
Diana's Diamonds
Building a prototype library to scale AI-driven design.
- Lead Designer, Design Engineer
- September–October 2026
- Designers, PMs, Frontenders
- Prototyping with AI, Design Ops, Developer handoff
- prototypes in the first 6 weeks, and growing
- 18
- deploy pipeline instead of 18
- 1
- faster handoff to developers
- 2x
Every company adopting AI runs into the same problem: prototypes get cheap to make, and just as easy to lose.
At Tactile I build prototypes on our real source code, so developers can ship them. They were becoming a team asset with nowhere to live, so I proposed one library and built it. Past work stays findable, PMs catch up in a minute, and developers start from the design, not from zero.
The problem
The question isn't how to make prototypes, it's where they live once the meeting is over.
The tool
It keeps itself up to date.
Every time a prototype changes, its ticket updates with the link on its own, and the diamond shows how far along it is.
Six Claude Code skills cover the whole loop, from setting up a laptop to building the design for real.
- Not started

- In progress

- Ready for review

“AI prototypes are worthless if they're not collected in one place.”
Key tradeoffs
Impact
What I learned
Nobody asked me to build this. People started using it because it made their work easier, not just mine.
AI prototypes are worthless if they're not collected in one place. Scattered, they lose their meaning.