The AI Takeover That Didn’t Kill Coding

For decades, computer science students learned through the same process: start with loops and variables, create algorithms, and after enough sleepless nights filled with syntax errors, emerge as coders. However, the era of laborious hand-coding might be coming to an end. Powerful AI systems like ChatGPT and Claude can now produce the same code in seconds. If machines can code quicker – and better, perhaps – what's left for humans?

This question is shaping classrooms, career prospects, and coding itself. The rise of AI hasn’t simply changed how students learn and write code, but is forcing educators to rethink the true meaning of teaching computer science. According to Harry Smith, a senior lecturer in computer science at the University of Pennsylvania, the answer isn’t to abandon coding as a field of study. Rather, he states that “to know when things are correct or incorrect, you still need to be able to read and understand the code. And really, the only way to do that is with learning the fundamentals of a programming language.” Without a strong foundation in these skills, students risk becoming passive consumers rather than actively controlling their codes.

However, AI has indeed transformed coding from a solitary struggle into a more collaborative experience, requiring computer scientists to develop the ability to work with AI. Instead of typing out lines word by word, students must now learn to guide AI effectively, verify the results, and recognise when the use of AI is appropriate. According to Professor Smith, computer science education is shifting from how to code to how to think about code.

Professor Smith structures his classes around this principle. “What we try to teach is structured problem solving and thinking in abstractions,” he said. “So the abstractions are maybe the most important part. It's the process of identifying the parts of the problem at hand, thinking about which details are important to consider and which ones can be ignored or thought about later.”

The ability to think in abstract highlights a broader truth: there are limits to what an AI can do. While AI can produce accurate code, it lacks the contextual understanding and flexibility that only humans can exhibit. “They're not very good at integrating pieces into more complicated systems. It can be challenging to make sure that it actually works with all of the other pieces of the system that exist,” Professor Smith said.

In fact, real-world problems do not happen in isolation, where solving a single problem solves everything. Rather, problems are solved by integrating multiple factors and then creating a solution. That's why the abstract thinking and integration skills of humans will remain the cornerstones of computer science even as AI gets quicker at writing code.

So, how will computer science education look in the next decade? Professor Smith believes that a change will take place, but not overnight.

“Yeah, I think it might. I mean, the ability of a large language model to generate something is like, really useful for education.” However, he remains concerned about “the suitability of examples or texts or demonstrations or things that it comes up with.”

Furthermore, Professor Smith sees the coming years as a period of experimentation. “People are kind of figuring out what the best practices actually are,” and “if it's good and worth keeping around, it will still be here.”

Professor Smith noted that even as AI finds a place in classrooms, it will not replace the human aspect of education. Though Massive Open Online Courses promised to revolutionise education a decade ago, Professor Smith said, “still, people come to universities to be taught by people.” According to Professor Smith, “human or soft skill, or like network and connection elements of education. And I don't know that AI will be able to replace those to a level that is satisfactory.”

For students just starting their journey in computer science, the rise of AI can feel daunting. However, Professor Smith offers a reassuring perspective: “Before AI, for a long time, bad code was cheap or free. However, thoughtfully designed, efficient, carefully tested, and considered code that has always come at a premium.” While AI might be shifting perspectives and ideas, the true value of code lies in dedication and refinement.

“So I would say, when you're trying to learn this and you're struggling, lean into that, take heart in the fact that it is hard to do, but because you are someone who is insisting on sort of figuring things out for yourself, that in fact, that's the kind of grit that ends up making you stronger and more worthwhile in the end,” Professor Smith said.