I once came across a comment about how fewer people seem to use Stack Overflow these days. The explanation was simple enough: with LLMs, it has become much easier to ask a question and get an immediate answer.
But the part that stayed with me was a metaphor. Learning to code with only AI assistance can feel like living in a building with elevators but no stairs.
I do not think the elevator is a bad thing. In many cases, it is useful, efficient, and sometimes exactly what we need. AI coding tools can help us move faster, reduce friction, and get unstuck from problems that might have taken hours to solve alone.
Still, I keep thinking about the missing stairs.
I started learning programming around the time pretrained language models like BERT were becoming widely discussed. For the first couple of years, my learning process was still fairly traditional. I searched for error messages, read documentation, copied small examples, broke things, fixed them, and slowly built an intuition for how code behaves.
That does not mean I learned in a better way. It only means I had some time with the stairs before the elevators became this good.
Now, with LLMs, the learning process feels different. It is easier to reach higher floors, but sometimes harder to remember how the building is structured. I can ask for working code, but I still need to understand why it works, where it might fail, and what assumptions are hidden inside it.
So I do not want to frame this as a choice between old and new ways of learning. I would rather think of it as a balance. Use the elevator when it helps, but take the stairs often enough to keep the muscles working.
For me, learning in the age of LLMs means trying not to skip the struggle entirely. The struggle is not always pleasant, but it is often where the real understanding begins.