Play over
Precision
A creative button generator. Curated ingredients give a direction, the model adds the surprise, and a button gets a personality. Built to encourage play instead of precision.
Can a button have character?
Most AI tools optimise for speed and output. Playground AI optimises for exploration and surprise. I wanted to bring that idea to the smallest component there is.
What if a button were a small moment?
Not just a functional element, but a moment of character. An interface that rewards experimenting over getting it right.
Guided inputs, expressive outcomes.
Curated ingredients are the spark: they show what is possible and move people past familiar conventions. The model is the second layer, pushing each result somewhere you would rarely design on purpose.
Motion becomes personality.
Movement ingredients define how a button behaves in space. Each one is a direction, not a setting.
Emotion becomes interaction.
Mood ingredients shape the emotional tone of the button, translated into how it looks and responds.
Make the prompt quieter.
Early versions also asked the model for creative labels. It sounded fun and pulled focus from what mattered: shape, style, motion, texture. Constraining the output let the model spend itself on visual character.
Refine through language, not modes.
There is no edit toggle. If a button exists, the model changes only what the prompt asks for; otherwise it starts fresh. Iteration stays fluid and the creative thread holds.
Not everything should be generated.
Ingredients give direction; the base controls give precision. The mix keeps it playful, and still usable.
The interaction is half the output.
Expressive cards, a transparent state, and a steady interface that lets the button evolve.
Variation isn't a bug. It's the material.
LLMs are probabilistic systems, not precision instruments. This was an experiment in turning that behaviour into a product quality, with guardrails that keep people in control and still leave room for surprise.