In week 2 of an AI accelerator with Overclock, we focused on rapid prototyping. The most surprising aspect has been working in 4 parallel streams when building with AI. Every app I built broke down into:

  • functional definition (what it does)

  • visual identity (how it looks)

  • code implementation (how it works)

  • external integrations (how it connects)

While I prototyped 3 apps in the past week, one was launched into production. I made an app to keep track of my baby's feeding, diaper, and medication scheduling!

I started with messy pen sketches and converted them in Excalidraw.

I fed those wireframes into Claude to generate a design spec.

Then moved into Claude Code to help scaffold, debug, and deploy.

design system created with Claude. Two columns contain colors, typography, examples of buttons, input, navigation, and different colored cards

Normally, these phases feel separate. Professionally, these actions are the work of a whole team!

This time, they blurred together.

I could jump from:

UX idea → UI design → frontend development → deployment

without much friction.

That said, this only worked because I kept the scope tight.

AI makes it very easy to keep going… which is where things get tricky (more on that next).

Curious if others are seeing this—are your workflows collapsing into fewer steps with AI?

Design of the home screen, built with Claude. It has a photo of a happy baby. A section called Log Activity with options for 6 or 8 oz of formula, diaper, and medication buttons.