I’ve been asked a question recently. Today, given what AI can do is it still worth getting a degree in software development or something similar?

It’s a good question and one that cannot be easily answered. It falls, under the “it depends” like all good questions. It depends on for how long you’re trying to forecast. At this very moment, yes, there is a future for software developers. The profession isn’t going anywhere anytime soon. I cannot be a fortune teller, so I can’t predict the distant future, but the demand for human expertise is still there.

Let’s talk about what AI can or cannot do

If you need an MVP or a simple application, vibe-coding is a viable solution. You write up the list of specifications and if you word it carefully, then, the AI has a decent chance that it will create the application for you, especially if it’s in what we call Agent mode (it writes code, builds the app, tests it, fixes issues, writes some more and keeps going). It will be able to pull off the MVP, the minimum viable product.

The problem is that if you want something really complex and instead of a simple MVP you’re looking at an entire application, then the chances of coming out with a finished product drop. There is also the risk that it won’t be able to fix it’s own mistakes. At this point if a real developer tries to intervene it’s often too late. By then, the AI-generated technical debt is so high that even a human developer can’t make head or tail of it. It’s like the old running joke:

When I wrote this only God and I knew how it works. Now only God knows.

What’s the point of AI then? Should we even bother learning to use it?

Absolutely, yes! When we think of AI, we think of LLMs, or Large Language Models, so it should not be a surprise that they are great at syntax. As a counter-point, AI is not great at efficiency and architecture. Sure, it might spit out something that someone else considered great, but it might simply not work for your scenario. So, while AI might not yet be able to make a large scale application for some time it is really good at doing some tasks. Let me list some of them.

  1. Writing snippets of code - you know what you want the method to do, it’s straight forward, you know what the inputs are and what you expect as a result. Ask AI to build the method. Just review the code afterwards, because it can sometimes misunderstand due to lack of context.
  2. Writing unit tests - writing tests is important to make sure that bugs don’t come at you, especially when dealing with AI to help you. AI can write up for you. It still makes mistakes, so review what it cooked up, but it’s a lot faster than a developer writing them from scratch.
  3. Code review - 5 developers might look at the same piece of code and miss an obvious mistake in the logic. AI will see right through and point out a case that you did not consider. In fact, it often suggests edge cases that might be overkill, but it’s excellent at sniffing out potential runtime errors. Again, double check it’s suggestions because it will want to validate the same input parameters on each function you write.
  4. Refactoring - You wrote a snippet of code with if/elseif/else and you realize that it’s overly complicated. It’s not ideal and hard to read. AI can help you transform it into a switch statement if possible. It will take an instant and it is a lot faster than you typing it yourself. Still check the before and after, but this is probably the place where it’s least likely to fail, but still, be careful.

There are many other use-cases, but as you can probably already see it can be quite helpful. What’s the common denominator? The developer has to review the work before accepting the results.

So, for the time being, a developer it’s still required to validate the results. If you want to know for how long, unfortunately, I cannot say, but for the foreseeable future it’s still necessary.

What about the degree? Is it still required?

Actually that degree is required more than ever. Given the explosion of so-called developers who believe vibe coding is the solution and that they can make the next Facebook using AI, the demand for good developers who have finished some form of training is higher than ever. Of course, fewer developers are required compared to the past, but still required nonetheless.

What will change, however, are the list of important skills for a developer:

  1. attention to detail, as you will have to review a lot of generated code (from AI or from your coworker who also got it from AI)
  2. patience to debug, because AI will surely introduce bugs, simply because it lacks the context of what you’re trying to build or the entirety of the application

So, to answer the original question, is it worth pursuing a degree? Yes, more than ever. I can’t say for how long, of course, but right now, the answer is yes. Just expect your job description to be different than what it was 5 years ago.

P.S.: This article war written entirely by me, but it has been proofread using AI