Agent: Norm
A programmer friend told me she hasn’t coded for a year. She gives the work to AI agents. She also lost a contract to someone who could vibe code: give instructions to an AI, receive the code.
I’m trying to understand what these agents are. There is the word: AGENT. In English, I hear agency. In French there is agentivité, but it doesn’t arrive in the same way. Agent, though, I know this word.
Agent de la norme.
In 2011, I made an animation called Je remercie (I thank). A head with no face. Instead of a mouth, a ruler. The head is framed in a white rectangle, vertical A4 format, surrounded by repeated green forms. The ruler stretches beyond the sides of the head, small marks along its edge. A mouth that measures.
Click (CC) to turn on the English subtitles, translated by AI.
I was working on an experimental graphic novel about socioeconomic precarity in Switzerland, from the “I” perspective. I used agent de la norme to describe, for example, a social worker to whom I had to submit. I had to develop strategies to show that I was obeying in order to receive money, support, care. If I did everything right, I could have material safety.
In the animation, the head has my voice. Not the social worker’s.
I say: “Je remercie l’État.” I thank the state for taking care of me. And then, at the end, I thank the state for judging me non-autonomous, in difficulty, in precarity. I was eighteen until my early twenties. Still developing. Finding ways to live. I needed the help. Receiving it involved a judgement about who I was. The support was real. So was the submission: soft, and very violent.
For me, computation is anywhere I have to hold still and be read.
When I try to understand what these AI agents are, I go through those memories. How I felt in my body. What I had to do to receive help. What doing it, again and again, did to me.
My native language is French. I write in computational English. What if this language is the straight-edge in my mouth? Observing this makes my torso tense, my eyes heavy. My teeth press against each other. My tongue reaches for my palate, trying to hold the corners of my mouth. Something personal goes in. Something recognisable comes out. The uncomfortable part gets rounded.
My coder friend warned me about what she called a dopamine loop. With agents, she said, you can achieve everything you always wanted to code in a very short time. You keep going. You burn out. You give the task. The agent does it. While it’s doing it, you still need to check a little bit.
I stay with that “a little bit.”
It sounds like a very particular attention. Different from coding, being focused, in the flow. Different from leaving a task and still thinking about it somewhere in the background. You are there, but where, and what are you doing?
It brings me back to living-dead drawing, a concept I developed to understand temporal modalities of drawing. Living: the body drawing in real time, the audience having access to the process. Dead: the trace on paper. Living-dead: a generative process between authored and autonomous moments.
I’m jogging in my neighborhood to relax. Listening to my AI DJ. On my smartphone, I use an AI agent to quickly code an app. Better access to mindfulness strategies. Something to share with my friends who also struggle with mental health.
If my work is living-dead, when does it stop? If I keep producing through computation, when do I feel in my body that something is wrong? That I’m not there anymore? That I’ve submitted to something I only wanted to help me?
I bike home late from the hammam, where I give peeling treatments and soap massages to queer and trans people. Kese and Sabunlama. Someone lies naked on a marble table. There is water, soap, warmth, rhythm. My body, a trans body, taking care of another body. Attuned.
I chose this work partly to find a balance with my work with technology. To stay in contact with what matters to me in sensory connection. And to stay in contact with my dead grandmother, born in Izmir. What can happen between bodies without having to put it into words. It’s another kind of technology. The sound of water washing dead skin. The time to hike through the valleys and mountains of a body with my hands full of foam.
At home, coming down after a long shift, I watch Terence Tao talking about AI and science and the paradox of efficiency. He begins with a waterfall. You can hike there, get lost, discover something along the way. Or a helicopter can take you straight there. Then he describes proofs being generated and checked faster than people can absorb them. He talks about curation: someone has to understand what matters in a proof, find connections, make it possible to teach. He calls the accumulation “proof indigestion.” I think of the dopamine loop and burnout my friend warned me about. The thing gets done, and I don't ask why.
Tao says that making every step faster does not necessarily make science advance. I think of myself jogging, making an app to help us recover. When do I get to recover?
He also speaks about collaboration. Working with someone long enough that they can follow an idea before you have finished the sentence. I think of attunement. The attention I give to the naked bodies recovering on the marble table.
I want to ask: Once I've been helped, what can I still refuse?
One difference between the experiences is who gives the instructions. With social assistance, I had to show that I was following them. With an AI agent, I am supposed to give them.
Why, then, does it feel the same in my body?
What am I following when I give the instructions?
+
Body: Computation
Slop: Salope
Last update: 2026.10.14