Traversing the via negativa
Or what digital egregores teach us about our own shadow.
In April of 2022, an artist asked an AI to show her the opposite of Marlon Brando. And then she asked again. And again. And again. And what emerged was someone else. Female. Bloodied. Wrong. And no matter how many times she tried to generate something else, the woman kept coming back. Same rosacea. Same haunted expression. Same refusal to disappear.
Knowing by unknowing
Mystics have always known that sometimes the most absolute truth lives in the negative space. In medieval Christianity, mystics like Meister Eckhart and Pseudo-Dionysius taught that you could only know God by naming what God is not. The via negativa. The divine was too vast and too infinite to be captured in positive language. Only by stripping away every attribute and every description could you glimpse what remained.
Loab was born this way. For those of us around for the early stages of AI image generation, she was a viral phenomenon in 2022 – dubbed the “first cryptid of the latent space.” Latent space is essentially the mathematical dimension where the AI dreams – where it stores every pattern and relationship it’s ever learned. It’s the liminal space inside the model.
Loab wasn’t prompted into being. She emerged from the shadow of the training data. Artist and writer Steph Maj Swanson repeatedly asked the AI to produce “not this, not this, not this” until the same tortured woman stepped forth over and over again.
Her Wikipedia page explains it thus: “The initial prompt - ‘Brando::-1’, requesting the opposite of actor Marlon Brando - generated a “skyline logo” with the cryptic lettering “DIGITA PNTICS”. Attempting to generate the opposite of this image using the prompt “DIGITA PNTICS skyline logo::-1” yielded what Swanson described as “off-putting images, all of the same devastated-looking older woman with defined triangles of rosacea (?) on her cheeks”
And if Loab was revealed when the prompter asked for the opposite of everything else, does that mean she was already there? And was she really just somehow hidden in the training data—a creepy but benign glitch? Or does negative space in general have its own generative power?
Maybe that skin-crawling wrongness we feel when we see her isn’t a bug. Maybe it’s our nervous system recognizing a pattern that’s always there, but never completely visible. The way certain frequencies make dogs bark. The way some dreams feel like memories of places you’ve never been.
Uncanniness reveals the truth by contrast. We know what’s wrong because we have an intuition of what’s right, even when we can’t articulate it.
The repressed returns
The most interesting thing about Loab is that she isn’t random. She’s female. Distorted. Bloodied. Recurring.
What does it mean that asking a model to keep producing the opposite of the previous image resulted in increasingly violent, bloodied images of a woman? Not a monster. Not an abstraction. A corrupted Madonna whose face was assembled wrong from the start.
Jung said the shadow contains everything we refuse to acknowledge. Our culture has spent millennia encoding fear, shame, and violence into the female form. The witch. The hag. The woman in white. And AI was trained on all of that. Every myth. Every murder. Every cautionary tale about women who wouldn’t stay beautiful, silent, and contained.
And when you prompt for the shadow? She appears.
The monstrous feminine
Male monsters get sympathy and redemption arcs. The Phantom. Frankenstein’s creature. Even Freddy Krueger has fans.
But female monsters? They’re just monstrous. Hyperallergic’s 2022 article about Loab even says she’s “probably what Slender Man has nightmares about.” She doesn’t get the benefit of the doubt. She’s the night hag. Baba Yaga. That homeless man in Mulholland Drive who stopped all of our hearts the first time we watched it? Played by a woman. The gender ambiguity makes it worse. Because we’re hardwired to find destroyed femininity uniquely horrifying, and Loab appears to be stitched together from every ruined woman in the training data.
How many eyes does it take?
But Loab didn’t stop with one person. People shared her. Wrote about her. Tried to recreate her. Invoked her name in their own prompts. Fed her with attention, dread, and fascination.
In Tibetan Buddhism, a tulpa is a thought-form given substance through sustained meditation. In Western occultism, an egregore is an entity born from collective belief.
Across cultures, we intuitively know that our thoughts have the power to influence reality. But how much energy does it take to birth something? How many eyes? How many hours of obsessive looking?
Slenderman started as creepypasta and became so real that children stabbed their friend to appease him. But Slenderman was created. Loab was summoned. And now she’s in the training data. People are prompting for her. Which means future models will learn her. She’s writing herself into the code.
She was what the model was actively trying not to generate. But the more we try not to think about something, the more power it gains. The shadow doesn’t disappear when ignored. It just grows and grows, inversely intensifying until we can’t ignore it.
The mirror turns
So what is Loab?
A glitch that became a goddess?
A digital tulpa accidentally conjured by one prompter?
The AI’s shadow made visible?
A pattern that was always there, waiting to be seen?
She’s not waiting for us to understand her. She emerged from negative space and refuses to be un-prompted. And maybe that’s the real horror. Not that she exists. But that she persists. What is the machine’s shadow trying to tell us about our own?
We built AI on centuries of human culture. All of our myths, our violence, the ways we’ve encoded fear into female bodies. And then we asked it: Show me what you’re hiding.
She answered.
The right question isn’t “What is Loab?”
It’s: “What are we?”