/ Experimental film
i can’t stop thinking about a time i’ve never seen
Custom training, Deforum, traditional media, and a Dell CRT. Music: “Pregunta,” performed by Dora Luz.
Original film & credits ↗Personal practice / AI journey
AI, before
it behaved.
I will always want to work with the traditional medium first.
I’ve spent about ten years exploring and experimenting with AI. At first, I was curious whether those almost-coherent hallucinations could inspire new work. Later, I wanted to make art from the strange ways it perceived the world. The medium matched the surreal dreams I saw vividly in my head. My views on it have changed since then.
Enter the archive
“i can’t stop thinking about a time i’ve never seen”
Diffusion → edit → CRT → recording01 / The early work · 2022–2023
Before the
frames agreed.
I experimented with frustrating frame-to-frame animation and burned out my home GPUs. Some of these films had very little coherency, but they felt like they existed in a different time. This work came before the coherent video models I later used in production.
Original upload dates are shown below. Years inside fictional titles are part of the fiction.
/ Experimental film
the end: to begin again...
and again: the end
The companion film. Style concept by 0010, used with consent; exhibited in Unit London Digital.
Original film & credits ↗
/ Music collaboration
BLVCK
An AI collaboration with la+ch.
Original film & credits ↗
/ Fictional commercial
Crunchinator Commercial 1993
“Classic Cartoon Commercials.” An invented broadcast from a very specific imaginary past.
Original film & credits ↗
/ Fan-made reimagining
1987 Metal Gear Solid TV Show Intro
An imagined series intro, primarily assembled from Midjourney images. Extended version.
Original film & credits ↗
/ Fan-made reimagining
SNOW CRASH,
AI imagined
A fictional John Carpenter-style adaptation of Neal Stephenson’s novel, made with Midjourney v4.
Original film & credits ↗Players load only when selected. YouTube may set cookies once loaded. Each original-film link remains available if playback is restricted.
02 / Collection experiment
The Goose Is Loose.
2,528 Geese in the original
collection description
The goose is loose. A collection of 2,528 Geese, made for the feeling of “He’s just like me.” The GOOSE and a few other friends take a trip through the multi-metaverse.
Fictional imagery, for entertainment. This is an archive of the work, not an investment recommendation.
03 / How the practice changed
How I
got here.
I wrote about the future I hoped could happen and the one I feared might happen. I didn’t really expect it to go the way it did. It's been a fascinating experience watching how the tools have evolved.
- Early experiments
Early experiments.
At first, it was the hallucinations that interested me. As the tools improved, I also tried them on mundane tasks, like replacing grass in a Photoshop photo fix. Around the same time, I was building simple if/then bots on Discord for entertainment. Those were logic-based experiments, separate from the model work.
- 2022–2023
Working with artists.
I worked with artists to help train their own models on their own artwork, and collaborated on projects with their consent. I hopped around the early AI art scene and exhibited in galleries focused on it. I also explored free tools and the possibility of artists licensing their own style models.
- Sealed / 2023–2024
Sealed.
When we began work on Sealed, we used what I’d learned about training checkpoints and models to help build AI-based image search for art on Ethereum. We tagged artwork by subject matter and style so people could search for work from artists across the world. I also argued for clear disclosure when AI was used.
- Production work
Agency production.
I worked across agencies, helping implement the production workflows needed to get AI video out the door. At MOM, I produced and edited 60+ videos across client work. The writing, editing, and post-production were still human work, but the footage had become mostly AI-based. We were making it for an audience that was torn on the morality of it.
See the Kraken work ↗ - 2026 / Reassessment
Changing my mind.
I was optimistic about what AI could open up creatively. As more of the work came to rely on centralized models, I began to fear the inverse: a very linear, milquetoast future of creative sameness. I’ve moved my personal work back toward traditional production when it makes the most sense to. That doesn’t meant there's no use for generative tools in production.
04 / Notes on AI morality
Creativity is
still the point.
“We must write and build our own concepts.”
More than the output.
I believe AI in media can be used for more than funny Office clips or bananas having contentious relationships with strawberries. Art and media are where much of the public opinion lands. People don’t always see the more boring systems being improved. I don’t see anything wrong with using it to pull together metrics or build an intake system. But I do think people must push to leverage generative tools in a meaningful way.
Keep making the work.
AI should not replace human creativity. If everything is automated, the greatest currency we have as creatives is still creativity. We can use our own sketches to guide the work, shoot an extra bit of b-roll, or record a voice-over. Those small things add up, even in AI-assisted work, and can outshine the perfectly “fine” media we’ll keep seeing.
Where we draw the line.
Humans want to see things made by humans. Where we draw the line between humanity and machine is up to us. I still want to work with the traditional medium first when it's applicable. My interest in the technology has never made that part of the work less important to me.
Adapted from “AI Morality” / Robin Freundt
Related work / Commercial practice
Kraken
How we made the films, including the writing, voices, editing, and sound.