How I Work

A year ago, my workflow looked like every other designer's: Figma files, handoff specs, waiting for eng capacity. I'd design something, explain it, wait weeks, then tweak it based on what actually got built. The feedback loop was measured in sprints.

Now I ship features in ten minutes.

The change wasn't learning to code — it was learning to direct AI agents. I write a GitHub issue describing what I want. I open a workspace in Conductor, paste the link, and talk through the task using SuperWhisper. The agent creates a feature branch, writes the code, and spins up a preview. I check it, give feedback, iterate. When it looks right, I merge.

The key insight is parallelism. While one agent builds a contact page, another is working on the CV page, and a third is experimenting with a mobile feed layout. Each workspace is isolated — its own branch, its own preview URL. I can context-switch between them, review what's ready, send back what needs work.

This changes what it means to design. I don't hand off static mockups anymore. I iterate on the real thing, in the browser, with actual data. When I don't like something, I describe the change and watch it happen. The gap between imagination and implementation has collapsed.

There's a design philosophy embedded in this workflow too. I prefer iteration over one-shotting. It's easier to see something, decide what I don't like, and refine — rather than trying to get it perfect upfront. The speed of this loop makes that practical in a way it never was before.

I'm still a designer. I still care about typography, hierarchy, interaction feel. But now I have leverage. The bottleneck isn't implementation capacity — it's my own clarity about what I want to build.