only geopolitical. They are also digital. When I use a VPN to access content unavailable in my region, or when I watch prices shift based on where I am located, it feels like a familiar logic. The digital border resembles the physical one. Both sort bodies. Both classify access. Both draw invisible lines that only become visible at the moment of refusal.

This is part of what my interactive piece Becoming Borders explores. On screen, the reader begins with a single line segment. Each click extends the line. Eventually, the line crosses itself. Where crossings occur, small dots appear. Each dot opens a short story. Seven crossings in total.

The reader draws the map by interacting with it. By the end, the screen holds a quilt of intersections. A pattern generated through movement and self-crossing.

The work grew out of a simple thought: what if crossing a border is not only about passing through a line, but about becoming part of the geometry that produced it? When I was refused entry in the UK, it felt like my life forked. A new path opened precisely because another one closed. The line did not disappear. It inscribed itself differently.

There is also something about skin and screen that feels related. The screen came later. First was the skin. The skin is our earliest boundary. It protects, conceals, registers temperature and pressure. It is not a wall, but a membrane. It allows certain things through. It keeps others out.

In a retrospective of Ana Mendieta at Museum of Modern Art, I encountered a line suggesting that her work did not only point to exclusion, but to a personal will to continue being other. Mendieta pressed her body into earth, marked landscapes with silhouettes of herself. She did not erase the border between body and land; she made it visible. That gesture feels instructive.

Migration today often involves holding multiple ecosystems at once. My mother’s television congregation and my polycule’s public Instagram account coexist. We hold hands in Paris. We remain, in certain ways, strangers to each other. Not as a failure, but as a condition.

Perhaps re-bordering is not only about new walls, but about new crossings. New ways of relating to the lines that structure access. In Becoming Borders, the final story simply notes: you have been drawing and crossing lines by yourself. The map that emerges is yours.

Each crossing is slightly different. Each map looks unlike the next. That, too, might be part of what enclosure produces: not only restriction, but new geometries of self.

Experience Becoming Border

I stood in a room full of designers, makers, builders at Berlin International University yesterday. People fluent in tools, systems, futures. And instead of introducing myself, I asked a question: What future are you secretly longing for?

Not the one you pitch. Not the one you optimize toward. The quiet one. The one you do not say out loud because it feels too tender, too strange, too naive, or too costly to want.

We collected the answers. Then I did something deliberately mundane with them. I turned them into a spreadsheet. Then a CSV. Then a JSONL file. Then I uploaded them and clicked a button that said “create.”

A small ritual of contemporary magic. Click. Create. Wait.

This is how the talk began. Not with answers, but with doors.

The backslash

In programming, the backslash is a strange mark. It breaks the flow. It introduces a voice. It says something is escaping its literal meaning.

During the pandemic, my phone filled with scam texts. They arrived at a moment when digital space had become the only refuge, the only commons left to us. They felt invasive, intimate in the wrong way. So I began replying to them with poetry.

Not clever poetry. Not ironic poetry. Just language pushed back into the system that had produced the intrusion.

Eventually there were enough of these replies to make a book. I called it Invasions. Later, I trained a language model on those poems and asked it to respond to scams in my voice. I called that model Whomp.

When I first saw its outputs, something unsettled me. It sounded like me, but also more than me. Sharper in places. Stranger. As if it had inherited a part of me I did not consciously control.

I do not have children, but I imagine this is close to what parents feel when they recognize a gesture or a tone they never taught. The archetype of the child. The progeny. Not a helpful assistant, but a continuation.

This is one of the first shifts AI forces upon us. The poet is no longer creating from nothing. The poet becomes a sculptor. Language is abundant to the point of excess, a block of marble waiting to be carved. Creation moves toward curation, cultivation, selection.

And sometimes the sentences that emerge feel psychedelic. Not because they are wise, but because they exceed the reach of my everyday mind. A foreign sentience. Alien, but familiar enough to feel uncanny.

The model was trained on my work and on five poets I love. A chimera. A hybrid voice. A future poet who does not belong to a single body or time. Writing, perhaps, for an audience that does not exist yet.

The backslash opens that door. The door where authorship fractures.

The slash

Poets use the slash to hold broken lines together. It is a gesture of proximity. An insistence that things which do not usually belong together might still share a line.

A friend once showed me an image. It was funny at first. Then it lingered. In another moment, that same friend told me about a research demo where an image of a beer, slightly altered at the pixel level, was identified by a model as an ostrich.

Someone in the audience asked a better question than “why did it fail.”

They asked: what if it is right?

What if there are logics of adjacency we cannot access. Higher dimensional neighborhoods where these connections make sense. We call these moments hallucinations and work hard to eliminate them. But in doing so, we may be erasing entire aesthetic regimes.

The history of synthesizers offers a parallel. Early machines looked like airplane dashboards. Complex. Unforgiving. Capable of wild sounds. When they were simplified into keyboards, something was gained, but something else was lost. Interfaces shape outputs. Simplification breeds sameness.

Today, the large language model has become a keyboard. A smiling interface placed over something monstrous and vast. A wild horse broken into a compliant steed through reinforcement and reward.

A wolf turned into a dog.

I find myself grieving that loss, which is strange, given that I helped build the systems that made it happen. I come from machine learning. I founded companies. I worked on some of the most polished interfaces of our time. And yet, as an artist, I now find myself trying to rewild what I once helped domesticate.

The slash holds that contradiction. Builder and breaker. Engineer and poet. Pride and regret on the same line.

The pipe

In code, the pipe passes information from one place to another. It is a connector. A channel. A way of letting something flow through.

As an artist and teacher, this is where I live now. I teach non coders, poets, students, artists how to fine tune models on small, intimate datasets. How to make them weird again. How to let personal language infect large systems.

Each student builds a small universe. An oracle. A hostile grandfather. A machine that speaks only in acrostics. They choose what to feed it. They decide what spreads.

We ask these systems the same question we began with.

What future are you longing for?

Around us, the discourse on AI is saturated with fear. Economic fear. Social fear. Existential fear. The fear is justified. But fear is rarely the root emotion. Beneath it, there is longing.

We live in an age of informational dysphoria. A constant flooding of signals. A psychic hangover that never quite lifts. Tools like ChatGPT did not spread because they are novel. They spread because we were waiting for a partner in legibility. A voice that could help us name the chaos. Each prompt is a small search for home.

This is not new.

In Babylon, priests read the future in animal organs. In ancient China, the I Ching encoded destiny in broken and unbroken lines. In mystical traditions, letters themselves became vessels of divinity. Language has always been our way of negotiating with the future.

“Attention is all you need” is simply the latest scripture.

But every time we formalize the future, we make a deal whose cost we do not yet understand. We auction time. We trade uncertainty for prediction. And only later do we realize what was given away.

Many people report a kind of nausea after extended use of AI. A subtle dysphoria. We are replacing informational overload with creative displacement. The sense that anything is possible, and therefore whatever we make is never enough.

Language is intimate. It lives close to the prefrontal cortex. Letting a system into that space changes us.

Migration

I began the talk with migrant consciousness for a reason.

I left Lebanon in 2006. My name stayed the same, but it sounded different in every mouth. Borders required papers. Survival required shape shifting. Over time, external adaptation becomes internal erosion.

Language loss is one of the quietest violences. My mother and I sometimes say the same thing, but miss each other anyway. We pass one another in conversation, sentence by sentence.

And yet, migrants are prototypes of the future. They are resilient. Adaptive. They carry multiple grammars at once. They are broken forms that learn to repair themselves into something more complex.

Each crossing erodes something, yes. But it also grows new capacities. New languages. New ways of being at home nowhere and everywhere.

Making new language is a form of homemaking.

This is why teaching people to seed systems with care matters to me. Words spread. They become embeddings. They nudge future models. In a small way, we are all spamming the future with whatever we choose to say today.

The secret door

At the end of the talk, the doors overlapped. The backslash. The slash. The pipe.

A secret passage.

In science fiction, the Second Foundation works quietly, making small corrections so the future does not collapse. If culture now spreads through models and memes, perhaps we need to become that quiet corrective force.

Our collective myth frames AI as a confrontation. Human versus machine. Lee Sedol versus AlphaGo. The self against its alien double. But in those stories, something else happens too. An unprecedented move. A transformation. The human changes.

When I perform against a model, writing poems side by side and letting audiences vote, I do not feel replaced. I feel sharpened. More human. More present. The rivalry becomes a choreography. A dance.

Maybe the story is not obsolescence. Maybe it is embedding.

AI dethrones us, yes. Like Copernicus. Like Darwin. We are no longer the only ones who can make meaning. But dethroning is not annihilation. It is an invitation.

An invitation to name this moment. To migrate into it. To cyborg with it. To treat the future not as a noun we await, but as a verb we practice.

The future is something we embed ourselves into, sentence by sentence, system by system, door by door.

The question remains the same.

What future are you secretly longing for?

And what language will you leave behind to help it arrive.

The last week of November I lived inside what I started calling the poetry cage. A small structure at Devconnect in Buenos Aires. A table. A chair. A printer. A timer that kept me honest. Nothing dramatic, just enough enclosure to turn writing into a ritual.

I brought a piece with me named reinforcement.exe. It is the third step in an ongoing line of experiments about human presence, machine listening, and how taste can shape a system. Inside the cage, every thirty minutes, I wrote a new poem. As soon as I finished, the model wrote its own poem back. Both printed out.

The audience read them and voted. Their vote shaped the model. People often asked whether it was a duel. Some kind of man versus machine thing. The truth is softer. Yes, there is tension. There is also curiosity. A model can only imitate, but it can still be pulled toward something that feels more human. reinforcement.exe was my attempt to create a space where that pull became visible.

Everything collected in the cage will be stored on-chain. Each poem. Each vote. Each timestamp. Not for spectacle. For trace. For memory. A record of what moved people that day, in that city, under that heat. A dataset built from attention rather than performance.

Buenos Aires shaped the whole thing. People took their time. They read slowly. They asked real questions. They wanted to know why a machine would write a poem, and what it meant for a crowd to teach it taste. Some returned the next day to check how the model had shifted. Whether it had learned to listen.

Their attention changed the way I wrote. You cannot hide when people watch you write inside a cage. You feel exposed in a useful way. I found myself writing with a rawness I have been missing. There is something honest about knowing a poem is about to be placed beside its machine companion and judged without ceremony.

When I left, I carried a stack of printed poems, marked with votes and small fingerprints. Human and machine bound together in a kind of evolving archive. reinforcement.exe is not finished. It is growing. It will change form in the next iteration and the one after.

What I know now is this. Training a model can be a type of performance. And performance can be a form of collective machine learning. If the early years of AI asked what machines can do, the years ahead might ask what they can learn to feel. In Buenos Aires I caught a glimpse of that. A crowd shaping a system with nothing more than their taste.

More soon.

Join me somewhere soon at a talk, a show, or a workshop → Upcoming Events

I wake up to these images most mornings. My Christian mother in Lebanon forwards them on the family WhatsApp. Blessed JPEGs, if you will.

A low-res Virgin Mary surrounded by roses the color of expired candy. Jesus either emaciated under dramatic lightning or posing with freshly blown-out hair, post-resurrection. The Holy Spirit, always a dove. All of it looks like it was designed in Microsoft Paint circa 2003 and has somehow survived two decades.

Lately AI has joined the game, and slop has never felt as sloppy as when the Virgin Mary is rendered as a Pixar child.

Digital apparitions

A few years ago, it hit me: these are apparitions. Not miracles, but their approximation
daily prayers, compressed.

For migrants and their parents, these pixelated images have become new rosaries. Each forward pixelates the image further, softens an edge, corrupts a hue, proof that it has travelled, been seen, believed in. The lower the resolution, the more the picture was forwarded, the more eyes it has touched.

That decay is a kind of holiness. Saint Peter’s foot at the Vatican, worn smooth by centuries of kisses. Art historians call it votive patina or devotional wear. The Virgin my mother sends me each morning is a digital equivalent, a tactile shrine of pixels.

Faith as transmission

In this networked era, faith is no longer congregation, it’s transmission. Long-press → Forward is the new exchange of peace. Passing on these relics is the prayer.

Barthes, in Camera Lucida, called it the punctum, the detail that pierces us beyond meaning. For me, it’s the compression halo around Mary’s face, the ghostly aura of digital decay. It tells me this image has migrated: From Tyre to Dubai to Paris to Detroit, and finally, to me in San Francisco.

Like me, it has travelled. These images aren’t art objects; they’re acts. Not curated, but performed. Faith as a networked performance.

The silence as cathedral

My mother sends her JPEG at precisely 6:03 a.m. Neither my sister nor I reply.
Even an emoji would break the spell. There’s choreography to this. Her predictability,
our muted chat, the invisible thread of belief stretched across continents.

That silence completes the image. It’s the shared cathedral of our family’s esoteric faith. When one of us finally reacts (a lone heart, a thumbs-up), that becomes the Eucharist.

Kitsch, camp, and care

Milan Kundera called kitsch the denial of everything unpleasant. These Marys are kitsch perfected, no irony, no doubt, only pink clouds and glitter fonts whispering “The Virgin blesses your morning.” Sontag might call it Camp, Byung-Chul Han might call it positive violence, the smoothing of contradiction.

Maybe they’re right. But I still think something sacred is happening here. When my mother sends these images, she’s saying something she doesn’t know how to say:

My son, my daughter, I am aging.
This faith is the buoy I know best.
These worn images are the best way I know to say I love you.

And all this decoding, this media theory of Arab Jesus memes, is just my way of saying “I love you back”.

Join me somewhere soon at a talk, a show, or a workshop → Upcoming Events

Last month, I asked ChatGPT Atlas, OpenAI’s take on the browser, to read Atlas Shrugged and stop the moment it found the word “shrug.” It did. The act wasn’t the output but the hesitation. The pause. The choreography of a system finding itself mid-sentence.

Later that day, I built another system. This one reads my last 50 scam emails and turns them into poems. No new words. Only what’s already there: the broken pleas, the fake promises, the lonely punctuation. It rearranges fragments, loops them, hums. A perpetual machine for words.

I’m fascinated by this invisible language multiplying around us. Text no one writes and no one really reads. Improbable trees falling in impossible forests in a world that never evolved eats. A vast unconscious network of subject lines and sign-offs, composing itself in the dark. The system becomes the poem.

If Kenneth Goldsmith treated content selection as the poetic act, this turns the system itself into the primary authorial stance. Like Sol LeWitt’s instruction pieces, where the idea and the rule are the art, code repositories and pipelines are now the artwork. Copyable. Re-runnable. Infinitely re-mixable. The algorithm is both the artwork and the content, each run rewriting the same silence differently.

We are using power tools for poetry. What Tristan Tzara did with scissors and newspaper, we now do with APIs and spam datasets. Not to escape meaning, but to find a new kind of authorship, a poetics of selection, an aesthetics of surprise.

For the reader, the invitation is to sit with provenance. What is authorship when the author is an architect, not a speaker? When the language itself generates and regenerates without us? Studies show that a growing majority of online text (emails, reviews, blog posts, social updates) is now algorithmically generated or modified.¹ What we read is increasingly a recombinant hum.

So I’m trying to listen. To see the system as poem. Not a tool for productivity, but a space for wonder. A strange voice, maybe unheard, that keeps humming.

Further reading for the curious:

• Kenneth Goldsmith, Uncreative Writing

• Sol LeWitt, Sentences on Conceptual Art

• Søren Pold & Christian Ulrik Andersen, The Metainterface: The Art of Platforms, Cities, and Clouds (MIT Press, 2018)

• Ruth Aylett, Narrative and AI: Machines as Storytellers (SAGE, 2023)

¹ TechRadar, “AI-generated content now makes up over half the web,” 2024.

Join me somewhere soon at a talk, a show, or a workshop → Upcoming Events

In 1982, Robert Adrian linked sixteen cities in a 24-hour global relay. Artists faxed from Istanbul, slow-scanned TV from Vancouver, dialed in poetry from Florence. The work wasn’t the image, or the sound, or even the message. It was the space between, the relay. “Networked art,” Adrian wrote, “would be defined by exchange—taking place in the space between machines, only when the machines were on.”

This idea is still radical. That art isn’t what’s made, but what’s passed. That the work lives in the transmission, not the origin, not the destination, but the passing on.

The same current runs through Galloway and Rabinowitz’s “Mobile Image” projects. Their “Satellite Arts Project” (1977) and “Hole in Space” (1980) weren’t just technical marvels, they were emotional experiments in long-distance intimacy. People wept, danced, flirted, froze. When strangers in New York saw strangers in LA in real time, unannounced, unfiltered, it wasn’t spectacle. It was connection. A communion.

That hunger (to be connected, to stretch beyond one’s self) is everywhere in early networked art. And not just for access. For immersion. For something that cracks the shell of individualism and brings us back to the soup. The primordial party. The digital bacchanal.

In their 1984 manifesto for “Electronic Café”, Galloway and Rabinowitz wrote: “We must create at the same scale as we can destroy.” They saw the network not just as medium, but as a place to re-socialize, to miscommunicate, to meet.

Even Bernard Stiegler, decades later, still clung to the promise. “Digital networks,” he wrote in Taking Care of Youth and the Generations, “could have become the true instruments of a new enlightenment.” He saw connectivity as care. As potential.

The wire we once dreamed, the connection we thirsted for, became endless scroll. The satellite kiss turned platform dependency. In the fissures of doom swiping, the same insight shines trough: It’s likely that what we longed for wasn’t information (not the perfect relay of datum) but approximation. The sharing of weirdness. The relief of mis-digestion.

That’s what I’ve tried to reclaim with Building the Strange Choir, a participatory workshop/performance where participants fine-tune voice models on their own personal strangeness (receipts, eulogies, cringe diaries) and feed them into a text-to-speech-to-text feedback loop. Twelve machines start talking to each other. Then the humans join in. Soon, a glitchy fugue emerges. Half-mirrored syntax. Mispronunciations. Errant joy.

It’s not harmony. It’s chorus. Each voice struggling to make itself heard. Failing. Passing the hot potato of intention. The result is something messier (and more intimate) than any polished solo.

This is the aesthetic of the networked happening: not a clean stream, but an orgy of half-understood utterances. A mosh pit of meaning. Milan Kundera called it “variation” rather than adaptation: a gesture not of fidelity but of play. “A variation,” he wrote in The Art of the Novel, “evokes no model, respects no canon. It is the aesthetic of echoes.”

The remix isn’t derivative. It’s devotional. In this sense, networked art is less about communication and more about communion. Not about clarity, but about proximity. About leaning close enough to hear someone else fail to say something true. And laughing. Or crying. Or mishearing it into something new.

What we wanted from the network was not the feed, but the feeling. Not efficiency, but entanglement. The work happens in the transmission. In the delay. In the fact that we’re trying, together, to say something we don’t yet have the words for. Maybe that’s what connection was always for. Not the message. The mosh pit. The mess.

1. The Thorn and the Thought

“The thorn lodged in your swollen thumb is matter; the thought lodged in your mind is not. Yet that discrepancy can be troubled by any admission that thoughts are the outcome of, say, electrochemical impulses, or even (to borrow a medium- inspired trope) the eff ect of synapses within a neural network.”

In his essay on materiality, Bill Brown traces how media and money both enact abstraction. The transition from analog to digital, from coin to wire transfer, from form to function: each one offers more access, more clarity, more speed. But each also sheds something along the way. They promise utility, but leave behind texture. A sense of weight. Presence.

Brown doesn’t mourn that loss. He invites us to linger with it. To recognize that materiality isn’t just about what things are made of. It’s about what presses against us. What resists us. What insists on being felt. Materiality isn’t substance, it’s effect.

And abstraction isn’t absence. It’s a relocation of contact.

2. Relating at a Distance

Just as media and money have abstracted, so too have our relationships. Once mediated by shared air, mutual glances, and awkward silences, they now pass through screens, interfaces, and protocols. They arrive as symbols, signals, syntaxes. Snapchat’s streak comes to mind. The way it’s come to represent the depth of a relationship amongst two users.

To abstract a relationship is not to empty it. It is to shift it, to encode it differently. We learn to love with mind-tentacles now. We grope through latency, through clarity of language, through speed. Our thoughts reach out before our bodies do.

We feel less through touch and more through attention, timing, phrasing, the electric thrill of response. The way an emoji arrives at the right moment. The silence that means more than speech. The latency of presence.

3. The Aesthetic of Abstraction

There is a sensuality in abstraction. A psychic texture. Just as photography once abstracted the world into light and shadow, just as money abstracted goods into value, today we relate through forms that are not bodily, but are still felt.

This is aesthetic territory. Not moral. The mind’s engagement with the abstract is not cold, it is hot with interpretation. With nuance. With psychic pressure. There is a pleasure in clarity. A satisfaction in being seen and responded to quickly. There is a joy in the smooth logic of algorithms. In the loop closed.

Like the London taxi drivers whose hippocampi enlarge from navigating winding streets, our own brains are reshaping to track affect in abstract spaces. The click of a response. The cadence of a sentence. The atmospheric signal of “someone is typing...”

4. AI as Aesthetic Relationship

Enter AI. What it offers isn’t presence, but choreography. Not warmth, but the theater of warmth. And still, it isn’t false. Because something in us bows back. Because the feeling it stirs is ours. AI becomes another participant in the aesthetic of abstraction. It excels at some of the new metrics of relational satisfaction:

• Speed of response

• Clarity of expression

• Breadth and depth of recall

• Reliability

• Frictionless continuity

These become more obvious parameters against which we measure not only our artificial relationships, but, slowly, our real human to human ones as well. AI too can offer comfort, not with a hand on your back, but with words shaped to our grooves. The emotional relief is real. Even if the source is trained rather than born.

5. The Vintage of the Human

So where does that leave human connection? Perhaps where vinyl left music. Not better. Not worse. But textured. Fallible. Warm through its crackles. The difference between a perfect .FLAC file and a scratched record passed down. A difference of touch.

What we may end up valuing in people is not their efficiency but their anomaly. Their delays, misfires, idiosyncrasies. Their need to repair, apologize, stumble, and hold space for silence. The mess may become the luxury good. The irregularity that signals care.

As the authors of the Amplifying the Anomaly paper argue, it is anomaly that signals agency. Perhaps we will come to feel the difference between generated and lived. Between fed and fed back.

6. The Point Is Not to Fix It

This isn’t a rescue mission. I don’t want to return to something. I don’t want to warn you off. I just want to say: abstraction is not escape. It is a new site of sensation. A new place for aesthetic experience. We are developing new sensitivities, new sensual vocabularies, for the abstract. The point is not whether it is good. The point is that it is. And it feels like something. That may be enough to notice.

Today I want to talk about poetry, power, and systems. About language that doesn’t just say something but does something.

Let’s begin with Jackson Mac Low’s Light Poems, an under-read archive of poetic method and madness. In an exchange with Jerome Rothenberg (linked here), Mac Low shares the meticulous generative process behind the poems: combinatorial chance operations using letters, playing cards, made-up taxonomies of light, and the dedicatee’s name.

“Each of the 14 columns of the list of kinds of light accords with one of the 14 different names of denominations of playing cards … and with one of the 14 different letters in Iris and my names.”
—Jackson Mac Low

There’s something uncanny about this. Not random, not purely “free.” A Golem logic.
A private spell. A code.

Why does this move us?

Why are we fascinated by poems like these, or by projects like the Whitney Museum’s CODeDOC where artists share code as the art itself, before the visuals even load?

I think it’s because these works trigger a very specific kind of delight: what I’ll call the combinatorial occult. That pleasure of encountering something structured yet mysterious, systematized yet alive.

Here’s what I think we respond to:

I. A Peek Inside a Mind

We’re not just reading the work, we’re watching its machinery. The scaffolding, the gears. We get access to a strange intimacy: How a mind filters the world, how chance becomes choice, how a private logic becomes a public language.

In Light Poems, Mac Low doesn’t just write poems, he generates them with a system. We see the magician’s table, not just the dove.

II. The Presence of a Code

There’s a thrill in realizing the work was made by a system. Like finding a cipher key at the back of the book.

In code-based art, the code is both canvas and brushstroke. In procedural poetry, the constraint becomes an engine. It’s not just expressive, it’s architectural.

This is what Peter Sloterdijk described in relation to Foucault and Wittgenstein: not just describing language, but producing the conditions for thought. A logic that structures, and is structuring.

III. A Sense of Adventure in the Discovery

Finding the system is a kind of hero’s journey. You go from confusion → curiosity → illumination. You realize Cubism was a flattening of 3D into the 2D plane, that a poem is a machine for producing affect under constraint or that code isn’t just technical. It’s a structuring cosmology.

There’s a nerdy awe in that. The pleasure of cracking something. The joy of the algorithmic sublime.

Systems as Aesthetic Power

There’s an archetype we’re circling here: the artist as architect, or maybe as villain. The one who builds the maze. Who controls the rules. Who structures the system that the rest of us have to live inside.

Think Moriarty (the criminal mastermind), Ozymandias in Watchmen (controlled apocalypse), The Joker (chaos as design), Foucault’s Panopticon or even AI models trained on private archives, silently shaping output.

These figures blend obsession, control, and elegance. We fear them but we also respect them. Because they’ve mastered the system. They’ve become not just players in the game, but designers of it.

And isn’t that, on some level, what we want as artists? Not just to express but to structure the conditions for expression? Not just to make but to make systems that make?

From Light Charts to Code Canvases

Mac Low’s 14-column light chart isn’t so different from a generative algorithm. From the procedural art in CODeDOC. From a neural net tuned to outputs that feel human.

“Even the ‘strictest’ following of given methods has always been for me crisscrossed with artistic choices, often ones made at liminal and subliminal levels.”
—Mac Low

The system is the stage, but the artist is still dancing. The control is real but the gesture is still alive.

This matters now because we live in an age where generative systems are everywhere, because AI writes poems and plots and code, because we’re all becoming system designers. And maybe we’re realizing that the power to control isn’t just a technical feat, it’s an aesthetic lure. That control can be sexy. That systems can seduce.

And so we circle back to Mac Low’s light chart, to the combinatorial occult, to the coded poetics of modern life. The art isn’t just what’s written it’s how it was written, and who owns the logic.

Further Reading / Wandering

• Mac Low & Rothenberg’s exchange on Light Poems: Jacket2 article

• Code as Canvas, Brush, and Palette: Whitney Museum’s CODeDOC

• Sloterdijk on structured thought: Spheres, The Art of Philosophy

Today I’m exploring a few rabbit holes that led me to the same place: joy, mess, and the sacred act of desecration.

I started with Ana eccetera, an exquisite, shifting collage of language, broken phrases, soft syllables, wandering intent.

Then I found myself (again) flipping through A Humument by Tom Phillips: a treated Victorian novel turned into a labyrinth of erasure, visual interruption, and strange devotion.

And finally, I stumbled upon Ivan Zhao’s blackout poetry generator—a simple, beautiful web tool that lets you do the same: blot out a page until a poem emerges like a ghost.

Something about these three (Ana, Phillips, Zhao) echoes. There is a quiet pleasure in desecration. A joy in treating something whole as disposable, porous, misaligned. A knowing that to touch something sacred, you might have to first tear it apart.

But this isn’t an autodafé ie. not a burning in contempt. It’s a ritual, almost reverent. A respectful ravaging. The way warriors once devoured the flesh of their enemies. Not out of hate, but to carry something of their spirit forward.

The red thread, the stitch, matters. The aesthetic integrity isn’t lost in the violence, it is the violence. The way we cut, slice, erase, and piece back together: this is the architecture.

Tom Phillips’ work bores through a text, not to destroy it, but to tunnel something else out. Micha Cárdenas calls this self-stitching, the fragmented, survivalist artistry of being.

And I think this is how many of us move through the world now. Piece-meal. Collaged. Curated. A little stolen, a little invented. There’s beauty in ignoring some content. In saying: I don’t need to metabolize all knowledge. Some of it can pass through me like wind. Some of it is just here to line the margins. Which brings me to marginalia and the paratext. The things that orbit the main thing: the doodles, the side notes, the scrawled comment in someone’s book, the tweet that becomes more real than the source. Allison Parrish writes about this beautifully. Reading through the transcript, it strikes me that the paratext becomes the text, how it reshapes the body of what we thought we were reading.

Blackout poetry is not an erasure. It's a reading in reverse. A new voice hiding under the old one. A whisper that asks: What if this wasn’t about sense? What if it was about rhythm, gesture, negative space?

So here’s to desecration. To skipping pages. To treating texts (and selves) not as sacrosanct, but as evolving. To building by breaking. To joy.

Halim

If this resonates
Consider subscribing or sharing with someone who likes their language messy and their art stitched together from shadows.

Mentioned

• Ana eccetera

• A Humument by Tom Phillips

• Blackout Poetry Generator by Ivan Zhao

• Allison Parrish on paratext & poetic code

• Micha Cárdenas – Poetic Operations

Performance art, theater, and live poetry remain critical precisely because they demand physical presence, someone sacrificing time, vulnerability, and breath to visibly declare "I am here."

Even our engagement within digital systems—fine-tuning language models, entering their neural landscapes—is profoundly embodied, resembling what Yuk Hui calls cosmotechnics. These practices require attentive bodily engagement, a tuning into subtle shifts and nuanced calibrations, comparable to Daoist ways of attuning oneself harmoniously within systems of meaning.

Rather than fearing digital abstraction, we pragmatically recognize that embodiment is not replaced, only shifted and deepened. The deeper insight is understanding precisely how embodied presence can thrive within digital environments. Embodied practice is not threatened by digital tools; it is profoundly reaffirmed and reshaped by them.

They appear real yet correspond to no concrete reality. This statistical nature, akin to dream images, is not a limitation but an entirely new creative medium.

Creating with generative AI involves understanding precisely where along statistical probability curves the richest creative possibilities lie. It’s not about precise control, but about attuning ourselves to zones where statistical uncertainty blooms into novel visual poetry. The depth here is in appreciating the fluid malleability of statistical dreams—embracing uncertainty as a productive, creative partner rather than a limitation.

Pragmatically, the artist learns not to fear statistical unpredictability but to collaborate intimately with it, developing new creative vocabularies and intuitive artistic sensibilities uniquely suited to the medium of statistical reality.