//\\//\\
ART of FABRICATION
Here’s a selection of inspirations we collected from projects we’ve had the opportunity to work on over the years, take a look see! read more about our process ->
On the one hand, machine learning models are exceptionally good at reproducing compositions because they’ve normalized training on massive datasets of “successful” visuals. They can generate layouts with convincing hierarchy, spacing, and flow—often instantly.
But there’s a catch: In the interference model, AI misperceives relative shapes in order to develop to a final perceivable output; statistically, it does nothing but imitate raw data. That means it can overfit patterns—producing designs that technically follow principles like proximity or similarity but lack intentional tension or conceptual clarity.
By reverse-searching output, an added benefit of a responsible strategy reveals itself as the independence of unregulated datasets become clear and the ‘perfect imperfection’ of conceptualism becomes closer within reach.
EARLY GENERATION
