A year ago, if someone told me there would be software engineering jobs where part of the job description was essentially "talk to AI and build things," I probably would've laughed.
Now?
I'm seeing more and more job postings asking for experience with AI-assisted development, prompt engineering, Cursor, Claude, GPT, Windsurf, and whatever tool gets released next week.
And honestly?
If someone wants to pay me a software engineer salary to spend all day collaborating with AI to build software, I don't see the downside.
Before You Grab Your Pitchfork
Let's get something out of the way.
I genuinely enjoy programming.
I started coding long before AI could write a single line of usable code. I enjoy solving problems. I enjoy debugging. I enjoy understanding how systems work.
If every AI tool disappeared tomorrow, I'd still be writing code.
A lot of developers hear "vibe coding" and immediately assume it means:
- Not understanding code
- Copy-pasting random AI output
- Shipping broken software
- Replacing engineering with prompts
And to be fair, sometimes that's exactly what it means.
But that's not what the best developers are doing.
The Calculator Didn't Kill Math
When calculators became mainstream, mathematicians didn't disappear.
When IDEs added autocomplete, programmers didn't disappear.
When Stack Overflow exploded, developers didn't disappear.
Every generation of engineers has some tool that older engineers claim is ruining the profession.
Then a few years later, everyone is using it.
AI feels like the next version of that cycle.
The developers who understand software fundamentals are now able to move significantly faster because they can offload some of the repetitive work.
The Reality Nobody Wants to Admit
Most software jobs aren't about writing clever algorithms all day.
A shocking amount of software development is:
- Writing CRUD endpoints
- Building forms
- Connecting APIs
- Fixing styling bugs
- Updating dependencies
- Writing tests
- Refactoring old code
- Reading documentation
If AI can eliminate some of that friction, why wouldn't I use it?
I don't get bonus points for suffering.
Vibe Coding Isn't the Skill
This is where I think both sides of the argument miss the point.
Typing prompts isn't the valuable skill.
Understanding software is.
The reason experienced developers get better results from AI isn't because they're better at writing prompts.
It's because they know when the AI is wrong.
They know when a security vulnerability was introduced.
They know when the architecture doesn't make sense.
They know when the generated code won't scale.
They know when the AI confidently hallucinated an entire API.
The real skill isn't generating code.
The real skill is reviewing it.
The New Job Description
I think we're heading toward a world where software engineering jobs look different.
Instead of:
Write every line of code yourself.
The expectation becomes:
Use every tool available to solve problems as efficiently as possible.
That includes AI.
Not because companies love AI.
Because companies love productivity.
If one engineer can accomplish the work of three because they're effectively leveraging AI tools, management is going to notice.
Whether developers like it or not.
My Completely Selfish Perspective
If I'm being practical, the situation seems pretty simple.
I can:
Option A
Refuse to use AI because "real programmers write everything themselves."
Or
Option B
Learn the tools, understand their strengths and weaknesses, and become more productive.
One of those options sounds a lot better for my career.
And if a company wants to hire me to spend my day building software with AI assistance?
Sign me up.
I'll still review every line.
I'll still understand the architecture.
I'll still be responsible for the final product.
I'll just get there faster.
The Developers Who Win
The developers who are going to thrive over the next few years aren't the ones blindly accepting every AI suggestion.
And they aren't the ones pretending AI doesn't exist.
They're the people in the middle.
The engineers who understand software deeply enough to leverage AI without becoming dependent on it.
The people who can build something from scratch if they need to—but choose not to when a tool can save them hours.
Because at the end of the day, companies don't pay us to type.
They pay us to solve problems.
And if talking to an AI helps me solve those problems faster?
I can't think of a more developer thing to do than using the best tool available.
Top comments (1)
This feels like a real shift in where engineering effort sits.
Less time writing code, more time evaluating and constraining it.
The bottleneck doesn’t go away — it moves from generation to judgment, especially around correctness and architecture.
The “middle ground” you describe seems like the only sustainable way to approach it.
Hey, Martin. I see you are so talented. I have an idea and need an expert like you to implement this. If you are interested, please let me know. Let's chat.