I went to Korea before I turned 30 and came back with half a suitcase of skincare. I went to Olive Young three times. Approximately 300 face masks were brought back. (I have a problem… lol).
So I decided to build a tracker app, partly practical, mostly just a good excuse to test AI design tools I’d been meaning to touch. Here’s what I tried.
A skincare tracker: inventory management, routine logging, a thing I can pull up when I’m standing in a store about to buy my ninth snail mucin product. Simple concept.
But as a marketer, I couldn’t let it look bad. Even if it’s just for my friends and me on face mask nights. Especially then, actually.
Variant is an AI design tool that lets you describe what you want and generates UI options instantly. No account is required to get started. The search bar is also right there, which makes it easy to try, a little like the LLMs.
I typed “skincare app” and went broad. Watched what came back. Then found the green one, but wanted it with a wine-red accent and knew immediately it was right. I couldn’t fully explain why. I just reacted.
Variant gets this right: it generates fast enough that you stop trying to invent from scratch and start reacting instead. Sometimes you need to see fifteen wrong options to find the one that works. Variant makes that fast.
With pointed direction, it also brings the vision you had in your head the fastest.
I took a screenshot of the design I liked. That became my reference for everything else.
v0 is more design-oriented than Lovable, which matters to me. But when I fed it my Variant screenshot as a reference, it didn’t quite land. Technically correct — all the right pieces — but missing whatever energy the original had.
This is the honest version of working with AI tools that nobody really talks about: the gap between “good output” and “the thing you had in your head” is still a you problem. The tools execute. You still have to know what you want.
I had credits to test building something, and the design on Anything came out better than v0, if I’m being honest. But I couldn’t get it to render on mobile, which is a dealbreaker for an app I want to use on my phone in a store. Testing this weekend. Will report back next week.
I didn’t get to talk about this, but pencil and paper are also great developer-focused design tools that you can connect Claude to via MCP. You can also import code from Anything or v0, continue iterating here, and then transfer it to Claude, or vice versa.
I am speaking at Women in Tech SEO in May. Used Claude to work out the hook and problem framing before touching any slides, then built the deck in Gamma. Claude wanted Canva. I said Gamma. We moved on.
The Gamma agent also helped me edit my slides. You can talk to it inside the editor to make changes, which makes the whole thing feel more like collaboration than production.
I am excited to try out Lovart, Flora, & Noon (when it drops! It’s more product design, and Laura, their head of marketing, is the coolest). I tried Pomelli (by Google Labs) not too long ago, and it’s great for B2C creatives!
The faster these tools generate, the more these things become the work:
Taste & intuition
Patience
Perseverance
Care
You’re not blocked by execution anymore. You’re just making calls, over and over, about what’s right and what isn’t. A muscle that matters more now.
Still no finished app. Still mid-process. That’s kind of the point of this series.
Subscribe if you want this in your inbox every week — honest, mid-process, no clean answers. Or reply and tell me what you’re building. Going to be doing this every Thursday!
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.