Pointless rant. Please ignore. I’m a software developer and we all know how AI has changed our industry. How we work or why we’re fired and why we can’t afford PCs.
Anyways, we’re already all forced to use AI already and we’re already atrophying the minds of our juniors. It’s great.
New team meeting and one of our managers tells us that we’re never going to write code anymore at all. The AI will read the JIRA ticket and create the pull request (change request to the codebase) on GitHub. Our job is to only review the code on GitHub and then rank how well AI did and then comment and then get AI to fix it. We have to do this so we can improve the AI process. Which is funny because none of the people who plan this AI shit are data scientists. The only way they can change things is by promoting, it’s not like we’re releasing our own coding models but anyways … He’s like, now you should be able to do much more work and just review PRs all day now and that we should never be doing only one thing. You can only tell AI through a GitHub comment to fix a mistake and then you can start reviewing the next thing.
We were like, if it’s a simple fix why can’t we just fix it?
“Because we need to improve the AI process”
But then, I have to context switch.
“Yes that’s the point you can come back to it later”
Why come back to it later when we can solve it now? We can even use AI to solve it now.
“No, we want you just comment on the PR so the bot can handle it”
Context switching is free apparently… It’s actually infuriating because apparently we’re not using IDEs any more. I personally use the GitHub plugin to review PRs in my IDE but no one else seems to do it so I don’t think they even took that into account.
These guys have auto merged AI code that’s taken us weeks to unravel and which we still haven’t fully been able to fix. They just merge shit all the time and a lot of it is fucking slip. AI merged hundreds of tests and no one cares when they break. They didn’t configure prettier because AI doesn’t use it so it breaks out formatting when humans do it.
I ranted to my own manager for 30 minutes about it today and he was just as upset because every developer is now asking what exactly are they doing. My manager asked me what I would do. I said the process sucks but what are we supposed to do as devs. If I review 20 PRs a day, how is the company going to ensure my skills are gonna be sharp? What are we doing about taking in ideas from regular devs? How do we ensure code ownership when we’re just merging tickets we don’t write and code we had no hand in shaping?
Sorry. I actually thought I had faith in my company with AI because they were coming up with thoughtful approaches but it seems like utter incompetence.
“Where’s the fix for ticket #6618?!?”
“Sorry boss, the AI hasn’t gotten it right yet.”
“You said it was a quick fix!”
“For a human it would be. But we’re on infinite monkey time now. It may never be done.”
“It was the best of times…it was the blurst of times”
In sociology, we call it McDonaldization. By reducing a job to a series of fixed, repeatable steps, they can pay less.
Form a small company with colleagues. Call it UnSlopify.
Now seems like the point for you to realise your boss doesn’t care about the code base, your company doesn’t, you shouldn’t either. Don’t get upset over work, it’s just work. If the company wants to fuck themselves over, let them, help them.
You can also do some malicious compliance stuff. You can’t write code? Great, take 3 weeks on every PR and do hundreds of prompts. Make the context as big as possible, burn as many tokens as you can.
Give the company what it wants, a shit code base and a massive bill.
I think it was the author Henry Miller who said of working,“Do exactly what they tell you and let them live to regret it.”
But get it in writing first.
The CIA work a book on sabotage that would be a good addition.
The summary of the book: use their rules and processes against them with plausible deniability. E.g. in this case, be nitpicking about the pull requests.
How do you make the context bigger? Duplicate text? Videos?
Since costs are token based now, and output tokens are more expensive, the best way would be to increase outputs by asking it to try 10 different ways and choose the best (or worst, if that’s your goal).
Wow, that’s a genius idea. Awesome!
Add “ceterum censeo Intelligentia Artificialis esse delendam” to the end of every query.
Sounds like a plan to make tech debt skyrocket and make productivity tank.
If I were you, I would start looking for another job. And, once you get an offer, tell them in your resignation letter why: you see the company is going to crash compared to its competitors due to wasteful use of AI and the loss of productivity that comes with it
Lol why are you training it? You’re a software dev with a product, not ai qa. Wild times we’re in. Should really just cut your losses and gut all ai and go back to where you were before it. If you can’t do that, then your company no longer has a purpose.
Seems like exactly the wrong way to use AI.
It sucks, I feel you. I’m also in tech but changed jobs and got lucky.
Anyhow, wing it. Nobody cares. Take care of yourself.
We have to do this so we can improve the AI process.
I’m curious what they mean by this. Are they just clueless about how anything works? Maybe AI companies are signing contracts with companies to upload all their code, data, traces, etc; so that the AI companies can improve their models? I.e. companies are acting like reinforcement tuning contractor firms for AI companies?
Are they just clueless about how anything works?
That’s certainly how I read it. They seem to think that if you feed the LLM a new prompt that tells it it made a mistake, it’ll update itself to avoid that mistake in the future. Which isn’t how any of this works, but I can understand why they wouldn’t know that given the marketing from the big AI companies.
A couple of years from now everyone will know this was really dumb, but right now only the engineers know that, and management won’t listen to them. I guess they’ll learn the hard way.
This is a “Cover your ass and wait for the fireworks” scenario. Get every stupid request in writing. Document everything you do. When it all blows up in their faces, be ready to roll up your sleeves and start unfucking the mess (but only if they’re paying for the overtime).
You will not convince them this is a bad idea as long as it appears to be working. In IT and software dev we’re all engineers; we like to fix things, so our instinct is always to try to at least make a bad process work better. Fight that instinct. Follow their stupid instructions to the letter. You want this to fail as quickly as possible so that you have hard data to point to.
This man is a straight shooter with upper management written all over him
We have to do this so we can improve the AI process.
…To be clear, does this manager think y’all are “improving” the model with feedback? Like it will learn or train off your comments?
Or what? Are they just talking about the intermediate steps to the LLM API?
I’m just trying to wrap my head around this.
Most of the things that try to address this do so by adding information to the context window. Doesn’t solve the fundamental issues, like LLMs still being solely dependent on word/token correlations and the assumption that if you throw enough conversations at it and do statistical analysis on the words used, you can encode the specific knowledge behind those words used into those statistics and then models can get back from those generalized statistics to discuss specific topics (which isn’t really how statistics work).
Well that’s often counterproductive, too. Even the largest models degrade if you fill their context with stuff they don’t need, or push them outside the patterns they were finetuned for.
Not that RAG isn’t really important.
Back when it was just “add a note about the specific thing in the hidden prompt”, it felt like an old-style chat bot was patched on top of the LLMs, where it had a series of specific responses to watch out for and fix. Now it feels more like there’s a large database of this shit and we just hope that the LLM will look up the correct stuff when it needs it (and probably prompt it to do that when demoing it for others and creating impressive examples that it probably won’t live up to very well in the real world unless each user knows how to direct it with follow up prompts when it inevitably goes in the wrong direction).
I don’t know because these guys aren’t data scientists … they’re devs so they can only do this by prompting so it’s just gonna be whatever steps they add to the agentic workflow.
By “these guys” you mean a separate team setting up an agentic workflow just for your workplace?
I’ve finetuned some models for a job, not just LLMs, and I can all but guarantee they aren’t finetuning some self hosted coding model. The APIs they use do not “learn” from mistakes, so if that’s what management is thinking, they are misinformed.
And interesting things like constrained output are often not even supported by APIs.
Modern LLM APIs and wrappers come with really good tool harnesses for coding, too. Like, good enough you don’t want to screw with it and pollute the context structure the original trainers tuned it for.
What’s more… not only do models change, but single model’s capabilities change under load, for more speed during peak usage times. So there is no guarantee some configuration they find that “works better” will work better later. See:
What I’m saying is… I wonder how long this mess last?
Because if you already have GH PR integration, most of what an agentic team can do is mess things up.
| they are misinformed
That’s very, very generous of you to say they are misinformed.
If it were my post it would be “they are obviously incapable of understanding the basic premise behind what they’re doing and shouldn’t be allowed to lead a lunch line let alone a team of highly skilled technical experts.”
Or maybe
“They are so gullible and have such low opinions of their own skill they’re willing to swallow whatever marketing bullshit they’ve most recently read despite not having the faculties to apply even basic logic to their approach.”
Anyway, I feel for you OP.
Run like fifty agents in parallel and get lauded for how “productive” you are while you find another job and hopefully before they get the bill.
This, but only after you have the other job’s offer in hand
Software developer w/ 25+ years of experience here. My condolences, this sounds like a shitty situation. But it’s definitely not the first time I’ve heard a story like this.
To be perfectly honest, my professional opinion is that if we can’t, as an industry, get away from the “all AI all the time” mode of operation you describe here, we are completely fucked.
I’m looking to retire early anyway, so I will likely get out soon. But I feel awful for young engineers who don’t have that luxury, and who will be expected to maintain AI-written spaghetti code without having the years of experience writing and understanding complex code using their own brains.
I was in industry for 10+ years until I got laid off. I’m looking to start a circus now. The job market is all slop and I want none of it.
I’m looking to start a circus now
That was a casual way to mention that lol
Haha I’ve also been a circus performer for around 10 years? It’s not entirely out of nowhere
Haha gotcha. I had to do a double take at first to see if you meant that metaphorically. Hope it goes well!
Your monkey, your circus.
I’m 40 something and I didn’t make FAANG level money. I still need to work for at least a little while but I am still planning to retire early as well.
I definitely feel for the next generation as well. Every new job I usually did my best to train the juniors and try to put them on a good career path. It seems like it was already difficult with the thirty million line problem and just how much sits in between the code can be a lot. Now add in AI and it’s just a mess :/
What’s up, fellow non-FAANG senior engineer? :)
Same here, always tried to go out of my way to help junior folks out. A lot of other people our age I know seem to be wanting to get out soon as well. In some ways I’m hoping the AI bubble bursts soon, but I realize that will likely be catastrophic for the economy as a whole. It’s a shame the bubble has gotten this big.
Shoulda bought bitcoin.
AI is trash, but bitcoin was a real life changer. It let me retire in my 30s
Wow, what a useless comment. Bitcoin is trash just like AI, and the fact that some people got rich from both does not change that.
My point is that they’re both not trash. Just AI
And not just monetary value. Permissionless means it always works, whereas cards are a confidentially and availability nightmare
Leave it to a crypto bro to brag about making money and then give the usual fake talking points about crypto, in a fucking thread about AI in software engineering. Get a life.
its going to be a law of diminishing returns if they hold on to more senior devs in the industry and wont hire any juniours down the line, creating a artificial shortage.
Ohhh, they’ll ditch the seniors too.
The seniors are already building chicken coops and/or automating small farm practices. We’ll come back in 2029 when all the big orgs get SoWs to greenfield all the broken shit. Then we’ll triple our rate and go back to work a for year then retire. Or maybe that’s just me.
I feel bad for the slew of juniors being ‘AI-promoted’ to maintain large production codebases and infra. They will have nobody to turn to when things go sideways other than their ever-worsening AI tooling. This means they will never actually be seniors in their mind or practice. But will have all of that responsibility.
I’m also wondering whether senior engineers will be able to come back in 5ish years to charge exorbitant rates to fix all the broken stuff. But for me, unless the rates are truly out of this world, I’ll probably be happy just hanging out with my goats and chickens. Their loss, I guess.
This happened at our shop fifteen years ago. We were instructed to outsource everything to offshore.
That didn’t work at all, because with a few rare , bright exceptions the people the offshore company have us could only achieve an outcome if they had a list of steps for that specific thing.
It’s going to be exactly the same thing with vibe coding. It kind of works in the hands of somebody with a deep understanding of the tech, but they expect to hand it to juniors and get good result.
So we’ll either have to pick up the pieces or let them flounder.
If I were you, I would burn soooo many tokens. This is a malicious compliance dream scenario.
Fuck the owner class, get paid, and watch it all burn.
Opus large 1m context. /Plan and thinking mode. Use cli. Converse with it Think deeply. Think step by step. Read the entire codebase every turn. Bloat context. Wait. Sharpen your skills and wait for the accountants to be the new heroes
If you’re the only one doing this - if only your token budget is through the damn roof - then you’re just volunteering to take the blame. Now it’s not “AI is really expensive”, instead it’s a “problem employee.”
True. You have to work together to unoptimize your tools
