Artificial intelligence sees architecture differently. It does not experience space; it processes form. It samples, categorizes, and reassembles, always searching for the most recognizable patterns, the most legible structures. AI is inherently object-oriented, privileging the foreground—the building, the landmark, the thing itself—over the field from which it emerges. This work resists that tendency. Instead of reinforcing architecture as an isolated object, it attempts to dissolve it into context, flattening hierarchies and erasing focal points. It is an image of architecture where everything is background.

By leveraging AI’s capacity for sampling, this process does not construct buildings as objects but instead as conditions. The architectural image is not defined by edges, figures, or clear markers of significance. Instead, it becomes a landscape of relationships, a diffuse field where depth and hierarchy break down. Like an aerial perspective with no fixed scale, or a drawing rendered from a viewpoint that doesn’t exist, these images propose architecture as something that is not seen but absorbed—something environmental rather than discrete.

This approach embraces AI’s tendency toward misalignment and misrecognition, pushing the technology away from its typical role as a generator of defined objects and instead toward an exploration of perception itself. These images are not depictions of space as it is but rather as it is felt—subjective, unstable, and without a center.

In Transductions, architecture is not the subject; it is the medium. The work refuses the clarity AI seeks to impose, favoring an architecture that is diffuse, ambient, and uncertain. It does not assert itself as an image of a building but rather as an impression of a place—an accumulation of space, flattened into a field, where nothing is more important than anything else.