I have been thinking a lot about learning in the era of AI. What does it really do to learning overall, and how do I adapt my learning to get the benefits rather than fooling myself that I’m developing in the process? Some studies show that students who use AI get better grades in the short term, but in the longer term, they develop less of an understanding of the subject at hand.

It is very appealing to use AI, I feel it myself sometimes, a rush when I complete the work, but then somehow I feel like I didn’t earn it, and I doubt myself. I do see, however, that being freed from tedious work is a huge benefit, not just for me but for anyone who can produce something way faster than before. The tedious work has become somewhat optional. There are plenty of honest reasons someone can’t spend years building a foundation but still wants to make something that helps them in their life.

The thing that gets missing when you skip the hours of learning is the friction. This friction is not always a “nice” feeling; we might feel dumb, question ourselves, or maybe eventually lose our energy. The thing is, though, this wrestling with the subject is important for learning a skill.

A counterpoint, though, and let’s use programming as an example, is that we’ve been climbing the abstraction ladder for decades in many domains, so what skills are actually needed? What I mean is that in the evolution of programming languages, the first ones were low-level, which meant that the programmer had to hand-manage memory allocation, for example. Often nowadays, this does get handled by more modern programming languages. There can be different viewpoints on whether something is lost or not, but it is worth questioning the relevance of that loss for the majority.

I would say that one thing to ponder is whether you are actively building yourself or just a system/tool? They aren’t mutually exclusive, and it is possible to do one without the other and both at the same time. The optionality of friction, I would argue, is what makes this more of a choice today than before. When you build or create today, you can have the thing with almost none of the self attached with AI-involvement. It is important to be aware that our imagination and inventiveness are built from problems we’ve actually wrestled with, and the friction you skip now is the understanding you won’t have in ten years.

The friction is what shapes us, because it exercises our brains and we form new connections that can last for a lifetime. Knowing that AI isn’t going anywhere, we probably should reflect on whether we want to be changed by our work or just get it done.

This blog that you are reading right now isn’t free of AI fingerprints. I am continuously experimenting with the degree of involvement to strike a balance. I can’t deny Claude’s usefulness in many areas of writing, but I’m also very aware of the costs. That said, I hope readers believe that these ideas and thoughts are my own. I keep challenging many things that come out of Claude, so that I can have some of the friction that builds me, just like I would with a colleague.

The images on the blog are another story. They’re made with Gemini, and I do not quarrel with that, because I never set out to become a digital illustrator. Nothing in me was supposed to be built by drawing it, so nothing for me is lost by generating it. There might be some who dislike AI-generated images, but it felt like a tradeoff I’m walking into eyes wide open, willingly.

After deeper reflection, the risk in my own learning lies in my choices. What I let grab my attention and time, and what I let form me. So the question for any new project I take on becomes: Do I keep this friction that can shape me, or quietly let it go?