GAN Poetics Does an unfinished image ask more of its viewer than one that looks complete? And if so, what did we lose when the AI image field moved on to diffusion models, whose entire premise runs in the opposite direction: resolution, coherence, an image that looks like it was never generated at all? For a few years, GANs left a sort of fingerprint: a smear, faces sliding into one another, unfinished images. It recurred across artists who had nothing else in common but the GAN models, so consistently critics gave this a name: GANism. Was that sameness a limit of the architecture, or a portrait of the people using it, the datasets being used? If a thousand artists bring a thousand intentions to the same generator and get back one look, where does authorship actually sit? With the artist, the dataset, or the machine? 10 years later, artists, with access to the most advanced diffusion models, are now going back to those GANs to explore them as expressive tools. Diffusion models generate with a polish and control that GANs never had, but that polish arrives bundled with cost, with licensing, with infrastructure few individuals own outright. So where do artists go who want to keep working at the level of the pixel, the glitch, the visible seam, who find more to say in an unstable image than a resolved one? Is it strange that the technique researchers left behind is the one that remains open, inspectable, free to misuse, free to run on a laptop rather than rented compute? Does a technology's obsolescence to the field that built it create exactly the kind of margin an artist needs, a space left alone long enough to be reoccupied on different terms? Is the artist's retreat toward the older, rougher, cheaper GAN a nostalgia, or a refusal... a way of insisting that an image should still show its work, still carry the evidence of a process rather than disappear into polish? And as diffusion gets faster, cleaner, and more expensive all at once, is the open, imperfect GAN becoming not a relic but a position? A deliberate place to stand and ask what we actually want images to do for us? Here are some artists from the Daily Program @dailydotxyz who have explored these tools and aesthetics, each with their own unique intentions. 📺 7303711398_artificial_intelligence_wikiart_029 by @twt_cha_