Sculpture and algorithms
Peer into any dimly lit window between Red Hook and Bushwick and you will probably catch a glimpse of an off-white rice paper lamp emitting a soft glow. Sometimes oversized, mini, orb-like or oblong, they are unmistakingly ubiquitous as they adorn the bedrooms, dining rooms, and trendy restaurants of nearly every multi-hypenate creative living in Brooklyn. A signifier of a generation who perhaps defines themselves more by what they like versus what they own.
The maker of the original icon, Isamu Noguchi, was an artist exceedingly concerned with materiality. At the core of his work is material honesty, letting materials guide the form factor versus imposing strict control. He embraces paper, stone, wood, and bronze for its natural textures and imperfections, working with the material, not against it.
Noguchi’s creative process was deeply responsive, often unfolding in direct response to his materials. He would sometimes work in open-air studios near quarries, and spend time simply observing a raw stone, absorbing its history, looking at its unique markings, and surface variations before deciding to make a cut. He viewed the process as a dialogue, going as far as to “stand in front of a rock” and listen for the “voice speaking from [within],” essentially talking to a rock everyday.
Noguchi’s approach of radical collaboration feels surprisingly relevant to building with AI. The tools are different, but the mindset is similar. You’re shaping something with its own behaviors and resistance. Try to control it too much, and it falls flat.
That’s been especially true for me while working on the application layer, creating scaffolding around the algorithms. Writing (*ahem* vibe coding) the software itself has been the easy part. What’s harder is prompting the foundation model, the part where you try to guide the system into submission without smothering it.
Right now, foundation models feel like giant blocks of marble. Alejandro Matamala Ortiz, cofounder of Runway ML, once described their design/product team’s work as uncovering the hidden qualities inside the models, letting a sculpture or form emerge. “You’re presented with this raw material,” he told a group of founders during a recent fireside chat a part of AI Residency (AIR) in Brooklyn, “and then you start carving possibilities out of it.” That approach, shaping from within, has quietly guided Runway’s evolution since its founding in 2018.
If model researchers are doing the heavy lifting to understand the physics of the medium, how it behaves and where it breaks, then it’s up to entrepreneurs to do the carving. I, like many others, am experimenting with tools and techniques to produce “better” outputs. The goal is to let the model’s strengths come through while still shaping it to your liking. It’s a delicate balance. Too little structure and the output is proper AI slop. Too much and you also sterilize it. I’ve found it more effective to create environments where the model naturally follows the behaviors I want, rather than enforce strict rules.
As Betaworks recently put it, “prompting is the UX for AI systems.” It is how we interact with the model’s inner logic. You are not just telling it what to do. You are designing conditions that bring its potential to the surface.
So when I sit under my own $15 knockoff Ikea rice-paper lantern, experimenting with prompt scaffolds and UX shells, I try to remember Noguchi standing in front of a rock. Not to conquer it, but to listen. Building with AI, like sculpture, is a quiet act of attention. Not about force, but about feel.
Emily Manges is creative technologist and Entrepreneur-in-Residence (EIR) inAI Residency (AIR)’s inaugural cohort. A curator-turned-operator, Emily is building an ambient AI wellness companion that blends biometrics and astrology. Subscribe to her newsletter here:
 | A guest post by
|
Emily! Emily! Emily!