Losing interest in SWE due to not feeling productive. Am I not cut out?
Posted by Sufficient-Year4640@reddit | ExperiencedDevs | View on Reddit | 32 comments
I work on the monetization stack at a FAANG. Lots of GPU training jobs, model iteration, that kind of thing. And honestly, the day-to-day developer experience is rough in ways that I don't think people outside this niche fully appreciate.
Reproducing issues is a nightmare This is the big one. Something goes wrong in a training job, and to reproduce it you need: a build (30+ minutes), available GPU capacity (good luck), and enough time on the cluster to actually run the thing. Chain those together and you're looking at half a day just to confirm a bug exists, let alone fix it. Sometimes capacity simply isn't available and you're just... waiting.
Dev servers are painfully slow. My devserver lags constantly. Basic editing and navigation feel like working through molasses. I don't know if it's resource contention or just undersized machines, but it makes everything take longer than it should.
PRs are full of AI-generated slop. More and more I'm reviewing code that's clearly Claude/Copilot output -- verbose, over-abstracted, weird variable names, unnecessary error handling. It takes longer to review than hand-written code because you can't trust that the author actually understands what they submitted. Sadly, the company is all in on AI and AI usage like probably even a metric for performance.
It's becoming impossible to understand the stack end-to-end. Everyone is writing AI-assisted diffs and being encouraged to do so. The deep knowledge that used to build up naturally through writing and reviewing code isn't accumulating anymore. We've had a record number of breakages recently and I don't think that's a coincidence -- but leadership is blind to it. By lines-of-code metrics AI is making us faster. By breakages, it's making us worse.
I like the problem space and the scale is genuinely interesting, but the tooling and infrastructure make the actual work feel like a slog.
Anyone else in a similar spot?
Iz4e@reddit
Sounds like Meta or Amazon. On top of that you now have spyware. Im personally am taking a sabbatical. I think AI burnout is going to be a huge problem going forward especially if they do more layoffs and push more work on the survivors.
obelix_dogmatix@reddit
Man the Meta people can’t code for shit on GPUs. I work at a GPU vendor and my goodness the PyTorch code is so poorly written as far as the GPU performance goes. We are consistently pushing PRs to improve the performance. There is 0 knowledge about distributed or parallel computing in that organization. Seems like they keep hiring people who know how to program pytorch, instead of hiring teams that know how to talk to large systems.
ultraDross@reddit
That's hilarious. It's almost as though the interview process can be gamed and has nothing to do with the actual job at FAANG.
obelix_dogmatix@reddit
Lol … what kind of a cluster y’all got at a FAANG that it’s lagging. And are you always debugging at scale? Do you not have GPUs on a local machine?
mxldevs@reddit
I wouldn't be able to work a job where I need to spend an hour just to even begin debugging, and then an hour just to test the fix.
Even if they're paying you million dollar FAANG salaries
normalmighty@reddit
Can't decide if you're a bot or just someone leaning so heavily on and LLM that you sound like a bot.
In case you are actually a real person hiding behind that bot post, I'll say that this sound like a problem with the org you're in, not an issue with your fit into the industry. Try looking outside of fang for roles at mid size companies that sound appealing to you. They'll have way less of the really painful slog that you're describing.
Sufficient-Year4640@reddit (OP)
sorry I used AI to "reformat" my thoughts. Apologies if it came off like a bot.
And yes the irony is not lost on me.
I think for usecases like rewriting posts it's much easier to verify the output and AI is quite useful if used properly. But I can see how it's possible to abuse AI even for this; I have seen truly sloppy posts that sound right but hard to make much sense of. I hope mine isn't taht bad
normalmighty@reddit
I sometimes see about how an llm rewrites my thoughts to get a better idea of how to articulate it, but never copy/paste. LLMs have a very specific way of writing which sticks out like a sore thumb to people, especially when you've got a default [word]-[word][number] username like almost all bots on reddit lol.
The only exception is in a lot of professional settings, especially enterprise, where for now using an AI to write all your messages is seen as a pro and not a con. An attitude I'm strongly opposed to personally, but I still do it sometimes because it's how the game is played.
EvilTables@reddit
AI is not a replacement for writing skills.
onFilm@reddit
AI is fine, but dont use it to contextualize your thoughts, because your meaning and intention get lost through that filter.
Smok3dSalmon@reddit
Could be non native speaker
Successful_Door1847@reddit
i had something like this before
Select-Field1668@reddit
maybe check out the related subreddit for more tips
CherryChokePart@reddit
Decisional control + high mental challenge = Happy.
Low decisional control + high mental challenge = stress.
High decisional control + low mental challenge = boredom
Low decisional control + low mental challenge = apathy.
TurnUp0rTransfer@reddit
Huh. I never thought hard about it but I think High decisional control + low mental challenge is the reason why I’m not motivated to do well in my current job.
Sounding more like I really have to leave in order to get more job satisfaction. My tech lead has full control of all technical decisions down to how we write our code and I just feel like an AI coding agent, they’re a good boss and a great engineer but I just don’t feel motivated in this job
trashacount12345@reddit
Decisional control + high mental challenge + not actually being able to pull it off = also stress
luckyincode@reddit
It’s like 1980.
superdurszlak@reddit
We've got AI agents but we never got time to build proper automation around them.
Our org pushes is to 10x our output simply because we've got rather sloppy AI agents to work with, even though we don't even have enough quality gates or spec-driven development practices to keep that under control.
It could have been better if we had access to MCP servers and could use skills for those agents, but due to policies we just work with the most basic setup.
Now, to add insult to injury, we simply started pumping out more and more slop just to satisfy this wishful thinking while increasingly giving up on quality. Is it our responsibility? Yes, but the issue is, if all devs except one push back, and that one person says "yeah of course no problem, we can do it in like one hour"... Well guess which response is taken for granted.
Sufficient-Year4640@reddit (OP)
how much additional gains do you think you'd have with agentic skills?
MoreRespectForQA@reddit
It was the same back when there was no AI. Most organizations tended to view any investments which indirectly contributed to overall quality and speed as a waste of time.
The idea that 5 parcels of work could be done by 2 people instead of 5 with investments in better tooling didnt compute. This is routinely true but they simply wont believe it.
The idea that 5 parcels of work can be done by 3 people with "virtual employees" providing the work output of 2 employees works in their head because AI is "employee-like".
superdurszlak@reddit
Yeah I agree. It's just that AI era made this frenzy snowball even more and at a much faster pace I'd say.
You can now accumulate decades of tech debt within weeks.
MoreRespectForQA@reddit
For sure. The last 10% of a project became 1000% harder. If you can skip that part of the project you can appear wildly productive. If you cant you just bought yourself more work with higher expectations.
aeroverra@reddit
I’m one of the remaining devs from a small 3 person team. We are now at like 30 devs and while I’m a manager of one of our teams there is just no way to get those guardrails in place without causing friction with execs. I gave up and will always remember our nice clean codebase. RIP
shifty_lifty_doodah@reddit
bigcos are 95% grunt work. It ain’t pretty. It ain’t fun. It’s wrangling hogs in the mud
penguinmandude@reddit
This entire sub is just AI slop posts and AI slop comments responding to the AI slop post now
Ok-Future-8420@reddit
Are you me? Because I’m exactly on the same boat as you. Our company laid off almost all of our iOS and android devs last fall because management thinks we “just use AI to rewrite our mobile code into react native so we can reuse it” and it’s been a disaster so far.
PayLegitimate7167@reddit
10yr plus experience eventually you get stuck in a rut. Same problem all over. Personally it doesn’t help when we have mentally taxing jobs
4ss4ssinscr33d@reddit
I work at a FAANG and I’m praying for a layoff so I can get some severance and take a few months off from working. Absolutely leaving the Big Tech world soon because dear god this place is a cluster fuck right now.
kevinossia@reddit
Be the staff engineer you were always meant to be and push for improvement in all of these things.
ElGuaco@reddit
Dude I don't even have a job right now, like many people, and execs are blaming your product for it. You probably make obscene money at a faang company that is also directly responsible for the AI bubble that is wreaking havoc with the economy and our jobs. And you have the gall to complain about being bored.
I can't tell you properly how I feel without getting banned.
PressureHumble3604@reddit
Similar spot with a bit less problems with AI and more with stack/resources
I just want to get out of directly touching the tech.
retroroar86@reddit
I also hate this AI-timeline. Fortunately it is not at bad with AI use where I work, but I certainly feel the pain and feel we are going fast in the wrong direction. Can't wait for the entire thing to implode, or explode, or both.