Table of Contents
I’ve been programming for over 20 years, and I have a project graveyard with more than 100 incomplete applications.
If you’re a developer, this probably sounds familiar: I get an idea, name the project (the most important part, of course), tackle the hardest architectural challenge to prove I can do it, and then… I lose interest. Building the login screen, building a mobile-responsive design, writing a mountain of automated tests? Those tasks never excited me enough in a personal project to overcome the effort required.
Many developers are drawn to the challenging work: architecture, software design, the theoretical and academic aspects of building systems. The rote stuff (CRUD operations, styling, matching designs exactly) just isn’t the source of our passion.
Generative AI has changed my relationship with side projects entirely.
What I Actually Get to Do Now
With tools like Claude Code, I can focus on ideation, design, and iteration. I speak concepts into being. Instead of spending days typing out syntax, I spend minutes reviewing code quality checks and automated test results. I use the software. I verify the work meets my specification and move on.
The valuable part of creating new software has always been validating an idea. Solving a business challenge, proving out a theoretical concept, or building something meaningful that didn’t exist before.
Now I can do those things without getting discouraged by the parts I care less about.
Why I’m Not Worried
Some people worry that AI will replace software developers. I don’t share that concern, because my value has not been cemented in typing syntax for a long time. That’s a secondary task. It consumed significant time, sure, but it’s not the most valuable use of time.
The value is in design, features, architecture, the things that enable people to do things. Code is a vehicle for that, but the value isn’t in knowing how to assemble the vehicle. It’s in producing a quality, usable vehicle.
I treat AI-generated code the same way I’d treat a teammate’s code. If it meets the specification, if it’s stable and maintainable, if we have good quality checks and they all pass, then I don’t need to criticize every line. That’s how it should be regardless of who or what produced the code.
A Caveat for Early-Career Developers
I do think it’s important to build foundational skills. New developers need to understand what it’s like to build common features using best practices, to follow code style guidelines, to debug and struggle through problems.
When you’re just starting out, those aspects of programming are interesting and challenging. That knowledge is background education that prepares you for higher-level thinking about software. Once you have that foundation, you now have tools in generative AI to handle the parts that are no longer the most interesting or challenging to you.
Innovation Is Still Human
Generative AI code is inherently derivative. It’s trained on existing codebases and produces code like existing codebases. True innovation remains a human activity.
We’re moving past memorizing syntax, debugging simple errors, and fighting with compilers. Those problems can be solved by machines now. Truly meaningful and impactful work are what humans get to focus on: designing meaningful and comprehensible systems, figuring out how to deliver concepts that haven’t been created before.
For me personally, someone who’s been on a management track for most of my career, this shift has been entirely positive. The sleepless nights debugging problems were never where I derived joy (though winning these battles did leave me feeling accomplished).
Producing something interesting, meaningful, or valuable? That was always the point. And that’s the part I get to keep doing, no matter how good AI-generated code becomes.