I'm entering my third year of a computer science bachelor's degree. I understand the theory well enough for exams and can write easy-to-moderate programs, but most of my projects over the past two years were produced with substantial help from AI. Now that I need to build stronger portfolio projects, I'm worried I won't be able to work independently or develop a good career in computer science. Should I stop using AI completely, revisit my earlier coursework, practice data structures and algorithms, or implement machine-learning models from scratch? What would be the most effective way to catch up?
4 Answers
A useful progression is to temporarily remove AI from your normal coding workflow. Rebuild familiar projects or course assignments from a blank file, then study the gaps you encounter. Once you can work independently, reintroduce AI in a limited role: code review, test generation, documentation, or hints. If the tool writes everything, you lose the practice needed to judge whether its output is correct.
The most direct fix is to start building without AI and accept that it will feel slow at first. Work back through your first- and second-year material, redo important assignments, and gradually move toward larger projects. There isn’t a shortcut for replacing practice you skipped. Use books, documentation, tutorials, and search when you get stuck, but make sure you understand and write the solution yourself.
Focus less on proving that you can memorize syntax and more on learning how to investigate problems. Read library documentation, trace errors, inspect existing code, write tests, and debug systematically. Real jobs often involve unfamiliar and messy codebases, where the important skill is understanding what the system is doing and making safe changes. AI can assist with that, but you still need enough knowledge to verify its suggestions.
Don’t confuse solving coding exercises with learning to build software. DSA is useful for problem-solving and interviews, but you should also create practical projects: plan the features, design the data model, implement them, test them, debug them, and explain every part. Pick something small enough to finish and add complexity over time. A portfolio project you genuinely understand is more valuable than a large AI-generated one.
You can still use AI later as a reviewer. First attempt the task yourself, then ask it to point out bugs, suggest tests, or explain an error rather than generating the whole solution.