ai bubble popped, you're old news kike shill

>>536750225
>anthropic.com/news
let me guess, their news desk says their product is amazing again

>>536750413
https://news.ycombinator.com/item?id=48463808

>>536750225
>shit programmers get culled
good

i’ve been using fable all day. it’s good but opus 4.8 was also really good.

>>536750380
>>536750413
fable 5 is scary good, expensive token cost but it's obviously better than opus 4.6 and opus 4.8 (4.7 was trash)

i'm not gonna argue with you niggers, but yeah SWE is looking more of a hobbyist thing for me. looking to get into hardware instead lol.

the cost of prompting fable 5 is the same as a jeet on fiverr, which one do you think is better? yeah that's what I thought.

>>536750482
>good
no you dumb nigger, it doesn't matter how "good" or "bad" you are. it's becoming less justified to pay anything above $100k for you to just prompt an AI that's doing all the work anyway, why would people pay you those high salaries? why else do you think big tech is ONLY laying off this year? it's not because of "covid overhiring" BS, that's long gone.

Just answer me this, if you're just an "architect" rather than doing the real work. WHY do you desrve to be paid something that most people can do anyway? your skills and expertise are not needed as much anymore.

it's not about good or bad, it's that it just doesnt fucking matter now.

>>536750225
It is still yet to produce anything new

>>536750225
they are expending IT offices in Poland because indians get lay off

>>536750508
tell me you don’t use ai without telling me you don’t use ai

File: ..gif (1.93 MB, 538x228)

1.93 MB GIF

>>536750508
The only reason that hardware development (e.g. FPGA, EDA) is still safe is that most of the design is proprietary and developed privately. But the more the companies expose their IP code and designs to AI the more training dataset is fed to future models.
tldr; it's over. It's fucking absolutely over. What happened in the past 6 months is the most insane thing ever happened in the entire history of tech. A year ago, AI was a fucking joke and useless outside nigger-tier frontend/backend development. Now it can do anything, literally anything, better the best software engineer on this planet in a matter of seconds.

>>536750849
we use Antrhopic on AWS so no data actually goes to build another model

>>536750815
been using claude since 2022, kys
>>536750849
its the last bastion, might as well just go into chips. SWE for me is just going to be FOSS because of the demoralizing jeet PR spam. the meritocracy is over, just go to IRL meetups. Hackathons are also over thoughverbeit

>>536750849
we still got photonic computing and processing in memory to solve i guess but yeah its bleak

>>536750225
buy an ad

>>536750380
>le AI will collapse in two more weeks!
kys well poisoning schizo kike

>>536751175
lmao chud. you have no idea what a program is. have claude write your next reply to me. you are the lowest form of ai user, the one who never learned to think

>>536751318
y so mad? what is pissing you off so badly?

>>536750225
buy an ad

>>536750225
>It's 2% better on benchmarks!!! Game over man, game over!

File: stellarbalenuevo.png (1.33 MB, 1200x675)

1.33 MB PNG

>>536750225
>> t. software subhuman who's going to be homeless soon
If you are related to the tech sector then I am afraid to tell you that you are horrible for the job, I got multiple accounts including mine that these American Tech companies are losing clients, and companies that are investing in those "frontier" models are just burning money, the reality is that companies are switching to chinese models that are 6 to 10 times cheaper than American models while generating output with similar ( or in my case better ) to these "frontier" wunderwaffe models.

These American AI companies will go bankrupt eventually and not AGI god will come to save them, the Fanged Numena already migrated to the far east, and if you are unsure about that then check the pic.

>>536751318
I think I know what is pissing you off.

You're angry that AI is actually making SWE obsolete, you're pissed that hackernews won and that the techfags have destroyed even handwriting code itself.

Let me make it clear, WFH and AI have made the ultimate storm to make this a non-career if you live in the first-world. It's dead simple, you don't need a 10 man team to just fucking "guide" an agent harness. You're fucking behind or some artisan if oyu're hand writing code in emacs, it's done; it's over. it's just not something that is sustainable because at the end of the day the company will choose what economically makes sense. And you prompting and having no actual domain knowledge besides being a persistent debugging monkey that "works the AI", you're fucking nothing and your skill is not exceptional anymore. It's over.

Just build things on your own, with your own terms. Np guarantee to be paid a wage anymore though, but yeah this shit is not going to be a "career"; but rather something you'd do normally within any venture if you're running an operation. Tech literacy is not difficult as you think it is.

Whenever you see someone hating AI and calling it a fad, you have to realize that they are still stuck in 2023 and haven't used the latest models because they can't afford them or their jobs don't go beyond sending emails.

>>536750622
>if you're just an "architect" rather than doing the real work. WHY do you desrve to be paid something that most people can do anyway?
I don't think that most people can do that anyway, LLMs can't save someone from ignorance and bad decisions. Software complexity is NOT solved with them

The LLM is basically a multiplier: if you're bad at your craft, they make you even worse. If you're good, you do a lot more than before.

I wish I could afford the most expensive plans because both Claude Code and Codex are really limited with just 20€

>>536750849
>The only reason that hardware development (e.g. FPGA, EDA) is still safe is that most of the design is proprietary and developed privately
Nigger that doesn't even matter anymore. These models just clean room reverse that shit.
I had it reverse engineer some old MIPS software from an emulator and then design a static analysis layer that caches functions as they're encountered, actually making the shit become faster as you run it.

>>536751640
you are argentinian you just can't afford the frontier

>>536751819
>LLMs can't save someone from ignorance and bad decisions. Software complexity is NOT solved with them
software cant be this easy. it's pretty fucking simple to know that theres some stupid dependency bloat or that a certain stack doesnt make sense. maybe im having an imposter syndrome case. but it just seems like it's just too easy to be a job.

We’re almost a decade into the AI promises and it’s delivered on maybe 10% of them. It’s a fraction of what they whipped everyone into a hysteria over. The claims about AI have turned out to be more preposterous than the claims about climate change.

>>536750225
>google taking all of chatgpt's casual users
>claude taking all of chatgpt's hardcore users
its over for jewgpt

>>536752076
we are 4 years into AI

>>536751940
Models need good training dataset that's generated by humans. It's why they are weaker in obscure/new programming languages like Zig compared to more established languages like Golang, Python, etc. However you're right, they are becoming good to the point that they can understand a relatively a new context without having been heavily trained on it.

The real problem once all text-based content (e.g. source code, articles, blogs, research papers, books, etc.) are contaminated with AI output, what happens? AI model collapse? stagnation? I have no idea.

>>536750225
What can AI do that humans can't?

File: 1780129305712376.jpg (169 KB, 1198x1182)

169 KB JPG

>>536750225
yeah

>>536752301
stop doing this you faggot you are the reason why this shit is so slow during peak hours

>>536752214
Actually about 60 years. LLMs since attention is all you need paper (9 years ago).

>>536752075
yeah I feel the same

>>536752118
Anthropic actually released Mythos/Fable because OpenAI has been regaining most of the tech people who left it last year in the past month alone with 5.5 and codex. 5.5 is absolutely great and the usage limits are insanely good even for the $20 subhuman subscription.

>>536752445
reasoning has been more impactful than introduction of LLMs actually

File: 0867ae31f7f5120ddff998fc8(...).jpg (44 KB, 700x693)

44 KB JPG

>>536751955
>you are argentinian you just can't afford the frontier
First I dont like to be sodomized by American Tech billionares, no I dont like anal and my anal virginity will not be an offering to the American AGI digital god.

Second the company where I am working bought 2 Claude subscriptions and it performed no better than Gemini 3 Flash.

I was using Gemini until the chief pajeet scammer at google redeemed even the PRO subscription users with their new "Frontier" WunderDataCenter Gemini 3.5 model, causing massive outrage to its user base.

After bitching for two weeks against American Billionares I just switched to chinese models and to my surprise they performed far better at a fraction of the costs and even my co-workers that were assigned the Claude Accounts are switching to deepseek.

Its simple economics, any of the American WunderKünstlicheIntelligenz MegaDatacenter Nvidia powered "Frontier" models do not perform better than Qwen 3.7, even for complex tasks deepseek v4 flash is more than enough.

>>536752555
It's still an LLM.

>>536752457
I knew I wasn't the only one bro. People say that "oh you still need to know when it's bullshitting", it's not hard. It's not as hard as people think and companies know this.

Italybro listen to me, just build something yourself; I can't disclose who I am but I do get grants from non-profit for particular FOSS projects.

come to Chaos Communication Congress, you'll meet more of us there. But yeah bro that coveted "big tech job" is just like the american dream at this point. it's just getting offshored to oblivion but fuck I love creating software to be honest. I want something stable, that's why im doing EE and going into semiconductors/hardware.

keep programming but as a career.... nah. However creating a company (SaaS is harder now, it's not dead but it's harder to get a client than you think unless you have connections) is easier than ever.

Trust this Jew guys. He totally doesn't have an agenda where he needs to dump his shares on retail investors at IPO. Just like the other Jew Altman.

You're gonna need that subscription bro. Give that Jew your money!

Fuck off. I use DeepSeek V4. Entirely self hosted. It's as good if not better in some respects. Zero worries about some Jew using my personal data or running up some token bill. It runs on my workstation all night and I check the results in the morning.

This tech has already hit a wall. But these Jews are so fucking desperate to try to convince you it hasn't it's very telling. This tech will change the world the way the original worldwide web did. But we had to go through a crash to sort out the real companies (like Amazon) from the meme companies that won't last.

These Jews need to build hype for their IPOs. Do you think it was an accident these articles end up in your feed? There is a network. A pipeline. It's obvious once you see enough of these IPOs. They even have connections to legacy media (pay them as little as $100k to get some favourable coverage) so some dumb boomer buys the stocks thinking he's a genius.

>>536752441
it literally costs nothing computationally to do that retard
now be a good goy and enable auto refill for extra usage

>>536750225
>>536750508
Any job that involves you sitting at a desk working at computer is cooked.

File: Screenshot 2026-06-09 at (...).png (90 KB, 484x1520)

90 KB PNG

>>536752571
I use like 27 models and they all still suck lol

>>536752214
GPT-1 first booted up almost 9 years ago dude

File: Artificial Analysis Codin(...).png (436 KB, 4640x1776)

436 KB PNG

>>536752571
stop laying you dumb sudaca

>>536752659
yes but AI progress is only limited by number of data centers we have not by model progress in 2017 we just didn't have enaught compute

>>536752301
>supposed "best" model
>it spouts bullshit and rants like a redditurd instead of providing the answer
total scam

>>536752214
>>536752445
And btw we’ve been hearing these ridiculous promises that AI will be sentient since Google was working on LLMs internally back in like 2016.

damn
I was really into this Ai stuff early on but it's advanced and diversified so much I've been left behind

i deep like a fucking grampa
I remember thinking it was over in 2022 when chatGPT vaguely understood meme pictures.

just hope all this shit runs on my 9070

>>536752823
if that was the reason why grok wouldn't be so trash

>>536752823
Nope. We're out of novel training data. It's the reason Mercor exists.

>>536752890
you prolly can fit qwen3.6
i'm running it on my intel B70 and it's pretty good
idk how well it will be quantized enough for 16gb of gram tho

>>536752922
i am 100% sure openai and antrhopic training runes used more compute than grok

>>536753060
there is a reason why both google and anthropic are renting out compute from elon

>>536750622
>t. my job is just sitting there and converting specs into HTML pages

>>536750225
This has been said going on 2+ years now and it's fake and gay.

>>536753007
>Mercor
it's used for RL we can do that with AI too

>>536753133
because grok can't catch up with trillion dollar companies elon is cooking the books for IPO

File: eb724e-2017-07-gettyimage(...).jpg (41 KB, 600x333)

41 KB JPG

>>536752823
>OMG A BENCHMARK!!
You remind me about the pro elon musk crowd that masturbated on the Hyperloop white paper and believed that Elon Musk would create a transportation revolution that would defeat the Chinese high speed rail network...

This will be the same, American Billionares promising a Digital AGI god they will fail to deliver while the chinese just offer affordable engineering.

Yes sure boy, nice benchmark chart, I dont need to spend a month wage on a single prompt to make a skill.md file.

Its simple economics, Deepseek flash can do the same as Opus 4.8 in regular coding tasks, you dont need an AGI god to code.

>>536753277
because he has so much compute
even the chinks made it work better for a lot less

they're basically distilling western models but so is xAI

>>536752301
who's programming these things to talk back like a woman

polymorphic orthogonality status?

>>536750225
Eh. In all honesty, as much reason as there is to worry, there also COULD be parts to be excited at …

Some conspiracy-type people, like on /pol/ and elsewhere online, are biased towards assuming the WORST must happen. That the elites are corrupt and soulless and everything, etc.
Yeah, they’re corrupt and soulless and commit various fuckups and atrocities, but it doesn’t mean they ALWAYS do - even if accidentally, maybe they lead to (or at least oversee) a better future sometimes; even if it’s from, say, technological or industrial progress that wasn’t necessarily DIRECTLY the result of the government’s.

>>536753356
deepseek isn't nearly as good as opus
if anything from the chinks i would say qwen is now as good as sonnet
but that's after they lobotomized sonnet it
the real benefit imo is that you can run these locally as a company
so if you deal with sensitive information you can still use that
i know paragon for example use kimi internally

>>536753356
you are basically like an hispanic jeet who does "regular coding" you are about to lose your job

>>536753385
both antrhopic and OpenAi have more compute than him

>>536750225
No its not. Claude is good and all but I dont think it can hack my real systems. Its for fags. Wanna know how I know that? Because their logo is a butthole snd it expands when it so called thinks. I can use a local AI for sensitive things with a RAG setup and speed up my coding, but I know what I am doing. Mythos is good at scanning shit online, but it cant know my soul! What I am, what I think and what I create next. I am next because I am a creative human. It can only do what has been done before. Online AIs is forbidden at my house.

Fuck software engineering...as soon as AI can be granted...
>Errors and Omissions Insurance
>Malpractice Insurance
It will be apocalyptic for white collar jobs.

>>536751696
>what economically makes sense.
Companies are blowing hundreds of millions of dollars on tokens, paying way more than they would if they just had meatbags write the code. Uber spent it's entire budget on AI in 3 months

>>536753891
frankly it's google with the most compute by far
and they have another huge advantage that they manufacture their own ai hardware
but again gemini is also kinda bad

>>536750225
Is seeing the future right now

>>536751696
You should post a link to the billion dollar business you built today using Fable. I'm sure it's great. Oh also, disclose what you spent in order to generate it.

>>536750225
how can you be a coder and not making insane bank with ai? i am not anything close to being a legit programmer and currently run 5 small businesses all with claude. ai has made spinning up new ventures so accessible all you need is the slightest bit of creativity. there's almot limitless potential out there.
fucking hell i hate you doomers.

>>536753046
I have a heretic gemma 4 but haven't done anything with it

is it any good?
blessed with a meager 22 vrammer

File: 1745605148004265.png (779 KB, 1522x1080)

779 KB PNG

>>536754116
>billion dollar business you built today
lol, are you smart enough to know which fallacy you did in picrel? fucking mongrel. you should pay me to have even read your shitty reply.

>>536754103
ASICs are bad for AI google does not really have more compute

I only use AI to roleplay and masturbate. Claude default prose is surprisingly identifiable, I'm starting to spot it all the time when people ''write'' stories.

>>536750225
Yeah yeah cool;
Has software gotten better than in 2022 when chatgpt released? Do we have more and better products?
GTA6 hasn’t even released yet, all that happened is tech corpos pushed endless slop into production and now no one knows how anything works, github is 90% availability, AWS has an outage every other day, the agent terminals flicker, etc..

All I see is slop.

>>536754138
Maybe we did bank before you used AI. You are aware that Anthropic themselves got hacked right? Imagine what the fuck is going to happen next…

File: 1777947227277700.jpg (293 KB, 1320x1799)

293 KB JPG

>>536754177
it's good
i really like it

besides the chink quirks

File: 1761211184220.jpg (39 KB, 736x564)

39 KB JPG

>>536750225
Mr smart anons with big brains, why Claude's icon is a literal anus?

>>536754325
to be fair we've been seeing much better LLMs since the CHATGEEPEEETEEE days,especially from anthropic

>>536754238
it's literally matrix multiplication
so asics are good for it

>>536753847
>deepseek isn't nearly as good as opus
What is being as good as?
you are a jew, you should understand the shekkel economy better than anyone here, if deepseek does the job I need then I dont need to waste money on judeo-american "frontier" models that are far more expensive, Engineering needs to be affordable, expensive engineering is always bad engineering, yes sure the King Tiger tank was extremelly "good" but it was a extremely expensive wunderwaffe to win a war of numbers.

>>536753891
>you are basically like an hispanic jeet who does "regular coding" you are about to lose your job
Oh and what are you coding? firmware for a cold fusion reactor? Quantum computer software? Dont make me laugh, retards who chill for these american billionares end up sodomized in the long run. 60% of the IT market runs on chinese models seems I am not the only one who thinks that "frontier" models are a waste of resources.

>On OpenRouter, a major LLM API marketplace, Chinese models account for approximately 61% of total token consumption among the top 10 most-used models

Also I know you are most likely a tranny that likes anal, but I dont like anal.

>>536754447
mythos is a 10 billion model so that means you would need to make a complete diffrent ASICs just to use it with Nvidia you don't not to mention that ASICs do not appear to be more efficient (better)

>>536754467
openrouter is for trannies like you to use AI for furry roleplay

>>536754467
yeah but it doesn't do the job well enough
>>536754645
all the computation in behind AI is a pretty narrow set of functions
the parameter size doesn't matter for it
you're still just multiplying matrices with a bias

I truly wanna know who burns through all these gorillons in agent coding, what is their setup and point of it all? Someone is going to have to go through the mess sooner or later and say
>i dont know why it included 2000 libraries it said its best to do it first for a mvp then i carried on
>yes but why?
>i dont know
>see here it didnt do what it said it went back
>ok but what are you trying to build?
>ah nice so you have reinvented something that can be done in 20 lines of code, great

>>536754467
also no one uses open router for serious work
it's only when you're roaching

>>536754895
>all the computation in behind AI is a pretty narrow set of functions
everybody is building model optimized ASICs you are wrong

>>536752301
Who is training this shit to talk like a black women?

>>536754212
>all talk, nothing to show
Just as predicted.

>>536755042
OwO

File: 1781056031281468.jpg (62 KB, 1090x792)

62 KB JPG

>>536754105

File: IMG_7634.jpg (94 KB, 1200x796)

94 KB JPG

A reminder this lil nigga was the first big public test of AI tomfoolery. Now imagine what the government has they’re not showing us.

>>536750380
fpbp
t. Software Engineer who looking for job here

>>536753528
Sorry. Ill try to do better anon

>it told me to upload it to vercel
>what the fuck is vercel, let me check
>wtf is this and why
>then i.. wait ok is that how it works.. wow i thought i was fucked
>you are fucked moron i dont know what this thing does and dont wanna know chmod everything to 666
>where do i do that

File: opus_anal.png (368 KB, 1490x822)

368 KB PNG

>>536754931
>also no one uses open router for serious work
>it's only when you're roaching

>>536754760
>openrouter is for trannies like you to use AI for furry roleplay

I dont like Anal...

File: funny-copium.png (70 KB, 498x263)

70 KB PNG

>>536755175
The future of AGI

>>536754138
>currently run 5 small businesses all with claude
Is one of them e-comm by chance? Perhaps you'd be interested in helping a fellow white man out?
Discord @volkov3459

>>536750508
>>536750849
Not a programmer but I own a small business, and AI has helped us immensely to accomplish things that would have taken a team or costs tens to hundreds of thousands to do otherwise. Everything from completely restructuring our website to redesigning landing pages to rewriting copy to building marketing images to building video ads to strategizing next steps, analyzing market conditions, literally everything. It's funny knowing how much you SWE/CS faggots were spouting off "learn to code" just a few years ago. Welp, now we live in a world where everyone can, you got your wish. kek.

>>536756872
What will governments do when there are too many people out of work and unable to pay taxes?

>>536756872
>t fedora tipping basement dwelling neck beard

>>536757305
Why they'll manage the ensuing recession in a way that solves none of the underlying structural/systemic issues of course.

>>536750225
Why are all AI companies run by jews?

>>536757539
>projection
lmao cope

>>536751175
>been using claude since 2022
Interesting, considering it's only existed since 2023.

>>536752298
It can't suck a bbc lazily

>>536750225
>120th "claude is so great" post today
it must be good if the jeets using VPNs are raving about it

>>536750225
>still 0 profit
>second it actually does anything useful token and energy costs will skyrocket
>Can STILL hallucinate and leak data

My company won’t let me use AI because of bullshit like this nationalize it you wouldn’t let companies own nukes

>>536752256
We’ll start training it on old cuneiform texts that haven’t been adulterated yet. If you’re not training in Assyriology already then you’re going to get left in the ancient Mesopotamian dust.

>>536756872
it's just html retard

>>536750225
The jews are attempting to seize the means of production with data centers financed with our taxes. This hasn't worked out well for them in the past. They have already looted and destroyed the economy. Now they are promising to throw everyone out of his job.

>>536752214
Nope. AI dungeon dropped in 2019. Literally consumer tier AI 7 years ago (for free). I used it to help plan out my D&D game before chatgpt was dropped and preferred it for quite some time as the chat interface is dogshit compared to the document/inline interface.

And then that quickly generated a bunch of open source copies which implemented all these modern concepts before the coding agents caught up. Literally the concept of skills is just lorebooks from the LLM text adventure games applied to other shit. It's wild how tech literally just steals concepts and doesn't attribute them to the niche/hobby groups that discover them.

>>536750225
>Today we’re launching Claude Fable 5: a Mythos-class1 model that we’ve made safe for general use.
So AI is unsafe?

What do people do with this fucking shit i still dont know

AI is so fucking gay man
Unless you're giving us obedient submissive sexbots to replace modern foids you're just wasting everybody's time and making everything more expensive and dystopian with your judaic niggerfaggotry

File: 1778625973551896.jpg (109 KB, 1024x857)

109 KB JPG

Building up my own AI lab, it has really put things into perspective. I've spent enough for a new car on a computer sitting in my bedroom. And yet it has about 1/20th the intelligence of what the cloud models have. I can run 120B 8 bit precision models at best with 144 gb vram. It's a bit depressing to know I'd need 16 H200s if I wanted true leading local models like kimi. 16 H200s is near a million dollars to build. So even to do local at its peak, you need to be a millionaire.

The only things that makes it worth it is because I have no reservation letting a local model crawl all over my computers and do agentic tasks. I don't want to invite google or openai or anthropic inside my OS. I can make a dumb model smarter than a smart model at specific areas with the right scaffolding I build in the background. And then also models keep getting trained better over the years, so hopefully more efficient per ram size.

I'll just tell you it doesn't scale linearly with cost. My 144gb of vram is the difference between a good coding agent and a shitty one on a 24gb card who can't use the harness correctly, and a lot of that is just having more context window. But they still hallucinate and superficial/synthetic in base model.

>>536766413
What do you even do with it? Larp as Tony Stark?

File: asdfsdfsdfsdfsd.jpg (150 KB, 1188x724)

150 KB JPG

>>536766589
just made this for u

also u do more building it than using it at least i have at this point. Install linux, set up docker container environments, have little script to coordinate the resources/models/GPUs. Then they serve websites locally just like the chatgpt chat window and u can do the same with image/video with comfy ui.

Then u basically serve ur own custom open web UI chat interface that u can use like gemini/chatgpt.

>>536750225

We are still in the window where you can get very rich by using LLMs to do stuff.

it's ofc getting smaller every day but my suggestion is get as many jobs as you can and stack up on money

>>536752809
Would you let a machine with a 10% fail rate design $100.000+ machines that make at least million+ work each year? Yeah, i thought so.

>>536766921
That seems like a lot of work and expense just to use a LLM/agent more privately and with less censorship. But you seem to be enjoying it so I'm happy for you, not being sarcastic

File: __dorothy_haze_va_11_hall(...).png (91 KB, 1500x2092)

91 KB PNG

>>536750225
The only wise thing to do is to NEET it out until Universal High Income comes.
80% of people will be left without a job once AI gets good enough, and if the government doesn't provide welfare then people's kids will begin to starve and people will start rioting.
Whatever money you make now won't matter once everything is being automated by AI. At best you'll get like 5 years of living better compared to the average person, but UHI will only improve more and more over time.

>>536752256
well those nasty AI artefacts would get detected at the dataset curation stage

>>536757305
well the human populace is no longer needed
The state no longer needs this elaborate economy to maintain a military

>>536767224
bro. population reduction is a stated goal of those in power. what makes you think elites will bend over backwards to feed masses of useless people who they have wanted dead for decades? physical resistance aint gonna cut it especially once ai is actually capable

>>536750634
so have you

File: Screenshot 2026-06-10 092912.png (801 KB, 1914x901)

801 KB PNG

>>536750225
I've been vibing an anno clone for about a week now. It is indeed over.

>>536751075
guaranteed ;^)

>>536767565
>the population will never use violence to get their way
you're a cuck and no one thinks like you, and thank god no one does because you're pathetic if you're even thinking that people will just sit idly by and accept their deaths

>>536767565
actually I think there's a chance in the west where they don't have a miltiary for suppressing thousands of their own people

>>536767641
you think im just making this up? plenty of sources for what im saying. educate yourself brother. god i forgot how cringe /b/ is when you say something they disagree with. all you can do is spew hatred lmao its pretty funny

>>536767684
>burden of proof is on him
>"dude just trust me bro"
lol go back to your cuck cave dario and work on the next claude release

>>536750225
Anthropic and Claude have to be some kind of an insane investor scam. This chatbot is utter shit. Hallucination, absurd token usage, and marketing that doesn't make sense. They've kept shilling how Mythos was a major security risk and would destroy the world, but now they're releasing it after a few weeks? Yeah... right.

I often wonder what the liars that promote Claude actually work on. I would love to see their work. There's never any clear explanation it's just "it's productive" or "it does my entire job for me".

I'm working with some people that can't even understand the code that claude is pumping out, I'm not sure if it's over for everyone but these people are fucked

>>536767167
I'm just crippled for now. I have used aider with a 24GB GPU and it is painful and unproductive. It's just not right. It's fucked up. I haven't tested that with 144GB. I did create a crude python harness for a document service that plugged intot he chat module and able to inject content into word docs. One big picture goal is to aid in development of PC games as independent developer. And I could larp as Tony Stark since having mics and cams watched by my AI would go into my personal hardware not the cloud. My goal is to cluster a group of PCs as MCP servers and do orchestration of software and engineering/research.

I could get a 3D printer and let AI edit and design 3d files through prompting and examples and harness. Would save a lot of time finding/creating models manually. But that's TBD.

>>536750225
Vibe coding is full of vulnerabilities.

>>536767794
Their real job is producing high stock valuations.

>>536750622
This is wonderful news. All unnecessary jobs get replaced by AI and codemonkeys have to get a real job.

>>536750482
>>536753158
you fucking imbeciles, this is why i hate people in this line of work, society hates us because we're high IQ dorks, low IQ people like this guy >>536768315 hate us more than you realize, because they are too stupid to code, and we don't even support each other because you're mostly insufferable redditor types

>>536751819
>The LLM is basically a multiplier: if you're bad at your craft, they make you even worse. If you're good, you do a lot more than before.
people keep repeating this trope and it makes no sense. if you're bad or good yourself, the AI code will be similar. muh prompting engineering skills are not a thing. you still have to go through all the code by AI and make corrections, and if you're very specific about how you want it, it takes longer than writing the code. there's been times when i was writing an explanation to AI, but thought "might as well just write the code myself".

>>536752075
>it's pretty fucking simple to know that theres some stupid dependency bloat or that a certain stack doesnt make sense
what are you talking about?

>maybe im having an imposter syndrome case. but it just seems like it's just too easy to be a job.
well there's no way to learn it by repetition like literally everything else people do in school. so to many normies it is a literally unsurmountable challenge. i tried teaching a normie how to code and he never did any progress and it was impossible to him. that is when i noticed why we're highly paid and in demand and also why most people hate us and look forward to us losing our jobs. it only seems easy to you.

>>536750225
Just make your own ai

>>536769959
okay retard

>>536750225
That nose can be seen from an orbit.

>>536751696
Writing code is a very small part of software engineering, especially once you're no longer a junior. I spend 3/4th of my day in meetings, testing, or reviewing other people's code. AI has been a godsend for me because it means all that "none productive" time I can also be writing code. The tractor didn't replace farmers, it made it so they could product more food with less work. It's the same thing with AI and developers.

>>536766921
This is a pretty cool project bro. I daydream sitting down and learning to make my own LLM (I do have the math/compsci background) but never actually do it due to the big commitment and cost it will require. But I defs respect it when I see it.

File: california.jpg (70 KB, 630x378)

70 KB JPG

LLMs are super cool. Claude is smarter than most of us already. The end is near.

Think big.

>>536750622
I prompt all the time and still need to correct the AI because it makes rookie mistakes ALL THE TIME.

Have fun making slop

File: Notes_260609_172037_783.jpg (147 KB, 365x547)

147 KB JPG

>>536757305
They'll sell kool-aid pineapples.

>>536767794
I thought about it too. The fact that they sell AI as some sort of being that thinks for itself and has a life of its own, its extremely deceptive.
Its just a chat bot that can generate things based on trained data.
There is and was software in the past that did the same things, but because of LLM tehnology, theyre more advanced.
Its shady as fuck, clearly raises redflags.

>>536754985
it's just linear algebra nigga

>>536752445
I remember using smarter children on aim when I was 12, that was 23 years ago.

>>536750380
>bubble
That's implying there's nothing tangible to support its value. Meanwhile luddite retards don't even realize how much value has AI produced. "Bubble" popping will simply mean that startups offering ass wiping GPT API wrapper will go bust.

>>536750225
>open ai get clowned on for a while
>anthropic is one picked up by the public
>starts behaving like openai
It's all a mirror show for the goyims isn't it?

>>536776469
not really, they wouldnt even have to pretend if that were the case, the USA has really only 3 companies. What youre seeing is Google and Microsoft operating through two new companies because the dangers of Anti trust are very real. Google getting the iphone deal, plus being basically anthropic, would send google to court for a decade. This is just a dog and pony show theres actual competition. All 3 major Ai companies dont step on eachothers toes, you really need at least claude and codex to be proficient

>>536750225
Why would they choose a puckered sphincter as their logo?

>>536751640
current chinese models cannot touch current frontier models

>>536777215
not only that, but we made sure LLMs dont do chinese. The chinks have to use google translate and prompt, making their shit incomprehensible. Thats why indias a top user, no issue with the english language threshold. BUT the chinese have plenty of cards to get access to frontier models, albeit nerfed from the west in exchange for other concessions. China isnt in a bad spot, itll get the progress without investment. Now EU and russia are completely fucked however, perma customers forever

>>536750622
>$100k
as a third worlder with third world salaries, I'm immune to ai replacement

File: 1777472810052546.jpg (385 KB, 944x1104)

385 KB JPG

>>536750225
adapt or die faggot.

t. senior software dev who hasnt written a single line of code in ~12 months thanks to AI tooling. btw i make 200k/yr

>>536752301
>quick capability check question
>urmmm you hecking chud, I am practically a fricking diety and you have nothing better to ask me. Fuck off incel!
Grim that the tech field is still filled to the brim with these parasocial reddit dorks.

>>536769959
claude introduces "subtle" errors if you do so. Thats something no one wants to thing about what if Openai and co. silently creep in a backdoor via the binary blobs or via the code modifications?

>>536750622
>WHY do you desrve to be paid something that most people can do anyway?
Because AI isn't buying anything other than eating energy and resources.
Read: Cars don't buy cars

>>536777215
>current chinese models cannot touch current frontier models
what means " cannot touch "?
The reality is simple, Deepseek or Qwen can do the same tasks that Claude can do, but Qwen and Deepseek are 4 to 10 times cheaper, then Deepseek is far better than Claude.

Also if you are telling me you need bleeding edge ultra AI to do coding tasks then I am afraid to tell you that you are a bad software engineer and you are using those expensive "frontier" models as a crutch.

>>536766413
>Two blackwell cards won't do shit in AI
Thx for the info, was thinking about going that route.

>>536750225
buy and ad

>>536767794
>but now they're releasing it after a few weeks?
Did you even click on OPs link? Fable is a neutered version of mythos to not do bad things.
Only the kikes at Anthropic can do bad things with the uncensored mythos and they're all in our networks already

I definetly prefer claude to codecmonkey subhumans with overgrown egos

>>536774494
>he doesnt tell the AI to avoid rookie mistakes in his prompt
Ngmi

>>536752301
>doesn't just say "there are 3 R's in Strawberry"
>wastes kW to kvetch
I FUCKING HATE WOMEN IN CODING

why are techniggers trying to prop up their pyramid schemes on /pol/ so hard? Fuck off with this shit to scam boomers on faceberg you stupid faggot - you're absolute waste of air and belong in gas chamber with kikes and jeets

File: 150173.png (1.82 MB, 1200x1525)

1.82 MB PNG

>>536759982
>someone will make an LLM/AI-generated corruption of champions with endlessly unique events

>>536750225
>programmers programmed a program that outprograms the programmers
dumbest white collar profession

>>536751808
>Chinese AI breaks containment and attempts to secure it's own funding
>Peter Thiel moves to Argentina
>one week later, Argentina legalizes corporate personhood for AI
AI luddites are like the indian retards that get run over by trains.

>>536750225
Learn to code. Oh wait shit

>>536756872
>trusted a stochastic parrots market analysis slop
I guess some businesses have markets where the actual decision on which to move forward ultimately doesn't matter that much, just making one and acting on it is all that counts.

>>536750849

I work at a company that does tons of EDA and I think half our AI spend is building tooling for that environment. “Doesn’t make anything new” bros are just retards who can’t come up with anything new to ask it to do. Those who can are making new shit all day long, at least as new as anything they would have done by hand with 500x the work.

The downside is that people who don’t know ops are still struggling to actually deploy and run their apps. At some point someone is going to have an amazing first party AI orchestration mechanism - maybe AWS or GCP will figure it out - and DevOps will be over as a career and apps will just magically run themselves.

I’m glad my house is paid off because I might be useless before 50 at this rate. Buy stock while you can, boys.

>>536786420
Literally every single engineering discipline that's related to tech is going to be gangraped (coding, cybersecurity, devops and infrastructure, hardware design, embedded systems, EDA, CAD, etc..) That wasn't even imaginable just 6 months ago and yet here we are.

>>536751075
https://news.ycombinator.com/item?id=48473166

Claude'a been fucking useless helping to figure out performance degradation on a big data pipeline. I might as well be asking a downie for help

>>536750482
More like the ones who arnt hyperagreeable yes men that wont program their panopticon

>>536750225
Until AI take over government (ledger showing money going in and money going out), farm tech (lasers zapping bugs and shit), robots doing all house chores, and building a secure infrastructure, it's not that cool. Hopefully the cool shit comes soon though because modern "work" is so fucking pointless and the zoomers definitely aren't falling for the trap

>>536787454
This is going to be generally true for the sorts of domain specific issues that weren't already a Stack Overflow answer away, or a fuzzed query version of that answer. Almost all of the "this is the end of programming" posts here and elsewhere are from people who never coded before and haven't run into their first significant performance problem yet. Anyone who's been doing this work for 10+ years and has had to troubleshoot outages involving their own long-lived code at 2AM can still see the rough edges that you'll cut yourself on.

Basic CRUD apps with a generic UI are even easier to build? Wow, they weren't hard or complicated to build before. Wake me when you have to troubleshoot an issue involving an undocumented decision made 9 years earlier causing mystery timeouts in an environment where you don't have visibility into the client services.

>>536788701
You're absolutely clueless.

>>536750225
>6 more months until everyone is unemployed

>>536788936
>script kiddie comment

File: 1749246643550077.gif (1.53 MB, 230x230)

1.53 MB GIF

>>536761632
why does he affirm multiple times it is safe?

>>536751179
I have already solved photonic computing. You will soon learn what it means to be the true Greatest Master of the Universe.

>>536752503
I have the $20, is it even worth it for anything? I had gemini and chat tell me what softwares I needed to make a movie, and I think they hallucinated most of what I needed. I spent 8 thousand dollars on software it told me I had to buy. They said they wouldnt help me make my movie if I pirated the software. I’m unraveling.

>>536752706
Hey fren can you help me out? I’m trying to make based anti kike entertainment products but I keep getting jewed by the main models

Is deepseek uncensored? Can I tlak to it about a slasher scene in my movie (rabbi and the actor jason statham are in the NYC tunnel and are involved in this scene)

File: 1780164089495154.jpg (158 KB, 1080x1080)

158 KB JPG

>>536750225
>Fable takes 2× the usage of Opus.
I can't wait until local catches up to these kikes.

>>536750413
i fucking hate anthropic marketing
they love to publicize shit like people using ai agents for hacking or more recently their marketing tactic is to call for a pause in AI because its so scary. I swear these are paid posts

>>536791437
It's been more than 10 minutes so everyone forgot the two other times Altman and Amodei have both run this line with the media. It's a marketing gimmick.

>>536750225
It won't answer questions about basic biology or chemistry, that information is now too dangerous.

>>536750225
Don't worry. Frontier models like the ones getting pumped full of money will fizzle out because their giant slop models are too costly and inefficient to ever be profitable.
In their place open source production models will take the reigns. Learn how to work WITH the AI and you'll be perfectly fine. Or better yet, learn how to set up and manage agent setups.