I've been around long enough to watch a few waves hit – mobile, apps, cloud, and now AI. I can say without hesitation: this one is different.

AI's ceiling feels genuinely open. The only bottleneck I keep running into is myself.

I wanted to write down how I actually use it, day to day. This post might change in a few weeks given how fast AI is moving, but I'll keep updating it. I'll be adding demos too.

Research and understanding concepts

I use AI to understand things fast. New concept, unfamiliar territory – I don't go to Google first anymore. I use AI to orient myself, then go deeper.

If I don't know a topic well, I can't easily tell when the answer is wrong. I feel infallible, but I might not be. It's easier to steer AI when you already know enough to push back. Something to stay honest about.

Pseudo prototypes

I use AI to make ideas tangible. Ideas move faster when you can see them, even roughly. You learn more from a rough version of something than from a detailed plan for it. AI gets me to rougher faster.

Internal tools and skills

I built an internal tool recently: a prototype playground. Designers can share their prototypes with the whole design team. This is like what Brian Lovin and Shopify created with Artifact. Something I would have had to beg an engineer friend for a year ago, so that has changed.

I also build skills used by our designers and PMs. They’re simple, but they get the work done. For example, I built a content review skill for Claude Code that automatically compares the copy in a design against our official content guidelines.

Shipping code

The newer thing is that I've started shipping code. I'm a designer with a developer background – but I haven't touched code in years. I understand the basics, and I'd always imagined pushing a PR on GitHub at work. Now I actually do.

These aren't full workflow PRs. They're small things – quality and craft improvements that usually get ignored at enterprises, like adding smooth transitions to side drawers and modals.

Thoughts

Most of my problem-solving still happens on a whiteboard. Citrix is complex – it requires rigorous human thinking before anything else. And yes, I still use Figma. I think it is unfairly questioned right now.

For me, Figma is still the best place to explore. It helps me figure out what I want to solve before I write a prompt. AI is where I start to build it. They're different stages, not competing tools. That's my take, anyway.

I've always had more ideas than I can hold. The problem was always execution. I could do the engineering myself – but it was slow. And for anything bigger, I had to rely on engineer friends, which meant waiting for their time.

The gap between an idea and a version of it is small now. I'm using AI to explore ideas, build things (check out tinkering), and develop skills I didn't have before. I'm having fun and I'll keep going.

AI is excellent for solo work, but solo isn't where the best ideas come from. Great work often includes friction. This means someone questions your ideas. They point out what you might miss and help improve your thinking before you choose a direction.

Right now, AI agrees too much. I'd love to see that change.

Until then, I have more ideas than I have time. The only things stopping me are time and sleep.

Glad that LLMs don't have either. :)