What AI speeds up, and what it quietly takes away

I use AI tools every day, including to build this website. The speed still surprises me. A first draft of a page takes minutes. A working prototype of an idea can exist by the end of an afternoon. When I started as a software engineer in 2014, some of those things would have taken a small team several weeks.

I don't want to be the person who complains about a tool while using it constantly. The speed is real, and it's good. More people can build the thing they've been imagining, and far less time goes into the tedious parts of making software.

But I've noticed something that's easy to miss while everything is moving this fast.

Everything is starting to look the same

Look at a few dozen new product websites and they begin to blur together. The same layout, the same soft gradient behind the headline, the same three feature cards, and the same words: seamless, effortless, powerful, built for modern teams. Apps have the same onboarding screens. Posts have the same rhythm.

That isn't a coincidence. An AI model is very good at producing the most likely answer, and the most likely answer is, more or less, the average of everything it has seen. Ask it for a landing page and you get the typical landing page. Ask it to write about your product and you get the way products are usually written about. It's competent, and it sounds like nobody in particular.

I saw this firsthand on this site. The first version of the copy came back fluent and tidy, and it could have belonged to any studio in any city. It took several rounds of rewriting before it said what I actually wanted to say, in a way that sounded like me.

Where the heart of a product comes from

The things that make a product feel made by someone are rarely the things that take the most time. They're small decisions that come from knowing who it's for: a feature left out on purpose because the users don't need it, a sentence only this founder would write, a default that shows someone thought about how the product is really used, a detail that's slightly unexpected and exactly right.

Those decisions come from spending time with the people you're building for and having an opinion about what matters. AI can help you carry out an opinion very quickly. It can't form yours for you, because it doesn't know your customers, and it doesn't care how the thing turns out.

How I try to use it

The approach that works for me is to let AI do the parts that are genuinely routine, which is a lot of them, and spend the time it saves on the parts that aren't. The setup, the first draft and the repetitive code can go fast. The thinking about who the product is for, what to cut and how it should feel gets more time than it used to, not less.

That means reading everything it produces as a starting point rather than an answer, rewriting in my own words when the words matter, and noticing when a result feels generic, which usually means I haven't told it enough about the people I'm building for.

Speed isn't an advantage any more

When everyone can build quickly, being quick stops setting you apart. What's left is the part that was always hardest: understanding a problem properly, making choices someone could disagree with, and caring about the details a user will feel even if they never notice them.

If you're building something now, I'd use every bit of speed AI gives you. Then I'd put the time it saves into the part only you can do, because that's the part people will remember.