I'm an artist, not a programmer. I build wallpaper apps and wellness apps as a solo developer, but my relationship with code is more like: code is the medium I use to make the thing in my head real. Writing code isn't the goal. The goal is the thing.
I started using Antigravity not to become a better programmer, but because I wanted to actually build what I could imagine. After a year of it, I've noticed some things that don't show up in the typical developer reviews — because they're coming from a fundamentally different starting point.
Where Artists Have an Unexpected Advantage
"Here's how it should look" is a surprisingly powerful prompt
Most developer tutorials teach you to specify what to implement. My instinct is always to describe how it should feel. And Antigravity turns out to handle that kind of instruction remarkably well.
Main color is a deep navy. Card corners should be rounded. I want a subtle press-down
feedback when tapped. As the user scrolls down, I'd like the background to gradually
lighten — a slow gradient shift, not abrupt.
That's not a technical spec. It's a mood description. Antigravity generates something that's maybe 80% of what I imagined. Not perfect — but close enough to react to. From there it's a cycle of "make the radius slightly larger," "slow down the gradient," "the shadow feels too heavy" — and that back-and-forth rhythm feels completely natural to me. It's how I work on a painting.
The aesthetic judgment that programmers sometimes skip
When Antigravity generates a UI, I can look at it and immediately feel whether it's right — not whether it works, but whether it belongs. That instinct is harder to train and easier for artists to just have.
Programmers often accept working code as correct. I accept working code as a draft. The difference in how we relate to an AI-generated first pass means I'm comfortable rejecting more, iterating faster, and being specific about what's wrong on a sensory level rather than a logical one.
This, paired with Antigravity's ability to execute, is a genuinely useful combination. The AI implements; the human judges aesthetics. The division of labor is natural.
What Was Actually Hard
The terror of error messages I couldn't read
For the first six months, a red terminal error would stop me completely. "I don't know what this means" is a helpless feeling, and asking the AI to fix it sometimes kicked off a chain reaction — fix one error, surface another, fix that one, find a third. I was scared of that chain.
The shift came when I learned to just paste the entire error message and ask: "What's happening here, why is it happening, and what should I do?" Asking for the why alongside the fix made the errors feel less like walls and more like information. I still don't understand everything happening in my codebase. But I no longer panic when something breaks.
Outsourcing architectural decisions leads to trouble later
Early on, I delegated everything — including structure. "What's a good way to organize this?" I'd ask, and build exactly what Antigravity suggested. Efficient in the moment. Painful later.
When I didn't understand why the structure was what it was, I couldn't give the AI useful context in future sessions. "How does this file relate to that file?" — if I don't know the answer, I can't frame my request accurately, and the AI generates something that conflicts with decisions it made two weeks ago.
Now, whenever Antigravity proposes a structure or approach, I ask one follow-up before proceeding: "Why this way?" I read the explanation. Then I decide whether I actually understand and agree with it. This one extra step has changed how much I feel in control of my own projects.
Why I Think Artists Fit This Tool Particularly Well
If programmers relate to coding as problem-solving, artists tend to relate to it as a means of expression. That difference shapes how you use an AI coding assistant.
Because expression is the goal, I'm comfortable with "good enough for the experience" — a working, beautiful thing at 80% beats a perfect implementation that isn't finished. Antigravity's speed and fluency make that tradeoff work. I ship things. They feel intentional.
The risk, which I have to actively manage, is accumulating code I don't understand. AI-written code that works but that I can't reason about is a liability that grows silently. The more I treat Antigravity as "magic code generation" rather than a collaborator I'm keeping pace with, the worse this gets.
I don't have a complete answer to that yet. The balance between going fast and staying grounded is something I'm still figuring out. But a year in, I'm more convinced than ever that you don't need to be a programmer to build things with a tool like this. What you need is clarity about what you're trying to make — and that, artists usually have.
The code is just how you get there.