I've been using agentic coding tools for months now, and honestly, I enjoy it. The speed is real. I can explore ideas in minutes that used to take hours.

But something's been nagging at me. Now that the agent handles so much, I'm noticing what gets squeezed out: the collaboration, the depth, the care for invisible details that make something actually good.

The work that doesn't show up

I've always believed that in the world of frontend development, my best stuff is the stuff nobody notices. A smooth interaction. A subtle animation that guides the eye. A button that does exactly what you expect. Spacing that just feels right. People don't consciously see these things, but they feel them. It's what separates "this works" from "this is a joy to use."

Now that AI can scaffold interfaces in seconds, there's a quiet pressure to skip all that. The generated code works. The time spent on polish becomes the expensive part. And when the fast shipping speed becomes the expectation, that expense gets harder to justify.

The question shifts from "how do we make this feel right?" to "does it work?" Accessibility - which requires slowing down to really sit with someone else's friction - starts to feel like a luxury. But I got into this work for the invisible stuff. There's a specific joy in building interfaces that delight people even when they can't say why.

When I'm building one thing at a time, iterating, refining an interaction, I have space to focus on polish. But when the agent can scaffold five features in the time it used to take me to build one, the instinct is to keep moving. Get to the next thing. Keep the momentum. Each thing gets good enough. Nothing gets great.

The cost of working alone

There's a loneliness to this that I didn't expect.

Engineering always felt social to me, even coding alone. I'd walk over to a teammate's desk to whiteboard something. We'd go back and forth in a thread about the right approach. Or just jump on a call and talk it through. There was a shared friction in the not-knowing.

Now when I hit a wall, I ask the chat box and the answer comes back near-instantly. The friction is gone - but so is the reason to reach out. We're all in our own corners, tweaking our agents. Shipping more, talking less.

The "aha!" moments used to happen in conversation. We'd be bouncing ideas, pushing back, building on each other's thinking until something clicks. Now, I can just get the answer. Fast, sure, but not as rewarding as building that understanding together.

The why behind the how

What worries me most is the bigger picture stuff.

The real work has always been the thinking: understanding why a system was built a certain way, spotting the pitfalls in a product idea before it becomes a roadmap nightmare. That requires depth. Sitting with complexity instead of moving past it.

AI is incredible at the how, but less so at the why. It can give me a hundred lines of code, but it can't tell me we shouldn't build the feature in the first place because it doesn't serve our users.

That deeper thinking is where I actually add value, not just in code, but in the perspective I bring to product decisions. I can spot when something won't work, or suggest a better path. If I stop sitting with problems, I may lose that perspective and while we ship faster, we may ship the wrong things.

The deliberate choice

I'm not sure where it's all going - all I know is that the landscape has changed.

I know I need to push back on speed when it means cutting corners. Protect the space for the invisible work - the stuff that makes the difference between functional and good.

AI hasn't taken away my ability to think deeply or care about craft. In fact, it has made those choices more deliberate and demonstrated their importance more than ever. Maybe it is actually a good thing.