What should I learn as a software engineer if AI is writing more code?

November 18, 2026 5 answers
Asked by 9w~coppervisor_634

I'm a computer science student trying to figure out where to spend my time. Is learning a language or framework deeply still worthwhile, and how much should I focus on problem solving, security, and system design?

Answers

3w+gasketcompass_681

Learn at least one language well. Not because you need to memorize every bit of syntax, but because you need to read code, understand its behavior, and spot when a generated solution is wrong.

Practice debugging and testing alongside writing code. Give yourself a small problem, implement it, then compare your version with an AI-generated one. Can you explain where they differ and write a test that catches a mistake? That exercise teaches more than accepting output just because it runs.

_p64~drelvani.7:

Reading generated code can be harder than reading code you wrote yourself, so tracing it through with a debugger is useful practice too.

Reply to this answer
Morlavi.7q

Think beyond individual functions. Learn how the pieces of an application fit together: interfaces, APIs, data storage, authentication, and the boundaries between them. A solution can look fine in isolation and still make the larger system difficult to change.

That doesn't mean AI can't help with system design. It means you need enough context to judge its suggestions and explain the tradeoffs.

Reply to this answer
_z74+plovari.6

Security is a good area to study, especially around authentication, access control, and handling data. Generated code can introduce vulnerabilities, so working code isn't the same thing as safe code. Learn to ask what inputs are trusted and who can access what.

Reply to this answer
_m83~zordavi.6

Build something you actually want to use. It doesn't have to be a giant portfolio project. A small app will give you concrete reasons to learn a language, a database, an API, and whatever else the project needs.

Use AI where it helps, but slow down when you can't explain what it produced. The project will show you which gaps are getting in your way.

9w~coppervisor_634:

That sounds more manageable than trying to pick every skill in advance.

Reply to this answer
_w48~velnixa.6

Problem solving still matters, including defining the problem clearly. Practice breaking a feature into requirements, constraints, and test cases before asking a tool to implement it. Clear direction makes the output easier to evaluate and maintain.

Also, don't assume every company has agents doing all its coding. How teams use these tools varies, and the work still includes deciding what to build and checking whether it meets the need.

Reply to this answer

Add your answer

Answers are reviewed before publication.

Related Discussions