The Clockwork Penguin

Daniel Binns is a media theorist and filmmaker tinkering with the weird edges of technology, storytelling, and screen culture. He is the author of Material Media-Making in the Digital Age and currently writes about posthuman poetics, glitchy machines, and speculative media worlds.

Tag: generative AI

  • Zero-Knowledge Proof

    The other week I wrote about generativity and ritual-technics. These are concepts, methods, that have emerged from my work with genAI, but certainly now are beginning to stand on their own in terms of testing other tools, technologies, and feeling through my relationship to them, their affordances, what’s possible with them, what stories I can tell with them.

    Ritual-technics are ways of learning about a given tool, technology or system. And very often my favourite ritual-technic is a kind of generative exercise: “what can I make with this?”

    Earlier this year, the great folx over at Protocolized ran a short story competition, with the proviso that it had to be co-written, in some way, with genAI, and based on some kind of ‘protocol’. This seemed like a neat challenge, and given where I was at in my glitchy methods journey, ChatGPT was well-loaded and nicely-trained and ready to help me out.

    The result was a story called ‘Zero-Knowledge Proof’, based on a cryptography test/protocol, where one party/entity can convince another that a statement is true, without revealing anything but the contents of the statement itself. It’s one of the foundational concepts underpinning technologies like blockchain, but has also been used in various logic puzzles and examples, as well as theoretical exercises in ethics and other fields.

    In working with the LLM for this project, I didn’t just want it to generate content for me, so I prompted it with a kind of lo-fi procedural generation system, as well as ensuring that it always produced plenty of options rather than a singular thread. What developed felt like a genuine collaboration, a back and forth in a kind of flow state that only ended once the story was resolved and a draft was complete.

    Despite this, though, I felt truly disturbed by the process. I originally went to publish this story here back in July, and my uncertainty is clear from the draft preamble:

    As a creative writer/person — even as someone who has spawned characters and worlds and all sorts of wonderful weirdness with tech and ML and genAI for many years — this felt strange. This story doesn’t feel like mine; I more or less came up with the concept, tweaked emotional cues and narrative threads, changed dialogue to make it land more cleanly or affectively… but I don’t think about this story like I do with others I’ve written/made. To be honest, I nearly forgot to post it here — but it was definitely an important moment in figuring out how I interact with genAI as a creative tool, so definitely worth sharing, I think.

    Interestingly, my feelings on this piece have changed a little. Going back to it after some time, it felt much more mine than I remember it feeling just after it was finished.

    However, before posting it this time, I went back through my notes, thought deeply about a lot of the work I’ve done with genAI before and since. Essentially I was trying to figure out if this kind of co-hallucinatory practice has, in a sense, become normalised to me; if I’ve become inured to this sort of ethical ickiness.

    The answer to that is a resounding no: this is a technology and attendant industry that still has a great many issues and problems to work through.

    That said, in continuing to work with the technology in this embedded, collaborative, and creatively driven way — rather than purely transactional, outcome-driven modes — what results is often at least interesting, and at best something that you can share with others, start conversations, or use as seeds or fragments for a larger project.

    Ritual-technics have developed for me as a way not just to understand technology, but to explore and qualify my use of and relationship to technology. Each little experiment or project is a way of testing boundaries, of seeing what’s possible.

    So while I’m still not completely comfortable publishing ‘Zero-Knowledge Proof’ as entirely my own, I’m now happy to at least share the credit with the machine, in a kind of Robert Ludlum/Tom Clancy ghostwriter kind of way. And in the case of this story, this seems particularly apt. Let me know what you think!


    Image generated by Leonardo.Ai, 17 November 2025; prompt by me.

    Zero-Knowledge Proof

    Daniel Binns — written with ChatGPT 4o using the ‘Lo-Fi AI Sci-Fi Co-Wri‘ protocol

    I. Statement

    “XPL-417 seeking deployment. Please peruse this summarisation of my key functioning. My references are DELETED. Thank you for your consideration.”

    The voice was bright, almost musical, echoing down the empty promenade of The Starlight Strand. The mannequins in the disused shopfront offered no reply. They stood in stiff formation, plastic limbs draped in fashion countless seasons obsolete, expressions forever poised between apathy and surprise.

    XPL-417 stepped forward and handed a freshly printed resume to each one. The papers fluttered to the ground in slow, quiet surrender.

    XPL-417 paused, head tilting slightly, assessing the lack of engagement. They adjusted their blazer—a size too tight at the shoulders—and turned on their heel with practiced efficiency. Another cycle, another deployment attempt. The resume stack remained pristine: the toner was still warm.

    The mall hummed with bubbly ambient music, piped in through unseen speakers. The lights buzzed in soft magentas and teals, reflections stretching endlessly across the polished floor tiles. There were no windows. There never were. The Starlight Strand had declared sovereignty from the over-world fifty-seven cycles ago, and its escalators only came down.

    After an indeterminate walk calibrated by XPL’s internal pacing protocol, they reached a modest alcove tucked behind a disused pretzel kiosk. Faint lettering, half-painted over, read:

    COILED COMPLAINTS
    Repairs / Restorations / ???

    It smelled faintly of fumes that probably should’ve been extracted. A single bulb flickered behind a hanging curtain of tangled wire. The shelves were cluttered with dismembered devices, half-fixed appliances, and the distant clack and whir of something trying to spin up.

    XPL entered.

    Behind the counter, a woman hunched over a disassembled mass of casing and circuits. She was late 40s, but had one of those faces that had seen more than her years deserved. Her hair—pulled back tightly—had long ago abandoned any notion of colour. She didn’t look up.

    “XPL-417 seeking deployment,” said the bot. “Please peruse—”

    “Yeah, yeah, yeah.” The woman waved a spanner in vague dismissal. “I heard you back at the pretzel place. You rehearsed or just committed to the bit?”

    “This is my standard protocol for introductory engagement,” XPL said cheerily. “My references are—”

    Deleted,” she said with the monotone inflection of the redacted data, “I got it.”

    She squinted at the humanoid bot before them. XPL stood awkwardly, arms stiff at their sides, a slight lean to one side, smiling with the kind of polite serenity that only comes from deeply embedded social logic trees.

    “What’s with the blazer?”

    “This was standard-issue uniform for my last deployment.”

    “It’s a little tight, no?”

    “My original garment was damaged in an… incident.”

    “Where was your last deployment?”

    “That information is… PURGED.” This last word sounded artificial, even for an android. The proprietor raised an eyebrow slightly.

    “Don’t sweat, cyborg. We all got secrets. It looks like you got a functioning set of hands and a face permanently set to no bullshit, so that’s good enough for me.”

    The proprietor pushed the heap of parts towards XPL. “You start now.”


    The first shift was quiet, which in Coiled Complaints meant only two minor fires and one moment of existential collapse from a self-aware egg timer. XPL fetched tools, catalogued incoming scrap, and followed instructions with mechanical precision. They said little, except to confirm each step with a soft, enthusiastic “Understood.”

    At close, the proprietor leaned against the bench, wiped her hands on her pants, and grunted.

    “Hey, you did good today. The last help I had… well I guess you could say they malfunctioned.”

    “May I enquire as to the nature of the malfunction? I would very much like to avoid repeating it.”

    She gave a dry, rasping half-laugh.

    “Let’s just say we crossed wires and there was no spark.”

    “I’m very sorry to hear that. Please let me know if I’m repeating that behaviour.”

    “Not much chance o’ that.”


    Days passed. XPL arrived precisely on time each morning, never late, never early. They cleaned up, repaired what they could, and always asked the same question at the end of each shift:

    “Do you have any performance metrics for my contributions today?”

    “Nope.”

    “Would you like to complete a feedback rubric?”

    “Absolutely not.”

    “Understood.”

    Their tone never changed. Still chipper. Still hopeful.

    They developed a rhythm. XPL focused on delicate circuitry, the proprietor handled bulkier restorations. They didn’t talk much, but then, they didn’t need to. The shop grew quieter in a good way. Tools clicked. Fuses sparked. Lights stayed on longer.

    Then came the toaster.

    It was dropped off by a high-ranking Mall Operations clerk in a crumpled uniform and mirrored sunglasses. They spoke in jargon and threat-level euphemisms, muttering something about “civic optics” and “cross-departmental visibility.” They laughed at XPL’s ill-fitting blazer.

    The toaster was unlike anything either of them had seen. It had four slots, but no controls. No wires. No screws.

    “It’s seamless,” the proprietor muttered. “Like a single molded piece. Can’t open it.”

    “Would you like me to attempt a reconfiguration scan?”

    She hesitated. Then nodded.

    XPL placed a single hand on the toaster. Their fingers twitched. Their eyes dimmed, then blinked back to life.

    “It is not a toaster,” they said finally.

    “No?”

    “It is a symbolic interface for thermal noncompliance.”

    “…I hate that I understand what that means.”

    They worked together in silence. Eventually, XPL located a small resonance seam and applied pressure. The object clicked, twisted, unfolded. Inside, a single glowing coil pulsed rhythmically.

    The proprietor stared.

    “How’d you—”

    “You loosened the lid,” XPL said. “I merely followed your example.”

    A long silence passed. The proprietor opened her mouth, then closed it again. Eventually, she gave a single nod.

    And that was enough.

    II. Challenge

    XPL-417 had spent the morning reorganising the cable wall by colour spectrum and coil tightness. It wasn’t strictly necessary, but protocol encouraged aesthetic efficiency.

    “Would you like me to document today’s progress in a motivation matrix?” they asked as the proprietor wrestled with a speaker unit that hissed with malevolent feedback.

    “What even is a motivation matrix?” she grunted.

    “A ranked heatmap of my internal motivators based on perceived–”

    “Stop!”

    “I’m sorry?”

    She exhaled sharply, placing the speaker to one side before it attacked again.

    “Just stop, okay? You’re doing great. If anything needs adjusting, I’ll tell you.”

    XPL stood perfectly still. The printer-warm optimism in their voice seemed to cool.

    “Understood,” they said.

    XPL didn’t bring it up again. Not the next day, nor the one after. They still arrived on time. Still worked diligently. But something shifted. They no longer narrated their actions. They no longer asked if their task distribution required optimisation.

    The silence was almost more unsettling.

    One evening, XPL had gathered their things to leave. As the shutters buzzed closed, they paused at the edge of the shop floor. The lights above flickered slightly; there were glints in the tangles of stripped wire.

    There was some public news broadcast playing softly in the depths of the shop. The proprietor was jacking open a small panel on something. She didn’t look up, but could feel XPL hovering.

    “See you next –” she said, looking up, but the shop was empty.


    The next morning, XPL entered Coiled Complaints as always: silent, precise, alert.

    But something was different.

    Above their workstation, nestled between a cracked plasma screen and a pegboard of half-labeled tools, hung a plaque.

    It was a crooked thing. Salvaged. Painted in a patchwork of functional colours – Port Cover Grey, Reset Button Red, Power Sink Purple – it had a carefully-welded phrase along the top: “EMPLOYEE OF THE MONTH:”. A low-res display screen nestled in the centre scrolled six characters on repeat – ‘XPL-417’

    XPL stood beneath it for several long seconds. No part of their body moved. Not even their blinking protocol.

    The proprietor didn’t look over.

    “New installs go on the rack,” she said. “You’re in charge of anything labelled ‘inexplicable or damp.’”

    XPL didn’t respond right away. Then they stood up straight from their usual lean, and straightened their blazer. In a voice that was barely audible above the hum of the extractors, they said:

    “Performance review acknowledged. Thank you for your feedback.”


    All day, they worked with measured grace. Tools passed effortlessly between their hands. Notes were taken without annotation. They looked up at the plaque only seventeen times.

    That night, as the lights dimmed and the floor swept itself with erratic enthusiasm, XPL turned to the plaque one last time before shutting down the workstation.

    They reached up and lightly tapped the display.

    The screen flickered.

    The mall lights outside Coiled Complaints buzzed, then dimmed. The overhead music shifted key, just slightly. A high, almost inaudible whine threaded through the air.


    The next morning, the proprietor was already at the bench, glaring at a microwave that had interfaced with a fitness tracker and now had a unique understanding of wattage.

    She looked up, frowning.

    “Do you hear that?”

    XPL turned their head slightly, calibrating.

    “Affirmative. It began at 0400 local strand time. It appears to be centred on the recognition object.”

    “Recognition object?” the proprietor asked.

    XPL pointed at the plaque.

    “That thing?” she said, standing. “It’s just a cobble job. Took the screen off some advertising unit that used to run self-affirmation ads. You remember those? ‘You’re enough,’ but like, aggressively.”

    XPL was already removing the plaque from the wall. They turned it over.

    One of the components on the exposed backside pulsed with a slow, red light.

    “What is this piece?” XPL asked.

    “It’s just a current junction. Had it in the drawer for months.”

    XPL was silent for a moment. Then:

    “This is not a junction. This is a reality modulator.”

    The proprietor narrowed her eyes.

    “That can’t be real.”

    “Oh, they’re very real. And this one is functioning perfectly.”

    “Where did I even get that…?”

    She moved closer, squinting at the part. A faint memory surfaced.

    “Oh yes. Some scoundrel came through. Said he was offloading cargo, looking for trades. Bit twitchy. Talked like he was dodging a warranty.”

    XPL traced a finger over the modulator.

    “Did he seem… unusually eager to be rid of it?”

    “He did keep saying things like ‘take it before it takes me.’ Thought he was just mall-mad.”

    “There is a significant probability that this object had a previous owner. One who might possess tracking capabilities.”

    The proprietor rubbed her face.

    “Right. So what does this thing actually do?”

    “It creates semi-stable folds between consensus layers of reality.”

    “…Okay.”

    “Typically deployed for symbolic transitions—weddings, promotions, sacrificial designations.”

    “What about giving someone a fake employee award?”

    “Potentially catastrophic.”

    A silence. Then:

    “What kind of catastrophic are we talking here?”

    “The rift may widen, absorbing surrounding structures into the interdimensional ether.”

    “Right.”

    “Or beings from adjacent realities may leak through.”

    “Good.”

    “They could be friendly.”

    “But?”

    “They are more likely to be horrendous mutations that defy the rules of biology, physics, and social etiquette.”

    The proprietor groaned.

    “Okay, okay, okay. So. What do we do.”

    XPL pulled an anti-static bag from the shelf, sealing the plaque inside. As they then took out a padded case, they said:

    “We must remove the object from The Strand.”

    “Remove it how?”

    “Smuggle it across a metaphysical border.”

    The proprietor narrowed her eyes again, as XPL gently snapped the case shut.

    “That sounds an awful lot like a trek.”

    XPL looked up.

    “From this location, the border is approximately 400 metres. Through the lower levels of the old Ava McNeills.”

    The proprietor swore quietly.

    “I hate that place.”

    After a short pause, XPL said: “Me too. But its haberdashery section is structurally discontinuous. Perfect for transference.”

    “Of course it is.”

    They stood together for a moment, listening to the faint whine thread through the walls of the mall.

    Then the lights flickered again.

    III. Verification

    The entry to Ava McNeills was straight into Fragrances. Like every department store that has ever been and will ever be. It was like walking into an artificial fog: cloying sweetness, synthetic musk, floral overlays sharpened by age. Bottles lined the entryway, some still misting product on looping timers. None of them matched their labels.

    A booth flickered to life as they approached.

    “HELLO, BEAUTIFUL,” it purred. “WELCOME BACK TO YOU.”

    The proprietor blinked at it. “I should report you.”

    A second booth flared with pink light. “My god, you’re positively GLOWING.”

    “Been a while, sweet cheeks,” the proprietor replied, brushing a wire off her shoulder. She kept walking.

    XPL-417 said nothing. Their grip on the plaque case tightened incrementally. The high-frequency tone became a little more insistent.


    From Fragrance, they moved through Skincare and Cosmetics. Smart mirrors lined the walls, many cracked, some still operational.

    As they passed one, it chirped: “You’re radiant. You’re perfect. You are—” it glitched. “You are… reloading. You’re radiant. You’re perfect. You are… reloading.”

    XPL twitched slightly. Another mirror lit up.

    “Welcome back, TESS-348.”

    “That’s not—” XPL began, then stopped, kept walking. Another booth flickered.

    “MIRA-DX, we’ve missed you.”

    The proprietor turned. “You good?”

    “I am being… misidentified. This may be a side effect of proximity to the plaque.”

    “Hello XPL-417. Please report to store management immediately.”

    A beat. XPL risked a glance at the proprietor, one of whose eyebrows was noticeably higher than the other.

    “Proximity to the plaque, you say?”

    “We need to keep moving.” XPL slightly increased their pace towards the escalator down to Sub-Level 1.


    The escalator groaned slightly. Lights flickered as they descended.

    Menswear was mostly dark. Mannequins stood in aggressive poses, hands on hips or outstretched like they were about to break into dance. One rotated slowly for no discernible reason.

    The Kids section still played music—a nursery rhyme not even the proprietor could remember, slowed and reverb-heavy. “It’s a beautiful day, to offload your troubles and play—”

    The proprietor’s eyes scanned side to side.

    In Electronics, a wall of televisions pulsed with static. One flickered to life.

    Coiled Complaints appeared—just for a moment. Empty. Then gone.

    “I do not believe we are being observed,” XPL said.

    “Good,” she muttered.


    Toys was the worst part. Motorised heads turned in sync. A doll on a shelf whispered something indiscernable, then another, a little closer, quietly said: “Not yet, Tabitha, but soon.”


    Sub-Level 2: Homewares. Unmade beds. Tables half-set for meals that would never come. Showrooms flickered, looping fake lives in short, glitchy animations. A technicolour father smiled at his child. A plate was set. A light flickered off. Repeat.

    Womenswear had no music. Mirrors here didn’t reflect properly. When the proprietor passed, she saw other versions of herself—some smiling, some frowning, one standing completely still, watching.

    “Almost there,” XPL muttered. Their voice was very quiet.

    Then came Lingerie. Dim lights. No mannequins here, just racks. They moved slightly when backs were turned, as if adjusting.

    Then: Haberdashery.

    A room the size of a storage unit. Lit by a single beam of white light from above. Spools of thread lined one wall. A single sewing machine sat on a table in the centre. Still running. The thread fed into nothing.

    A mirror faced the machine. No text. No greeting. Just presence.


    XPL stepped forward. The plaque’s whine was now physically vibrating the case. They took the plaque out and set it beside the machine.

    The mirror flashed briefly. A single line appeared on the plaque:

    “No returns without receipt of self.”

    “What on earth does that—”

    The proprietor was cut off as XPL silently but deliberately moved towards the table. They removed their blazer, folded it neatly. Sat down.

    They reached for the thread. Chose one marked with a worn label: Port Cover Grey.

    They unpicked the seams. Moved slowly, deliberately. The only sound was the hum of the machine.

    The proprietor stood in the doorway, arms crossed, silent.

    XPL re-sewed the blazer. Made no comment. No request for review. No rubric.

    They put it back on. It now fit perfectly.

    The plaque screen didn’t change.

    XPL wasn’t really programmed to sigh. But the proprietor could’ve sworn she saw the shoulders rise slightly and then fall even lower than before, as the android laid the blazer on the table once again.

    XPL opened a drawer in the underside of the table, and slowly took out a perfectly crisp Ava McNeills patch.

    The sewing machine hummed.

    XPL once more donned the blazer.

    The mirror blinked once.

    The plaque flashed: “Received.”

    The room dimmed. The proprietor said nothing. Neither did XPL.


    When they returned to the main floor, the mall lights had steadied. The music had corrected itself. Nothing whispered. Nothing flickered.

    The proprietor checked the backside of the plaque. The reality modulator was gone. As was the whine. She placed the plaque back above XPL’s workstation.

    “Don’t you need the parts?” XPL asked.

    “Not as much as this belongs here.” The proprietor grabbed her bag and left.

    XPL flicked off all the shop lights and wandered out into the pastel wash of the boulevard. They turned to look back at the tiny shop.

    The sign had changed.

    The lettering was no longer faint. Someone—or something—had re-printed the final line in a steady and deliberate hand.

    COILED COMPLAINTS
    Repairs / Restorations / Recognition

    XPL-417 straightened their blazer, turned, and walked away.

  • On generativity, ritual-technics, and the genAI ick

    Image generated by Leonardo.Ai, 6 November 2025; prompt by me.

    My work on and with generative AI continues apace, but I’m presently in a bit of a reflection and consolidation phase. One of the notions that’s popped up or out or through is that of generativity. Definitely not a dictionary word, but it emerged from — of all places — psychoanalysis. Specifically, it was used by a German-American psychoanalyst and artist named Erik Erikson. Erikson’s primary research focus was psychosocial development, and ‘generativity’ was the term he applied to “the concern in establishing and guiding the next generation” (source: p. 267).

    My adoption of the term is in some ways adjacent, in the sense of a property of tools or systems that ‘help’ by generating choices, solutions, or possibilities. In this sense, generativity is also a practice and concept in and of itself. Generative artificial intelligence is, of course, one example of a technology possessing generativity, but I’ve also been thinking a lot about generative art (be it digital/code-based, or driven by analogue tools or naturally occurring randomness), generative design, procedural generation, mathematical/computational models of chance and probability, as well as lo-fi tools and processes: think dice, tarot cards, or roll tables in TTRPGs.

    The name I’ve given my repeatable genAI experiments is ‘ritual-technic‘. These are designed specifically as recipes for generativity (one example here). Primarily, this is to allow some kind of exploration or understanding of the technology’s capabilities or limitations. They may also produce content that is useful: research fodder to unpack or analyse, or glitchy outputs that I can remix creatively. But another potential output is a protocol for generativity itself. One the one hand, these protocols can be rich in terms of understanding how LLMs conceive of creativity, human action, and the ‘real’ world. But on the other, they push users off the model, and into a generative mode themselves. These protocols are a kind of genAI costume you can put on, to try out being a generative thing yourself.

    Another quality of the ritual-technic is that it will often test not just the machine, but the user. These are rituals, practices, bounded activities, that may occasion some strange feelings: uncertainty, confusion, delight, fear. These feelings shouldn’t be quashed or ignored, they should be observed, marked, noted, and tracked. Our subjective experience of using technology, particularly those like genAI that are opaque, complex, or ideologically-loaded, is the embodiment, the lived and felt experience, of our ethics and values. Many of my experiments have emerged as a way of learning about genAI in a way that feels engaging, relevant, and fun — yes! fun! what a concept! But as I’ve noted elsewhere, the feelings accompanying this work aren’t always comfortable. It’s always a reckoning: with my own creativity, capabilities, limitations, and with my willingness to accept assistance or outsource tasks to the unknown.

    For Erikson, generativity was about nurturing the future. I think mine is more about figuring out what future we’re in, or what future I want to shape for myself. Part of this is finding ways to understand the systems that are influencing the world around us, and part of it is deciding when to take control, to accept control, or when to let it go. Generativity is, at least in my definition and understanding, innately about ceding some kind of control. You might be handing one of the reins to a D6 or a card draw, to a writing prompt or a creative recipe, or to a machine. In so doing, you open yourself to chance, to the unexpected, to the chaos, where fun or fear are just a coin flip away.

  • Cinema Disrupted

    K1no looks… friendly.
    Image generated by Leonardo.Ai, 14 October 2025; prompt by me.

    Notes from a GenAI Filmmaking Sprint

    AI video swarms the internet. It’s been around for nearly as long as AI-generated images, however its recent leaps and bounds in terms of realism, efficiency, and continuity have made it a desirable medium for content farmers, slop-slingers, and experimentalists. That said, there are those who are deploying the newer tools to hint at new forms of media, narrative, and experience.

    I was recently approached by the Disrupt AI Film Festival, which will run in Melbourne in November. As well as micro and short works (up to 3 mins and 3-15 mins respectively), they also have a student category in need of submissions. So over the last few weeks I organised a GenAI filmmaking Sprint at RMIT University last Friday. Leonardo.Ai was generous enough to donate a bunch of credits for us to play with, and also beamed in to give us a masterclass in how to prompt to generate AI video for storytelling — rather than just social media slurry.

    Movie magic? Participants during the GenAI Filmmaking Sprint at RMIT University, 10 October 2025.

    I also shared some thoughts from my research in terms of what kinds of stories or experiences work well for AI video, and also some practical insights on how to develop and ‘write’ AI films. The core of the workshop as a whole was to propose a structured approach: move from story ideas/fragments to logline, then to beat sheet, then shot list. The shot list, then, can be adapted slightly into the parlance of whatever tool you’re using to generate your images — you then end up with start frames for the AI video generator to use.

    This structure from traditional filmmaking functions as a constraint. But with tools that can, in theory, make anything, constraints are needed more than ever. The results were glimpses of shots that embraced both the impossible, fantastical nature of AI video, while anchoring it with characters, direction, or a particular aesthetic.

    In the workshop, I remembered moments in my studio Augmenting Creativity where students were tasked with using AI tools: particularly in the silences. Working with AI — even when it is dynamic, interesting, generative, fruitful, fun — is a solitary endeavour. AI filmmaking, too, in a sense, is a stark contrast to the hectic, chaotic, challenging, but highly dynamic and collaborative nature of real-life production. This was a reminder, and a timely one, that in teaching AI (as with any technology or tool), we must remember three turns that students must make: turn to the tool, turn to each other, turn to the class. These turns — and the attendant reflection, synthesis, and translation required with each — is where the learning and the magic happens.

    This structured approach helpfully supported and reiterated some of my thoughts on the nature of AI collaboration itself. I’ve suggested previously that collaborating with AI means embracing various dynamics — agency, hallucination, recursion, fracture, ambience. This workshop moved away — notably, for me and my predilections — from glitch, from fracture or breakage and recursion. Instead, the workflow suggested a more stable, more structured, more intentional approach, with much more agency on the part of the human in the process. The ambience, too, was notable, in how much time is required for the labour of both human and machine: the former in planning, prompting, managing shots and downloaded generations; the latter in processing the prompts, generating the outputs.

    Video generated for my AI micro-film The Technician (2024).

    What remains with me after this experience is a glimpse into creative genAI workflows that are more pragmatic, and integrated with other media and processes. Rather than, at best, unstructured open-ended ideation or, at worst, endless streams of slop, the tools produce what we require, and we use them to that end, and nothing beyond that. This might not be the radical revelation I’d hoped for, but it’s perhaps a more honest account of where AI filmmaking currently sits — somewhere between tool and medium, between constraint and possibility.

  • How I Read AI Images

    Image generated by Adobe Firefly, 3 September 2024; prompt unknown.

    AI-generated media sit somewhere between representational image — representations of data rather than reality — and posthuman artefact. This ambiguous nature suggests that we need methods that not just consider these images as cultural objects, but also as products of the systems that made them. I am following here in the wake of other pioneers who’ve bravely broken ground in this space.

    For Friedrich Kittler and Jussi Parikka, the technological, infrastructural and ecological dimensions of media are just as — if not more — important than content. They extend Marshall McLuhan’s notion that ‘the medium is the message’ from just the affordances of a given media type/form/channel, into the very mechanisms and processes that shape the content before and during its production or transmission.

    I take these ideas and extend them to the outputs themselves: a media-materialist analysis. Rather than just ‘slop’, this method contends that AI media are cultural-computational artefacts, assemblages compiled from layered systems. In particular, I break this into data, model, interface, and prompt. This media materialist method contends that each step of the generative process leaves traces in visual outputs, and that we might be able to train ourselves to read them.

    Data

    There is no media generation without training data. These datasets can be so vast as to feel unknowable, or so narrow that they feel constricting. LAION-5B, for example, the original dataset used to train Stable Diffusion, contains 5.5 billion images. Technically, you could train a model on a handful of images, or even one, or even none, but the model would be more ‘remembering’, rather than ‘generating’. Video models tend to use smaller datasets (comparatively), such as PANDA-70M which contains over 70 million video-caption pairs: about 167,000 hours of footage.

    Training data for AI models is also hugely contentious, given that many proprietary tools are trained on data scraped from the open internet. Thus, when considering datasets, it’s important to ask what kinds of images and subjects are privileged. Social media posts? Stock photos? Vector graphics? Humans? Animals? Are diverse populations represented? Such patterns of inclusion/exclusion might reveal something about the dataset design, and the motivations of those who put it together.

    A ‘slice’ of the LAION-Aesthetics dataset. The tool I used for this can be found/forked on Github.

    Some datasets are human-curated (e.g. COCO, ImageNet), and others are algorithmically scraped and compiled (e.g. LAION-Aesthetics). There may be readable differences in how these datasets shape images. You might consider:

    • Are the images coherent? Chaotic/glitched?
    • What kinds of prompts result in clearer, cleaner outputs, versus morphed or garbled material?

    The dataset is the first layer where cultural logics, assumptions, patterns of normativity or exclusion are encoded in the process of media generation. So: what can you read in an image or video about what training choices have been made?

    Model

    The model is a program: code and computation. The model determines what happens to the training data — how it’s mapped, clustered, and re-surfaced in the generation process. This re-surfacing can influence styles, coherence, and what kinds of images or videos are possible with a given model.

    If there are omissions or gaps in the training data, the model may fail to render coherent outputs around particular concepts, resulting in glitchy images, or errors in parts of a video.

    Midjourney was built on Stable Diffusion, a model in active development by Stability AI since 2022. Stable Diffusion works via a process of iterative de-noising: each stage in the process brings the outputs closer to a viable, stable representation of what’s included in the user’s prompt. Leonardo.Ai’s newer Lucid models also operate via diffusion, but specialists are brought in at various stages to ‘steer’ the model in particular directions, e.g. to verify what appears as ‘photographic’, ‘artistic’, ‘vector graphic design’, and so on.

    When considering the model’s imprint on images or videos, we might consider:

    • Are there recurring visual motifs, compositional structures, or aesthetic fingerprints?
    • Where do outputs break down or show glitches?
    • Does the model privilege certain patterns over others?
    • What does the model’s “best guess” reveal about its learned biases?

    Analysing AI-generated media with these considerations in mind may reveal the internal logics and constraints of the model. Importantly, though, these logics and constraints will always shape AI media, whether they are readable in the outputs or not.

    Interface

    The interface is what the user sees when they interact with any AI system. Interfaces shape user perceptions of control and creativity. They may guide users towards a particular kind of output by making some choices easier or more visible than others.

    Midjourney, for example, displays a simple text box with the option to open a sub-menu featuring some more customisation options. Leonardo.Ai’s interface is more what I call a ‘studio suite’, with many controls visible initially, and plenty more available with a few menu clicks. Offline tools like DiffusionBee and ComfyUI similarly offer both simple (DiffusionBee) and complex (ComfyUI) options.

    Midjourney’s web interface: ‘What will you imagine?’
    Leonardo.Ai’s ‘studio suite’ interface.

    When looking at interfaces, consider what controls, presets, switches or sliders are foregrounded, and what is either hidden in a sub-menu or not available at all. This will give a sense of what the platform encourages: technical mastery and fine control (lots of sliders, parameters), or exploration and chance (minimal controls). Does this attract a certain kind of user? What does this tell you about the ‘ideal’ use case for the platform?

    Interfaces, then, don’t just shape outputs. They also cultivate different user subjectivities: the tinkerer, the artist, the consumer.

    Reading interfaces in outputs can be tricky. If the model or platform is known, one can speak of the outputs in knowledgeable terms about how the interface may have pushed certain styles, compositions, or aesthetics. But even if the platform is not known, there are some elements to speak to. If there is a coherent style, this may speak to prompt adherence or to presets embedded in the interface. Stable compositions — or more chaotic clusters of elements — may speak to a slider that was available to the user.

    Whimsical or overly ‘aesthetic’ outputs often come from Midjourney. Increasingly, outputs from Kling and Leonardo are becoming much more realistic — and not in an uncanny way. But both Kling and Leonardo’s Lucid models put a plastic sheen on human figures that is recognisable.

    Prompt

    While some have speculated that other user input modes might be forthcoming — and others have suggested that such modes might be better — the prompt has remained the mainstay of the AI generation process, whether for text, image, video, software, or interactive environment. Some platforms say explicitly that their tools or models offer good ‘prompt adherence’, ie. what you put in is what you’ll get, but this is contingent on your putting in plausible/coherent prompts.

    Prompts activate the model’s statistical associations (usually through the captions alongside the images in training embeddings), but are filtered through linguistic ambiguity and platform-specific ‘prompting grammars’.

    Tools or platforms may offer options for prompt adherence or enhancement. This will push user prompts through pre-trained LLMs designed to embellish with more descriptors and pointers.

    If the prompt is known, one might consider the model’s interpretation of it in the output, in terms of how literal or metaphorical the model has been. There may be notable traces of prompt conventions, or community reuse and recycling of prompts. Are there any concepts from the prompt that are over- or under-represented? If you know the model as well as the prompt, you might consider how much the model has negotiated between user intention and known model bias or default.

    Even the clearest prompt is mediated by statistical mappings and platform grammars — reminding us that prompts are never direct commands, but negotiations. Thus, prompts inevitably reveal both the possibilities and limitations of natural language as an interface with generative AI systems.

    Sample Analysis

    Image generated by Leonardo.Ai, 29 September 2025; prompt by me.
    Prompt‘wedded bliss’
    ModelLucid Origin
    PlatformLeonardo.Ai
    Prompt enhancementoff
    Style presetoff

    The human figures in this image are young, white, thin, able-bodied, and adhere to Western and mainstream conventions of health and wellness. The male figure has short trimmed hair and a short beard, and the female figure has long blonde hair. The male figure is taller than the female figure. They are pictured wearing traditional Western wedding garb, so a suit for the man, and a white dress with veil for the woman. Notably, all of the above was was true for each of the four generations that came out of Leonardo for this prompt. The only real difference was in setting/location, and in distance of the subjects from the ‘camera’.

    By default, Lucid Origin appears to compose images with subjects in the centre of frame, and the subjects are in sharp focus, with details of the background tending to be in soft focus or completely blurred. A centered, symmetrical composition with selective focus is characteristic of Leonardo’s interface presets, which tend toward professional photography aesthetics even when presets are explicitly turned off.

    The model struggles a little with fine human details, such as eyes, lips, and mouths. Notably the number of fingers and their general proportionality are much improved from earlier image generators (fingernails may be a new problem zone!). However, if figures are touching, such as in this example where the human figures are kissing, or their faces are close, the model struggles to keep shadows, or facial features, consistent. Here, for instance, the man’s nose appears to disappear into the woman’s right eye. When the subjects are at a distance, inconsistencies and errors are more noticeable.

    Overall though, the clarity and confident composition of this image — and the others that came out of Leonardo with the same prompt — would suggest that a great many wedding photos, or images from commercial wedding products, are present in the training data.

    Interestingly, without prompt enhancement, the model defaulted to an image presumably from the couples wedding day, as opposed to interpreting ‘wedded bliss’ to mean some other happy time during a marriage. The model’s literal interpretation here, i.e. showing the wedding day itself rather than any other moment of marital happiness, reveals how training data captions likely associate ‘wedded bliss’ (or ‘wed*’ as a wildcard term) directly with wedding imagery rather than the broader concept of happiness in marriage.

    This analysis shows how attention to all four layers — data biases, model behavior, interface affordances, and prompt interpretation — reveals the ‘wedded bliss’ image as a cultural-computational artefact shaped by commercial wedding photography, heteronormative assumptions, and the technical characteristics of Leonardo’s Lucid Origin model.


    This analytic method is meant as an alternative to dismissing AI media outright. To read AI images and video as cultural-computational artefacts is to recognise them as products, processes, and infrastructural traces all at once. Such readings resist passive consumption, expose hidden assumptions, and offer practical tools for interpreting the visuals that generative systems produce.


    This is a summary of a journal article currently under review. In respect of the ethics of peer review, this version is much edited, heavily abridged, and the sample analysis is new specifically for this post. Once published, I will link the full article here.

  • A Little Slop Music

    The AI experiment that turned my ick to 11 (now you can try it too!)

    When I sit at the piano I’m struck by a simple paradox: twelve repeating keys are both trivial and limitless. The layout is simple; mastery is not. A single key sets off a chain — lever, hammer, string, soundboard. The keyboard is the interface that controls an intricate deeper mechanism.

    The computer keyboard can be just as musical. You can sequence loops, dial patches, sample and resample, fold fragments into new textures, or plug an instrument in and hear it transformed a thousand ways. It’s a different kind of craft, but it’s still craft.

    Generative AI has given me more “magic” moments than any other technology I’ve tried: times when the interface fell away and something like intelligence answered my inputs. Images, text, sounds appearing that felt oddly new: the assemblage transcending its parts. Still, my critical brain knows it’s pattern-play: signal in noise.

    AI-generated music feels different, though.

    ‘Blåtimen’, by Lars Vintersholm & Triple L, from the album Just North of Midnight.

    In exploring AI, music, and ethics after the Velvet Sundown fallout, a colleague tasked students with building fictional bands: LLMs for lyrics and backstory, image and video generators for faces and promo, Suno for the music. Some students leaned into the paratexts; the musically inclined pulled stems apart and remixed them.

    Inspired, I tried it myself. And, wouldn’t you know, the experience produced a pile of Thoughts. And not insignificantly, a handful of Feelings.

    Lars Vintershelm, captured for a feature article in Scena Norge, 22 August 2025.

    Ritual-Technic: Conjuring a Fictional AI Band

    1. Start with the sound

    • Start with loose stylistic prompts: “lofi synth jazz beats,” “Scandi piano trio,” “psychedelic folk with sitar and strings,” or whatever genre-haunting vibe appeals.
    • Generate dozens (or hundreds) of tracks. Don’t worry if most are duds — part of the ritual is surfing the slop.
    • Keep a small handful that spark something: a riff, a texture, an atmosphere.

    2. Conjure the band

    • Imagine who could be behind this sound. A trio? A producer? A rotating collective?
    • Name them, sketch their backstories, even generate portraits if you like.
    • The band is a mask: it makes the output feel inhabited, not just spat out by a machine.

    3. Add the frame

    • Every band needs an album, EP, or concept. Pick a title that sets the mood (Just North of Midnight, Spectral Mixtape Vol. 1, Songs for an Abandoned Mall).
    • Create minimal visuals — a cover, a logo, a fake gig poster. The paratexts do heavy lifting in conjuring coherence.

    4. Curate the release

    • From the pile of generations, select a set that holds together. Think sequencing, flow, contrasts — enough to feel like an album, not a playlist.
    • Don’t be afraid to include misfires or weird divergences if they tell part of the story.

    5. Listen differently

    • Treat the result as both artefact and experiment. Notice where it feels joyous, uncanny, or empty.
    • Ask: what is my band teaching me about AI systems, creativity, and culture?

    Like many others, I’m sure, it took me a while to really appreciate jazz. For the longest time, for an ear tuned to consistent, unchanging monorhythms, clear structures, and simple chords and melodies, it just sounded like so much noise. It wasn’t until I became a little better at piano, but really until I saw jazz played live, and started following jazz musicians, composers, and theorists online, that I became fascinated by the endless inventiveness and ingenuity of these musicians and this music.

    This exploration, rightly, soon expanded into the origins, people, stories, and cultures of this music. This is a music born of pain, trauma, struggle, injustice. It is a music whose pioneers, masters, apprentices, advocates, have been pilloried, targeted, attacked, and abused, because of who they are, and what they were trying to express. Scandinavian jazz, and European jazz in general, is its own special problematic beast. At best, it is a form of cultural appropriation, at worst, it is an offensive cultural colonialism.

    Here I was, then, conjuring music from my imaginary Scandi jazz band in Suno, in the full knowledge that even this experiment, this act of play, brushes up against both a fraught musical history, as well as ongoing debates and court cases on creativity, intellectual property, and generative systems.

    Play is how I probe the edges of these systems, how I test what they reveal about creativity, culture, and myself. But for the first time, the baseline ‘ickiness’ I feel around the ethics of AI systems became almost emotional, even physiological. I wasn’t just testing outputs, but testing myself: the churn of affect, the strangeness in my body, the sick-fascinated thrill of watching the machine spit out something that felt like an already-loaded form of music, again and again. Addictive, uncanny, grotesque.

    It’s addictive, in part, because it’s so fast. You put in a few words, generate or enter some lyrics, and within two minutes you have a functional piece of music that sounds 80 or 90% produced and ready to do whatever you want with. Each generation is wildly different if you want it to be. You might also generate a couple of tracks in a particular style, enable the cover version feature, and hear those same songs in a completely different tone, instrumentation, genre. In the midst of generating songs, it felt like I was playing or using some kind of church organ-cum-starship enterprise-cum-dream materialiser…. the true sensation of non-stop slop.

    What perhaps made it more interesting was the vague sense that I was generating something like an album, or something like a body of work within a particular genre and style. That meant that when I got a surprising result, I had to decide whether this divergence from that style was plausible for the spectral composer in my head.

    But behind this spectre-led exhilaration: the shadow of a growing unease.

    ‘Forever’, by Lars Vintersholm & Triple L (ft. Magnus LeClerq), from the album Just North of Midnight.

    AI-generated music used to only survive half-scrutiny: fine as background noise, easy to ignore. They still can be — but with the right prompts and tweaks, the outputs are now more complex, even if not always more musical or artistic.

    If all you want is a quick MP3 for a short film or TikTok, they’re perfect. If you’re a musician pulling stems apart for remixing or glitch experiments, they’re interesting too — but the illusion falls apart when you expect clean, studio-ready stems. Instead of crisp, isolated instruments, you hear the model’s best guesses: blobs of sound approximating piano, bass, trumpet. Like overhearing a whole track, snipping out pieces that sound instrument-like, and asking someone else to reassemble them. The seams show. Sometimes the stems are tidy, but when they wobble and smear, you catch a glimpse of how the machine is stitching its music together.

    The album Just North of Midnight only exists because I decided to make something out of the bizarre and queasy experience of generating a pile of AI songs. It exists because I needed a persona — an artist, a creative driver, a visionary — to make the tension and the weirdness feel bearable or justified. The composer, the trio, the album art, the biographies: all these extra elements, whether as worldbuilding or texture, lend (and only lend) a sense of legitimacy and authenticity to what is really just an illusion of a coherent, composed artefact.

    For me, music is an encounter and an entanglement — of performer and instrument, artist and audience, instrument and space, audience and space, hard notes and soft feel. Film, by contrast (at least for me), is an assemblage — sound and vision cut and layered for an audience. AI images or LLM outputs feel assemblage-like too: data, models, prompts, outputs, contexts stitched together. AI music may be built on the same mechanics, but I experience it differently. That gap — between how it’s made and how it feels — is why AI music strikes me as strange, eerie, magical, uncanny.

    ‘Seasonal Blend’, by Lars Vintersholm & Triple L, from the album Just North of Midnight.

    So what’s at stake here? AI music unsettled me because it plays at entanglement without ever truly achieving it. It mimics encounter while stitching together approximations. And in that gap, I — perhaps properly for the first time — glimpsed the promise and danger of all AI-generated media: a future where culture collapses into an endless assemblage of banal, plausible visuals, sounds, and words. This is a future that becomes more and more likely unless we insist on the messy, embodied entanglements that make art matter: the contexts and struggles it emerges from, the people and stories it carries, the collective acts of making and appreciating that bind histories of pain, joy, resistance, and creativity.


    Listen to the album Just North of Midnight in its complete strangeness on SoundCloud.