I'm More Productive Than Ever. So Why Am I Anxious?

I ship more features in a week now than I used to in a month. I’ve built more side projects in the last six months than in the previous three years combined. Last month I described a webapp to an AI (3D visualization, complex menu logic, session management) and had a working prototype before lunch.

And I’m not sleeping well.

The Skeptic

At first I was skeptical that AI tools, specifically code generation, were merely hyped up, and this seemed to be backed up by the wild valuations of the frontier model providers. I’d spent years treating coding as a craft. Something you get better at slowly, through repetition and bad decisions and the scar tissue of production incidents. The idea that a language model could participate in that process felt premature, maybe insulting.

The early tools didn’t help their case. Autocomplete on steroids, hallucinated APIs, code that looked right until you ran it. Every bad suggestion reinforced what I already believed: these tools are toys for demos, not instruments for real work.

I didn’t question that belief for a long time.

The Turn

Then the tools got much better.

I don’t remember when AI-assisted coding became something I relied on. It felt more like a slow concession. I’d ask it to scaffold a REST API in a framework I barely knew, then to debug something I’d been staring at too long.

I dropped my skepticism. The output was undeniable. I finished work faster, explored ideas I would have dismissed as too time-consuming. I enjoyed development again. The years of boilerplate and meeting-heavy sprints had worn something down, and these tools gave some of it back.

And then I got anxious.

The Anxiety

Becoming more productive feels like something is expected of you now.

When you can build faster, the voice in your head says “you should be building something important right now, before the window closes.”

Before what window closes? I’m not sure.

I’m six years into my career. Experienced enough to build real systems, to know when something smells wrong in an architecture. But I’m not yet at a level where my technical direction carries its own weight, the kind of gravity that comes with a decade-plus of accumulated reputation.

I’m in the dangerous middle. And AI compressed the timeline on everything.

So I feel this compulsive urge to build and ship and carve out a niche. But I don’t know where to direct it. I open my laptop on a Saturday morning with all this creative energy, open a Claude Code session, stare at the prompt, alt-tab to Twitter, scroll through other people shipping things, feel worse, go back to the terminal, type something, delete it. I could build anything. I should build something. I don’t know what.

The Dopamine Problem

For most of my career, I’ve wired my sense of professional identity to one thing: solving hard problems. Tracing a bug through four layers of abstraction. Finding the solution that’s correct and clean. That’s what satisfies me.

LLMs broke that reward loop.

When I pair with a model and it generates a solution to something I would have spent an hour puzzling through, I get the outcome (the problem is solved, the feature ships) but the hit is different. Muted, like eating a meal someone else cooked.

Recently I started building database systems, something outside my core expertise, because AI made it feasible. A year ago I wouldn’t have attempted it. Now I can move through unfamiliar territory at a pace that would have been impossible alone. I’m more capable than I’ve ever been, and I’m less sure than ever which parts of that capability are mine.

The Whiplash

AI broke my coping mechanism for imposter syndrome.

It used to be a steady hum: “they’re going to figure out I don’t know what I’m doing.” You learn to manage that. It fades as evidence accumulates.

Now it contradicts itself within the same hour. I fed a model a bare minimum bug report (hardly any reproduction steps, a stack trace and some context) and it identified the root cause of a bug my team had been circling for days. My stomach dropped. Why do they need me?

Then, twenty minutes later, the same model hallucinated a function signature that doesn’t exist. Recommended an approach that would have fallen apart at scale. And I’m the expert again, the one who catches the mistake, who knows what the system does.

One hour I’m obsolete. The next I’m essential. Back and forth, all day, and it never settles into something I can get used to.

My imposter syndrome hasn’t gotten better or worse. It’s gotten weird. It oscillates at a frequency I can’t keep up with, and the contradictions don’t cancel out. They compound.

Making Sense of the Contradictions

I’ve spent months trying to resolve these feelings into something coherent, and they don’t resolve.

Some mornings I’m excited about what these tools make possible. The same afternoon I’ll feel a real grief for the parts of the craft they’re replacing. I’ll know my judgment matters, that I bring something the model can’t. Then I’ll watch it solve a problem that would have taken me a day, and feel that knowledge wobble underneath me. All of it in the same day. Sometimes in the same hour.

I think the anxiety is a signal that you’re paying attention. If your career is built on intellectual problem-solving and a new tool changes how that problem-solving works, you should feel something. I’d worry more about the developers who don’t.

What matters is what you do with it.

Choosing Forward

This is a working position, not a final answer.

I’ve chosen to embrace these tools. Ignoring them would be the same mistake as ignoring containerization in 2014 or CI/CD in 2018. The tools change. You either change with them or you spend a lot of energy convincing yourself this time is different. Though I’ll admit the comparison only goes so far. Docker didn’t threaten to do the thinking for you.

But I’m trying to be deliberate about how I use them. Scaffolding, exploration, first drafts. That’s where they make me faster without making me dumber. I pull back when I catch myself accepting generated code without understanding why it works. I keep misjudging the line between leverage and dependency, and I don’t always notice when I’ve crossed it.

There’s a version of software development, the version I fell in love with, that’s changing. I’m still figuring out what the new version looks like, and whether I can love that too.

The Uncomfortable Optimism

I don’t have a neat bow to tie on this. I’m still anxious. I still open my laptop on Saturdays and stare at the cursor.

But I keep going back to the terminal. I keep building things I couldn’t have built alone. Some days the anxiety wins and I close the laptop. Some days I ship something real and forget about the anxiety entirely until I’m trying to fall asleep.

I don’t think the anxiety goes away. I think it becomes the background noise of working in a profession where the rules change in real time.

If you’re a developer feeling something like this, I don’t have answers. But I wanted to write this down because I think a lot of us are carrying some version of it around and not saying anything.