[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

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


Janitor applications are now open. Apply here!


[Advertise on 4chan]


File: 1775373028101333.png (97 KB, 811x402)
97 KB PNG
They're all celebrating the power and value of generative AI!
>>
File: 1765812317522319.png (3.88 MB, 2048x1536)
3.88 MB PNG
>>108958295
vibe sloppers on suicide watch lmao
>>
I understand the whole goal was to make people reliant on it(AI) and then turn them into addicts, but people really became reliant on it and are now addicts?
>>
File: 1778516083247702.png (54 KB, 699x312)
54 KB PNG
>>
File: 1754922649848010.png (374 KB, 778x967)
374 KB PNG
$1000 a month for your hello world webshit app
>>
I can't really tell, are they being sarcastic?
>>
>>108958358
It's actually over for vibesisters
>>
>>108958295
So it's (YOU) redditors that are spamming AI threads
>>
oops, looks like the snailcats said was 100% correct, almost all improvements in AI were from just throwing more and more power at the problem! oopsies, now its too expensive to use for even basic shit! oops! luddites, like usual, win again, like always! this makes ya mad!
>>
>>108958464
hey buddy shut the FUCK up
>>
>>108958295
HAHAHAHAHAHAHA where's the guy with the github pricing projections?
How are we feeling vibeGODS?
>>
>>108958490
hungry... im so very hungry and cold...
>>
>>108958350
>changed the good boy points system again award
>aiCHADS and vibeGODS cheering
>>
>>108958322
>but people really became reliant on it and are now addicts?
Yes.
>>
File: 1678968503957865.jpg (149 KB, 1473x1149)
149 KB JPG
>>108958295
>>108958333
>>108958350
>>
How are the tokens counted?
>>
>>108958464
wrong, you can just use your own api key
https://docs.github.com/en/copilot/how-tos/administer-copilot/manage-for-enterprise/use-your-own-api-keys
point it at deepseek and it's even cheaper than before
>>
>>108958522
Need a COPE CODE for
>just use a shittier model BYOK quantized deepslop bro
>>
>>108958529
>Quantized
>DeepSeek
Retard-sama...
>>
>>108958535
>Yeah bro it's definitely the full unquantized model no demand spikes bro no tiering keep putting the tokens in the bag
>>
>>108958540
anthropic is doing such things, not deepseek:
https://www.anthropic.com/engineering/april-23-postmortem
>1. On March 4, we changed Claude Code's default reasoning effort from high to medium to reduce the very long latency—enough to make the UI appear frozen—some users were seeing in high mode. This was the wrong tradeoff. We reverted this change on April 7 after users told us they'd prefer to default to higher intelligence and opt into lower effort for simple tasks. This impacted Sonnet 4.6 and Opus 4.6.
>2. On March 26, we shipped a change to clear Claude's older thinking from sessions that had been idle for over an hour, to reduce latency when users resumed those sessions. A bug caused this to keep happening every turn for the rest of the session instead of just once, which made Claude seem forgetful and repetitive. We fixed it on April 10. This affected Sonnet 4.6 and Opus 4.6.
>3. On April 16, we added a system prompt instruction to reduce verbosity. In combination with other prompt changes, it hurt coding quality and was reverted on April 20. This impacted Sonnet 4.6, Opus 4.6, and Opus 4.7.
>>
>>108958522
>>108958535
>Just use Sonnet 4.5 bro
>Just use Haiku 4.5 bro
>Just use gpt-oss 120B bro
>Just use Mistral Small bro
>>
>>108958518
In the most jewish way possible to encourage customers to grossly over-run their budget.
>>
>>108958295
>hell yeah bro vibe coding is awesome
>w-what do you mean it went up 100x in price I don't actually know how to code without it I'm going to lose my job
lmaoing
>>
>>108958295
Ive vibed complete features on my web app for an entire month and didnt come close to maximg out my Claude code.
>>
File: vibe.png (135 KB, 2496x908)
135 KB PNG
Snailcat bros this is hilarious.
>>
>>108958610
>hahah, i'm abusing the prompt-based pricing loophole
>nooo, ms cannot just close that loophole
>>
>>108958610
kek there it is
>vibeGODS, it's just.... um.... the market conditions, yeah? eat up!
>>
>>108958648
she clearly was abusing the prompt-based pricing. prompting your agent to never ask you anything but just deliver a final prodcut never was good practice.
>>
>>108958295
I remember anon several weeks ago pushing this pay-per-prompt model to the limits, by prompting it to spawn 20 sub agents that talked among themselves to push the project forward. His agents were like Tolkien, Myiamoto, Carmack, etc, and were assigned specific roles accordingly. The whole shit was hilarious! He could one shot tiny games with a single prompt, iirc.
>>
>>108958529
We need a VibeGOD Copedex, stat
>>
File: 1777741756899420.gif (184 KB, 244x248)
184 KB GIF
>>108958596
>>
>>108958648
Companies like OpenAI and Anthropic have spent the past few years trying to make their models bigger and smarter. Now it's time for them to focus on efficiency. I'm sure there are gains to be made.
>>
File: 1756750973692693.png (510 KB, 736x560)
510 KB PNG
>>108958610
feels good to be a snailcat
>>
>>108958596
Fucking KEK! I told you guys to bus factor the agent, but nooooo, you couldn't listen to this old snailcat.
>>
File: Society.jpg (2.91 MB, 2501x1407)
2.91 MB JPG
>>108958295
In the 80s, home computers put the means of production into the hands of a proletariat. AI is giving the bourgeoisie that control back. Local models are the main defense against this, which is why the AI industry isn't content with the already accumulated gains. They want a wide, deep, expensive moat that anyone developing software will have to pay rent to cross.
>>
>>108958682
microsoft's build conference literally is tomorrow. microsoft will announce their own efficient models for github copilot.
>>
>>108958322
You have to see AIjeets try to code without the auto complete bullshit. It's hilarious, they have no idea what to do
>>
>>108958522
>>108958682
>>108958710
Cope type: Pricing Efficiency
>The quality of the new Efficient Models is on par with Gemma 4 31B, but it's the only model I can afford after getting shafted by Micro$hit
>>
>>108958350
>copilot student unlimted
it's like selling crack on school yard
>>
Vibebros, did the Luddites win?
>>
File: 1779163843550838.png (683 KB, 3456x2234)
683 KB PNG
>>108958306
>>108958333
>>108958350
>>108958464
>>108958596
>>108958768
Not if you're a local Mac-Chad
>>
>>108958774
Hardware specs? Are we talking $20k Mac Studio here? Can it just about run a Qwen then?
>>
File: 1764917855774703.png (1.05 MB, 800x1000)
1.05 MB PNG
hey Vibechuds, it's time to invest your money into something useful now instead of tokens.
>>
>>108958785
Need a "So you're a recovering vibecuck..." infographic lmao
>>
>>108958783
M3 256 GB ram
>>
>>108958644
>>108958659
>abusing the prompt-based pricing
LOL
people were being baited-and-switched and you're lying as though they were abusing something
>um actually those were only the introductory prices, it's industry standard to increase prices by 1000x after the honeymoon phase
it's probably illegal but the US government is so corrupt and in bed with the AI jews that nothing will ever be done about it
>>
>>108958796
we knew this was coming for a while
it was called as soon as people were bragging on X about running 15 agents in parallel
>>
>>108958794
Which model? The 8 GB, 16 GB, or 24 GB RAM?
>>
>>108958610
>60x
lol lmao even
>>
>>108958801
if the AI companies advertise agentic AI as the future and offer initially low prices and then a few months later jack up prices by 1000x then that's a bait and switch
what are AI bros (Indians) supposed to do now that they can't afford it? they literally cannot code without it (or with it, probably)
>>
>>108958817
this pricing is not new to you if you've been using API directly from OpenAI/Anthropic. I prefer it actually to avoid usage limits
>>
File: 1764445745855421.png (1.63 MB, 1562x675)
1.63 MB PNG
>>108958817
>what are AI bros (Indians) supposed to do now
back to work
>>
>>108958785
what a shit list of random books. design patterns is for java but there's no book teaching you java. ...
>>
>>108958852
>design patterns is for java
no it's not Vibecuck
>>
>>108958863
do tell which patterns do apply to c
>>
>>108958868
here, scroll to the bottom:
https://www.adamtornhill.com/articles.htm
>>
>>108958333
>>108958295
sarcasm is not picked up by some LLMs so the hive will see these posts and think meatbags enjoy the wallet rape. lol & lmao
>>
Why don't these retards just set up deepseek LLM with agents and call it a day?
>>
>>108958464
I’m a Luddite in far more ways than this one (I hate phone apps too), and all the AI companies need to do is tell companies to abandon every way of using their service that isn’t AI. You can’t phone companies and speak to people any more, even though that was better. You can’t print off plane tickets at home and take them to the airport, at least not for the flight I need to take in a couple of weeks. You can’t even buy things in person if you need anything other than groceries or the occasional item of clothing now. AI just needs to fuck up my life even harder, and the normies will lap it up.
>>
>>108958306
That man is experiencing levels of absolute comfort I will never achieve
>>
>>108959250
he really probably is not
>>
>>108959250
He's dead.
>>
It's not worth it unless you can run it locally
>>
File: 1754418993393612.png (909 KB, 3456x1026)
909 KB PNG
>>108958783
See Pic rel. Bought it in December 2025 for just over six grand.
>>
>>108958794
>>108958806
See >>108959334
>>
>>108959201
>ChatGPT, what do our customers think of the price increases?

It's actually pretty funny, though, because many goyim genuinely seem to love being financially devastated. So its response might not be as incorrect as you think.
>>
>>108959221
>Why don't these retards
>Retards
You answered your own question
>>
>>108958295
Maybe stop being fucking poor you waste of air. If you have the best tools at your disposal and cannot make a buck out of it then the problem is you
>>
>>108958768
Luddites won when the Apollo program turned to be an expensive nothingburger and the shuttle program was even worse of a failure, so much it cancelled the successor program with no replacement. Just like we went back to single stack rockets (mostly) in the way soviets did it, we're going back to proper human coding with powerful auto-complete with local llms. The kind that detects you're writing boilerplate in 2 seconds and you train in your personal habits, not this delusional robo-vibe-coding bullshit that's worse than copy-pasting templates (which is free) and horridly faulty, dangerous (it melts your brain) and expensive.
>>
>>108958295
Does github copilot use the same compute sources as desktop copilot?
>>
>>108959687
>Compute
No such thing. Stop using buzzwords
>>
>>108958295
Luddite cope thread
>>
>>108959704
Really? So would it be like, server farms or whatever?
>>
>>108958295
whats the issue?
consoomers love spending money
>>
>>108959709
you dont know what cope means
>>
>>108959709
You don't know what luddite means either.
>>
>>108958295
Been telling you people for over a year that the free ride was ending. Did you listen? No.
>>
Coding is on borrowed time regardless of the cost. No matter how greedy corpos get, cost of tokens to run current models drops to zero given enough time (not even a lot, just a couple of years) In 2 years, open source models will be as good as opus 4.8, and in 5 years they'll be runnable on inexpensive consumer hardware.
>>
>>108958306
Maybe NOW they'll learn to code. :)
>>
File: file.png (993 KB, 1414x1256)
993 KB PNG
>>108959391
Americas unironcally do love to get an opportunity to talk about how poor and broke they are and how everything getting more expensive and they can't afford anything before before going out and spending more money they claimed they didn't have 5 mins ago
>>
>>108958350
Where's the premium crystals? The super mega awesome ultra shards? The code infused special pancakes? Github should be implementing every manner of gacha currency tricks to drain every cent from these literal retards
>>
>>108960587
No lol we will continue to insult coders. Learning is for pussies.
>>
>>108961791
What a retarded title, byline and top blurb. Luxury items are bought by the rich, who benefit from inflation (it all goes back to their pocket). The normal people pay a lot more for things that are mandatory to survive, such as housing, phones (can't get a job without one, ergo mandatory to survive), computers (in many professions same as phone, you need a personal device), etc.
Budget options ARE significantly up in profit the past several years, despite the fact that "budget options" don't actually exist anymore due to inflation (they cost nearly the same as standard-grade options). And retail sales ex-essentials are way down.
>>
>>108961949
nope
most popualr car on the road? oversized luxury suv. Either americans are all rich as you say or they are all unbale to stop spending on junk like that. It's just a false strawman to say that the article is talking about basic necessities like food and rent when it's clearly not. Americans but a lot of luxury goods, it's not only the rich unless your definition of rich is anyone above the bare minimum poverty line. even within the realm of food they're all spending exorbitant amounts of doordash and crap and eating out daily.
>>
>>108958322
It starts off with free samples, to get people addicted and crave for premium products. Like free cigarettes and vapes.
>>
>>108961962
Are they new cars? What was their purchase price? Who drives them? When were they purchased? By whom?
Retard.
>>
>>108958295
>deepseek v4 pro
>one million tokens for less than $0.5
Thank you Xi Chingping
>>
>>108958295
>company stops subsidizing product
>poors revolt for having to pay a fair price
Go use llama 3, faggot
>>
>>108961791
They are broke and poor, even with a big SUV and luxury goods. It's called credit, the availability of which helps drive inflation in the first place, as poorfags somehow get 96 month loans to buy their car from shylock auto mall.
>>
>>108958295
gotta love bait & switch
are these companies preparing for the big IPOs? is this temporary, or is this the step at which they recover all those billions "invested" in this scam?
>>
>>108958295
these retards still subscribe to github copilot to pay openai betrayal tax?
>>
>>108962090
Well now they can get a loan for their AI sloppa. You don’t want to miss out on making millions off of your vibecoded viral app, right?
>>
>>108962042
>10x times as much token usage due to it being a shit model
thanks Xi
>>
>>108962102
They locked down enoguh enterprise users that there’s no point catering to proles anymore. A poorfag who suddenly went from $50 to $800 will leave for some chink model, but in enterprise it’s a year long process for approval (or even more if a local buildout is an option) so why not extract as much money as you can?
>>
>>108960585
>cost of tokens to run current models drops to zero given enough time
No because they discovered that adding more layers ("thinking" mode, agentic mode) can increase token usage by orders of magnitude. And it's still not good enough, so they'll no doubt come up with more ways to use electricity at an increasing rate.
>>
>>108962120
The amount of tokens is irrelevant when the final output is decent and the tokens are pretty much free. I have yet to chew through the $50 I paid deepseek last september. The only real downside is speed but I’m not a wagie with strict deadlines so whatever
>>
>>108962129
All you ever did was ask the weather, you wouldn't know what it's like to use these retarded tools to try to write any code, even fizzbuzz, correctly.
>>
>>108962122
Enterprise are the hugest of jews, making 'AI investment'/layoff decisions with prices going atmospheric is going to fuck shit up. MBA consultants already made their money and fucked off though, so good for them.
>>
>>108962132
Whatever helps you cope, sweaty
>>
>>108962134
AI is just a cover for layoffs done for other reasons. "We’re laying people off because AI" makes you sound like a forward looking company. "We’re laying people off cause the economy kinda sucks atm" makes you sound like a loser which makes share price go down.
>>
>>108958295
I don't get why a former colleague is telling me he vibe codes all day on claude with no issues but on copilot we're about to struggle to get past the first week, despite claude's API being the most expensive
>>
File: 1772448410151540.png (773 KB, 1417x1417)
773 KB PNG
AI is the future whether you like it or not.
Even if you say AI stocks are in a bubble, that didn't stop the internet from changing every aspect of daily life after the dot com bubble burst.
>>
File: 1754733627514357.png (48 KB, 923x341)
48 KB PNG
>>108958774
Might as well just get a Gemini Ultra subscription at that point
>>
>>108962068
>15 minutes of compute for $20
>a fair price
lol, but Im happy theyre stimulating on-prem/self-hosted solutions
>>
What exactly is Microsoft expecting to happen? Like the plebs will keep spending 20/100 bucks on their slop getting less then if they sent that money directly to Anthropic or Open Ai or even fucking Grok?
Like these prices are almost API level.
>>
>>108962510
expecting users stop using slop so they can shut down "ai"-project and stop the money bleed
>>
>>108962547
All they are going to do is just move to the competition. And the competition being investment based will get to brag about higher sign up numbers and get more money.
>>
>>108958868
I like agner fog, I don't read it cover to cover but do flip some pages for some micro ops.
>>
>>108962557
It's been four years of LLM slop push.
The infinite venture capital is only infinite as long as there is a prospect of becoming a monopoly. Since "the competition" is plenty, you don't get that here. Even if all the competition dies, you still don't have a monopoly, because any mid-size business has enough money to buy hardware capable of running equal-capability models for free.
>>
>>108962164
This
>>
File: 1761700696247832.png (136 KB, 728x514)
136 KB PNG
We won
>>
>>108962589
I mean Microsoft isn't even real competition. All they got is that shitty little Raptor model. They are just a reseller.
>>
>>108958295
>Just one prompt!
But it looks like they triggered an entire huge workflow with dozens of agents. No wonder that ate tokens like crazy.
>>
File: p9p.jpg (53 KB, 500x500)
53 KB JPG
>ai isn't profitable
>ai becomes profitable
kyek, luddite cope thread
>>
>>108962448
My M4 Max was "just" $6K when I bought it back in December
>>
>>108958518
They're vibe-counted.
>>
>>108958868
You can implement your own objects in C and then use all the patterns with them. Its dynamic languages that eliminate the need for many of the patterns.
>>
>>108958757
I just realised this is so fucking evil, many niggas will be jobless and without future
>>
>>108962510
Plebs are irrelevant. They have no money to spend and all they buy are cars and phones on credit. The actual money is in enterprise, which is why massgravel has been up on github for years and years and microslop doesn’t care because it literally doesn’t matter if some faggot buys or pirates their shitass OS. The whole point of winblows and office is to keep people there so mac / linux never becomes the default, and use this familiarity as sales funnel for companies that aren’t desperately poor. A single enterprise company tends to be worth more (and is less troublesome to keep) than tens of thousands of garden variety customers.
>>
snailgods eatin' good tonight
>>
>>108962761
I mean yes, but also this pic is older than 4chan itself I bet. So hold your horses.
>>
File: 1766876668693685.jpg (97 KB, 802x840)
97 KB JPG
>>108958596
>>
File: 1771781456615704.jpg (56 KB, 342x342)
56 KB JPG
>>108958350
>Context Size
...are they unironically charging higher than reseller API prices for context cucked model versions?
>>
>>108963702
Do you expect them to charge less? The whole point of copepilot is that microslop is already treated like trusted. Go ask whether you can use some boiler room reseller in a non-joke company and you’ll get some raised eyebrows at best.
You get microslop, aws and MAYBE direct openai / antrophic / google model use (all at exorbitant prices under enterprise licensing) and that’s it. It’s why the prices went to the moon recently, actual competition simply isn’t a concern.
>>
Good. It should be a tool for programmers to use, not autopilot for non-programmers to code.
>>
Real AI Chads are just "borrowing" published API keys and letting google cloud customers pay the bill.
>Google added AI permissions to API keys, even ones they previously told users were okay to share publicly. Hilarity ensues.
https://www.tomshardware.com/tech-industry/artificial-intelligence/google-cloud-customer-wakes-up-to-usd18-000-bill-despite-usd7-budget-thanks-to-forgotten-public-api-key-attacker-put-in-60-000-requests-and-blasted-through-usd1-400-spending-cap
>>
>>108964021
Real ai chads are reselling api access for literally free money
>>
Why not just subscribe to anthropic directly? I mean, they're probably going to rugpull soon too (and already have to some extent) but right now it's still ok for me using sonnet for most of my work.
>>
>>108964057
>Why not just subscribe to anthropic directly?
Aint paying $200 to dario goldberg
>sonnet
gtfo
>muh $100 / $20 tier
unusable
>>
>>108964090
>Aint paying $200 to dario goldberg
Why?
>>
>>108964090
>100/20 unusable
I think 20 for either Codex or Claude get's you more than 60 or even 100 gets you on copilot right now.
>>
>>108958757
>cuck students out of healthy brain development
based
>>
File: 1760300898475044.png (248 KB, 512x512)
248 KB PNG
time to get the O'Reilly subscription now and read some programming book, Vibechuds.
https://www.oreilly.com/online-learning/pricing.html
>>
>>108962042
I tried using deepseek v4 pro to revive an old c++ project
I burned like 20 dollars in a day
I think it's better to get dario's subsidies at this rate
>>
>>108963925
Sucks to suck. I'm still using reseller APIs. Price increase my ass. Imagine unironically paying for a model with a worse context window that what you can get for fucking free on openrouter.
>but they sell your data!
Yeah I'm sure M$ copepilot doesn't.
>>
Deepseek api + opencode go has been the poorfag meta for over a month. People on this website are really behind the curve.
>>
>>108964265
>deepseek
>I burned like 20 dollars in a day
This is more than 40 million tokens woth zero cache hits (which is… astronomically implausible), so this sounds rather doubtful
>>
>>108958295
>>108958350
>>108958610
this is what it looks like just before the bubble bursts
I give it another 4 months tops
>>
File: act.jpg (84 KB, 1924x571)
84 KB JPG
>>108964757
my bad, it was closer to 17 usd, I was working with kimi 2.6 before
>>
>>108964865
You’re a know nothing retard though. Spacex, openai and antrophic bought themselves an exception to SEC rules and did a bunch of financial wizardry that will cause retirement indexes to automatically buy their stocks at inflated IPO prices, which means they (and all insiders from previous series) will be flush with cash for years at the cost of grandma’s retirement.
>>
>>108964890
>56M tokens
>$17
Do you get it from deepseek or some third party that charges 5x as much? I have over 200m tokens since v4 released and spent less than you
Or are you doing some weird shit with caching? A significant majority should hit cache, which is borderline free on v4
>>
>>108964914
openrouter
but I read somewhere that they had issues with caching which might be driving that
>>
>>108964914
You can stop asking deepseek for the weather once a day and pretend that this means using AI to code anytime, retard.
>>
>>108958757
Github and microsoft has always done this. They give free access to github private repositories (when you had to pay for that) and microsoft services like microsoft word, excel, powerpoint, etc to students. They then rely on it and it becomes the norm at companies as well. You then have to depend on it.
>>
>>108965086
Autodesk also worked this way (3DS Max, also bought up Maya). Companies in turn started throwing money at Blender in an attempt to break the cycle.
>>
>>108958518
In the same way like asking Chat how many 'r' letters there are in word 'token'.
>>
>>108959334
>$6000
>chad
lmao, you paid extremely more for less performance than the average vibecoder had access to.
>>
>>108965229
>Implying vibe coding is the only thing I use it for

I was surprised at your lack of imagination. Then I remembered this place is full of spergs
>>
>>108958350
>$1000 a month
What's stopping me from staring my own "small scale" AI company, and charging 1/2 of that to a dozen companies/people? But secretly the AI is just me doing all the coding the old fashioned way?

is AI just a branding thing? do people pay premium for it because of the novelty effect?
Can I make fake or "off brand" AI that's just me responding to prompts, and charge people $100 per prompt reply?
>>
>>108965501
>Can I make fake or "off brand" AI that's just me responding to prompts, and charge people $100 per prompt reply?
not unless you can answer dozens of prompts in seconds
>>
>>108965501
Nothing's stopping you just don't get caught like Builder.ai

https://en.wikipedia.org/wiki/Builder.ai



[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.