The AI tools I actually use as a solo developer

There's no shortage of "AI tools for developers" posts right now. Most of them read like someone dumped a Product Hunt feed into a blender. This isn't that.

I run Pomona App Studio alone. Eight apps across education, finance, news, sports, books, and travel. Design, code, marketing, support -- all me. AI has changed what one person can ship, though the way it helps is more boring than the hype suggests.

What I use every day

Claude Code

This is the main one. I use it as my coding partner, and it's made me noticeably faster. The reason it works is project context. It knows my codebase, my SwiftUI patterns, my test setup. That matters when you're across eight apps and can't hold every detail in your head.

It's good at multi-file edits where a change in one service needs to ripple through view models, views, and tests. It writes tests that follow my existing conventions (Swift Testing, my fixtures, my mock patterns). And when I paste an error, I get a fix that accounts for my specific architecture instead of a generic Stack Overflow answer.

It's more like a fast collaborator who has read every file in the project than a junior developer I'm supervising.

Design iteration

I use image generation to explore directions fast. App icon concepts, App Store screenshot layouts, social media graphics. The output is almost never final, but it compresses brainstorming from days to hours. I used to agonize over icon variations. Now I generate twenty, react to them, refine the two that feel right.

App Store copy

Writing App Store descriptions is tedious and important. AI is good at generating keyword-dense copy that still reads naturally, and at producing variants I can A/B test. I also use it for keyword research -- it surfaces terms I wouldn't have thought of by analyzing competitor listings and review language.

Review analysis

With eight apps, reviews and support emails pile up. I use AI to categorize and summarize feedback across all my apps weekly. It catches patterns I'd miss reading them one at a time. Like "four users this month said the onboarding flow is confusing." Easy to miss at small scale, but it adds up.

What I don't use AI for

Architecture decisions. AI is good at implementing patterns but mediocre at choosing between them. When I'm deciding whether to use SwiftData or a custom persistence layer, I need to weigh tradeoffs that require experience. AI gives confident answers to architecture questions. They're often subtly wrong in ways that show up weeks later.

Final UX judgment. AI can generate layouts and suggest flows, but it can't tell me whether something feels right. That instinct comes from years of building and using apps.

I also won't use it to learn new frameworks. If I'm picking up something unfamiliar, I work through it the slow way. AI speeds up what you already understand. It doesn't replace understanding.

The economics

About $200/month on AI tools. Less than a single hour of contract iOS dev work. If they save me five hours a month -- and they save more than that -- the math is obvious. Best return on any expense I have besides my Apple Developer account.

Hype vs. what I've seen

AI has made me roughly 30-40% more productive on the tasks where it helps. That's real. It is not the 10x that people claim on Twitter. The bottleneck in shipping good software has never been typing speed. It's knowing what to build and having the taste to build it well. AI doesn't help with either of those.

What it does is shrink the grunt work that used to make solo studios impractical. Boilerplate, test coverage, marketing assets, feedback analysis. That overhead is way lower now, and it changes what one person can realistically ship.

-- Daniel