Inside My AI Prototyping Environment
A video tour of the AI-assisted prototyping tool I built at Sublime Security.
Welcome to Unknown Arts — I’m Patrick, your field guide to tech’s creative frontier. Join thousands of builders around the world navigating what’s next.
Last week, I shared a brief introduction to the AI-assisted prototyping environment I’ve been building over the past several months at Sublime Security.
This week, I get to show you how it works.
I joined Michael Riddering on the Dive Club podcast to give a tour of the system and talk through the ideas behind it.
The starting problem was simple. I could make fast prototypes with Claude, but they were disconnected from the product, had no shared home, and were difficult for the team to build on over time. Prototyping directly in the production application created the opposite problem: too much complexity and not enough room to explore.
I built Design Studio as a useful middle ground.
The workflow I demo in the show starts with “blueprints”: reference screens based on Sublime’s real product. From there, an agent can set up a new prototype and handle the execution, while a designer supervises and guides its work from a canvas overview.
We also talk about how I use the environment day to day: exploring low-fidelity directions, guiding the agent as the work evolves, and creating guardrails that help the agent and other designers work more consistently and effectively. We get into collaboration and handoff, too, including how I bring relevant patterns from our production code into prototypes and how our brand team uses Design Studio to build internal tools.
There was more to share than we could fit into one conversation, so I’ll be unpacking the ideas and systems behind Design Studio in the weeks ahead — here in the newsletter, and on X and LinkedIn.
For now, watch the full conversation on YouTube to see it in action!
Until next time,
Patrick
Find this useful? Share it with someone who might also get value.
Not subscribed yet? Get my writing directly — 7k+ already do.





