Building software with AI is addictive.
I don't mean that as a warning about shortcuts or tech debt. I don't mean it as a metaphor. I mean I cannot stop.
The Trade-Off Nobody Talks About
Here's the thing about moving up in management that nobody prepares you for: building software stops being your job. Slowly, then all at once. You're in meetings. You're planning. You're reviewing architecture, mentoring engineers, building a team. That's the job. That's what you're supposed to be doing.
But I love building software. I always have. It's why I got into this career in the first place. And for years I've watched that part of the work slowly disappear from my day, replaced by the organizational gravity that pulls every senior technical leader away from the keyboard.
You accumulate ideas. Side projects. Architectural experiments you want to prove out. Features you wish you had time to prototype. They sit in notebooks and notes apps, a growing pile of things that will probably never get built because the math doesn't work. There aren't enough nights and weekends to turn ideas into software when you're doing everything by hand.
Then the Math Changed
AI development tools gave me something back that I thought I'd lost to seniority. Now I can take an idea, plan the architecture, write tests, and actually ship something. Not in weekends and nights stretched over months. In a reasonable amount of time.
The work shifted in the best possible way. The planning, the architecture review, the test design, the code review—that's the bulk of what I do now. That's where the real thinking happens. The parts of building software that I was always best at are now the parts that matter most. The tedious middle—the boilerplate, the scaffolding, the repetitive implementation—that's what the AI handles.
Where I used to have a pile of half-baked ideas collecting dust, things are actually getting built. Features are shipping. Prototypes are becoming real systems. The backlog of "someday" ideas is shrinking for the first time in years.
It feels incredible.
The Problem Is the Rush
And that's the problem.
I used to leave my desk at a reasonable time because a few more hours of coding just meant incremental progress. You'd make a little headway on a feature, but the dopamine rush of completion was days or weeks away. The friction of manual development was, in retrospect, a natural stopping point. You hit a wall, you go home, you pick it up tomorrow.
That friction is gone.
Now I can see things come together quickly. I can watch an idea go from concept to working code in a single session. And right when I should be shutting down for the night, the thought hits: one more feature. One more idea I've been sitting on for months. I'm so close.
I'm at my computer more hours than I've ever been. More than when I was a junior developer pulling late nights to prove myself. More than during crunch time on critical projects. And this isn't crunch. Nobody is asking me to do this. There's no deadline driving it.
It's just that the completion loop is so tight now that stopping feels wrong. The gap between "I have an idea" and "I shipped it" collapsed, and my brain hasn't recalibrated.
What I'm Actually Saying
AI gave me back the part of the job I loved most. That's real. That's valuable. I don't want to frame this as purely negative because it isn't.
But the tool also removed the friction that used to force me to stop. And I'm being honest with myself that I haven't replaced it with anything yet. The natural governor—the fact that building was slow—kept me in balance. Now that governor is gone, and I'm still figuring out where the new boundaries are.
This isn't a productivity post. It's not a humble brag about how much I'm shipping. It's an honest observation from someone nine months into this transition who has realized that the conversation about AI and work-life balance isn't just about jobs being replaced. It's also about what happens when the work becomes so rewarding that you lose the signal to stop.
Is anyone else experiencing this? Have you found where the new boundaries are?
I'd genuinely like to know. Because right now, I'm still looking for the off switch.