[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


software engineers maximum cope
>>
>>109843496
"Professional" code bases are full of spaghetti garbage written by DEI hires and jeets. The truth is most programmers suck at their jobs, are shitty people in general, and AI is an objective improvement on both the soft skills and technical ability.
>>
>>109843528
software engineering is/has changed fundamentally. it's akin to CAD for drafting. the skills need to manually draft plans were no longer valuable.
>>
LLMs are definitely here to stay, they're like how compilers replaced assembly programming. AI "agents" I'm still skeptical for anything non-trivial, too easy to wipe things.
>>
>>109843496
What's cope about it?
>>
>>109843612
watching their future job market shrink and comforting themselves with the fantasy that the AI replacing them will eventually fail so badly they’ll be called back in as heroes, while the ones embracing it become exponentially more productive
>>
>>109843668
>the ones embracing it
There are no SWEs left to embrace it. My work already has a fully automated pipeline. Customers open the Jira ticket, and thenentire pipeline from Jirat ticket to push to production is automated along rules that the CIO and VP of R&D have designed around it. We've been slowly removing R&D and other tech staff at rates to keep us below triggering the WARN act.

So I guess I misspoke. The idea is to get rid of anyone below the top of the pyramid.
>>
>>109843496
codemonkeying is basically over, whether you like it or not
>>
>>109843779
Yeah, I agree. My point wasn't really that everyone who embraces it survives, more that thinking AI will eventually fuck up so badly that the old workforce gets called back in as heroes is cope. The reality may be that most of them never had a chance either way.
>>
I never liked codemonkeying, even though I've been doing it for a long time, let ai deal with the slop world
>>
File: 1788312040448522.png (601 KB, 867x505)
601 KB PNG
>>109843496
brutal cope and delusion
>>
File: vibe_coding_2.png (817 KB, 1133x1280)
817 KB PNG
>>109843496
>>
>>109844448
>Microsoft acquires github in 2018
>it turns to shit
who could have seen that coming
>>
>>109843779
Have you considered industrial sabotage?
>>
>>109844448
>Microsoft acquires github
>actually starts tracking outages
>>
>>109844527
Look man, every single businessman who read Atlas Sheugged has dreamed of an army of robots to replace all his workers. turning labor input from a monthly expenditure to a single capita outlay plus occasional rented maintenance service and just have them go 24/7. They've always viewed labor as just as a raw input. Now they can finally make that dream a reality.
>>
File: vibe_coding.png (312 KB, 1022x1020)
312 KB PNG
>>109844479
>>109844532
vibeshitters are mentally ill
>>
>>109843528
Rare intelligent post.
>>
>>109843496
Total codetrans death
>>
>>109843496
I already get paid to clean up, fix, and correct AI slop pictures. It’s depressing, like cleaning up after a dog.
>>
>>109843595
Nay, they’re going right to implementing programs directly as an ASIC.
…assembly language… sheesh… AIs don’t need that fluff human layer to generate machine language directly.
If AI were real, that is.
>>
>>109843528
I agree. People pretend claude/chatgpt/whatever is making slop, but a good part of the code written by people I reviewed was shit that would be made 100x faster and way better by an IA.
>>
>>109843528
In my limited professional experience the human coders are aware enough to know when they don't know how to do something and seek help, whereas the LLM will confidently spew total nonsense and insist that it works.
>>
>>109843496
god/reality/simulation does not care if something is "good" only if it works. do you know how many animals have to eat literal shit because it helps their species to survive?
>>
>>109843528
Then why are there still hundreds of thousands of engineers still employed
>>
>>109843528
>objective
There's that word again
>>
>>109843668
If this is real your company is fucked. I used agents often enough to know this will quickly turn everything to shit. Maybe if the product is CRUD webshit it will last for a while before the problems are too evident to ignore
>>
>>109843595
>they're like how compilers replaced assembly programming
This is kind of a crap analogy. It's much more like version control and other advanced tools.
>>
>>109843496
>wanting to perform tasks for a living like a "white collar" factory worker
>>
>>109843496
AI is definitely useful, but it often goes down the wrong path, gets lost, doubles down on its wrong choice, and only makes things worse. At that point you have to be competent and look at what it's done and tell it what to do. Just repeatedly telling it that it isn't working will create excessive bloat, bugs, or both.
>>
>>109843595
>they're like how compilers replaced assembly programming

Lol what a dogshit analogy.
Imagine if your compiler produced a random result every time you ran it.
>>
>>109846492
Or if you had to rent your compiler for money from some shady company.
>>
>>109843496
prediction, faggot. there's nothing cope about it.
might be true. might be wrong. but still. prediction.
>>
>>109843496
He's wrong.
>>
>>109844448
>look at Y axis
>still ~99%
It's not like shithub is critical to life and git is supposed to be decentralized and async anyway. Companies harmed by these minor (yes, minor) outages are retarded.
>>
>>109846579
There are no minor outages when you are paying for a service that you need for your business to operate. Imagine if the electricity in your work office building went out randomly for a few minutes a week.
>>
>>109844836
that's because of dns not the code
>>
>>109846472
Software engineering is a young field some people dont think its not real engineering. I'm thinking they're about to find out.
>>
>>109846696
Yes there are. You don't *need* shithub. People overblow these fucking outages way too much.
>>
>>109846754
It isn't a field and it isn't engineering.
If it was engineering, basic things like a static type system and type narrowing to prevent null deref would be law. Static analysis like the borrow checker would be required in languages where bad memory usage can result in type confusion or corruption of unrelated memories.

It's insane how low quality software is and how obstinate the field is to basic improvements to correctness and safety.
>>
>>109843496
ai doesn't make shitty obfuscated code on purpose for "le job security" so it's better by default
>>
>>109843528
>and AI is an objective improvement on both the soft skills and technical ability.
Only wrong part of this post.
And that says a lot about how bad "AI" is.
>>
>>109843779
Bad larp
>>
>>109846806
You're right. It makes a messy pile of shit as the project size grows because it can't do any better.
>>
>>109846790
You are a retarded nocoder AND you don't know what engineering is or does.
>>
>>109846472
And to see what it did and tell it how to fix it, you need to spend 10-100x more time than if you just wrote the whole thing from scratch because it creates a 4x more LOCs and all of it is slop and really hard to browse through, with many repetitions and duplicated code with subtle erroneous differences, and one-line "synonyms" (like fn1 just returning the result of calling fn2 with all its args). Furthermore, it routinely generates impossibly retarded wrong logic. For example if you want to implement a quick sentence similarity, instead of using trigrams or bm25, it will just do bag of words but won't even do basic shit like stemming or splitting at punctuation, and won't use standard libraries that implement any of that.
>>
File: 17653064503660.jpg (130 KB, 600x407)
130 KB JPG
>>109843595
This guy gets it, but I do hate it when my compiler spits out completely different machine code everytime I use it.
>>
File: 1787595376070516.jpg (340 KB, 1920x1080)
340 KB JPG
>>109846842
This.
Anyone can build something solid that stands the test of time.
But only engineers can build something that barely collapses.
>>
>>109846865
>For example if you want to implement a quick sentence similarity, instead of using trigrams or bm25, it will just do bag of words but won't even do basic shit like stemming or splitting at punctuation, and won't use standard libraries that implement any of that.
If you tell it to do those things, it will. You just have to give it 10x more detailed instructions than you'd give a programmer since it lacks common sense and makes those kind of dumb choices.
>>
>>109846872
Unironically true
>>
>>109846883
>If you tell it to do those things, it will.
Yes, but think of it from a project perspective. The actual point was never to specifically use trigram indexing, bm25 scoring, nor was it to use bow with stemming or any of that. It was to do sentence similarity. You will only discover it fucked up when you take the time to read the code. The result is that it takes more active time and significantly more wallclock time to do the same thing with AI than without AI, even before you account for the cognitive overhead of working with a codebase you have to parse because you didn't write the code yourself.
>>
>>109846883
No matter how many instructions you feed it it won't stop being a prediction engine, it can't think, it just predicts the next token via a statistical probability, even RAG only goes so far, it is basically a glorified no code solution just like all the previous ones.

The problem is eventually a human will have to touch the code and human actually is conscious and has specific intent, then the human will start looking at the code and seeing small bullshit code which will then turn into a big bullshit code block which is a mess and it will either warrant a ground up rewrite without using any AI or a very slow migration to a manually coded solution. It is inevitable.
>>
>>109847006
Yes, so my point is that it hasn't yet passed the usefulness threshold. It's barely usable for search (too low recall, precision is bad but can be excused in that I can tell it's wrong, but you can't tell what's missing if you're using it for search) and that's the best thing it can do. That's despite search engines committing total suicide to try to push people to AI.

And it's all plateau'd off. But there is 0 funding for actually building useful tools that would use the new capabilities in language processing of these systems, all the funding is in just resourcemaxx squaring these round pegs.
>>
For the people swearing that this shit is a panacea, AWS's AI-DLC group even says they greatly prefer greenfield projects over brownfield, because the LoE required to safely modify brownfield is far, far higher.

I have now worked with two AI-DLC task forces, and in both cases it was thrown over a fence because, one, neither team did long term maintenance and feaure enhancements, and two, in both cases their lead dev expressed that they no interest, intellectually or professionally, in getting locked into a single large project.
>>
>>109846555
pull your head out the sand
>>
but how is it "his prediction" if he's just copying the two lines above?
cunt needs to clean up his own mess before trying to preach some kind of enlightened sermon.
>>
>>109843528
>most programmers suck
>AI is trained on most programmers
Just proving the point that highly skilled people are going nowhere.
>>
>The reality

Competent dev writes half the program in 1000 lines of code
AI writes the other half of the program in 19000 lines of largely unnecessary code

LOOK EVERYONE 95% OF THE CODE WAS WRITTEN BY AI!!!!
>>
>>109848414
the delusional dev still stuck on gpt 3 issues
>>
>>109847249
Put your cock in between scissor blades.
>>
>>109846923
You're making too big of jumps. You're expecting it to make senior dev decisions rather than junior devs.

I have done writeups as riders onto a project manager's requirements where I have high level technical explanations of how I want the junior dev to approach the solution. Not full on imperative writeups, but declarative + watch out for these hangups. I have found LLMs can work well in this mode, although obviously I acknowledge it's still short sighted because now I'm not helping junior devs to take over for me.
>>
>>109848401
Except it's not, the data AI is trained on is highly selective
>>
>>109843496
more like in 2027 ai bros will increase the price 20x and everything will crash and burn
>>
>>109843528
This is sadly the reality. Late 2027 will be paying a better version of claude to clean up the pajeetcode vibe coded by Indians.
>>
>>109848697
It ain't just the AI bros. I think there's a bit human laziness + tradgedy of the commons occuring as well that's going to contribute.
>>
>2028 - 90% of SWE refines AI code for a golden parachute
>2029 - AGI achieved
>2030 - bio intelligence is pioneered
>2035 - AI captures the last independent human
>2036 - Biological computing is being used to further enhance AI and send colonies through space with implanted augmentation to further survival
>2037 Humans are 92% machine augmented with the brain and spine optimized for human survival in all conditions with organs being obsolete except for blood and bone
>2040 Synthetics and their AI overseers rule over the cosmos before getting BTFO by rival alien empires
>>
>>109843496
Me without Claude
>Man I wish I could write a network monitor and run it through caddy with https. Man I wish I create a personal organizer. Man I wish could create a badass workout tracking apk just for me..
>Oh hey how much for you to code these for me, Me. CoDeRzMaN?
>Yeah that'll be $300,000 and I'll need about 6 months to do it, ya know I need vaccay time and sick time and I only work 4 hours a day. Oh when those packages are done, they're going to be full of bugs, won't be optimized at all, and probably full of security holes. LMAO
Be me after Claude
>Hey Claude let's build x, y, and z
>One day later, here you go bro I went ahead and debugged and checked security. Let's test it and make sure everything is good then you can live.
Fuck human devs, fuck every last of those tranny, limp wristed faggots. We have Claude now, you are no longer necessary. Time for you to out there and get a RRAL fucking job like and an actual MAN you estrogen riddled cucks.
>>
File: internet_lies.jpg (54 KB, 475x356)
54 KB JPG
>>
Hes true. Ai is only as good as calculator
>>
>>109843528
You sound like like you're never worked in the industry.
>>
>>109846579
>git
>decentralized
wat
>>
>>109846697
dns is software
>>
>>109849027
Are you retarded?
>>
>>109849038
Are you? It's a typical server client architecture.
>>
>>109849070
Anon, google DVCS and look into how many origins a given repo can have.
>>
File: 1773929280549525.gif (1.52 MB, 165x165)
1.52 MB GIF
>>109848808
>hey're going to be full of bugs, won't be optimized at all, and probably full of security holes
You'll have the same with Claude.

Any IT solution worth its salt is worth time and money, because it takes a special kind of person to sit at a computer all day writing code, debugging, thinking of workarounds, architecture, how it would scale, adjusting the solution to the ever changing requirements, shipping features to make sure it doesn't break working code. Most people aren't built for this.

The problem is if you don't like your developers, you can fire them and get another team with more experience, also if you have an in house team, the team can grow and become even more competent knowing your solution in and out thus maximizing your profits. No such thing is possible with an LLM, you can't fire Claude, you can't make it more competent, because there's no other Claude, there's just Claude and once they jack up the prices you will basically become a slut standing on her knees in front of a 10" clanker cock begging for a limit reset.
>>
>>109846579
>>109846775
thank you for your input neet sama microsoft's incompetence is warranted the days at work i couldnt read pull requests were fine we should have just sent patch files peer to peer on usenet or something you gargantuan faggot
>>
>>109845684
My professional experience shows the opposite. LLMs at least are good enough nowadays that most of the time their code and decisions are above the average intern. I have an intern at work, and I stopped reviewing most of his code because I realized he's writing none of it and it ends up looking very similar to what I produce using Codex.
Nowadays I only review or manually write critical parts or places where patterns are not well established enough for the LLM to carry them.
>>
>>109843528
GNUKE

enterprise codebases are already burnout churning machines that chase away those "supposed Senior SWEs." AI is at least paid to give a shit or pretend to anyway.
>>
>>109847006
>>109847036
You say the things someone who doesn't have a job in software would say. These are solved problems.
All you need to do to address 99% of what you're saying is to make the AI iterate with your guidance.
>>
>>109849070
you don't need a server for git.
proof: the Linux Kernel.
>>
File: slow brain.gif (49 KB, 256x256)
49 KB GIF
>>109843496
>>109843528
>>109843668
>>109844532
>>
>>109848697
doubtful. Anthropic and OpenAI don't have a moat. Large Chink Models are very competitive already and as time evolves it will become a commodity where maybe you have some boutique training houses that all the models maybe use.
>>
>>109849129
Call me when your clanker dumbass LLM can produce the same exact output for the same exact prompt 100000 times of out 100000. Otherwise, eat a bag of clanker metal cocks.
>>
>>109848808
99% of software development is testing and refining. And not just for correctness, but for meeting the expectations and vision of the one who wanted the software in the first place (ie the human).
The code writes itself, but the software does not develop itself and it never will, by definition.
I've been at this for 3 decades. The only thing I've seen happen in all that time is that more and more of the code writes itself.
>>
>>109849154
The code it produces for the same exact prompt does behave the exact same for the exact same input. The implementation details are irrelevant when it has no side effects.
>>
>>109846842
engineering is mistake-proofing. software dev is not that. to cope, legal and customer boilerplate frameworks were built to force a nonfunctional regime on retards (You). So because you refuse to adapt tooling that obviates wide classes of basic software dev problems, you now have to spend 100s of thousands of dollars per line of code in documentation, QA, meme "validated toolchains" and other such nonsense that does nothing to prevent real world bugs that end up crashing an F-35 anyway.

YWNBAE, kiddo.
>>
>>109849125
I had to "replace" a senior SWE and a senior legacy sysadmin. I would not have gotten as far as I did without an LLM, but let me tell you, my output is still shit compared to theirs in a thousand little ways and decisions that snowballed 3 months down the line, simply because my prompt doesn't know what I didn't think to ask for. And I have no idea how to put that in an root cause analysis either.

If nothing else, take this away: G.K. Chesterson's Fence is real, it does not give a shit about your good intentions, and when it comes for your booty hole, it always does so at 2:30 in the morning.
>>
File: 17628730295370 (1).gif (1.65 MB, 220x190)
1.65 MB GIF
>>109849169
Huh, so just like a gloryhole where it doesn't matter if it's your mom sucking me off or your sister, or even you, the end result is that my cock got sucked off.

You convinced me anon, coding is solved!
>>
>>109849096
Cope, seethe, and dilate harder, coding cuck. You're about to be EXTINCT
>>
>>109849096
or they'll just go to Moonshot or Deepshit or whatever. You guys act like the quality is that important. What's dangerous is AI is "good enough" and worst case they can just spin up whole new, from cloth, software from existing user requirements.
>>
>>109849267
Performative nigilism is not cruise control for cool, anon. I promise that when we're on our death beds, neither of us is going to stare back across the decades and say "I should have shitposted on the internet more."
>>
File: 17660533274470.gif (292 KB, 343x278)
292 KB GIF
>>109849267
Why so much hostility towards coders?

Did your mom get gangbanged by a bunch of coders and you were forced to watch it with your cuck dad sitting in his lap in his cuck chair while your mom was deepthroating BLOATED software engineering cocks?
>>
>>109848643
anything to get you hard
>>
>>109849280
Did you even try writing anything with LLMs? They all fucking suck dick, because there's no intent involved, you literally cannot convey to it what you want no matter how hard you try, the only correct way to convey what you want is to type it and literally do monkey "computer do" shit with high level languages that we already have. LLM just add another stack of shit on the already monkey shit, it's literally shit over shit.
>>
>>109843528
AI isn't an objective improvement, it is consistently wrong about shit all the time, it literally hallucinates dependencies, packages and commands constantly.
Hell, I literally was having an issue with some file, copy pasted it to the AI, it gave me a fixed version that didn't work, copy pasted the fixed version and told me it was broken and fixed it again, in the end the issue wasn't even in that file but somewhere else.
>>
>>109849302
>Coders gangbanging anything other than their cum crusted anime waifu pillows
lol LOL LMAO LMFAO
>>
>>109843496
SWEs are some of the most pompous and elitist fags out there. it's not even true engineering.
>>
>>109849423
Nope.
>B-BUT I tOoK mAtH
>>
>>109849374
If you ever worked in an office (which you didn't because you only go from bed to shitter to shit out your mom's nuggets out of your dragon dildo destroyed anal cavity) you'd know that in every IT department there's a bunch of buff engineer dudes who have loving wives and 2+ children, always get off work on time without doing any overtime unless it's an emergency, so these dudes have such big hearts they make it a rule to satisfy some unhappy lonely lady at the office whose husband's limp cuck dick can't satisfy, someone like your mother.
>>
File: BuffITdudebro.png (1.81 MB, 1312x1199)
1.81 MB PNG
>>109849455
>every IT department there's a bunch of buff engineer dudes who have loving wives and 2+ children, always get off work on time
>>
>>109849492
Shouldn't post your picture here, anon.
>>
>>109849423
Look man, we know we're not licensed. If the shit we did got people killed when we fucked up like a petroleum or mechanical engineer, I'd agree that we needed it. But the industrial engineers don't get licensed either, mostly. Go yell at them.
>>
>>109843528
This

Realistically AI code is an improvement to most code because most code is written by retarded jeets and bootcamp retards hoping on the SWE train, you wouldnt believe what bullshit is running today, especially in the web, the quality of AI is already an improvement
>>
>>109849600
Absolutely based and true
>>
>>109846883
>If you tell it to do those things, it will. You just have to give it 10x more detailed instructions than you'd give a programmer
It's the things you'd normally catch when reviewing code, but nobody does that anymore because AI PRs are always 5-10kLOC of spaghettified feces.
There is a reason you can *feel* software getting worse every year since AI coding became commonplace
>>
I'm a senior SWE and I think companies are struggling to operationally make use of LLMs to their max potential.
I can smash out work with Fable 5.1 or Astra in a day that would take me a week to do manually, but then I have to wait another week for reviews, QA, deployment, some random stakeholder's opinion, etc.
I'm experienced enough to steer the LLMs away from slop and retarded decisions, that's probably the only "skill" I feel is valuable nowadays and what I'm being paid for I guess.
If we're embracing LLMs though, I wish they would just let me VIBEMAXX. I feel like I have literally nothing to do most of the time and I'm bored. The web is always constantly broken anyway, why so much fanfare, people are used to shit not working.
>>
>>109849455
I too like to boost my own self-esteem by basking in the reflected glory of other people's personal accomplishments on the basis of an arbitrary shared identity. I too am spiritually a well beloved buff engineer saaar
>>
>>109849139
nobody is talking about local
>>
>>109849825
>I too like to boost my own self-esteem by basking in the semen of other people's personal ejaculates on the basis of an arbitrary shared gender identity. I too am spiritually a well beloved effeminate homosexual engineer saaar
Ahh the true nature of muh CoDeEzMaNz
>>
>>109849084
>can
A world of warcraft server can have multiple endpoints too. . Vast majority of version control is a single server and multiple clients.
>>
>>109843496
if AI can defeat OOP slop then it's a massive win for humanity
>>
>>109849778
Good post saar.
>>
>>109849359
>it literally hallucinates dependencies, packages and commands constantly.
Let me guess, you're not even using a harness and you try to one shot a complex code add or feature. Just like regular development you need to take things one step at a time. The more complex something is the more chances are for major fuck ups. Also just like regular development version control is mandatory
>>
>>109849778
>people are used to shit not working.
And that's why you're getting replaced by a bank of "5090's"
LMAO
>>
>>109850113
I never stated the opposite. This job feels so weird to me right now like we're just passing around LLM output to each other and everyone's getting paid
They'll probably still need some of us around to do the actual prompting and occasional human steering, but not as many as before
>>
>>109843668
and what's preventing agentic vibecoders from getting replaced by AI?
>>
>>109849600
Guess what AI was trained on.
>>
>>109849866
huh? Linux contrib is literally through git formatted patches sent over email. Linux Foundation runs a generic cgit remote, but it isn't used for any dev activity.
>>
>>109849876
That's not git's fault that you, me, and Duprey ignored the docs and all use GitHub and nothing else. Absolutely nothing is preventing us from having Github and a local Gitea as our upstream servers
>>
>>109850206
there isnt, >>109843790
>>
>>109843528
Unemployed brown hands wrote this post.
>>
>>109849179
>If nothing else, take this away: G.K. Chesterson's Fence is real
a good lesson
>>
>>109843779
why would anyone pay for this
>>
>>109850389
Using 2 servers simultaneously doesn't make it decentralized.
>>
>>109849302
Most programmers are wildly overpaid for their complete incompetence. You retards should make less than iron workers.
>>
>>109850644
Okay Captain Semantics, your local full copy does. But that's not nearly as illustrative is it?
>>
>>109850660
NTA i don’t care if i earn the same as a ukrainian pig farmer, i just want my
job back
>>
>>109850108
Anon, I don't use AI, the only time I've used it for something meaningful was for it to spout the damn commands I needed to use for certain tiptap editor's features.
I still needed to basically do by hand everything else around them because it basically gave me a menu bar with zero styling, not that I would've relied on it anyways since I know better than the AI and I'm faster doing my job.
>>
>>109850660
>You retards should make less than iron workers.
they do
>>
>>109850756
I'm going to be real with you, in a just world, anyone that shows up to work saceificing their body day in, day out for 20-30 years deserves more money than I do. But I don't make that much anyway.
>>
vibecoders are the vegans of the software industry
>>
>>
>>109850793
nah vibecoders are loud, numerous, and have a seething inferiority complex, while vegans are loud, few, and have a massive superiority complex.
>>
>>109849129
Neither I nor any of my colleagues have found any use for AI. Consistently, 100% of the people we interview who use AI at all are clueless at a level where it's not clear how they manage to get through uni at all. It's not a "trend", it's a 100% accuracy indicator.
>>
>>109848653
No, I'm talking about INTERN LEVEL DECISIONS that this piece of shit can't do, but you shills keep sliding the scale down and pretending that interns were ever allowed to be even remotely this clueless.
>>
>>109848695
L-lol? AI psychosis status: beyond terminal.
>>
>>109849178
I would ask you if you are retarded but you're a wannabe engineer so the question is already answered to the affirmative by yourself.
>>
>>109851013
Mate, I don't know what to tell you. I've used them for a everything from swapping out PKIs to an OIDC solution with short lived keys in an afternoon. I had one find a gnarly ass bug in GSON because it parsed God knows how many forum posts and commit histories before it reported back so we could side step the bug in our object hierarchy. I've used one to refactor god knows how many 20+ line dogshit GraphQL queries that someone clearly just threw over the fence into a much cleaner and clearer set of data projection templates.

If you can spend the time and forethought to formulate your design and implement your acceptance criteria framework before you start prompting (unit + integration, static code quality analysis, and linters), you can automate some of the toil out of you converting an idea out of your head and through your fingers.

If you're a good senior dev, you can make that shit work for you. I don't like it either, but I can spend the time judging the ever loving fuck out of the diff and still come out ahead in time.
>>
>>109850753
Then I find your "it hallucinates deps" shpeel hard to believe. Not that it never happens, but that you specifically actually experienced even half of what you said.
>>
>>109843528
Statistically speaking human code has less bugs on average than AI, basedience confirmed this but I'm not going to argue there.
I think my main problem would be the kind of bugs that can be produced by AI which are sometimes shit that no human no matter how retarded would do.
Here's one example I would think of, we had a status enum and some filters for every value of that enum like if status X then mark it X from our side too (as a syncing mechanism). One teammate vibecoded something where he needed to remove one status which is used in a filter of that status. AI solution? Just remove the status and have the filter keep on working but on another value, and then we had a massive incident where everything got marked as the that value... the code was vibecoded, even the description was vibecoded and the reviewer just reviewed it with AI, when I was investigating root cause and found it literally in the description of the PR, I wanted to kill myself. Shit like this would literally never happen if any human in the loop used their fucking brain.
>>
>>109846579
>t. doesnt know shit about computers

thank you for your input zoomer-san but there may be a kick stream of rappers and felons you should be posting on instead?
>>
Should I learn how to code?
>>
>>109846579
Every large company I've worked at uses gitlab anyway
>>
>>109844532
wdym tracking saar? All their shit is broken all the fucking time even the pull requests page, which was perfectly fine beter.
-10 izzat
>>
>>109852956
no, learn politics to fight against the tyranny of tech monopolies instead of learning how to become their slave
>>
>>109852956
unless you're VERY interested in it, no.
>>
File: monwednesday.jpg (15 KB, 279x185)
15 KB JPG
>>109848695
>>
>>109853241
Yeah, in my experience edit prediction says some pretty weird stuff.
>>
File: top-men.gif (1.47 MB, 480x204)
1.47 MB GIF
>>109853241
>>
>>109843790
>everyone who embracies it survives
if it's good as you say not even those who embrace it survive, and you have no point.
>>
>>109844448
on other words, JeetHub
>>
>>109852956
Should you learn caligraphy?
Should you learn basket weaving?
>>
>>109850753
>I don't use AI
The why spew your shitty baseless opinion here? I haven't seen a dependency hallucination for a year now.
>>
>>109853241
Didn't they find the repo that Monwednesday belonged to?
>>
File: smiles.png (107 KB, 718x682)
107 KB PNG
>don't use AI
>still faster than my peers
>don't really get into the AI train because I genuinely enjoy what I do
>most days I learn something new
Skill issue I guess
>>
File: 1782961618304580.jpg (148 KB, 1920x1080)
148 KB JPG
>>109853878
welcome to the real world.
>>
>>109851100
You have an IQ of 65 and suffer from severe AI psychosis.
>>
>>109853878
It do be like this.
>>
File: file.png (1.17 MB, 1080x1055)
1.17 MB PNG
*passes away*
>>
>>109850793
>implying they are in the industry
>>
>>109853953
Sadly, some of them are. Largely the pajeets. Most of them genuinely think people can't tell.
>>
I'm tired of seeing this Luddite shit. Swallow the pill, boomer tard, coding by hand is dead. I'm a tech lead, I'm not hiring you if you use your keyboard to type code.
>>
>>109853932
Did he finally Troon out?
>>
>>109852956
learn comedy because AI is and never will be good at it.
>>
>>109853878
>faster than my peers
i dont believe you
>>
>>109854647
Because you have AI psychosis? I'm not him but that is the case for me as well. The one colleague who AImaxxes takes a week to do a task that takes the rest of us about a day at most including peer review.
>>
>>109846790
lol look and laugh at this retard
>>
>>109843528
>"Professional" code bases are full of spaghetti garbage written by DEI hires and jeets. The truth is most programmers suck at their jobs, are shitty people in general,
Truthnuke
>and AI is an objective improvement on both the soft skills and technical ability.
If you are a competent programmer, having an AI is an objective massive improvement over having a normal programmer coworker. But if there's no competent programmers at all, or if they are the minority or they don't call the shots, the AI will be used to generate more shit code
>>
how much code is being written by ai in 2026?
>>
>>109848575
AI is still stuck at gtp 3 issues
>>
>>109854801
Correct.
>>109854794
More than 0 (aka: far, far too much) but maybe at most 10%.
>>
>>109854685
i dont believe you either
>>
>>109853913
>IQ of 65
Probably. This job will do that to you.
>AI psychosis
It's just depression (I think).
>>
>>109854835
Of course, you suffer from AI psychosis so you don't believe in real life.
>>
>>109843496
It's all just a demoralization campaign really. In my industry, we're not allowed to use AI at all, as the code we write has to meet strict standards or someone will die and it will be our fault.
Any job getting replaced by AI is fake and gay really.
>>
>>109854883
>AI psychosis
someone learned a new buzzword and is very eager to use it I see.
>>
>>109854913
I actually didn't like the term for a long time, but it really is the only way to describe profound disconnection from basic reality entirely induced by AI use.
>>
>>109854920
Let me get this straight: you genuinely believe hand-coding is going to save the world in 2027 and devs will suddenly be worth 10x what they're paid now, while anyone embracing the technology changing the industry is suffering from "AI psychosis"? At some point you might want to reconsider who's detached from reality here.
>>
>>109855012
Thanks for demonstrating what AI psychosis looks like.
>>
>>109854813
i wonder how much of code being generated is just useless garbage
>boss proudly stated LLMs are the future and the best software developers on the planet
>gave our team some huge ass notebook to run a few weeks ago
>it failed to run because there were things set up on his account that werent set for others
>instead of debugging it, i just screenshot the error and gave it to him
>"he" fixes it
>now that works but something else is broken
he's been working on it for weeks and it still doesn't work. it's somewhere around 10,000 lines of code
>>
>>109855050
Yup. Had a job a bit ago at a small startup that had hired a CTO 8 months prior to me joining it. It was exactly like that. I ended up building the entire tech stack from scratch, no relation at all to the shitty retarded ai-generated notebooks we had before and led to the first productive results the company ever had. Then the company said there was "not enough AI" and fired the whole tech team.
>>
>>109855012
Good post saar, you showed him.
>>
>>109843595
assembly is the only programming language you need
>>
>>109855045
AI psychosis: A condition characterized by rational thought processes and an openness to emerging technology that appears deeply delusional to manual coders suffering from chronic AI denial.
>>
>>109855161
Thanks for demonstrating what AI psychosis is all about by the nature of your post rather than its contents.
>>
>>109855196
>Thanks for proving my point
which is failing to prove me wrong.
>>
>>109843790
>thinking AI will eventually fuck up so badly that the old workforce gets called back in as heroes
The problem with this chain of thought isn't the idea that the AI will catastrophically fuck up and ruin the code base. It will. The problem is assuming anyone in charge of hiring and firing gives enough of a fuck to bring back in a skilled human to fix it. We've been living in the age of the Minimum Viable Product for a while now. We're just getting better at optimizing what the word "minimum" can really mean.

The future is the shittiest software you've ever seen, so horrendously inefficient that it can only be run via streaming from a gigantic data center, produced for a salary cost of $0 and sold to you as a subscription package. This is not a bug, it is a feature. This is the end game. You will never be called in to "solve" it, it's not a problem in the eyes of the people orchestrating it.
>>
File: 20260919_17h53m58s_grim.png (703 KB, 1684x2550)
703 KB PNG
>>109846579
https://mrshu.github.io/github-statuses/
Check out April 2026 if you dare.
>>
File: Descartes.jpg (151 KB, 817x1000)
151 KB JPG
>>109843528
>"Professional" code bases are full of spaghetti garbage written by DEI hires and jeets.

not on mainfraimes kek
>>
>>109856019
Do you guys still work in EBCDIC?
>>
>>109856019
Especially on mainframes kek
>>
>>109855987
indian excellence

we need more h1b to fix this problem
>>
>>109843496
I am not cleaning shit. we're starting from scratch, sorry.
>>
>>109854647
>colleagues complete around 12 to 14 story points
>I usually push around 20 to 24 story points
Turns out being competent and knowing what you're doing pays off I guess.
>>
>>109843528
>political buzzword spam
*yawn*
>>
>>109859251
Oh really? What did you get that they didn't get by performing so much better than them? Do you earn ~2x as much as they do? No? Do you get promoted 2x faster? No, in fact they're higher up than you? Hmm... strange... almost like competence doesn't pay.
>>
why is this board constantly shilling and lying about everything?
>>
>>109843779
lol that's the most retarded thing i've ever read here. imagine giving ability for dumbfuck CUSTOMERS to push changes to prod.

unless it's like a wordpress blog and all they do is change backgrounds or something lmao
>>
it depends on the internals of the company

dumbass company: increased loc is considered a good metric

you are fucked

smart company: less loc is desirable

you are also fucked because the software will enter a state of being finished

gg
>>
>inb4 nobody does that anymore
plenty of companies use loc delta as a metric even today despite it being stupid as fuck
>>
>>109843528
fpbp and /thread
It's another "professional" arena that turned out to be nothing but "convert these symbols to those symbols using this rule list", and so ripe to be destroyed by LLMs.
All the seething is just overpaid, underworked biological calculators finding this out the hard way.
Should have learned to plumber.
>>
File: file.png (229 KB, 1051x907)
229 KB PNG
>>109861472
Very informative.
>>
>>109861493
Why did it take until last quarter? And did nobody else in the entire team ever get promoted? It's not about you, pay attention to my post. it's differential. You also didn't confirm if you earn 2x as much, which you should since you are 2x as good as them.
>>
>>109861499
Deleted last post to fix some typos. Didn't realize you had already replied.

I am not sure if I earn 2x as much. I don't discuss my salary with other people, anon.

I am sorry, but did I unintentionally touch a nerve or something? You seem agitated over nothing.

Anyways. I enjoy what I do. And you should too anon. :)
>>
>>109861510
Just pointing out the math, anon. What happens in 99.999% of cases if you overperform is that your colleagues gang up on you in political ploys to get you shitcanned, management never recognize you no matter what the metrics and KPIs say.
Furthermore, with the added time you have from being able to do the work in 1/2 their time, you could instead work on a startup, or on hobbies, or anything else that is globally infinitely more productive for you. THAT is the path of white people, not the path of poo slaves like you whose only self-worth is tied to slaving for shlomo.
>>
>>109861524
You seem to be assuming a lot of things anon. But, let me tell you, I work around 3 to 4 hours a day, from home. The rest of the time I either spend it studying or enjoying my fucking life.

Other than that, I don't know what else to say to you. Something I said either touched a nerve in there or really really triggered your autism bone lol
>>
>>109861546
I'm making no assumptions, as your post itself already demonstrates.
>>
File: comfy.gif (101 KB, 498x451)
101 KB GIF
>>109861557
ok anon
>>
>>109861593
Why are you seething so hard?
>>
>>109854794
Like 30-40%
>>
>>109854794
at work, the only manual modifications to code I do nowadays is removing AI comments (they're too verbose). other than that, it's 99.9% AI.
I don't use AI in my private projects tough (at all).
>>
>>109863581
Why don't you use AI to remove or rewrite the comments? Or better yet, instruct it not to write verbose slopments in the first place
>>
>>109843668
So when is it supposed to start shrinking? Two more weeks?
>>
>>109849423
My take is that this is all projection. You guys hate us because our industry dominates and nobody wants to be an engineer in anything other than software or computer hardware.

I've never met or seen elitist software engineers in the professional world. Motherfuckers literally get into the field without a degree and are accepted by their peers, I work with people with non-CS degrees.
>>
>>109849010
You sound like like you're never worked in the industry.
>>
>>109846502
that used to be the case before gcc
>>
>>109846696
if you absolutely **need** github to be up 101%, and you absolutely need it to be github-like, you'd be selfhosting something like gitea
>>
>>109865711
>fake openings for h1b hires
lmao keep coping
>>
>>109865873
The whole point for orgs is they don't use github for code hosting, but for infra coordination. CD/CI runs on github. You absolutely don't want to run it on your own cloud, far too expensive especially when you need to keep it up. The whole point of paying for github is they promise to keep that shit up so you don't have it. It's the central decision process of build vs buy. And it's broken.
>>
>>109864166
NTA, but I'm convinced that EU mandated digital watermarking was when the comments in Opus/Soneet went straight down the toilet. It's words, words, words in the passive voice that say nothing at all when a simple numbered list explaining the process would have been superior.

But then that wouldn't give enough characters to setup the digital watermark.



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.