I see skilled developers (a lot of game devs) in complete denial over AI. They think this is a "bubble" that will pop any day now. Do they not want to accept that their careers are literally over?

A lot of people aren't ready to see their incoming cashflow dashed on the rocks and really are in denial. Imagine making 200-300k a year, and you take out big loans to buy a nice big house in an expensive area, expensive cars, send your kids to nice schools, so on and so forth, and then suddenly, your job starts putting thumbscrews into you because they know they can, just to squeeze a little more out of you before replacing you like a burnt out lightbulb with an Indian. You can't just up and leave and find a new gig, since there's massive layoffs across the industry and companies are inundated with potential hires. Yeah, people would rather stick their head in a hole than acknowledge that.

>>105962420 (OP)
What makes you so sure they're not right and/or that you're not retarded?

>>105962523
AI slop.

>>105962549
Not OP but what does it matter if there's a bubble? The tech to replace them is already here and just needs to be more efficiently put together for boomers to user in order to wreak full havoc.

>>105962549
I think lots of very good devs will probably be fine, but pareto principle is at work in the industry. Tons of people are passable code monkeys who can do the needful, and they're going to be fucked by these tools. If you can do more with less people, if there's less toil overall, why would you pay for as many people?

>>105962420 (OP)
AI is just another endpoint for devs to pipe into other things. It's not good enough to complete replace a dev yet. You still need a good understanding of a projects low level details in order to competently improve/rework it with LLMs.

>>105962554
It just turned into 'babysit the LLM' initiatives at work for me. The leadership is fanatical about trying to get people to just shift their workflows to generation of code. The company was acquired by private equity, and I'm pretty sure the intent is just to thin the herd as it were, see how much 'code velocity' can be gained through the tools and remove the deadwood.

>>105962420 (OP)
I may be the oldest person on /g/. I remember this exact discussion happening when higher level programming languages were becoming more prominent. I've spent decades hoping to be replaced and it's never happened.

It's an indictment on the industry that so many software engineers are worried about this situation, given that anyone with a modicum of experience knows that bashing out code is only a single aspect of their job.

>>105962566
That doesn't matter. If CEOs and managers can get tools working on a surface level, that's all that will matter to them. I'm already seeing clear AI code in new SAAS tool updates and those are likely being pushed by devs.

>>105962584
Good, if the CEOs and managers are that retarded then their company deserves to be run into the ground by shitty code

>>105962578
It's just underachievers hoping everything crashes so they can feel like they're smart for not trying.

>>105962578
I see where you're coming from but this comparison is no longer valid. I would be surprised if even 10% of white collar workers are still employed in their field within 5 years. Like the other guy said, the Pareto principle will go into overdrive here. It will be so bad that the Fed/Treasury will have to print trillions to prop up the debt and especially mortgages.

>>105962612
No we're just being realistic here.

>>105962578
>ohhh we're not code bashers
massive cope

>>105962556
>the tech to replace software engineers is already here
>...it just needs to be implemented
*never gets implemented*

>>105962697
knowing which direction to send the LLM in is pretty important. the only thing they can really replace is H1Bs which is fine by me.

>>105962420 (OP)
Complacency
Most brogrammers had a long time learning their tech stack, learn the trade, learn the interview and corporate culture, it's unreasonable even for most senior of programmers to learn something outside of work unless it clearly benefits them of is pushed by middle/upper managment

The reasonable programmers that I talked to all say that programming is dead and will be completely revamped in next 2 years
Me personally, I am pivoting

>>105962420 (OP)
In the US, Trump's bill will allow AI to, basically, unchecked for 10 years... This, it's going to obliterate a fuck ton of occupations.
>Finance
>Accounting
>HR
>Programming
>Various parts of IT
>Energy
The list goes on.
Now, the tradies, cheer because they think they're safe. But, when you lose all those white collar jobs, there isn't going to people paying to have them fix their AC, put plumbing in, etc... so a of blue collars are going to be out of work as well.
Ok, now, I'll wait while some idiot 19 year old explains why they think this won't happen.

>>105962834
>I am pivoting
to?

>>105962569
this is pretty much exactly what i'm seeing at my company, and it's rolling up into a proper shitshow.

at my level, i spend about 20% of my time actually writing code, and a local LLM agent speeds up drafting it by maybe a third. in return for this, i receive PRD that are plainly written by chatgpt and review code that may never been read or executed by a human being.

it's no wonder the talented people in the mid-upper levels are walking out.

>>105962885
NTA but I left the field and became a Bitcoiner

>>105962885
Using my skillset and knowledge in tech unexplored fields, infusing tech into them and solving issues that weren't solved for last 40+ years due to lack of knowledge

Currently medicine, future will be chemistry and electrical engineering. I have a technical model already and will have a patent soon and then will go for a VC funding. Then academia (math or physics, we will see)

>>105962936
>Currently medicine, future will be chemistry and electrical engineering.
Yeah, AI won't touch that at all.
LMAO

What the fuck do you guys do that can be done entirely by LLMs? Agents basically just make buttons and auto type. They can't even implement very minor features in an actual, real product. The only thing it's replaced is the laziest junior devs that fuck up even the simplest bitchwork

>>105962957
It will, but not in a way I will. Think that 15 year old kid that found a way to test for detecting pancreatic cancer by adding enzymes to graphene voltage sensors. Cheap, easy and quick to manufacture. And idiots at the lab looked at him dumbfounded.

>>105962967
>They can't even implement very minor features in an actual, real product
They can if your code is in some design pattern that's reasonable.

>The only thing it's replaced is the laziest junior devs that fuck up even the simplest bitchwork
With large enough context window and RAG you could rewrite the whole codebase in a new design pattern, write unit tests for all functions and dockerize it, shiny it up and make it better. Something that took 3 to 6 months of good programmers time

>>105962967
>They can't even implement very minor features in an actual, real product.

If anything, that's all they really can do. Though sometimes they can tackle things a little more complex.

>>105962991
>write unit tests for all functions
Ok I'll give you this but it still takes a lot of work
>shiny it up and make it better
Ok yes this is bitchwork that used to be passed to interns and junior devs
>Something that took 3 to 6 months of good programmers time
That work should have never been given to good programmers.

>>105962991
> With large enough context window and RAG you could rewrite the whole codebase in a new design pattern, write unit tests for all functions and dockerize it, shiny it up and make it better. Something that took 3 to 6 months of good programmers time

no one who has actual experience with both the tools and the problem believes this. i got to supervise multiple studies at my job and even a dedicated cluster for an agent with state of the art MCP server integrations, and they found that even a straight refactor of a single microservice- which trivially fit in the context window of several trial models- was a complete disaster. mind you, this is literally the best in class, since the model can actually start the service and directly integration test it, and it still generates shit that doesn't build, writes test cases that prove its own shitty logic (that's out of business spec), thrashes its own OS, etc.

>>105962420 (OP)
>doomposting this hard
There's plenty of work, ironically much of it due to ai
It's coming for 3 billion other jobs first

>>105962420 (OP)
all you need to do is look at AI art to see what will happen when AI comes to a field near you
entire art teams laid off, film crews canned, design departments downsizing to single digit headcounts
youtube and tiktok flooded with AIslop, but the common cattle doesn't care because it's good enough to fire cheap dopamine hits so it captures all the attention and drowns out genuine human work
you think AI has to be as good as human work, it doesn't, it just has to be good enough for the average consoomer and it already is

>>105962420 (OP)
a lot of game devs are surrounded by tranny "artists" who freak out about AI and lynchmob anybody who doesn't agree with them so they have to blend in.

>>105963103
"With the to-be-refactored Java documents as input, ChatGPT and Gemini identified only 28 and 7 respectively out of the 180 refactoring opportunities. However, explaining the expected refactoring subcategories and narrowing the search space in the prompts substantially increased the success rate of ChatGPT from 15.6% to 86.7%. Concerning the recommendation of refactoring solutions, ChatGPT recommended 176 refactoring solutions for the 180 refactorings, and 63.6% of the recommended solutions were comparable to (even better than) those constructed by human experts. However, 13 out of the 176 solutions suggested by ChatGPT and 9 out of the 137 solutions suggested by Gemini were unsafe in that they either changed the functionality of the source code or introduced syntax errors, which indicate the risk of LLM-based refactoring"

https://doi.org/10.48550/arXiv.2411.04444

Whatever you are using, whatever prompt engineering and context engineering you are using, must be shit. Don't blame the tool for the lack of your own knowledge and skill.

Mind you, this is a November 2024 study, just to note that. We are light years ahead of that.

>>105963208
>but the common cattle doesn't care because it's good enough to fire cheap dopamine hits

eh, there's already fatigue to AIslop amongst normies

>>105963103
>>105963274

>>105963302
Nice meme, I think I'll post this on facepage

You can't have sex with AI.

>>105963103
Just the simple fact that you called MCP server integration "state of the art" means you are full of shit, when MCP servers at their core functionality are basic API or CLI interface for the LLM. That's like calling your light switch "state of the art", makes you look dumb upper managment drone who doesn't know what he is talking about.

Just because you put "state of the art" in front of it doesn't automatically make it cool or make your case valid, neither does padding your opinion with imaginary crowd that caters to your belief

If you were to actually do something, instead of actually delegate your work, you would see what that word "synergy" really means

>>105962420 (OP)
evidence
no I didn't say 'speculation', 'prediction' or 'hope', I said evidence

>>105963311
It was a first image response of an old adage, you literal drone

>>105962420 (OP)
It can't code something that hasn't been coded before.

>>105963274
>>105963302
>>105963327
>>105963342
uh oh, tripnigger having a melty

>>105963358
well xcuse me for wanting to help :(

>>105963327
holy shit man, you replied to me 3 times? how fucking mad are you?

first of all, i didn't run the studies myself, i was just part of the team that had to argue for the funding, review the papers, and then deliver the results to leadership. second, if you think an 85% "success rate" on a refactor is acceptable, then i am going to underline my previous claim that you have absolutely no idea how actual software shops work, and you are talking out of your ass. finally, i did not say that MCP is state of the art, i said that our integrations were. our studies were working directly with vendors who write and maintain them, so we got patched versions of bleeding edge releases from every vendor in our supply chain. we also had consultants and an open line of communication with each model vendor, so it's not like we were flying blind. the tools are what they are-- improving, but just not ready for enterprise software.

hope this clears things up for you.

>>105963398
>if you think an 85% "success rate" on a refactor is acceptable
Yea, it gives you a good framework to work with. as I said 85% was last year

>bleeding edge
can you stop using the cringe tech corp jargon, its making me sick, you literally said nothing really technical in a span of a paragraph other than "we da best" and "our people da best" and "our business partners da best" jesus fucking christ

>hope this clears things up for you
same with this passive aggressive corpo talk, dumb faggot

>replying 3 times in a row == mad
your ego must be really appeased you corpo monkey

>>105962420 (OP)
coding is dead. I'm going to kill myself.

>>105963480
no just become a vibe coder. start vibing.

>>105963509
It's not the same bro... it's over

>>105963446
>Yea, it gives you a good framework to work with. as I said 85% was last year

ok then. see my statement in full. :^)

>can you stop using the cringe tech corp jargon, its making me sick, you literally said nothing really technical in a span of a paragraph other than "we da best" and "our people da best" and "our business partners da best" jesus fucking christ

in my defense, you haven't said anything particularly technical either, so i may have misread the tone of this conversation. my point was that we really did give this a sincere shot. my technical concern was creating a reproducible sandbox environment that exposed the MCP integrations, had appropriate resource spin up / teardown, provided safe debugging for both our engineers and our partners to inspect and replay actions taken by the agent, etc. i studied AI extensively like 15 years ago, but i'm really just a compute and networking guy. :^)

>>105963553
Okay, can you give me more in depth of what you guys used. Like what LLM, what RAG, what kind of vector database or memory, what MCP servers were in use, what prompts and chain of thought, did you use MoE or any context engineering?

>>105962420 (OP)
It's not any day now, but it's extremely obvious that it's a bubble that will pop. Hop on LinkedIn if you don't believe me. There are C-level executives literally posting about how they used AI to help them choose the ripest watermelon at Costco. There has never been misinformation and hysteria on this scale in tech.

I'm not sure if it's worse for everyone to be lying, or to genuinely be this ignorant of what the technology is and what it does. Either way it can only end in disaster.

If you don't believe me, show me the profits. Nobody doing AI is making money. Not Microsoft. Certainly not Apple. Not even OpenAI. Nvidia is raking in cash, but they have no profitable AI service; it's all hardware and CPU.

but my artisan functions...

vibe coding is kinda great for me as I was never very good at coding despite doing it for 10 years, I am however pretty good at designing programs from requirements etc.

>>105963480
smart people already figured this out when gpts went public, it will take some time yet for the leftcurvers to catch up
luckily there is still time for you to pivot to something else, unless coding was your only passion, then I guess you really do have nothing to live for
I would avoid art and math or anything that can be done exclusively on a computer really, just do something that requires physical dexterity

As someone who never went to university because I'm too lazy to study, I'm hoping that AI replaces all of these smartfag geniuses so I can laugh at their misfortune.

>>105964039
According to the newest studies from Apple on novel thinking, LLMs can't do even the most obvious of abstract thinking, it can just find patterns in what it already saw, namely language

>>105963667
That's my exact skillset. I'm good at explaining things and all LLMs like that a lot. I got my degree with fucking chatgpt 3.5 back in 2022. I can't help but laugh at the retards here who can't even get them write hello world. You're ngmi

>>105964090
According to everyone tripfags are attention seeking faggots