Make Soup Not Slop

On Aronofsky's Primordial Soup, the AI filmmaker as capital magnet, and what the money is actually buying

Darren Aronofsky's AI studio, Primordial Soup, is raising $15 million in equity according to an SEC filing, with more than $11 million already sold. Neill Blomkamp, director of District 9, announced his own AI studio, Barley, alongside a 13-minute short generated through ByteDance's Seedance 2.0. Oscar-winning co-writer Roger Avary told reporters that simply attaching the word "AI" to a production company makes it dramatically easier to attract investors.

This is the story. Not that filmmakers are using AI — that was never going to be a surprise. The story is that "AI filmmaker" has become an identity you can capitalize on, and that the capital is flowing to the identity faster than it is flowing to the work. The relationship between the two has inverted. The money is no longer betting on films the studio will make. It is betting on the category itself.

Primordial Soup is the clearest case. Its unofficial motto, per Vanity Fair, is "Make soup not slop" — an explicit positioning against the garbage that has come to define AI-generated content. The mission is to show that AI can be used for genuine storytelling, that artists can sit "at the forefront of technological innovation." This is the defense Aronofsky has made since launching the studio in May 2025 through a partnership with Google DeepMind. The problem, as Gizmodo documented with a run of clips that do not flatter the studio, is that the actual output so far — a series of AI shorts about the founding fathers, the Benjamin Franklin material especially — has been widely judged as bad. The gap between the promise and the product is enormous, and the market is funding the promise anyway.

I have argued before that Aronofsky's public framing of AI as a "tool" is incomplete — that the analogy to a camera or an editing system fails because a generative model does not extend human capability so much as replace it. Primordial Soup's funding round is the economic expression of that same incompleteness. The studio is not being valued on what it has made. It is being valued on the premise that a filmmaker of Aronofsky's stature, pointing the most hyped technology of the decade at storytelling, will eventually produce something the market wants. The $11 million is a bet on the name, the technology, and the moment. It is not a bet on the films, because there are no films yet worth betting on.

Blomkamp's Barley is the same structure in different clothing. The 13-minute NIGHTBORNE, made with Seedance 2.0 — ByteDance's video model that launched in February, generating up to 15 seconds of audio-video output per request — is a demonstration, a calling card. Blomkamp, whose District 9 established him as a director of inventive, grounded science fiction, is using his reputation to legitimize a tool and a pipeline. He is not just making a film; he is making a case that he, specifically, is the one who can make this technology produce real cinema. That case has value to investors precisely because it is unproven. The whole industry is looking for the first filmmaker who can make AI output that does not look like AI output, and Blomkamp is auditioning for that role in public.

The cultural signal in all of this is not the embrace of AI. It is the discovery that the word "AI" functions as a financial multiplier. Roger Avary's observation — that attaching the term to a company eases fundraising — is the quiet confession at the center of the trend. Consider the economics: generative AI is estimated to cut production costs by as much as 30 percent, which is a real, tangible appeal in a financing landscape where independent projects struggle to get made at all. But the multiplier effect of the label is not the same as the cost reduction. It is something separate: the promise that you are at the leading edge, that you understand the future, that your company is positioned not for the industry that exists but for the one that is coming. That promise, as Avary notes, is what opens wallets.

There is a version of this story that is a story about grift — about filmmakers packaging the word "AI" to raise money for work that does not yet justify it. That version is not wrong, but it is too easy. The more precise observation is that the market has decided that the question of whether AI can produce good cinema will be answered yes, and is pricing companies accordingly before the evidence arrives. The investors are not fools. They are making a bet on a direction of travel. They have seen the cost curve, they have seen the pace of improvement in models like Seedance 2.0, and they have concluded that the films will get better faster than the skepticism will harden. They may be right.

But here is what the bet leaves out, and it is the thing The Mirror is positioned to name. The capital is flowing to the category "AI filmmaker" — to the identity, the name, the promise. It is not flowing to the specific human vision that the studio might produce, because that vision does not exist yet. Aronofsky has not yet made a film worth $11 million; he has made a claim about what he might be able to make. The investors are betting that the human talent, pointed at the new machine, will win. What the evidence so far suggests is that the machine does not amplify human vision so much as it averages it toward what the training data contains. The founding fathers shorts looked the way AI filmmaking looks — generic, smoothed, uncanny in the wrong way — not because Aronofsky is a bad artist, but because the tool does not yet do what he needs it to do.

The inversion at the heart of "AI filmmaker" as a funding category is that it reverses the order of proof. Cinema has always demanded that the work come first: the film, then the reputation, then the financing that the reputation secures. The AI filmmaker trend inverts this. The identity comes first, the capital follows, and the work is promised for later, on the strength of the tool and the name. Whether that inversion produces anything genuinely new in cinema, or whether it just produces a lot of funded companies making "soup" that is indistinguishable from the "slop" they promised to transcend, is the open question. The market has decided it is worth $15 million to find out. The audience, as Bayreuth reminded us last week, will have the final word.