Caught! is a small app I've been building for Pokemon card collectors, specifically for people who are master setting, which means chasing every card ever printed for a single Pokemon. The moment the app turns on is marking a card as collected. You found it, you paid too much for it, it's yours, and the app should make you feel something about that.
On my desktop, in the simulator, it looked great. The timing read fine and the celebration did what I wanted it to do.
Then I built it to my phone, held it, tapped the button, and it felt cheap.
Nothing on the screen had changed. What was missing was haptics, which you can't evaluate anywhere except on a device in your hand, and timing, which I had tuned against how it looked instead of how it felt. There was a third thing I hadn't considered even slightly on desktop, which is that this is an action people take back to back to back. Especially at the beginning, when you're loading in the collection you already own. A celebration that feels great once can feel like an obstacle the eleventh time. I didn't notice any of it until I was sitting on the couch with my thumb, forty cards deep into my Snorlax collection.
So I fixed it there. Tightened the timing, brought in haptics, kept running it back until the fortieth one still felt okay.
I was on the Near Future podcast a couple of weeks ago with Tom and Jonny, and at some point I told them I rarely open Figma anymore. That's true, and it also sounds more dramatic than it is.
Jonny pushed back, gently. He'd expected AI design tools to feel the way writing code felt to him years ago: about as slow, better at the end, and slow enough that you still got to try things. What he got instead was output arriving so fast he couldn't be bothered to change it. Not how he would have done it, but fine, and there was always a next thing waiting. So he went back to Figma on purpose.
I went back to Figma so I could sit with a button for a bit.
He wanted to see what different inner shadows did to it. Just for a hot second, as he put it.
I've thought about that a lot since, and he's probably getting something I'm not. Those are subtle calls and I don't make all of them anymore. I've made my peace with that for most of what I design, partly because I think most of them are invisible to a user and immediately obvious to another designer. Whether that trade is worth taking depends on what you're building and who it's for.
But I said something on that podcast that isn't quite right, which is that I'd stopped doing the last five percent. I still do it, it's just somewhere else now.
It used to happen in a canvas at 400 percent zoom with a shadow panel open. Now it happens on a phone, in my hand, on the fortieth Pokemon card, and the things I'm fussing over are different: how long a celebration runs, how heavy the haptic is, whether an animation is still charming the third time you trigger it. The work is as fussy as it ever was. I can't see any of it until the thing is running.
What I do instead of opening Figma is crude enough to be embarrassing. When I want to block something out, I drop colored rectangles into a frame. Nav here. Text here. This one is big, this one is small. Calling them wireframes would be an insult to wireframes. Then I export the whole thing as an image, drop it into my IDE, and ask Claude to tell me what it thinks it's looking at, which is how I find out whether the shape in my head survived contact with the blocks I drew.
Then I build it, put it on a phone, and start feeling around for what's wrong.
The other thing pulling me away from the canvas is that a lot of what I'm designing now won't sit still long enough to be drawn. At Klay I've been working on a feed. Klay is a music app that lets you hear songs you love in new ways, and the feed is where you see what other people have made with it, so the whole premise is that it moves with you. It reads the playlists you imported, what you've listened to in the app, what you reacted to, what you've mixed before. I could draw you one version of that feed, for one person, on one afternoon, and learn close to nothing from it.
Where Figma still wins outright is brand and identity. I've tried to move that work and I can't, and I haven't seen anyone else move it in a way I'd want to copy. The place AI has earned a spot for me there is naming. I have a skill that generates candidates, digs up the visual ideas a name could turn into, and checks whether the one I like is already owned by someone, or reads badly in some way I'm too close to see. The mark itself is still me, in Figma, for a long time.
Working this way has a cost. Shipping at good-enough has bitten me, though rarely with users, at least in my experience. It's teammates, clients, and founders. When someone has put their money and their nights into a product, they see it the way you'd see a kid, and "let's ship it and find out" can land like you don't care about it. I do care about it. There's a real tension between wanting to learn something this week and wanting it right before anyone looks at it, and every designer and every founder I've worked with sits somewhere on that line. Sometimes I'm the one arguing for another week.
I guess this is all to say that "I stopped using Figma" is a bad description of what happened to me, and I'm the one who said it on a podcast, so this is me taking it back. The fussing is all still there. I can't get at it from a desk anymore, so I go find out on the couch, forty cards deep.