romano.io
All posts
AIDeveloper Productivity.NETEngineering LeadershipAgentic Development

I Haven't Typed Code in Months. I've Never Shipped More.

A .NET architect's honest look at what happened when AI replaced the keyboard — and why the skills that made you senior five years ago are depreciating fast.

Doug Romano··5 min read

Last Tuesday, a junior dev on my team asked me to pair on a bug. I said sure, pulled up the project, and realized something embarrassing: I couldn't remember the keyboard shortcut to set a breakpoint.

Not because I'd forgotten how debugging works. Because I hadn't opened an IDE in over two weeks. The bug was already fixed by the time I found the shortcut—Claude Code had diagnosed it, written the patch, and generated the regression test while I was still fumbling with my editor.

That moment crystalized something I'd been feeling for a while: the job I trained for no longer exists. And what replaced it is so much better.

Three Eras in One Career

I started building software in the VB6 days. You dragged a button onto a form, double-clicked it, typed some code, hit F5, and watched it run. It felt like magic. It also felt like the beginning of something—like software development was on an irreversible march toward easier.

It wasn't. WinForms gave way to web apps. Web apps brought build chains, package managers, and YAML files that nobody fully understood. The GUI designers disappeared. Terminals came back, ironically, as the primary interface for building graphical interfaces. We were writing more code, in more languages, across more files, to accomplish what used to take a drag and a drop.

For two decades, the complexity only went up. We just got faster at managing it. Better IDE features, better debugging tools, better muscle memory. I could navigate a 200-file C# solution blindfolded. I was proud of that.

Then the third era arrived and made all of that irrelevant.

What My Days Actually Look Like Now

I run five or six terminal sessions simultaneously. Each one has Claude Code working on a different piece of a feature—one handling the API layer, another writing the EF Core migrations, another building out the Blazor components, another writing xUnit tests against what the first three are producing.

My role in this isn't typing. It's thinking.

I describe what I want at the architecture level. I sketch the data flow on a whiteboard, snap a photo, and feed it in. I write design decisions in plain English. I review what comes back—not line by line like a code review, but structurally. Does this solve the right problem? Does the abstraction hold? Will this be maintainable in six months when I'm not thinking about it?

Then I iterate. Tighten the prompt. Reject a pattern. Ask for an alternative. It's a conversation, not a dictation.

The strange part: I'm directly responsible for more shipped code per month than at any point in my twenty-plus-year career. And the quality is higher, because I'm spending my time on the decisions that actually matter instead of wrestling with syntax.

The Uncomfortable Truth

Here's what nobody says out loud at conferences: the skills that made you a senior developer five years ago are depreciating fast. Speed in a code editor. Memorized API surfaces. The ability to hold a mental model of a complex codebase in your head and step through it with a debugger.

Those aren't worthless—they're context, and context still matters. But they're no longer the bottleneck. The bottleneck now is the ability to clearly articulate intent, to evaluate architectural tradeoffs quickly, and to know what good looks like without needing to have typed every line yourself.

I can spin up Azure infrastructure, deploy a .NET app, and have it running in a staging environment in minutes. I remember when that sentence would have described a six-month project involving purchase orders, hardware procurement, and a guy named Steve from IT. The entire chain from idea to running software has compressed by orders of magnitude, and AI is the final multiplier that made it all collapse.

Small Teams, Outsized Impact

What excites me most isn't personal productivity—it's what this means for small teams. A three-person squad with Claude Code, Cursor, and a clear architectural vision can deliver what used to require ten or fifteen people. Not by cutting corners, but by eliminating the mechanical overhead that used to consume 70% of everyone's time.

The documentation gets written—because AI writes it. The tests get written—because AI writes them. The boilerplate gets handled. The edge cases get considered. And the humans on the team get to spend their energy on the hard problems that actually require human judgment.

The velocity is something I genuinely did not think was possible two years ago.

The Widening Gap

If you're leading a technical team and you haven't fundamentally rethought how your people build software in the last twelve months, I'd argue you're already behind. Not a little behind—a lot behind. The gap between teams who've embraced this shift and those still operating like it's 2023 is growing faster than most leaders realize.

Gone are the days of breakpoints. I don't miss them.