I gave GPT-5.6 Sol, Claude Opus 4.8, and Grok 4.5 the same 100 frontend briefs—here are all 300 results

After generating enough websites with coding models, I started noticing that each model seemed to reach for the same handful of visual ideas. A single impressive screenshot can’t tell you whether that’s actually true, so I tried testing it at a larger scale. I gave GPT-5.6 Sol, Claude Opus 4.8, and