The demise of software engineers due to AI is greatly exaggerated(techleader.pro)
techleader.pro
The demise of software engineers due to AI is greatly exaggerated
https://techleader.pro/a/679-The-demise-of-software-engineers-due-to-AI-is-greatly-exaggerated-(TLP-2025w6)
30 comments
I watched one of the new hires burn 8 hours in chatGPT trying to make excel do something that took me five minutes. Not worried for anything but the economy and the environment.
Exactly this. In their current state, ChatGPT (and Claude, etc) are a great boost for people who already have good foundational knowledge.
They'll probably speed up learning in general, but are no replacement for an in-depth understanding.
They'll probably speed up learning in general, but are no replacement for an in-depth understanding.
While I have seen similar (engineer struggle to get claud to write some TF that could have been done in a few min) I also have experienced the opposite.
I was attempting to put together some basic statistics for my oncall shift, so I downloaded the CSV from our system and used DuckDB to inspect. I had never used DuckDB before so using an LLM to help me refine my queries was great. Even with a few turds of a reply it still mostly helped save time.
Knowing when to stop and try a different approach is the same underlying skill that some devs don't exercise and others do.
I was attempting to put together some basic statistics for my oncall shift, so I downloaded the CSV from our system and used DuckDB to inspect. I had never used DuckDB before so using an LLM to help me refine my queries was great. Even with a few turds of a reply it still mostly helped save time.
Knowing when to stop and try a different approach is the same underlying skill that some devs don't exercise and others do.
It's the same as using Google search instead of programming skills, then Stack Overflow instead of programming skills, and in the meantime there was also WYSIWYG programming.
> and in the meantime there was also WYSIWYG programming.
Oh, that one's perennial; it periodically shows up, promising to obsolete programmers, and has since the 1980s.
Oh, that one's perennial; it periodically shows up, promising to obsolete programmers, and has since the 1980s.
Then let AI arrange together the boxes and arrows? Should work.
I think it's great if you know what you want it to do and you prompt appropriately. Can save quite a lot of time if you already have some experience.
I saw a guy on twitter boasting about launching some micro saas with 28,000+ prompts and doing "no coding". I'd pay to see that codebase.
These last 2-3 years have made it very tangibly clear how much of our industry's job market is tied to whims of the investor class. I have zero belief that any softening in the software job market is based on anything other than a shift in capital and a fad for layoffs and hiring freezes. But that doesn't mean folks in leadership won't use AI as a talking point to force a rebalancing in power between employer and employee and drive down wages.
I often (but not always) do chat driven development and I can see AI is not ready to replace actual engineers yet.
They're only useful for writing code. Id liken the strongest one I've used so far, o3-mini, to be like a super smart junior engineer.
It knows it all, can often handle the more boring parts of writing code, but doesn't have the experience to not make some elementary mistakes, so you need to review everything it generates and it can't be let loose without strict supervision and guideance.
They're only useful for writing code. Id liken the strongest one I've used so far, o3-mini, to be like a super smart junior engineer.
It knows it all, can often handle the more boring parts of writing code, but doesn't have the experience to not make some elementary mistakes, so you need to review everything it generates and it can't be let loose without strict supervision and guideance.
The underlying problem here is best fixed by engineers learning just enough about business to start their own companies, without the people that think of them as overhead.
The problem is that some tasks really do require cooperation from a large group of people working on the same thing. And that requires capital, and that capital is going to be asking "Why did you hire so many engineers? Accenture says you can replace 40% of them with AI, I read it in a white paper, so you'd better start showing more efficiency or we'll find a CEO who can move with the times."
I take replacing myself as the ultimate challenge. Does anyone know of a workflow/tool that can from a prompt find which relevant files in a directory to include, and then update existing files and create new ones?
All the the tools I've tried require me to select which files to include in the context, and/or then require manually copying each snippet to where it should be... this makes me feel like a neanderthal in some bronze age cave.
All the the tools I've tried require me to select which files to include in the context, and/or then require manually copying each snippet to where it should be... this makes me feel like a neanderthal in some bronze age cave.
The sample of engineers they surveyed for this were using copilot. Get them to use cursor + Claude for 6 months and see if they think the same.
I've been trialling Cursor + Claude, I don't think it's that much better?
Try again if it’s been a while.
Cursor since agent mode was introduced a few months ago is a crazy leap forward.
Cursor since agent mode was introduced a few months ago is a crazy leap forward.
I'm actively trialing it now, it is good but yeah, not feeling anything revolutionary.
So sick of all this now.
Honestly any time I go online to learn about AI writing code, I end up knowing less.
One person will say 'Lets not fire all our devs, AI is another tool not a replacement'.
Then you get another guy saying 'Lol yeah right just wait till SuperAI v3.6 comes out, then we'll all be fired'
What the hell is wrong with you people? These are people's livelihoods.
What the hell is wrong with you people? These are people's livelihoods.
Well I guess I'm in the former category but what do you expect us "the people" to do about it?
I don't think you should halt progress just because people would lose jobs because people create and do jobs because there is something that needs to be done.
If we keep them around just so people have jobs then we're running an adult kindergarten which makes no sense to me.
Besides, there will be jobs that can't be done by AI for a long time to come and the current AIs are relatively crappy compared any average software dev.
I wonder why this is such a popular topic even, to me it feels like if you'd use one of these LLMs for more than a week you'd start to see why it's incapable of replacing anything major.
> What the hell is wrong with you people? These are people's livelihoods
Do you think about lamp lighters the same way?
Do you think about lamp lighters the same way?
If you're able to compare software devs with lamp lighters, you've truly bought in to the AI hype.
If lamp lighters also build and maintain the lamps, produce the town gas, build and maintain the gas pipes, clean the streets, you sure as hell shouldn't fire them, because they invented self lighting lamps.
Yes? And train wagon brake operators.
Social upheaval sucks for those involved.
Social upheaval sucks for those involved.
Shame on you.