Except it never fully becomes a car bc the shifting of priorities, the experiments and breakdown of how and when features are shipped creates a Frankensteined product that never fully delivers on the promise or the original vision (if there was one).

But in hindsight these were just bandaids. Once the gaps were filled the experiments and AB testing started again and gaps reappeared.

I'm starting to see a flaw in that process. It's dysfunctional and justified by imagining their skateboard eventually turning into a bicycle, then a car....

Thanks! Happy to share the code to make it!

When design leaders ask teams to leverage AI, it can be overwhelming, too ambiguous, and abstract. Thinking in terms that can be more relatable help make the workflows more concrete.

The first design was produced without AI help. Took 6 hours appx.

The second design was made with AI help. I wanted to use AI to reproduce this. Most AI tools produce React. I asked Lovable to make a card with a gyroscopic effect, and used Claude to recreate in SwiftUI. This took less than 45m.

I get this often from junior designers. Just because you haven't found this in the wild, doesn't mean it shouldn't exist. Pattern recognition is important but also, patterns evolve and so do human behavior.

Using AI I experimented with 2 approaches (same starting point):

1. Vibe prompting with casual English (18 prompts)
2. Using correct SwiftUI terms (4 prompts)

Highlighting another reason designers should get familiar with code tokens (what models are trained on). It'll get you results faster.

The point of this is post isn't to say one is better than the other. But rather lean on your strength and decide which tool to use. Do you need code for devs? Are you fast with Figma? Is this a conceptual prototype or something close to production? Who is the audience?

#prototyping #AI #design