I'm seeing junior developers at work build surprisingly impressive projects with AI tools, including things I wouldn't have been able to create on my own. My current approach is slower and more deliberate: I read books, work on small projects, write code myself, make sure I understand it, and use chatbots as a kind of search engine or programming companion. I'm worried that unless I fully embrace AI-assisted development, I'll become outdated. How are others learning these days, both with and without language models?
4 Answers
The difficult part of programming has never really been typing syntax. It’s understanding the problem, choosing an appropriate design, and building a solution that actually works. AI can produce a lot of code, but it won’t reliably notice that you’re solving the wrong problem or make all the architectural decisions for you. Learning through curiosity, building things, studying how others solve problems, and keeping a strong mental model of how systems work is still valuable. Release notes, technical articles, and hands-on projects are all useful, and AI can complement that process rather than replace it.
Learn because you enjoy programming, solving problems, or building things—not only because it looks like a safe career choice. AI may make routine implementation faster, but it also makes judgment, curiosity, debugging, and understanding systems more important. If you like figuring things out, keep learning and use AI where it helps. If you only chose programming for the job market, the current changes may make the work feel especially frustrating.
That’s the part I’m unsure about too. Some workplaces are pushing people toward supervising AI output instead of writing code directly, which can feel different from the kind of programming many people originally wanted to do.
Your slower learning style is not obsolete. Reading, experimenting, and writing code yourself help you build intuition that can’t be outsourced easily. Use AI to explain unfamiliar concepts, suggest alternatives, generate tests, or help investigate an error, but don’t skip the effort of understanding the result. Large AI-assisted projects still require experienced engineers to plan the work, coordinate the tools, review the output, and fix regressions. It’s too early to assume that strong software engineering skills will stop mattering.
You still need enough knowledge to recognize when an AI-generated answer is wrong. Otherwise it can give you insecure, inefficient, or completely inappropriate code and you may not notice until it causes a serious problem. AI is a productivity tool, but someone still has to provide direction, review the result, understand the tradeoffs, and take responsibility for what gets shipped.
A good test is to ask the developer to explain the generated code and its design choices. That’s often where shallow understanding becomes obvious.
Exactly. Engineers need to learn how to reason about and solve problems, not just memorize syntax or tool features. The tools change; the underlying engineering skills remain useful.