The Cost of Speed

For most of my life, I’ve written software.

Mostly web, but not only. I’ve done the whole tour: DevOps, UI, UX, design, illustration, backend, frontend, Infra, ML. I’ve shipped products, fixed outages, and fought bugs that had no right to exist. I’ve jumped between languages, frameworks, skill sets, and roles, sometimes in the same week.

Twenty years of building in pure chaos—and still shipping.

And then I started using AI.

I don’t mean “I tried it once.” I mean I adopted it. It moved into my workflow the way a good editor moves into a writer’s life: quietly at first, then suddenly you wonder how you lived without it. I used to hate spending an entire day on Google and Stack Overflow just to track down one bug—one missing detail—one edge-case comment buried in a forum thread from 2014.

Now it’s fast. Almost unfairly fast.

You ask, you get an answer. You paste an error, you get a diagnosis. You sketch an idea, you get a scaffold. Momentum is no longer fragile. You stay in flow. You build.

So why does it still feel… expensive?

Because speed isn’t free. It never was.

The Costs

The obvious cost is dependency. If a tool is always there to suggest the next step, the temptation is to stop holding the full system in your head. You can still deliver results—sometimes better results—but your mind changes shape. You begin to trust the assistant’s confidence. You start to accept code you didn’t truly reason about, because it looks right and tests pass.

But the deeper cost is more personal: it changes how you learn.

Before, the pain of searching was part of the training. You’d dig through docs, compare answers, read arguments, and accidentally absorb context. You’d build a mental map of why things work, not just that they work. The process was slower, yes, but it left fingerprints on your understanding.

AI can skip that whole staircase.

And when you skip stairs long enough, your legs forget what they’re for.

Another cost is creativity. Not because the tool is uncreative—it can generate plenty of ideas—but because it nudges you toward the most likely answer. The safe pattern. The standard architecture. The “best practice” you didn’t ask for. Over time, you can start building what the internet agrees with instead of what your problem actually needs.

And then there’s the quiet cost: attention.

When solutions arrive instantly, patience shrinks. You get less comfortable sitting with uncertainty. You reach for an answer before you’ve fully formed the question. You ask the tool to think so you don’t have to. Not always—but often enough to notice.

Why I Still Use It

Still, I’m not going back.

I don’t miss the hours of low-quality searching. I don’t miss the performance of expertise: pretending I’m fine while my day disappears into ten open tabs and three contradictory answers. I like that I can stay in the editor. I like that debugging feels like a conversation instead of a scavenger hunt.

So I keep AI—but I set rules.

• If I can’t explain the code back to myself, I don’t ship it.
• If it suggests a “magic” fix, I ask for the why.
• I prefer small diffs over big rewrites.
• I force myself to read the docs sometimes, on purpose, even when the tool could summarize.
• And when something matters—security, data correctness, money—I slow down and verify like it’s 2009 again.

AI didn’t remove the craft. It changed where the craft lives.

Maybe the real question isn’t “What does this tool cost?” but “What do I want to protect?” My curiosity. My standards. My ability to reason. My patience. My taste.

Speed is incredible. I’m grateful for it.

I just don’t want to wake up one day and realize I traded away the part of the job that made me feel free.