Which AI model do you build against?
Every prompt is written and validated against Claude Opus 4.8, the strongest model available. Prompts work in other models, but Opus holds the design and motion language most faithfully.
We test every prompt against Opus 4.8 and don't publish until the output is recognisably the same page as the layer.
Other models will still run these prompts, and most produce something good. What weaker models lose first is the motion timing, the depth and layering, the restraint — you keep the structure and lose some of the craft.
If the output doesn't match the preview, check the model before anything else. A wrong or outdated model is behind most mismatches.
When to use it
Check this if your output looks structurally right but flat, or if the animation feels off compared to the preview.