At my talk on Claude Code at the KSE winter school, a great question from the students was: does AI coding help or hinder learning to be an engineer?
I think most of us think about AI in learning (both as students and as continuous learning at any age) with the wrong bias — toward cheating, or “now no one will ever learn to program.”
Here are a few AI-positive affirmations:
(1) AI coding as a skill should not replace ordinary coding but complement it. Students who ignore Claude Code and learn to write code purely by hand will arrive at a job where everyone already works with AI coding agents and won’t have even minimal experience with them. Students who never even try to write code by hand are building a castle on sand — it won’t stand for long without fundamental knowledge.
(2) It’s the same process as having to learn to multiply and divide by hand before moving on to algebra and calculus — and once you’ve mastered that, you’re better off using a calculator. While studying, I’d advise splitting your time 50/50 between AI coding and craft, hand-written coding.
(3) Very few people pay attention to the new possibilities. 15 years ago, to understand how databases work I could either ask a professor or look for a book in the library — in the best case, an article online. Today I can fire up Claude Code and ask it about anything down to the smallest detail. I can ask it to pull the postgres/postgres repository from GitHub and build me a personalized presentation (or a book, or a podcast), and then quiz me to check what I’ve learned.
(4) Practical tools: the explanatory plugin for Claude Code doesn’t just do the work for you but also explains each step; the learning plugin for Claude Code will, on top of that, ask you to write individual pieces of code yourself.
Access to knowledge is no longer the bottleneck; the bottleneck is motivation and curation — what exactly to learn (besides, of course, the fact that access to AI is relatively expensive on a student’s budget).