I started learning Python, SQL, and VBA a little over a year ago, and I now use them in my job. I recently began using Claude Code and was surprised by how quickly it can generate or modify applications. I can work on parts of a .NET web app and improve Python or VBA scripts without knowing C#, JavaScript, CSS, or HTML in depth.
That has made me question whether learning programming or paying for a computer science degree still makes sense. AI can produce working software much faster than I could on my own, and sometimes it feels as though traditional programming skills are becoming obsolete. At the same time, I realize I may not have enough experience to judge whether the generated code is maintainable, secure, well-designed, or appropriate for the problem.
I am considering starting an applied computer science program this autumn because self-study has been difficult for me. Is formal education still worthwhile for someone hoping to work in software? If so, how should I learn programming in a world where AI can generate an application almost immediately? How can I use these tools without becoming dependent on them or skipping the fundamentals?
5 Answers
AI is a powerful tool, but it doesn't automatically give you the judgment of an experienced developer. A useful way to think about the rankings is: a capable developer who uses AI will often outperform one who refuses to use it, but a beginner who uses AI still has to develop the knowledge needed to become capable. Use AI as a tutor: ask it to explain design choices, review your code, identify tradeoffs, and walk through bugs. Simply accepting whatever it generates may help you ship something, but it won't teach you how to build reliable software.
You don't have to choose between learning programming and using AI. AI is particularly good at repetitive, well-documented tasks, but it is much less reliable at deciding what should be built, understanding unusual requirements, handling edge cases, and debugging complicated real-world systems. A degree may not be necessary solely to qualify for a job, but fundamentals are still valuable because they let you judge the output and fix it when the generated solution goes wrong. Learn the concepts, build projects, and use AI to explain and challenge your thinking rather than doing all the thinking for you.
The biggest issue is that you may not yet have enough context to recognize bad output. Ask an AI for a bridge design and it can produce something that looks convincing to a non-engineer, while a bridge specialist would immediately spot serious problems. Software has the same issue. Code can run and still have poor architecture, security vulnerabilities, performance problems, or maintenance costs. AI is excellent for a knowledgeable programmer, but it is not a substitute for one.
There is a difference between making something work once and building it properly. Generated code may achieve the immediate goal while being disorganized, fragile, difficult to test, or full of strange design decisions. Professional development also involves requirements gathering, communicating with nontechnical people, choosing tradeoffs, maintaining systems, and taking responsibility when something breaks. Those skills are not reduced to writing code in a prompt.
AI-generated code can look better when you are new, but that perception changes as you gain experience. One junior developer described seeing far more mistakes in generated code after a year on the job—not because the AI became worse, but because their ability to notice mistakes improved. Software teams may eventually need fewer developers for some tasks, but increased productivity can also lead companies to build more. The safest approach is to learn the fundamentals and treat AI as a productivity tool, not as proof that understanding the underlying work is unnecessary.
Experience still matters because you can't replace years of learning how systems fail by blindly accepting generated suggestions. AI can help you move faster, but you need enough understanding to direct and evaluate it.