The New Skill in Software Development isn't Coding
This article was first published on LinkedIn and is archived here for reference
Something has shifted in how software gets built.
This week I shipped a web app to the Android and Apple app stores. Then I ported that same React app to native Swift and shipped to Apple TV.
It started as a joke. A throwaway prompt to see if AI could build a playable game. I didn't expect it to become anything — just a quick demo to satisfy my curiosity.
I've never built a mobile app before. Never used React. Never written a line of Swift. The tvOS port took 26 hours.
I'm not saying this to brag. I'm saying it because two years ago, this wouldn't have been worth the effort for a pet project and now it's a casual week.
What AI actually changes
Here's what I think people get wrong about AI and development: it's not replacing programmers. It's leverage.
The people who thrive are the ones who already know how to ship. Those who can spot when something's broken, articulate why, and know what "done" looks like.
I've been building software for over 25 years. Seen frameworks come and go, hype cycles rise and crash. And I can tell you: this moment is different. Not because AI writes code; but because it changes what the job actually is.
The real skill is requirements
The bottleneck with agentic coding isn't prompting. It's knowing what you want before you ask for it.
When I ported my game to Apple TV, I wasn't writing Swift. I was describing behavior: "This enemy moves in a sine wave pattern. This weapon fires in a spread of five projectiles. When the player dies, there should be a 500ms delay before the respawn menu appears."
That's not coding. That's product management.
The developers who will thrive in this environment are the ones who can write a clear PRD. Who can break a project into discrete, testable requirements. Who can look at something that technically works and say "no, that's not right" and explain exactly why.
These are skills most engineers never had to develop. The code was the hard part, so we optimized for writing code. Now the code is the easy part. Clear thinking is the bottleneck.
Developers need to become product managers
This is uncomfortable for a lot of engineers. We got into this field because we liked building things, not writing documents. But the game has changed.
If you can't articulate what you want, AI will give you something. It'll be functional. It might even be clever. But it won't be right and you won't be able to explain why.
This is where experience becomes a multiplier. Not experience writing code, but experience shipping products. Knowing what good feels like. Having strong opinions about quality.
The cost of generating code is dropping
It's not zero (yet) — my tvOS port cost around $100 in tokens. But compare that to hiring a Swift contractor or spending weeks learning a new language for a side project.
The economics have changed. Ideas that weren't worth pursuing suddenly are. Projects that would've died on the backlog are now shippable in a weekend.
This doesn't mean developers are obsolete. It means the value is shifting upstream. Toward clarity. Toward taste. Toward the ability to define what should exist and hold the quality bar until it does.
A junior developer with clear thinking and strong product instincts will outship a senior developer who can only execute specs written by others. That's a wild sentence to write, but I believe it's true.
Where this is heading
My guess? We're about 2-3 years away from this being normal. Not "AI-assisted development" as a novelty, but as the default way software gets built.
The teams that figure this out early will move at a different speed. They'll ship more, iterate faster, and explore ideas that used to be too expensive to try.
The individuals who figure this out will have outsized impact. One person with clear thinking, strong taste, and the ability to orchestrate AI will output what used to require a small team.
I don't know exactly what the industry looks like on the other side of this. But I know the skills that matter are shifting and most people haven't caught up yet.
The weird, exciting part
Here's the thing nobody talks about: this is genuinely fun.
Watching a project come together through pure orchestration — describing what you want, testing relentlessly, iterating until it feels right — is satisfying in a way I didn't expect.
We're in a weird, exciting moment where the bottleneck isn't what you know. It's how clearly you can think.
For those curious, I wrote up the full technical process in two parts if you want the details: first, From Prompt to App Store in 48 Hours, then From Web to Living Room: Porting to Apple TV.
But the meta-lesson is simpler: learn to write requirements. Develop taste. Get good at knowing what "done" looks like.
The code will write itself.