AI makes us faster. It doesn't make us better.

Do you have a website or an app that was built in a hurry, and you don't know what is behind it? Book a free call: in 30 minutes you leave with a plan and three quick wins.

Does AI make developers faster?

Probably yes, with today's tools, but nobody has measured well yet by how much.

METR, a research institute, published in February 2026 the results of a study with 57 developers and over 800 tasks, on their own projects. The researchers believe developers are probably faster with AI than without it. They also say the data is very weak evidence for how big the time saving is.

One detail says a lot: more and more developers refused to take part in the study, because they no longer wanted to work without AI.

I take one thing from this. Speed is the part that is easy to see. Quality shows later, and it gets talked about too little.

Does quality drop when people work with AI?

It can, and the first signs already show in the way code is written. We have no proof that AI is the cause, but the numbers point in the same direction.

GitClear, a company that analyses code, looked at 211 million lines of code changed between 2020 and 2024. Two things moved at the same time: code copied from one place to another went up, and the work of looking after existing code went down.

Why does this matter to someone who doesn't write code? When the same thing is written in five places, every change has to be made five times. One of the places gets forgotten. That is how the mistakes that surface after launch appear.

The report shows a link, not a cause: in the same years, more and more developers started using AI. That is as much as can honestly be said.

What the numbers show

Between 2021 and 2024, copied lines of code grew from 8.3% to 12.3% of everything that changed. Lines tied to looking after existing code fell from 25% to under 10%.

GitClear's 2025 report, on 211 million lines of code.

Why does quality drop when nobody checks?

Because AI does what it is asked, quickly, and doesn't care what happens to the result a year from now. That is where a person has to be.

A 2025 report from Google's DORA team says it in one sentence: AI's primary role is as an amplifier, magnifying the strengths and weaknesses a team already has. A careful team gets faster. A team that worked without care still works without care, only faster.

I have seen this with texts. In the article about GEO I wrote about a company that lost around 20 positions on Google after publishing AI-made articles in bulk that nobody had read. The problem was not AI. The problem was that nobody checked.

What does this mean when you buy a website or an app?

It means a short deadline tells you nothing about quality. A website can be ready in a week and built well. It can be ready in a week and built badly. The tool doesn't make the difference, the person who checked does.

What you pay for later, when nobody checked: small changes that take a long time, mistakes that appear after launch and, sometimes, a website, an automation or an app that has to be rebuilt from scratch.

How can you tell the work was done with care?

Ask. You don't need to understand code to ask five questions and listen to how they are answered.

1. Who checks what AI produces, before it reaches me? You want a name, not "the team".
2. Who answers if something doesn't work after launch? And how quickly.
3. Can I change something a year from now without rebuilding everything? A yes or a no, with an explanation.
4. What happens to my company's data? Where it goes and who sees it.
5. Will you show me how it was checked, before I pay? Whoever worked with care is not afraid of this question.

A clear answer to all five is worth more than any promise of speed.

What do we promise?

We use AI and we don't hide it. It helps us deliver quickly. What we don't do is leave it on its own.

• A person answers for everything that goes live. Nothing reaches you just because AI wrote it.
• You work with the same two people. From the first audit to launch day.
• We tell you openly what we did and what comes next. No complicated terms.

If you need a new website, Let's Web has the Let's build it package. If you already have a site and don't know what is behind it, Let's check it is a full audit, in 4 to 7 working days, in which we check everything, including what others built. For an app made to fit your business, see Let's App.

If you are wondering what is worth leaving to a program and what is not, David wrote about it in What to automate in a small business, and what not.

Frequently asked questions

Will AI replace developers?

We don't have a sure answer, and we don't believe anyone who says they do. What the studies show: AI writes more and more of the code, but someone has to decide what gets built and answer for the result.

Is a website built with AI worse?

Not necessarily. What matters is who checked the result. A website built with AI and checked by a careful person can be very good. One built with AI and published without anyone reading it is a lottery.

How do I find out if the company I work with uses AI?

Ask them. Today, most developers use AI in one form or another. The question that matters is who checks and who answers.

Do you use AI on your projects?

Yes. It helps us deliver faster. A person checks and answers for everything that goes live.

Is a website built with AI cheaper?

At the start, maybe. The real cost shows later, when you want to change something or when a mistake appears. That is why you should ask who checked, not only how much it costs.