Artificial intelligence and computation are becoming more entrenched in how we work, design, and make decisions. What happens, then, at the climax of current trends, when computation infiltrates into all aspects of design? What is computational architecture, and what is its relationship to humans?
Currently we use computation as a tool. Its strengths lie in data aggregation and large data sets. Unlike humans, who abstract and generalize through things like mathematics, computation is able to operate procedurally with originals, or direct precedents.
As proposed by Mario Carpo, the “science of searching” is a new and different means of designing futures, prompting new aesthetics and different approaches to architecture and design. This project takes the critical stance that new computational design procedures which operate using different intelligences will not be understandable to humans. The aesthetics of AI will be an aesthetic unlike humans have seen before based in “patterns” totally dissimilar to the ways in which the human mind works. As such, the systems which would govern the future of farming are not designed by humans and ultimately not understood by humans, manifesting instead as chaos and complexity.
Looking to London, our analysis has made it clear that current systems of farming are insufficient. These systems are either too large and centralized, producing a bare jugular vein whose rupture would mean widespread starvation, or too small and decentralized, focused on variable individual efforts and inefficient land use. Innovations in computation and AI promise new forms of farming, ultimately meaning new forms of architecture.
This project looks to the abandoned Highgate Station as the site of the future of farming and computational design. Instead of proposing the structures and systems with which this farming would take place, this project treats those as given, and devises architectural interventions which would elucidate and bring understanding to these un-understandable systems. It would be a didactic farm.

Looking externally, the farm is organized along the ridge North of Archway Road. Following the slope from Archway Road to Priory Gardens, the farm is tiered and enclosed by a greenhouse-like, topologically-optimized structure. Plants are organized within this structure based on sunlight and symbiotic relationships, optimized through seemingly random placement, manifesting as clumps instead of rows and columns. Approaching from Archway Road, the visitor would be met by an uncanny diptych with rational Victorian Row houses to the right and computational complexity to the left. Approaching from Priory Gardens, the visitor would be met with the same uncanny juxtaposition, but in this case something that blends more with the surrounding landscape, connected to Archway Road through a man-made ”Valley.” This Valley operates not only as a connector but as a civic space, incorporating a public performance area along with the entrances for the Tube and Computation Experience. The linear connection between above and below – private and public – is mediated by planted crop beds – the computation tainting the human space.
This contested space at the boundary between human and computation, understandable and incomprehensible, is explored through the model. When viewed with augmentation, one begins to understand the ways in which the enclosure from the complexity below and how that complexity below interacts with both the valley and the human experience all while drones busy themselves overhead.
With the knowledge of above and below, understandable and incomprehensible, we return to the Computational Farm Experience, but this time as an architectural object.The visitor emerges from the Tube station on Archway Road and is greeted by the uncanny contrast between their right and left. As they walk east, they see a gap in the strange greenhouse to their left, and turn into the valley. Seeing signs for the Computational Experience below, they take the stairs into the gift shop, emerging at a view into the dark recesses of computational systems located under the Farm surface - a glimpse of what is to come. After scanning their ticket, the visitor moves into a coatroom and storage area where they deposit their belongings. They move from this storage space to a lift located on the east side of the farm, transporting the visitor to their first stop - a platform raised 5 meters above the farm surface in-between farm and enclosure. Here they can view drones flying overhead along with complicated arrays of sensors and structural systems with plants growing all around.​​​​​​​
The visitor then moves towards the north to take another lift to ground level, and their second stop - the farm surface. A soil layer on the floor connects the visitor to the farm directly outside, while a mirrored ceiling constricts their view to the plants directly ahead. Continuing straight, the visitor turns a bend and is greeted by a long ramp as they slowly walk under the soil. As the visitor proceeds down this ramp, they see their surrounding environments change from the conventional farm surface to the complexity of computation. Reaching the end of the ramp ,the visitor turns around and enters a dimly lit hallway with offshoots into artificially- greenhouses. The spaces here are smaller to match the size of the greenhouses - located a few meters below the bottom of the soil. Exiting the greenhouse area, the visitor proceeds down a ramp once more towards the server room – the center of computation. This is the deepest underground and the farthest from the surface they will get - a means of keeping the computational equipment safe. Finally, the visitor proceeds to the ”bang” - the final experience which adds context to the systems and environments they have seen. This takes the form of a cylindrical void capped by a mirror allowing views onto the farm above. It is here that the visitor begins to come to terms with the power and beauty of computational aesthetics. It is a vision of a computational architecture.

The environments of this didactic farm are formed through AI images made in MidJourney. They represent the most important human perceptions of the environments in the computational farm.
The process of creating these environments of computation occurred over 5 steps: Pre-prompting, Prompting, Generation, Human Selection, and Integration.
We began with context. We used ChatGPT to isolate and describe 6 research/design areas: site context, computational design, computational farming, topological optimization, the project brief, and the project description. For each area, we asked ChatGPT to generate a series of descriptive phrases that were then collected and collated into a pre-prompt. To establish a hierarchy of importance, we added weights to these design areas.
The next step was prompting, which began with a deeper dive into food distribution in London, computational farming, and topological optimization. From the research into farming, we discovered issues at large and small scales in terms of food availability and security. From research into Computational farming, we learned that the computational farm would look similar to a conventional farm on the surface with its complexity hidden from view. From topological optimization, we found that complexity would operate at all scales simultaneously, transforming the most banal connections into complicated forms. Based on this research, we developed 6 environments which represent the most important areas of the Computational Farm - the areas that the human would attempt to perceive. We created prompts for each of these environments, ultimately adding the pre-prompt to each and adding weights which heavily favored the environment descriptions.
From these prompts, we generated 100 images for each environment using MidJourney AI.
With these image-sets, we intervened as designers in narrowing the 100 to 1. This was done based on four criteria that ensured site specificity and aesthetic consistency with our research on computational design.
With our final 6 images, we moved from surface to volume, using the images to create the spatial environments of the computational farm. It was in these environments that we intervened as human designers, integrating a human experience that would elucidate their complexity.