Metldown at a new job
Posted by MrCallicles@reddit | ExperiencedDevs | View on Reddit | 23 comments
Hey ! sorry for the clickbaity title...
I'm a software engineer with 8yoe.
3 months ago I changed my job from cybersecurity to a sustainability startup.
I'm starting to wonder if it was a good choice, and even if my skills are relevant anymore or will be in the futur.
I always considered that my edges were software architecture, being a "clean architecture"/DDD advocate of some kind and being always the guy that try to make tidy and understandable changes so that everyone can review, understand the code in the futur, even in the architecture standpoint. Also trying to model the business in the code in the most relevant way, trying to understand the business and model it so that it's easy for everyone etc.
But right now things are starting to melt. CEO push to being "AI first" and there is one guy in the team that does anything with AI etc...
The code produced is kinda shitty, doesn't respect conventions and yeah, I'm not sure a human could understand the code as easy as mine etc. But I can't say it doesn't work, I can see that he is faster than everyone and I see that upper management is quite happy with that.
Also, I tried agents and stuffs and I can't say it doesn't work, even if the code is shitty, and that it's kinda impossible to make it exactly the way I would have written, but I feel like nobody care...
I'm quite afraid for my job and for the futur right know, I don't really now what to do. I talked a little bit with my manager and other colleagues and I don't really know if It helped me. Like I feel like everyone is lost right know. Obviously everything I read on the internet doesn't help either I feel like everybody is dooming and that I'm stuck in a bad dream that I can't wake up from.
My family told me that yeah in the worst case I could just move to another industry (I already done that so yep I'm "able to") but it sucks.
Also, I'm afraid that the way I respond to "let's AI do that" would cost my job. I mean, yesterday a PM proposed to do unit conversion with AI so I was like "wtf man". But in this company I feel like they are on the right side to propose that, and from my linkedin inbox I feel like it's the same everywhere. And even in my former company they were begining to hire people more keen to use AI so...
Tbh I feel like I also have a bad case of anxiety so I took a appointment to a councelor.
But yeah, my question is just, how do you all cope with all this ? I feel like I can't...
notnoteworthyatall@reddit
You need to jump ship. AI is great for helping execute small pieces of code. It is not great for architectural decisions.
roodammy44@reddit
We are living in troubled times in software. Pretty much every change in how we work has come from the bottom up until now. If you have a look on LinkedIn you will see the kind of rabid hype managers and CEOs are exposed to every day. Wild claims about 100x productivity increases and how anyone who doesn’t follow will be left behind.
I think this has created a crisis in the profession. Impossible standards and imposed workflows usually enforced by micromanagement. I chose my current job carefully to get away from all that.
Within a couple of years the hype will die down and reality will hit. There’s been no vibe coded smash hit software released. The only increase in velocity I’ve seen from big companies is the number of bugs in production. But damn, it’s going to be some tough years fighting with management.
shimroot@reddit
I see this argument in a lot of places: there’s no vibe coded smash hit. And while it’s true to some extent it also misses the picture.
Having a smash hit is difficult regardless of vibe coding or not. Building something is the first step. You then have to distribute it, promote it, support it, upgrade it, and so on.
That being said: afaik OpenClaw was vibecoded - so there’s at least one smash hit.
roodammy44@reddit
I think there is the point that if creating software takes basically 100x less time and effort, there should be 100x more software out there. It should be taking every industry by storm. The fact that it’s not is a good indicator that vibe coding has not achieved its claims.
You make a good point with openclaw, though I have heard that it’s rapidly losing users at the same pace that it gained them.
shimroot@reddit
Well yes and no. Let me give you an real life example.
I’m vibe coding a small web racing game. I have some users but not that many. As I said: distribution and promotion are difficult regardless how fast I can build.
But! For this game that I’ve vibe coded I also made 3 tools that help me with it: a mapmaker, an ai runner to check the maps, an auto gameplay recorder to make videos to share on tik tok.
So esentially I vibe coded 4 pieces of software out of which just 1 is public and the rest is internal tooling to help me.
Are others doing the same? Is vibe coding a nice tool to use for internal products? I don’t have an answer to that, I’m just saying that just because there aren’t visible products out there doesn’t mean they don’t exist.
mynameisDockie@reddit
I like the way you said this. I feel like our ability to advocate for quality has gone out the window with AI.
I've never seen management more convinced that they know better than their technical staff.
yellowmonkeyzx93@reddit
The problem with producing these kinds of results is the technical debt incurred with AI code. It works now, but when something on production goes down, new features or bug fixes are needed, this will be a nightmare for any vibe coding dev to handle.
If you're able to make sense of the madness, then you're gonna be the one that everyone depends on later. Alternatively, if its not worth the effort at all (which I feel is this), then its best to just do the minimum and get out soon to find another job. The insanity of it isn't worth it.
Either-Jelly-Pudding@reddit
Just something you need to accept it. NGL i do feel shitty nowadays coz ive been relying on AI but again, doesnt mean everything AI generated code you should accept it. Thats something you can amend accordingly by having conversation with the agent.
And talking about shitty code quality - thus why alot of company adopting AI SDLC and AI Spec Driven. Where you define your standard, your best practices, your spec basically for what your agent you adopt when they are generating your code. It's just a learning curve here. AI can be exciting and dreading at the same time. You can do this, and do more research on this. Trust me, the moment you know what you're doing, things will be okay. and NO, your skills will be relevant. ALWAYS. You'll be fine 😄 cheers!
Typhon_Vex@reddit
Yeah this will no be everywhere Manual coding will be like an artisanal blacksmith
TheCommieDuck@reddit
oh god
pierre_lev@reddit
There are some companies that are less AI enthusiasts, and have a bigger interest for the quality of code I am sure.
I would look into that before to switch domain.
padetn@reddit
You can make agents follow proper architectural patterns, it’s just a bit more involved than simply telling them to.
InterestingBoard67@reddit
That's true, but also makes your own skills atrophy, because you no longer code yourself. You think you do when reviewing the LLM's code, but you don't.
padetn@reddit
True. I code one in 5 tasks myself to avoid this but sometimes I wonder if I still should.
programmerman9000@reddit
I don’t think that AI is the problem here. It’s certainly making the problem worse but it’s not the problem in itself.
In my first job, well before AI, we had an environment that rewarded “cowboy coders”. Management loved the individuals who said yes to everything and were happy to deliver everything within a 2 week timeline, literally everything and anything. And these individuals created MASSIVE chaos.
The sad thing is that in that environment it actually “worked”. Quality was not needed, only flashy demos (sometimes hard-coded) was enough to impress the boss’s boss, who would anyway get replaced within 1 year and the whole process of shit-shoveling could be repeated.
So my advice is to recognize that if a team is okay with shit code then it’s not your responsibility to be the stop gap. Find a way to deliver value in that environment or try to get out. You won’t be able to “fix” them.
Intrepid-Ostrich2226@reddit
Agree. Some people like to set things on fire and then heroically extinguish it. While preventing seems like something boring.
grimr5@reddit
Can you not tell AI to make professional code? When Carmack and Torvalds say they are using AI.
Use AI as a collaborator, tell it you want professional level code, plan with it, switch scopes: get it to plan, get it to analyse the plan - AI is task resolution oriented, analysis result is to analyse well. Iterate. I get where you are coming from, however if you learn to use AI, it can be a massive force multiply. You also need to find an AI you gel with. I prefer Claude.
PerformanceThick2232@reddit
Confirm with management that responsibility if on AI now, and work as they ask. That is all.
If management forces you to use LLM and be responsible for LLM code, then this is shit hole.
If management forces you to use LLM and they are OK with quality drop, then this is ok place, they understand trade offs, and you will no be blamed for LLM slop code.
ares623@reddit
use up your daily budget and say "sorry I'm over quota" and move on
QuitTypical3210@reddit
Just do ur job with AI until ur u cannot be hired anymore then switch careers
MrCallicles@reddit (OP)
I mean you're right that the sanest thing to do I guess haha.
liminalbrit@reddit
I find llms make effective people more effective and the mediocre wash out. It sucks but maybe there's another field for you?
PileOGunz@reddit
I think you’re just going to have to roll with it. Yes a.i codes not great but it’s it generates it so fast your going to be left behind, it was only the devs that cared about clean code, I think our role now will be quality control for the robot.
At the moment there is still need for seniors even if it is just baby sitting a.i.