this post was submitted on 18 Sep 2026
109 points (96.6% liked)
Programming
28538 readers
345 users here now
Welcome to the main community in programming.dev! Feel free to post anything relating to programming here!
Cross posting is strongly encouraged in the instance. If you feel your post or another person's post makes sense in another community cross post into it.
Hope you enjoy the instance!
Rules
Rules
- Follow the programming.dev instance rules
- Keep content related to programming in some way
- If you're posting long videos try to add in some form of tldr for those who don't want to watch videos
Wormhole
Follow the wormhole through a path of communities !webdev@programming.dev
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
It's normalized now but I don't see it lasting. I'm retiring at the end of this year cause I've had enough of it and I don't want to wait anymore for companies to finally figure it out. I've told so many what needs to change, what they need to do in order to fix their builds (mainly consisting of rebuild it, rehire the devs you laid off, use AI as a rubber duck only) and none of them listen. They thank me for my time, cut a cheque, and send me on my way. I had the idiotic notion of trying to change things, hopefully make some people see the light but they've all drunk the kool-aid and have all pushed all their poker chips in and are currently hoping it works out. it won't.
Any dev worth their salt can look at what's been produced and everything is garbage. everything. I've yet to meet a single developer of any level that has looked at something that's been vibe coded and said "hey, that's not bad!" It's all crap. the ONLY things that would ever work is if the AI somehow managed to churn something out to be 100% complete and working on the first attempt. something that will never need to be upgraded or scaled or something that would never interact with the user space. Name me something that does that.
The issue as it stands right now is it's taking too long to fail. it will fail, I guarantee you that, but it's taking too long. Developers and programmers and engineers can't wait anymore. They're not keeping their skills up because many just don't see a point. Many have been unemployed from the field for so long now they've taken different career paths. We're simply losing developers due to time.
It should have happened by now, should have happened months ago. the gaps in training data are increasing. you get way more hallucinations than solutions. the tool is absolutely worthless but the problem is the people utilizing the tool don't know any of this. thus token use increases.
Basically what I feel like, except I’m too young and I’m not even close to not needing a salary. I’ve been pushing back; I bought some time. I’ve spent all my willpower on this now and the cancer is still growing. I’m pretty much done. The uncertainty that comes with career change doesn’t look so bad anymore compared to whatever this "work" feels like.
It's honestly a shame and on behalf of other devs my age and older I apologize. We let you down. I'd say start looking into FOSS projects or contributing to them but man even I'm getting out of those as most are also drinking the AI Kool-Aid especially in the Linux FOSS space. it's gotten REALLY bad for Linux based projects within the past year. they're actively killing any progress made on "The year of the linux desktop" I don't see that happening, ever.
I’ve switched to Linux on my personal computer a few years back and I was so happy I did. Seeing pillar distributions like Debian open up to the slop is immensely depressing. As you’ve said FOSS is unfortunately not a sanctuary.
OS maintenance, like Debian, is tons of tedious busywork, most of it is quite literally just pulling from upstream packages, making sure there is no conflicts or certain types of labeled bugs, and that it meets the licencing and policy standards. So long as they are checking the code, I could see open-source projects benefiting substantially from LLM usage. The one thing that FOSS projects have is a lack of manpower.
What are your thoughts for using AI to colour in & do the tedious bits? So you plan out the structure of what you're trying to do, work out the individual pieces and get the AI to do a bunch of the (well specified) leg work. I've found it fairly reliable like this, especially for things that involve looking up function parameters in unfamiliar APIs. But this does depend on a thorough look over what comes out as well. No doubt there's still issues that can pop up but generally have found it to be useful overall
If the API is unfamiliar how do you know what it puts out is the best way to use the API without going and reading the docs anyway?
Sure if you need it to work with that level of detail/efficiency then you'll need to understand the code thoroughly as well. If you need it to draw a dashed line and you get a dashed line then that is fine by me. Looking up the argument differences between matplotlib plt.plot and plt.scatter for example can take quite a bit of time for fairly trivial things
Personally I don't allow the AI to write any code. I do use AI but I mainly use it for debugging and as a sort of glorified rubber duck that I can bounce problems off. check my code, see what's wrong with it, see what it recommends, and based on my own knowledge know the right direction to proceed. It still gets things wrong most of the time but at least it can point me down a path that I know will be the correct one.
I used to use it to craft boilerplates/bootstraps and at first it did it well enough but lately it even fucks that up. so I don't bother. I honestly don't even trust it to do the tedious writing. It's a useful idiot at this point.
I'm thinking of taking a sabbatical for a few years and hoping it's less insane after... I have the money to do it but I'm worried it either won't get better or I'll find it hard to get back into work afterwards.