Nicolai Vittrup

← All thoughts

Thoughts, 15 November 2026

With AI, the tech isn’t the limit any more. The ideas are

Ballpoint sketch on graph paper: two drafts of the Amplify A logo with eyes

Last Thursday the whole office went out for dinner. Across the table, Jens asked whether our logo shouldn’t have a bit more personality. He’s just become our new Head of Search & AI, by the way.

The next morning I sat down with a pen and a notepad. I drew the logo, and I drew it the wrong way round. It wasn’t a pretty drawing.

That drawing is now an animation on every article on our site. The logo stands still, looks around, smiles, winks and hops back into place.

It took an afternoon.

The drawing as it ended up. The A jumps, lands as the smiley and winks.

A whole phase has disappeared

Everyone says things go faster with AI. That’s true, and it’s also the least interesting thing you can say about it.

What has changed for us isn’t that the work goes faster. It’s that a whole phase has disappeared. The phase where we put the idea aside.

The tech isn’t the limit any more. The ideas are.

The idea used to die on the notepad

We’ve had ideas like that for eleven years. A fun detail. Something creative. Something that would make a page ours instead of looking like everyone else’s.

And then we did the maths. A designer would have to draw it properly. A developer would have to build it. Someone would have to check whether it hurt our page speed, because it almost always did. The more creative, the more elements, the heavier the page. Creativity cost page speed, and we often went with the fast page.

All in, a couple of days’ work for an idea no client had asked for.

So we gave up on the idea and came up with something else. Almost every time.

The worst part wasn’t losing the idea. It was that we never found out whether it was any good. We judged it on what it would cost to find out.

And no, this doesn’t contradict what I wrote in October

In October I wrote that the hard part isn’t building, it’s finishing. That you have to choose what not to build. I stand by that.

The two things sit at different points in the process. You still have to choose, and you still have to finish. The difference is that now you can see the idea before you choose.

Before, we chose on a hunch. Now we choose based on something we’ve seen work.

That makes choosing easier, not harder. It’s easier to bin an idea you’ve seen than one you’ve only imagined.

What I do now

  • Build the sketch before you cost it. It’s often quicker to see the idea than to assess it.
  • Judge the finished thing, not the estimate. The estimate was always what killed the idea.
  • Throw it away without feeling guilty. When it took a couple of days to try, an idea had to be good. Now it’s allowed to just be worth seeing.
  • Let a person approve it. That doesn’t change. AI does the work, and a person makes the decision.

It comes at a price too

When it costs nothing to try, you also try the things that shouldn’t be tried. That’s the bill nobody talks about.

We have 482 written rules for how our AI is allowed to work. 427 of them apply to every client. Each rule exists because a person said no to a suggestion and wrote down why, instead of just moving on.

That’s the price of that freedom. It takes someone who dares to say stop, and it has to be a person with a name. Otherwise you build a lot and none of it can be trusted.

The ceiling has moved

We’re limited by our imagination, not by the technical side.

And that’s harder. Tech was something we could buy, and something we could blame. We can’t do that with imagination.

It feels a bit like playing a computer game at the moment. There are no limits apart from my own creativity. It doesn’t really feel like real life any more, and it’s a good feeling to take to work in the morning.

The drawing on the notepad wasn’t pretty. It’s on the site now, and that’s the whole point.

And as we say at Amplify: Let’s Amplify!

More thoughts