Produce a six-icon weather set on a shared grid, in one SVG. Tests visual consistency across repeated elements — the place where most models drift.
Draw a lighthouse on a rocky outcrop during a storm, as a single self-contained SVG. Tests composition, layering, and whether a model can suggest weather without raster effects.
Design a dense data table with sorting affordances and a selected row state. Tests whether a model can make information density feel calm rather than cramped.




Write an original dad joke about programming. Tests whether a model can be genuinely funny rather than merely recognisably joke-shaped.
Announce a fictional product in exactly one paragraph, without marketing clichés. Tests prose discipline under an explicit vocabulary ban.
Build a complete three-tier pricing page as one self-contained HTML document. Tests layout, typographic hierarchy, and restraint under a familiar brief.

