Xavier Bertels

I’ve always approached my martial arts practice as something to learn. Something to study. Yes, experience too. But also: books, YouTube videos, drills. Yet recently I came across a different way of learning martial arts: by turning it into small games with win conditions, you can experience through creation, rather than pure repetition. It’s so much fun and it allows your body to just explore things, without rationalising them too much. I feel I learn faster that way.

When we approach a challenge as something to learn, we might learn something. When we approach a challenge as a creative endeavour, we might create something, and learn something as a byproduct. Whatever the thing is you’ve learned, you’ll have a deeper connection with forever.

I made a bet in 2012. A friend was strongly convinced that fully autonomous cars were five years away. I didn’t believe it could work in Europe.

Fast forward to 2025 and we still don’t have fully autonomous cars in Europe. I won that bet. It turns out that even as you put in more and more rules, the terrain was never really like the map.

With the shift to AI however, the approach fundamentally changed from deterministic to probabilistic. Turns out that this is much closer to our experienced reality. Rather than a set of rules, navigating the world is a chain of probabilities. Technological systems that take this approach are much better at being our copilots. And they will eventually be better pilots, influencing reality at levels indistinguishable from fellow humans. The terrain is increasingly becoming one with the map.

A very entertaining podcast: Boomtown. It also happens to be the inspiration for one of my favourite series of the past year: Landman.

It’s all very American. But in a way we rarely see here in Europe. In Boomtown, they tell a different story and I like that. It feels like a fresh wind of unapologetic Americanness.

In the workplace most misinterpretations come from people who fill in missing context with their own, personal context. Without first trying to understand the other person’s context and view, we’ll nod “yes, consider it done” and then go off to do something entirely different than what the other person thinks he’s asked for. That scenario rarely ends in a fun conversation.

It’s remarkable how LLM’s struggle with context in the same way. Remarkable because the flaw is so deeply human.

Lucky for us, there’s a little hack. We can ask “what assumptions are you making right now?” Also lucky for us, using this question in ChatGPT’s custom instructions will help it give much more contextually useful answers.

The general design vector in AI powered apps is either chat, or voice. That seems like bad news for designers. But I think it opens up an opportunity.

Imagine you’re on a busy bus. You’re in a waiting room. Your spouse is sleeping. Voice works in some scenarios, but it sure doesn’t work all the time. It doesn’t jive with how a lot of humans live and work.

Then, there is chat. If you think about it: text is the ultimate design system. We’ve come up with a system that allows us to convey a lot of—but not all—meaning by combining a very limited set of little drawings. And language models are incredibly good at generating interfaces on the fly that allow users to “do stuff” by sending these little drawings back and forth.

But text is also an incredibly cumbersome way to interact with computers. When was the last time you typed something on Instagram? It’s a billion dollar business model built around swipe, tap, swipe, double tap.

Text is a stopgap. Eventually, we’ll want shorthands again. We’ll want something visual. Perhaps even something tactile. The interface may be generated on the fly though. And this is where designers will shine once again. We’ll need people to think about new systems, new shorthand, new ways of communicating visually.

It’s a very exciting time to be a designer.