I'm fairly young and have about a year before starting at a university in Denmark, where I'm considering studying Computer Science. I've been programming in some form for more than eight years and can move between languages and frameworks comfortably, although I don't have much professional experience yet.
I'm currently working with an artist friend on a game project. I'm designing and writing many of the classes and structures that will form the foundation of the project, so coherence and maintainability matter a lot. So far, my AI use has mostly been limited to asking questions in a chat window, similar to using a programming Q&A site. I haven't tried AI coding agents or "vibe coding."
I enjoy solving problems, writing the solutions myself, and understanding exactly what my code does. At the same time, I'm worried about falling behind if AI-assisted development becomes an expected part of software jobs. Should I experiment with tools such as Claude Code, or would it be better to keep developing my traditional skills and only use AI in limited ways? In particular, would you trust an AI agent with an important game framework, or is it better suited to smaller, isolated tasks?
4 Answers
The key is whether you understand the domain and can evaluate the result. AI-generated code can save time when you already know what good code should look like, especially for repetitive boilerplate or small experiments. But if you don’t understand the subject or the generated implementation, you’re left with code you can’t properly explain, maintain, or debug.
For the game framework, I’d keep the architecture, interfaces, and core systems under your own control. You can still ask AI questions, request alternative designs, generate minor boilerplate, or use completion for small sections—but review and test everything.
Keep writing the important parts yourself for now. Anything an AI completes for you is also something you didn’t fully practice, and that matters while you’re still building your fundamentals. If you eventually use AI, having strong programming and debugging skills will make it much easier to recognize when the generated code is wrong.
Reviewing AI output isn’t free. A large generated change can take as much time to understand as writing a smaller, clearer solution yourself, and agents can introduce inconsistencies across many classes. Use them where the time saved is obvious, not because you feel pressured by hype. There’s no need to sacrifice the parts of programming you actually enjoy.
You don’t have to choose one extreme. Build some projects entirely yourself so you continue practicing design, implementation, and debugging, while using AI on other projects for exploration or routine tasks. Most professional developers will probably need experience with both approaches, so learning where the tool helps—and where it creates problems—is more valuable than blindly adopting or rejecting it.
That makes sense. I’d be more comfortable trying an agent on a small, low-stakes project rather than handing it the framework I’m currently building.
There are exceptions for genuinely disposable or tedious work. For example, AI can generate a small compatibility patch if you read it carefully and write your own tests, but that’s different from letting it design the core of a project.