I currently work at an “AI First” 🤢 fintech company, and we’re mandated to program agentically. I was fortunate enough to start working before the AI hype, but the environment right now is daunting.
There’s too much unmaintainable code slop and throw away scripts committed that I can’t keep track anymore. Each PR has so much diffs that it feels like someone had diarrhea and shat all over.
I try to manually review the code, but it feels like best effort at most. Like I may only understand 70% of the thing that I’m reviewing. What ends up happening usually is that I just ask the model and verify the output(which is horrifying I know)
I spend the whole day arguing with AI, reviewing, and pushing back against its output that sometimes I wonder if I am the one driving AI or is it driving me.
–
Recently, I noticed that I’ve becoming more reliant on the model and that my skills have been atrophying. What would’ve taken me an afternoon if I program it by hand, would take me a week now if I do it now. I also find my head hurting and myself glossing over the code in code reviews and prompting claude to do even the basic of commands.
I get that there’s more to programming in software engineering (collaboration, dealing with people, etc etc), but I don’t feel like I am in control of the systems I am building anymore with the engineering aspect being replaced with babysitting and self lobotomy.
–
Off work hours, I try to brush up by reading on system design and architecture books, typing by hand if I’m learning a new technology or stack, but I feel like it’s not enough to prevent from skill atrophy.
How are y’all coping and how do you recover from this? I am afraid on what skills that I’ll be able to bring at my next place of employment if I resign and that I’m becoming a jack of all trades master of none type of thing - having only shallow understanding.
Malicious compliance, to some extent. The only people I can complain to are 3 levels down the AI train. Never let the AI update the git repo, review all changes it makes. I end up removing like 90% of any comments added as useless or too confusing to live; especially comments that have DO NOT REMOVE in them. Ends up taking about the same amount of time as if I just did the work myself. Quality is better, because the tool is good at what it does best.
Where I work we can still push back to big PRs. Our guideline is unchanged that it shouldn’t be more than 700 lines of change. The less the better. With agents it’s even simpler to deny a big PR. A manually written big PR would have meant lots of manual effort in reworking it. With agents I basically just demand the author to tell their agent to split their fucking PR up.
How do you push back if that PR comes from your boss or a more tenured eng
I am a Staff Engineer. So I am high enough in the chain to even close PRs at will.
Other engineers are encouraged to reach out to us staff engineers if they feel powerless on their own.
We can only do our best to not end up with a culture where people fear each other.
Here they just say “ohno so much effort, i already split it in 4, no use in split for the sake of splitting”
I made a horizontal career move and accepted as a trade-off that at this time I won’t write code.
Old company: mildly toxic environment, out of touch CEO, poor leadership. Increasing AI push and wild gaps between different people’s competence. Constant arguing. The dreaded 5k loc generated PRs started appearing. My role: ~5% web programming tasks (that I actually signed up for originally) rest was design, review, and people-related work. I was not a manager. Hated it.
The project: 1 year of goal-oriented homelabbing, from a 50€ raspberry pi to hybrid cloud, gitops, kubernetes, and all the good bleeding edge stuff you need to run your own cloud. Did this completely without AI, built the resume, applied to a few places. Happens that one of the companies was looking for the exact stack I built my homelab around.
New company: Platform team, 90% hands on, competent manager, capped AI budgets, people treat each other respectfully. I quite like it. When I do coding it’s simple enough that it stays manageable even though I ask some bot to write it 90% of the time. It gets the job done, I wasn’t hired to write code anyway.
It’s probably a patch, but I’m hoping to run the wave long enough by this move. For the first time in a couple years I feel clarity. I trust that I’ll be able to pick up programming by hand if needed. Most people are in the same boat anyway, so the expectation won’t be there to perform on a similar level instantly. I trust that my ability to learn new things keeps being valuable.
On my free time I don’t currently write code, non-technical hobbies and maintaining the homelab I conjured eat it up. I don’t mind too much. I have enough interesting stuff to do. I think most of all it’s important that I get to create something. If I’m not exactly paid to do that it’s not the end of the world. Much worse situations exist, I’m still paid similar €€ than in my software position.
I’m lead in my team and AI is being used to varying degrees. One colleague is a problem, but was a problem before AI as well. Me and my other three colleagues have a common understanding of you have to take ownership, understand what you’re doing, and document it.
As a reviewer, I ask for the information it lacks.
We (our company) has another project started with “AI first” too, and now still works mainly through “document features and behavior and stuff in md files and AI generates it” - so mostly about spec writing and AI generating. I don’t know how well that works, but I’ve heard bad more than good things (which may be common anyway) and it certainly hasn’t lead to meeting timelines (now more than half a year behind the original planned release).
Personally, I’m still skeptical about AI use, but do use it occasionally. Our project is simply too complex (and legacy and long history of changed dotnet technology as well). It’s more feasible on newer projects, but Blazor isn’t a straight forward framework either, mixing razor, html, css, component isolation, SAP lifecycle, render modes and whatnot.
I haven’t used AI to a degree where I feel it significantly [negatively] impacts my capabilities. Thankfully, my team hasn’t pivoted into that completely, either. I see the risk and hope pushes though. The promise of AI speeding things up. The customer saying “a bit more risk is fine”; but it’s on us to communicate the difference between risks (the bug fix missed one additional unknown place vs long term maintainability and its associated risks).
In a similar boat. I’ve been programming since my family got a IBM 386 in the 90s and I started dabbling in the dark arts. I’ve been a career programmer for 20+ years now. I’ve loved it, all the way up until the past year or two.
Here are some of my rambling thoughts and perspectives, for whatever they are worth.
-
“Don’t try to fix someone else’s mistake while they are still in the process of making it.”. AI is both a useful tool and a giant shit show, but if your company wants the shit show, then let them have the shit show, do not carry their burden.
-
Lean into hobbies that aren’t on computers. I do woodwork, gardening, arts and crafts, sewing, home repairs, camping, etc. “Go touch grass” as it were.
-
“Fight fire with fire”. (I don’t understand that saying) But the point of it is to say, don’t manually review AI slop, if they are gonna slop 800 file PRs at you, AI slop review them and move on. The Titanic is sinking, you aren’t going to save it by frantically bailing with a bucket.
-
Programming is both an art form and a skill and it is also like riding a bike. You may not ride for years, but if you get on a bike it will come right back to you. Don’t stress or worry that your skillset is deteriorating, it will always be there. It may get rusty and need some lube, but it won’t go away.
-
When the AI bubble finishes popping, AI will still be around, but it may not be shoved into literally everything as much, and there will still be a place for talented people to do quality work. Examples from 2 of my top hobbies is that we have both automated farming and assembly line furniture factories, but handmade solid wood furniture is still around and worth doing, and there are still gardens in a lot of back yards.
-
Most of the people in the top positions in our governments and companies are completely and utterly unqualified to be there. You’ll be fine. Focus on being your best self, being happy with YOU. AI is here to stay, but talented people have not lost any value at all, know what you are worth, have some confidence, don’t let the slop get you down, there is literally slop in every aspect and career, try not to worry about it, ya know? Mass produced, low quality clothing and furniture exists, but people still sew and woodwork. It’s ok, be better than the baseline slop and you will have value to those who understand.
-
Being a “Jack of all Trades / Master of None” isn’t a bad thing, neither is the inverse a “Master of one thing and nothing else”. All of that is also very black and white thinking, it’s not reality. People are more than just their skills.
Final thoughts and summary: The AI slop is here to stay. As I learned in therapy “Don’t fight the current/flood/rising tide, learn to swim and float above it.”. The river is going to go where it is going to go, you can either be swept away and dragged under or you can sail along the surface and enjoy the breeze. I, personally, recommend the latter.
Good luck out there! You got this!
Very wise words.
Thanks. This is a bit reassuring.
Fight forward with fire- My only gripe is if a prod issue arises, the ball will bubble up to my court and I hate to be caught with my trousers down
That is valid. So be the one to calmly point out that “800 file PRs are dangerous and will lead to bugs in production” and stand firm. You don’t have to refuse to review an 800 PR, but you can point out the position you are being put in.
Just because THEY chose to edit 800 files with an AI prompt doesn’t mean you have to be 100% on the hook for if/when it breaks. Drag that other person back into the spotlight with you and have them own it.
*Cartoon style shepherds cane hook their neck and pull them back on stage*
I have also been doing this since the 90s. All of this rings true.
I would say your whole thought process kind of floats above lemmy. Hopefully you won’t get a flood of douchevotes from zero-tolerance AI haters.
People are 100% valid in hating “AI” at present because currently it is causing a LOT of negatives and not many positives for the average person.
Think whatever you want about “AI” (machine learning/LLMs/chat bots) it is a tech that isn’t going away. Pandora’s box is wide open, no going back.
Wanting it to be used ethically and responsibly is good common sense.
I like the parallel of nuclear energy. We unlocked the secrets of the atoms and fission. We can either make nukes or we can make clean energy. People, rightfully, protested and “freaked out” because everything at the time was trending towards nukes, and look at that, they were right, however we did finally learn and now use fission as a power source and not for destruction. Yes the capability still exists to drop nukes, but we haven’t in a long time because of all the downsides.
I feel like LLMs and this wave of tech is going to be similar.
We will still have data centers, duh, but they won’t be in everyone’s damn back yard. We will still have LLMs, but we will train them responsibly and have guardrails and what constitutes fair use. We can have neat AI tools that assist us or speed up tedious tasks or automate boring things, that’s fine, but we can do that in healthy ways.
At least, that’s what we all hope for, so let’s help push for that to happen.
I agree with the sensible approach you’re talking about, but not with getting angry at every mention of AI as if it’s leprosy in Middle Ages Europe.
-
stop using AI?
Same situation as OP, when I do peer reviews using AI and it reports a list of flaws I just ignore it. If the Coderabbit doesn’t catch it then I’m just allowing the human error back into the code. It still sucks but there are future bugs to fix.
use AI to learn new stuff making sure to check stuff that smells fishy.
By being unemployed
Find an open-source project (better if it’s noAI), then work on it manually.
Personal projects done without AI. It’s like artisanal coding. People will pay tons of money for it one day. We will be able to sit in museums implementing simple scripts without AI so kids on school trips can see how human brain works without assistance. Of course the AI will have to explain to them what’s happening.
People will pay tons of money for it one day.
I hate to say it, but this sounds like copium. Do you still see loads of well-paid furniture woodworkers? Nah, it’s all replaced with factory IKEA stuff. Sorry, but nobody will pay for “artisanal coding” when they can get whatever they need from the AI for free.
But they will organize school trips to see us in museums, right?
(in case it’s not obvious enough, I was joking)
Loads? No, but the ones who do good work are well compensated. Just need rich clients.
Not a reasonable analogy. Coding isn’t like making the furniture, it’s like making the assembly line that makes the furniture (or an assembly line that makes assembly lines that make furniture, etc.).
It’s true that a factory doesn’t (and probably shouldn’t) care about deeply understanding each individual screw in each individual item. But it does (and should) care about how its assembly line is laid out — or at least the process of how that layout was determined, or the process of how that process was determined, etc.
Ah, object oriented object fabrication! :p
I don’t know but I’m in the same boat except its just killing all of my motivation. I really love software but not like this and it’s making me want to switch fields
Can relate
which other fields have you found not related to CS
Maybe electrician? I don’t know if that field is even safe though excluding the hands on parts. I have been finding I’m leaning more into server hardware and networking lately though just for my hobby setup. There’s no AI involved when swapping server components or running SFP+ cables to replace ethernet and it makes me happy
If job wants slop, I just queue up work back to back in the slop generator. While the slop generator is running, I’m reading a book or building something by hand to make sure I stay sharp. If my engineering leadership doesn’t care about quality, then why should I? Give em what they want and take care of your own needs
My job is 75% maintenance and 25% developing new app features. That 25% is really interesting and is what drew me to my current company, but maintenance is what management cares about most and what our customer service team is constantly pushing us to do.
The problem is that it’s boring and repetitive work. Ever since the company started promoting the use of AI, I’ve been delegating these tedious tasks to agents who work on separate Git worktrees, while I focus on learning new things and implementing them alongside new features that we previously didn’t have the time or budget for. Management is satisfied, and I’m learning a lot thanks to the time I’ve saved. Most importantly, I’m still writing code myself, code that I actually enjoy writing.
At work? Do what they want you to do I guess. If they are mandating you mass produce slop, do exactly that. I’ve tried fighting against it. There’s nothing you can do when your coworker opens a 785 file PR (not an exaggeration) and it needs to be reviewed today. Your carefully crafted code can never keep up with the constant waterfall of whatever the fuck Claude/ChatGPT/whatever calls acceptable these days.
Your best bet at work is to direct the agents others will use. Use a strict linter. Mandate a strict formatter. Add to the instructions that tests must be written and must pass. Do what you can to force it to emit semi-acceptable code (or fail if it can’t).
Outside of work? What’s a coding agent?
Oh my god, the slop PRs with dozens of files abd tens of thousands of loc are insane.
I don’t even know how to review it.
My answer is usually telling them to go back and rethink the design, LLMs hate well encapsulated modules and they’re very good at spaghettifying your dependencies.
They can do a decent job understand the spaghetti, but they have no concept of what exists that didn’t turn up in the last 5 commits or a list of file names. So they reinvent the wheel every time.
