AI OR DIE

Executive Summary

Mind Wank spent years making other people's bad movies and commercials before burning out and relegating filmmaking to hobby status. Boey studied film and animation but quickly learned that even experimental work required crews, equipment, and institutional support. Bengt Tibert spent twenty years as advertising creative director in Warsaw, channelling creative energy into commercial work while personal artistic visions remained unrealised. All three maintained artistic identities while existing in creative limbo – possessing vision and skill but lacking means of autonomous production.

When AI generative tools emerged around 2022, they describe it in survival language:

a creative lifeboat that we clung to.

This isn't metaphor but lived experience. AI didn't enhance existing practice – it enabled practice that had been structurally impossible.

Radical Iteration and the Writer's Room Model

The AIOD practice centres on what Mind Wank identifies as AI's "beautiful thing" – the "ability to iterate at every stage of the process." Traditional film production locks creative decisions progressively: script determines shooting, shooting determines editing, editing determines final form. Major revisions require returning to earlier stages – prohibitively expensive for independent work.

AI collapses this linear progression.

You can rewrite your entire movie at the end. It used to be that the only people who could do that were extremely well-established filmmakers like Woody Allen or Terrence Malick. But now all of us can decide at the last second that actually this story is better told from perspective B rather than perspective A.

Their collaborative structure amplifies this. Rather than each member handling a discrete role – director, cinematographer, editor – AIOD works fluidly. Someone throws out an idea, the others iterate on it, each bringing their own set of perspectives.

It allows us to all iterate on each other's ideas very very quickly – almost like a writer's room, but instead of the end product being a script, the end product is actually a full piece.

The AI makes this democratic structure possible by eliminating infrastructure bottlenecks that traditionally concentrate creative control in a single director.

Taste as the Only Differentiator

When asked what separates their work given AI's accessibility, Bengt's answer is immediate: **taste.** The average person playing with Luma or Sora produces something that looks the same as everything else – not because the tools are bad but because they lack the accumulated viewing experience to make the million small decisions that constitute a coherent aesthetic vision.

Bengt grew up in former Communist Poland consuming massive amounts of international cinema – Italian, French, German, American – which created what he calls sophisticated taste for film. This shapes how he deploys AI generation in ways that can't be replicated by someone without that foundation.

When you give one of these prompts to someone who has some kind of taste, they will choose things, they will orient themselves within that world.

Mind Wank extends this into a precise formulation: AI can usefully make a hundred decisions, maybe even five hundred or a thousand. Beyond that, it produces generic output requiring human intervention. The creative challenge involves determining appropriate decision-making scale – enough AI autonomy to generate unexpected possibilities, enough human control to maintain coherent vision. Their collaborative practice navigates this through rapid iteration: generate options, evaluate collectively, refine promising directions, regenerate with tighter constraints.

Embracing the Glitch

A consistent thread through AIOD's practice involves rejecting photorealism and commercial polish in favour of aesthetic distinctiveness. Early in their AI work, they experimented with achieving realistic outputs –

everyone was super hyped about the fingers and oh they finally can do fingers.

But they found this pursuit aesthetically deadening. Photorealism pushes toward generic commercial aesthetics. Stylisation creates a distinctive voice.

Their breakthrough came through **embracing AI's glitches and limitations** as aesthetic resources rather than problems to solve.

The fact that it's not perfect always kind of adds to my aesthetic. I like that there are some glitches. I like that you see that it was generated because it kind of adds to whatever the aesthetic.

Mind Wank describes preference for "slightly glitchy" outputs where "you can kind of feel the way a shot is sort of melting and warping." This temporal instability becomes expressive quality when embraced rather than fought.

This positions them against dominant AI video discourse emphasising realism and technical perfection. They're not pursuing AI video that passes as human-made traditional cinematography. They're developing aesthetic language specific to AI's affordances and constraints – natively digital cinema that acknowledges rather than hides its generative origins.

The Slot Machine Problem

Despite enthusiasm about AI's creative possibilities, the trio articulates clear frustrations with current platform limitations. Mind Wank identifies the core issue:

Right now you just sort of have to run the slot machine and hope for the best.

AI video generation produces complete outputs – you either accept or regenerate entirely. Missing is the capacity to identify successful elements and modify surrounding context.

It would be nice to have that ability within the platform itself to be able to say: I like this, this, this, and this. Let's use that, but we're not using any of the rest. Redo it. But I want this exact character in this exact same setting.

They work around this by regenerating multiple times, compositing successful elements, and manually editing to compensate for AI limitations. But integrated control would transform AI from batch generation system into interactive creative environment.

Bengt expresses a different attitude toward technological advancement pace. While Mind Wank and Boey anticipate continued rapid improvement, Bengt is satisfied with current capabilities:

It could stay as it is and I would be happy.

His focus remains on aesthetic development rather than technical capability – a divergence that reveals a genuine tension in AI creative practice between riding the cutting edge and developing deep facility with existing tools.

Authorship Through Curation

The trio's thinking about authorship is nuanced. Mind Wank frames it through an editing metaphor:

At the point where you've edited every other sentence to the point where it feels like your own voice, then I think you can claim authorship – at least colloquially.

Authorship derives not from generating raw material but from shaping it into coherent expression.

Their filmmaking backgrounds provide physical reference points – understanding of camera movement, spatial blocking, performance gesture – that inform how they prompt and evaluate AI outputs. They can identify when AI motion feels wrong even if technically coherent, drawing on embodied knowledge from traditional practice. This cinematic literacy, developed through years of working with physical media, guides curatorial decisions even when generation occurs algorithmically.

Boey's concluding assessment captures the trio's position with dark clarity:

It's going to merge into games, into film and TV and everything. We're cooked. We're all cooked.

This isn't despair but pragmatic acceptance – riding a wave they didn't create, deriving creative benefit while remaining alert to uncertain futures. They're not defending human creativity against AI encroachment. They're developing distinctive voices within an AI-augmented landscape, betting that as platforms flood with generic content, curatorially sophisticated work becomes more valuable by contrast.