AI agents, a quiet job market & future jitters — how are you dealing with it?
Posted by milmo00@reddit | ExperiencedDevs | View on Reddit | 23 comments
Hey folks,
I used to think “Copilot just saves a few keystrokes, no big deal.” Then I played with Claude Code: it spits out whole features, tests and refactors almost on autopilot. Sure, it still chokes on the crusty legacy monsters we have at work, but this tech is brand-new. Where does that leave us in five or ten years?
At the same time the hiring scene feels ice-cold: fewer postings, lower salaries. Companies seem to want people who wire up or guardrail AI, not pure CRUD coders. I’m honestly wondering whether to pivot — maybe into security, DevOps, or some domain-heavy niche. But are any roles really “AI-proof” anymore?
Your thoughts: 1. Are you feeling the hiring freeze too? 2. Which roles or skill-sets still look semi-safe to you? 3. What are you doing to stay relevant over the next few years?
wolff_james@reddit
These posts make up half this sub now.
Empanatacion@reddit
Em-dash
Idea-Aggressive@reddit
I have to stop using em-dashes because of people like you
DjBonadoobie@reddit
Ironically, my biggest beef with AI right now is AI slop posts like OP. It's getting harder and harder to reddit...
PetroarZed@reddit
Especially trash like this, marketing disguised as a concern post.
DjBonadoobie@reddit
Yea, and serves to further push FUD and the current narrative of companies laying off engineers because they "replaced them with AI"... smdh
tikhonjelvis@reddit
I like em-dashes and refuse to stop using them, so I'm pretty peeved that they got chat-gpted.
OneCosmicOwl@reddit
We're all going to go crazy with this.
Mo-42@reddit
Sigh.
micseydel@reddit
Also, what are "pure CRUD coders"?
Software_Engineer09@reddit
Job market is total s*** right now. I’m gainfully employed but still apply for jobs every now and then…these are jobs I’m a perfect match for and easily qualified for. I literally don’t remember the last time I was contacted for an interview. I’m convinced all these postings are fake.
Honestly at this point, I’m saving every penny I can and I’m looking to exit the industry in the next 6-10 years. I’m pretty confident my job will still be relevant at my current company that long.
I’m not afraid of AI tooling though, it just supercharges my efficiency.
bluetrust@reddit
I'm in the process of retraining into ML Engineering. There's a huge math prerequisite which is a bit of a slog to get through. It feels good though to have agency though and be working to give myself options.
lolikroli@reddit
What resources for math do you use?
bluetrust@reddit
Good question. From what I understand I need a solid foundation in:
Linear Algebra → Calculus → Probability → Statistics
My thought was that a car wash manager doesn't need to be an accountant, they just need basic bookkeeping skills, and so a ML engineer doesn't need to become a math professor, they just need to know the topics they'll run into. So that made it easy to rule out the really comprehensive courses like MIT OpenCourseWare.
So I focused on more condensed, targeted material:
Seems like with backtracking and learning from other resources when I get stuck it's gonna take about 6 months and I'm six weeks in.
false79@reddit
I remember there was a time back when people did not take mobile very seriously. Desktop/laptop makers at the time were like real users will want real power. Then after 2010 or 2011 I think, there was a point of inflection where mobile devices outsold both laptops and desktops.
Command line LLM tooling like Claude is pretty much that point of inflection where are lot of no coders are seeing the value of not hiring professional coders. And I am not surprised to some coders here dismiss AI as AI slop or not a real threat.
Those Laptop/Desktop makers went the way of the dodo and I'm pretty sure devs who haven't followed with the trends will find themselves with a much smaller market than when they started.
The main hiring surges in the very near future for existing/experienced software developers will be exposing tooling to LLMs by adhering to MCP or whatever standard comes out on top. Those devs will need to bring their proven domain experience as well as have knowledge how to cater to AI clients.
opakvostana@reddit
Went the way of the dodo? Right, all those massive OEM hardware manufacturers that went bankrupt in the 2010s: Dell, Fujitsu, HP, Lenovo...
false79@reddit
was thinking more like Gateway, Toshiba (no longer making comps) and the corps no one hears these days.
annoying_cyclist@reddit
Having given LLMs and agent driven coding the old college try, I'm not worried about them replacing me from a functional perspective. I do worry that the job of a senior engineer will slowly turn into thankless unfucking of AI slop as folks who came of age with ChatGPT become established in the workforce, become more senior, and eventually gain decisionmaking authority in spite of never learning how to write code or build systems beyond asking the model.
I'm far enough into my career that I could either retire outright on investments or take a long break to reskill into something else, which is how I deal with it.
opakvostana@reddit
I don't know what it is people work on that they find AI so massively useful. I make dozens of queries to 3 different AI applications ( ChatGPT, Co-Pilot and JetBrains Junie ) throughout my day, and like 80% of the answers I throw away after reading a few lines. We had questionnaires a couple of weeks ago to fill out about how we used the AI and so on, and I wrote I hope they're getting a good corporate discount on the license because the way I see it, the money is mostly wasted.
low_slearner@reddit
The prevailing wisdom seems to be that AI only gets you so far. It’s great at the start, then quickly starts to choke as the complexity increases. Maybe that will change as the size of the context window increases, but I have my doubts.
AI is largely trained on the simple examples you see in API docs, blog posts, and Stack Overflow posts. Real software engineering is taming complex systems, and simplifying overly complex ones. There’s plenty of material out there advising on how to do this, but it’s generally abstract. Applying it requires understanding, and I’m not aware of any real evidence that current approaches to AI is capable of real understanding.
If AI does reach that level, we’ve pretty much hit the singularity and all bets are off.
dagamer34@reddit
True, it’s only doing examples it has seen already, and it has not seen any significantly quality of complex codebases, too proprietary.
The more you work on unique products and features, the less it’s probably going to help you. If you are making yet another crud app though…
Beautiful-Salary-191@reddit
I am concerned like you but here is my answer for your questions:
1. Yes, it slowed down very hard. Beginning of 2024 we were hiring a senior SWE but all candidates were mediocre-average. Beginning of these year, I decided to switch jobs and the competition was very hard but I ended up getting hired at a respectable company.
If your job is just pattern matching, you should pivot (CRUD is one example of that, no offense). But if it requires a lot of thinking and domain knowledge and requires different skills combined, nothing will change for the next 5 to 10 years. This is my opinion based on what I observed in the corporate finance sector (big players are still using old school on-prem VM they didn't migrate to cloud yet... and I am talking about the central booking systems that all the other entities rely on...). Also, we work on huge code bases, ones that exceed the LLMs context windows by a long shot, unless there is a major breakthrough, it's hard to get the LLMs to do our work. Also, the costs of running LLMs are very high while the output is not that great.
I stay relevant by keeping up with the latest and greatest, doing some projects on the side (I have picked up old hobbies: I created small bot the displays emotions : happy, angry... and gives vocal orders to my daughters to pick up their stuff from the living room) It helped me work on my prompt engineering and AI automation skills. But I also have other automations and AI assisted personal software...
Let's imagine in your company, they decide to start optimizing processes with AI, who do you think can do that? Current employers will have to do that, especially when it comes to programming.
Sheldor5@reddit
nobody is safe if your boss/management is stupid ...