Why Designers Should Sample Models the Way Painters Grind Pigment
A designer's argument that AI models, like paint or HTML before them, must be learned by hands-on testing rather than from a mockup.
Pierre Silva1 min read
Why Learn AI by Building With It?
Jan Neeman, writing on UX Design Central, argues that AI models are the newest design material, and that designers can only understand them by running them repeatedly. The piece places this next to older lessons from painting, the Bauhaus workshops and early web editing, where makers learned their medium by working with it directly. It shows how a single static mockup hides the spread of answers a model gives, and why sampling a prompt many times reveals habits that a polished screen never will.
For brand builders, product designers and strategists, this is a useful reframe of the coding debate. The author separates throwaway sampling code, built only to understand the material, from production code that ships. Teams planning AI features will find practical grounding for testing variation and error before committing to an interface. It is worth a read for anyone who designs with, or briefs, AI-powered experiences.
Source: uxdesign.cc · Originally published
