In the wake of rogue LLMs, autonomous agents, opaque algorithms, and AI slop, architectural tools are undergoing a profound reconfiguration. Generative design workflows and computational surrogates offer immense modeling power but also contribute to homogenization, "slopification," and erosion of critical authorship. This studio investigates how AI might "design for itself," aiming to reclaim agency through the design of data and computational infrastructure. Rather than treating AI as a self-effacing technical container, we approach it as a social, civic, and ethical instrument, embedded in territorial, ecological, and architectural contexts.
We examine the Urban Transformer (UT) as a novel typology: a possible architecture for responsible and integrated AI infrastructure, and an inevitable one, whether we like it or not. These structures may take the form of transparent data farms, distributed urban brains, or sentient infrastructural nodes. They can appear as cohesive or decentralized systems, embedded within existing urban and suburban fabrics or as isolated techno-enclaves. The goals of the studio are threefold: (1) to critically explore the use of generative AI in architectural design, tracing its risks and potentials beyond aesthetic automation; (2) to invent an Urban Transformer prototypology that counters the tendency toward isolated, opaque, extractive techno-capitalist systems; (3) to articulate a new architectural process and language for intelligence where form is shaped as much by energy demands, cooling logic, and physical storage as by the need for commons, symbolic resonance, public trust, and data transparency. As AI's societal role remains contested, this studio positions architectural design as a crucial arena for negotiating how flows of intelligence become spatial, and how such architectures can mediate between technology and humanity.
This intensive studio will employ advanced digital tools and generative AI for "vibecoding" architecture. Experimental and remote LLMs and MCP agents will be introduced as exploratory digital modeling tools. A range of software, scripts, and plugins for mapping and open geodata analysis (e.g., Rhino & Grasshopper, QGIS), along with GenAI tools for representation (such as Stable Diffusion, Midjourney, and DALL-E), may act as co-design agents throughout the studio's successive phases.