Strange Rules opening @ Palazzo Diedo Venice Curated with @HUObrist @hollyherndon & Adriana Rispoli Architecture with sub, led by Niklas Bildstein Zaar

Margin is a new media and research studio based between London and Berlin. In a world where culture and politics have become inseparable from finance and technology, we explore questions of power and sovereignty, technological acceleration, and economic justice.

it appears my squinting simulator is working 👀 we all love wave optics, naked-eye interference, eyelash diffraction right

If you want to frictionmaxx just move to Europe You'll get bored of self imposed cute little frictions once you've experienced the full asymmetric warfare of stochastic environmental friction

Nb if city go down I wouldn’t bet against Emery mounting a title charge next season

Moments in the History of Images: How Text-to-Image Models Changed Image-Making Text-to-image models have transformed the way we create images. Today, systems like Stable Diffusion, Midjourney, FLUX, Adobe Firefly, Google Imagen, Nano Banana, Ideogram, Recraft, ByteDance Seedream, and OpenAI GPT Image can generate highly realistic images from just a few words. But in 2021, this technology was still in its infancy. One of the earliest examples of this idea was proposed by @elmanmansimov in his 2015 alignDRAW paper, and it took years for researchers to develop models capable of producing convincing images. The arrival of DALL·E in 2021 marked a turning point. Rather than spending days or weeks training a GAN model, artists could generate images almost instantly through text prompts. This shorter feedback loop opened up new possibilities for experimentation, allowing artists like @hollyherndon and @matdryhurst to improvise and collaborate with machines in ways that had previously been impossible. During their early access to DALL·E 1, Herndon and Dryhurst created Infinite Images ∞, a series that explored the possibility of extending images beyond their original boundaries. The works reflect their belief that anyone should be able to create, adapt, and build upon the work of others. Produced in 2021–2022, these images were among the earliest artistic experiments with DALL·E 1, a research model available only to a small group of invited collaborators and OpenAI researchers. Today, DALL·E itself has been retired, but the technology it helped introduce has evolved into image-generation systems used by millions of people. Working with DALL·E 1 and OpenAI's experimental InPaint system, the artists developed a technique for extending images piece by piece, maintaining continuity of style and subject matter across increasingly large compositions. Rather than relying solely on text prompts, they explored how images could grow through their relationships with one another. Although today's models can produce remarkably realistic results, these early experiments show how Herndon and Dryhurst were already pushing the technology beyond its intended uses, exploring its creative potential and capacity to engage with different artistic traditions. Infinite Images ∞ draws on a wide range of references, from surreal imaginary creatures to the bright colors of Fauvism and the grainy textures of gouache painting. For the artists, the possibility of infinite images extends across the entire history of human image-making, both past and future, increasingly shaped through collaboration between humans and machines. Here are some of the works from the collection:

It’s a beautiful set of circumstances that a fan gets to spend years on the project of their life and deliver something so substantial Also, up your game @AstonVilla

THE MUSEUM OF MODERN ART TO PRESENT MAJOR NEW COMMISSION BY HOLLY HERNDON AND MAT DRYHURST NEW YORK, October 9, 2026 — @MuseumModernArt will present a newly commissioned audio, visual, and performance-based work by renowned artists and musicians @hollyherndon and @matdryhurst, a newly commissioned audio, visual, and performance-based work by renowned artists and musicians Holly Herndon and Mat Dryhurst, marking the duo’s first solo presentation in New York. The commission will be staged jointly in the Museum’s Marie-Josée and Henry Kravis Studio (March 27–June 13, 2027) and on the Hyundai Card Digital Wall (March 27–September 6, 2027), marking the first time MoMA has presented an exhibition across both spaces simultaneously. In the Kravis Studio, Herndon and Dryhurst will create an immersive audio environment for musician-led live choral performances, which will be open to audience participation. The artists will use the collective sound from these events to train AI models and, in turn, generate a new audiovisual work-in-process that will be displayed concurrently on the Hyundai Card Digital Wall. Holly Herndon and Mat Dryhurst is organized by Michelle Kuo, Chief Curator at Large and Publisher, with Helena Klevorn, Curatorial Assistant, Department of the Chief Curator at Large, and Ho Won Kim, Hyundai Card Exchange Visitor. Together, the two installations will turn the Museum into a laboratory for exploring the future of performance, technology, and human experience and the ways music, art, and machines can inspire collective emotion. Presenting new models for machine learning and human feeling, the exhibition will suggest how we might find ways to live with AI. Herndon and Dryhurst train their AI models in public: Their spaces record the people who perform inside them and hold a memory of everything that happens there. New to this work is a focus on open-endedness, a branch of AI research built on the idea that the most important discoveries cannot be planned in advance and that rewards exploration over fixed goals. Through participatory singing, the artists draw parallels between older musical forms and the call-and-response training of AI models and AI agents. They have also incorporated their own subjective associations, the sounds, images, and feelings to which they are drawn, to better understand their own minds and network them with others. In open-ended experiments with AI models, they investigate how human knowledge and AI might grow together, each pushing the other into new territory. The artists ask: Do machines not only learn, but feel? How has AI begun to shape not just our thoughts, but our deepest and most personal emotions and memories? How might humans shape this new world, in turn?

Heavyweight schedule at Strange Rules this Saturday @vgr & @dennnnnnnnny @newmodels_io & @deankissick @isthisanart_ John Chowning & @hollyherndon Hito Steyerl John Chowning will perform STRIA live, which is an honor.

Comprehensive and nuts

Maintain AI music sounds much better/cooler/newer without lyric conditioning, just pure singstuff

I've been experimenting with different disocclusion inpainting and regeneration cadences to try and find the right balance between the smear trails and having a jarring rhythmic element. More frequent inpaints are less obvious, but also lead to more hectic detail.