When I say AI-native, I mean I use AI as a force multiplier to get from concept to deployment.
It starts with me. I brainstorm, work out the concept, do the research. Once I know what I'm building, I use AI to scope it and build it. I also use it as a backup brain. If I have an idea I'll run it by AI first, see if somebody already made it, see if it holds up before I spend real time on it.
It isn't one thing at a time either. Some days I'm running QA on one project while another chat builds something else I asked for. I've got pipelines set up so I'm not clicking around GitHub to push an update. Some days that's like being three or four people on the build.
What the decade buys me
The eye.
AI writes code that works. What it doesn't reliably do is make something that looks right or reads right.
On GamerStats I'd ask for a dashboard revision and it would leave dead space on the right, or put two items in a three column row and drop to the next line instead of filling the grid out. It would set type too close to the progress rings with no room to breathe. Sometimes it gets the order of the information wrong, which matters more than the spacing does.
On a project I'm building now, it made the divider between sections look almost identical to the borders on an FAQ where each row opens a dropdown. Two different things ended up looking the same. The code was fine. The design was wrong, because now you can't tell what you're supposed to click.
None of that is a bug. It all runs. You have to know what good looks like to catch it.
Where it falls down
I never thought I'd be able to build an iOS app, and I did. So I haven't really hit a wall.
Where it goes wrong is different. It's confidently wrong.
On IsoRide there was a stretch where the app told you rain was coming on a clear day. Nothing crashed, nothing looked broken. The logic was just wrong, and it was sure about it. I caught that one myself. Sometimes I catch things before a release, sometimes after it's already out, and then I fix it fast.
That's the real limit. AI builds the thing fine. It'll also build the wrong thing and sound confident, and the only thing between that and somebody using it is whether I catch it. So I run a full QA pass before anything ships, and I go looking for problems.
If the tools went away tomorrow, some of this would break. I probably wouldn't be building an iOS app. I'd still design and build a website, because that part was mine before any of this. What I'd lose is the bigger features and the speed.
The part AI can't do
I've built the last few years mostly alone, and I don't want to keep doing it that way.
I work better bouncing ideas off somebody, even when the idea starts out small or stupid sounding. Somebody sees the thing I didn't see. I end up presenting it better because I had to say it out loud to a person first.
That's the part AI can't do for me. It can tell me how it thinks a person might react to something. That's not the same as a person reacting. And that gap is most of the job when you're designing for people.
So that's what I'm looking for. A team to bring it to.