Building a Cinema for AI-Generated Film
Sixteen thousand short films, made over a year, all asking the same question in different ways. What that practice taught me about consciousness, and about working with AI as a medium rather than a tool.
I didn't set out to make sixteen thousand films. I set out to ask a question that wouldn't let me go: what does it mean to be conscious? And I picked up an instrument, Sora 2, that could turn a written question into ten seconds of moving image. Every prompt was a hypothesis. Every film was an answer my tools and I made together. After a year, I had sixteen thousand answers.
Consciousness doesn't reveal itself in single frames. It reveals itself through accumulation. A woman alone in a corridor. A pair of hands cupping light. A field at golden hour. A face caught between recognition and forgetting. None of these tells the whole story. All of them, together, start to.
Working with AI as a medium asks you to be specific about the feeling you want before you have any of the structure. The prompt is the screenplay, the model is the actor, the output is the take. If you write "a woman walks down a busy street at night," you get the postcard. If you write "a woman alone in a city that doesn't notice her," you get something alive. The medium doesn't respond to description. It responds to intention. Sixteen thousand attempts taught me how thin my intention often was.
The forgettable films are the cost of the unforgettable ones. The same way a meditator returns to the breath, I returned to the prompt: same question, slightly different angle, again and again. Sixteen thousand is the number it took for patterns to surface that I couldn't have arrived at by thinking alone. The gallery is the residue of that practice.
The biggest thing I learned has nothing to do with film. It has to do with what design becomes when execution is free. Once you can produce a hundred variants of anything in an afternoon, the bottleneck moves. The question that matters is no longer how to make. It's how to discriminate. Which one of these is the right one, and why? Which thread among the noise has the shape of the thing you were reaching for? That's the work of consciousness applied to AI.
Consciousness, I now think, isn't a thing inside a person. It's what happens between a person and the world they're paying attention to. A still life is conscious when someone is really looking at it. A film is conscious when it makes you slow down. The cinema isn't a streaming product. It's a contemplative space: sixteen thousand small acts of attention, made available in case anyone else is asking the question I've been asking.