DEV Community

Cover image for Is It Still Engineering If AI Writes the Code?
Andrew Chadwick
Andrew Chadwick

Posted on

Is It Still Engineering If AI Writes the Code?

It seems like everyone is using AI to generate code these days. Which begs the question: if an LLM is typing out the syntax, are we actually engineering anymore?

​Like any tool in this era or any era before it the answer depends entirely on how you use it. I’ve written before about AI being like a sword; in the hands of a skilled master, it’s an incredibly powerful weapon. But a sword doesn't win a battle on its own. Today, I want to talk about the skill of the engineer wielding it.

The "Screenshot to App" Trap

​We have all seen the viral demos. Someone shows an AI an image of a UI, writes a basic prompt, and the AI spits out a functioning workflow. It looks like magic.
​But look under the hood of those generated apps. You almost always get a massive, single-file monolith with zero security, tightly coupled logic, and a fundamentally fragile flow.

​What happens when you want to extend that app? You ask the AI to move a feature or reuse a component. To accommodate, the AI writes more code on top of the pile, inevitably breaking a dependency you established 15 prompts ago. Suddenly, the app is a spaghetti mess, and the AI gets stuck in an endless apology loop "You are correct, here is the updated code" progressively breaking three other things to fix one.

​It is not really the AI’s fault. It is just providing the most probabilistic solution to a surface-level prompt.

The Engineer’s Mindset: Systems over Syntax

​An engineer approaches the exact same tool completely differently. We don't just ask for "an app." We define the problem, break down the required steps, and figure out how different technologies need to gel together to make the system resilient.

​When an engineer uses AI, the prompt isn't a wish list; it is an architectural blueprint. It looks more like this:
​Structural Boundaries: "I want this app to consist of a main body, but split the three core workflows into discrete, decoupled services."
​Enforced Standards: "Every method call requires unit testing. Please add this rule to your project memory/instructions file so you don't forget it on the next prompt."
​Data Architecture: "Store the data using this specific repository pattern, and separate the data layer into its own isolated project."

The Power of Pushback

​The real engineering happens when the AI gets it wrong.
​Because an engineer understands the underlying mechanics whether they are structuring a complex backend or designing a relational database they know when to challenge the output. If the AI returns a probabilistic guess that violates the core project philosophy or introduces a massive bottleneck, the engineer catches it, rejects it, and forces a course correction.

The Verdict

​Typing syntax was never the hardest part of our jobs. Problem-solving is.

​AI does not replace engineering; it simply shifts the focus from writing boilerplate to pure system design and architecture. As long as the core fundamentals of building scalable, maintainable software are not forgotten, AI is just the newest tool in the engineer's belt to build better systems faster.

Top comments (4)

Collapse
 
acaciaman profile image
Karlis

I agree, AI is a new, powerfull tool. But for me it is still addition to previous known software engineering. I still develop in VS Code in Windows using Chrome browser. Yes, You should continue to follow good engineering principles as do analysis, coding and testing.

Collapse
 
chadders13 profile image
Andrew Chadwick

Yer Ai is definitely evolving and the pace at which you can build stuff now is scary but that dose means we do actually have time to put in the safe guards and testing 😊

Collapse
 
chadders13 profile image
Andrew Chadwick

It really is. I've seen Jr's not know that and try and get code through PRs 😅😂