>3 prompts on Fable 5.1>that will be $218where is all this AI insanity going?
>>541839273looks like its going towards gatekeeping processing power and information for money
>>541839267>turned on extra usage if using the plan>or was using the API unattended like a moronSelf inflicted injury, it takes it ages to setup 3D scenes.
>>541839267>where is all this AI insanity going?the same place the NFT and Crypto insanity is going
>>541839273>make me GTA 6>keep going until it looks good>wtf, you spent my money!Lol
>>541839267Nobody has any idea what that means. I'm the (probably) 1% of the population that knows what "vibe coding" means and even I don't get the significance of your post
>>541839267meanwhile Deepseek will code all day for a few dolllars lol
>>541839560Yeah I don't think the endgame for the big gated AI companies looks too good. Keep your stonks money in hardware producers.
>>541839267These are API rate costs, only fucking idiots pay at API rateAnd the most degenerate, braindead way to use AI is to 'one-shot' terrible ripoffs of popular AAA games, so you can get social media clout
>>541839676I wonder if they're good enough to mod assets for Arma now. New tanks etc with damage models.
>>541839717Almost certainly yes, the answer to can AI do it is usually yes now, as long as you're using frontier models like Fable. If you don't want to pay Fable prices, GLM or Kimi is a good alternative
>>541839267Corporations will pay that to avoid spending 1% of that money on hiring another person
>look mom i proomted another skeleton prototype of a game!we're like years in with this ai shit and it's all prototypes everywhere
>>541839555checked and kekd
>>541839833Other way around, those prompts we probably like 1 month of manpower each
>>541839267I've been noticing that all those commericial western AI cloud services have became extremelly bloated. Technically the quality is going up, but the amount of tokens they consume is growning disproportionally faster.All the self double checks, all the extra "thinking" on top of thinking, all the extra context online research, etc that they add is making them consuming much more tokens and time than before, for the exact same tasks. It's particularly annoying when they start talking about shit you didn't even explicity ask for. That consumes tokens I don't want to pay for.Luckly I'm mostly a hardware guy and use AI very sparingly during my small coding sessions. But I can totally see heavy AI users are probably getting robbed.
>>541839267I still haven't seen a uptick in improved software or games it all just lazy slop. The biggest trend is just new software to interface with AI and hvae more data to consume.
>>541839267>asking AI to make you a full game>not submitting 1 million of token minimum long request, multi-vetted gay
>>541839267Fable built me a house search tool that incorporates US census data to help me understand the racial composition of a neighborhood.
>>541839959This. Still on subscription but I burn through my 7 day quota by day 3-4 and have to get extra every week. But in those 3-4 days I’m deploying features that would have taken 3-4 weeks before. So what’s another $200 when they already pay me 6 figs. My API rate is close to $4K/mo I think which is nothing considering I am virtually running a team of 6 expert engineers all day every day
>>541840123It even compiled the last three censuses to provide me with a block wise ethnicity gradient - which neighborhoods are trending whiter, blacker, asianer, etc
>>541840123>>541840211And it incorporated blockwise household income historicals to identify gentrifying neighborhoods.
>>541839267that's bullshit, but I believe it
>>541839267>>541839273Desserts him right.
>>541839676API gives you full ownership and confidentiality. Subscription keeps things proprietary for you, but they scan and retain data for a set period of time. Free tier, you own nothing.
>>541839273>we'ren't
>>541839267If the AI bubble bursts, it could be big enough to trigger a global reset.
>>541839560>implying "coding all day has scalable value"same argument as saying "a private auto has no value because it sit idle"
>>541840211That’s amazing actually. Why wouldn’t there be need for a fine-grained tool like this. I congratulate you. Is it a web app or desktop? I guess all it needs to do is poll those APIs periodically (once per day or week?) and cache it and store it as a serialised lookup dictionary. How you visualise is then up to the user.
>>541840424Do you really think confidentiality matters when you're making bullshit counterstrike clones?
>>541839267And just think that they are still operating at a loss at those prices.
>>541840906Anthropic made a $560m profit in Q2 2026. Inference alone is disgustingly profitable, it's AI researcher salaries that are expensive
>>541840980>anthropic made a $560m profityou are literally retarded if you believe thatnot a single major ai company has made any profit to datehttps://isaiprofitable.com/
>>541840644[heavy breathing intensifies]
>>541840811The reason this doesn't exist as a commercial product already is because it would be a violation of the Fair Housing Act to overlay housing listings with racial data. It's absurd - but there's no longer a barrier to building one yourself besides a frontier model subscription and forty minutes of time refining it.
>it's still preferable to waste resources on this garbage noise filter software than to hire h1bs >a chink can still code it all, correctly, by hand for you later Lol bloody bastards
>>541840814No, that kid is retarded.. but that's why some people opt for API costs is all I'm saying.
>>541839527Making GTA 6 that way would be way cheaper than it is in reality.
>>541839267Hiring people is still cheaper than faggot AI. Everyone not a boomer already knew this.
>>541839267>convince yourself you don't need to learn anything because the computer will do it for you>never learn/forget how to learn skills>few years later, need to pay a subscription to be told how to cook your own dinnerits just training people to be retarded so that they can be sold common sense for the rest of their life.
>>541839267Don't forget to tip, faget.
>>541839847meet mr. proompter
>>541841093real chads don't use api because it's far too expensive. you just grab multiple subscriptions and you can have multiple adversarial agents working simultaneously. that's the newest meta is multiple adversarial agents on the same problem, that are focusing on least token usage.
>>541840261how would you ever know, if the data is correct?
>>541841199>Hiring people is still cheaper than faggot AIDepends on the application, some AI stuff since last year do really take out some jobs but not the final decision of someone to make everything not look convoluted
>>541839847Anthropic suddenly announced AGI won't exist until 2028.
>>541840450It should be we'rn't
>>541841263The data reduction is in the source code and nicely commented. You could review it, like any rational person would do, prior to pushing to production. Not very different than a pull request. Don't be lazy and review the shit.
>>541839555Auto approval+shitty prompts and ai goes generative over and over to the lowest common denominator of what you’re asking it. Deterministically shit. There’s a correct way to do this, even using local AI to offset token burn on repetitive stuff, but most people aren’t doing it right. Also Claude writes code so bloated you can’t navigate it, so it becomes a sunk cost fallacy.
>>541841358the problem is if you can write code then the time it takes you to understand the code is longer than it takes to just write iti've been writing software since the 90's and writing it is the easy part, understanding what you or what someone else wrote is the hard part.Kernighan's Law, and the ten to one reading to writing phenomenon
>>541841044The irony of that website is that it is clearly AI generated. Also the Q2 2026 results just came out in August, and the site's last update was in July, so it's out of date. Maybe google before running your dumb mouthhttps://www.forbes.com/sites/jonmarkman/2026/08/17/anthropics-groundbreaking-second-quarter-delivers-115b-in-revenue/
>>541841480>omg look for a single quarter they got a bunch of money >please ignore them burning billions of dollars >also ignore they are in debt still>please!
>>541841519You are so fucking dumb man. They posted a profitable quarter. If they are now profitable they are by definition no longer burning billions of dollars. They have to demonstrate consistent profitability now that their IPO is later this year.
>>541841452These models are like interns or junior engineers. They need correction, they need to be given tight leashes with regards to libraries, methods, etc or they'll deliver over-engineered jeetish garbage. But if you already know how to micromanage less experienced engineers, you can milk these tools for 10-50x speedup
>>541839527thisclear skill issue
>>541840980And they blow through it on training and all the dev work on routing and harnesses. Then they get cucked by the Chinese who give it away for free, albeit on a lesser scale. >>541841044This
>>541841668They're still ahead of the Chinese, but only barely now. There is definitely capability in Anthropic models that doesn't show up on benchmarks. I used GLM-5.3 the other day which is supposed to be Fable 5 quality and it's kind of dumb compared to Fable 5. And local models that are meant to be Opus killers like Qwen 3.8 are so damn slow and far too complex for most people to setup properly
>>541839267>order the entire menu>get served the entire menu>then ask to change things>get served again>play with the food a bit>get the billWtf! I only ordered once!Also did this retard use company resources? Planning to show off what he "made" on shitter but turned it into outrage bait?
>>541841798do you happen to use Tikkie?
>>541841732The Chinese train on the American models, there is no way the Chinese will surpass the U.S. but they’ll get good enough. It’s just funny that these companies spend $100’s of millions and China can distill on offshore gpu farms for a fraction of the cost. They’re slow because they don’t/can’t load all the parameters at once, limitation of not being able to run on a compute cluster. But well worth it because it gets good enough, and you can offset that slowness with other techniques like rag, LoRA, and fine tuning so you actually get usefulness out of one shot requests.
>>541839267>>541839273AI usage will just create an entirely new field of experts whose primary job is to write prompts in such a way that they get the desired outcome on the first or latest second try.AI can still help a lot with productivity, but it will also quickly rake up an insane bill if you just go trial & error and fire 20 prompts in a row instead of writing a proper prompt to begin with.
>>541839267How long until we can rent out compute from the home PC to an AI agent? I want that passive income.
>>541842268How big is your VRAM?
>>541841572This.LLMs are lower tools for experts, i.e. People who know what they are doing and exactly what they want.These are power tools the same way to inexperienced engineers - you are going to slice their finger off or nail their toe to a board because they don't know how to use them.LLMs are as smart & useful as the user.They fuck up allllll the time.If you aren't catching it fucking up, you probably aren't the smart one in the room.
>>541842195and that expert is me>helo bootiful pls be maiking Gta 7 pls :)
>>541839267Inevitable enshitification of anything aas. People should have supported open source.
i can code for less money than Claudesorry claude, you are being laid off soon
>>541839267this is why dario's employees leak trade secrets to china. its the right thing to do
Im on a company plan and use it to vibe code some tools for personal use like a money spending analyser, yeah those things exist but i want it my way, so a few days later i have a KI supported money analyzer that directly retrieves transaction from my bank accounts, runs classifications and statistics
>>54184233412GB 5070. Had to look that up, I don't game anymore so no idea why I bothered with a good graphics card. Think I still have illusions i'll start gaming again but modern gaming sucks balls and nostalgia gaming is like 1 session of something from the 90s then move on.
>>541839267I don't really know much about finances or LLMs, but aren't all the AI companies currently unprofitable? That effectively means the market can't evaluate the cost of using LLMs properly. And potentially, the usage of all non-local LLMs will cost much more.
>>541843068Funny thing is i know at least two of my colleagues did exactly the same thing, its truly a new era of personal software, going away from open source projects being the main lead for such things
>>541843068>stupid german thinks the sysadmin can't see what he does with company money
>>541843180None cares
>>541842195That basically already exists, but I wouldn't say 'expert' - it's like knowing how to google something properly to get a good result and not trash.
>>541839267He's a fucking noob and should have used Fable to make him custom PCG shaders using HLSL nodes with an asset pack he had already approved where all he has to do is fine tune parameters drop PCG volumes, but then again you could just use the free tier of Gemini and she will do it just fine
>>541843137the problem is privatized personal software release it online for others to not use.
>>541842195Prompt engineers is already a thing since chatgpt 1.0, got more relevant then image and video generation got a thing, i think now its less relevant again because models are less prompt sensitive
>>541841054> it would be a violation of the Fair Housing Act to overlay housing listings with racial data.Ridiculous but not unexpected.Just as an employer needs to know if the bastard he's interviewing intends to steal from him, file a bogus claim, go on long term sick, go on strike, cause friction etc I want to know if the neighbourhood in which I'm making a large personal and financial long-term commitment is suitable for me, and racial profiling is but one part of that large data set.Having that information is better than not having it.
>>541839273>using manners with clankers
>>541843452What
>>541839267Do retarded prompt that you know is going to burn loads of tokens.It burns loads of tokens.WTF HOW COULD THIS HAPPEN
>>541839527They already do this but with pajeets and a handful of trannies, which is why every game these days suck, looks and runs like shit, and is gay and retarded.
>>541839267Lol,lmao. I'll stay with my chinglish bot.
>>541842195>AI usage will just create an entirely new field of experts whose primary job is to write prompts in such a way that they get the desired outcome on the first or latest second try.>AI can still help a lot with productivity, but it will also quickly rake up an insane bill if you just go trial & error and fire 20 prompts in a row instead of writing a proper prompt to begin with.I'd just use deepseek. I could afford to run it simultaneously on 20 machines for a week and if probably only spent $100 while getting good results.
could have dropped some shitty assets in unity/unreal for free
>>541843075I have the same, plus 96GB RAM. I haven't found a local model suitable for coding yet, not for anything more complicated than some scripts or plugins.To get back to your question, the trouble is that parallelizing LLMs like that is infeasible. One GPU with 32GB RAM is more than twice faster than two 16GB GPUs on the same system bus, even with twice as much cores, when it comes to multiplying large matrices.So, never.
Man, they really wanna recoup those billions of dollars they blew
None of these people are shipping a single line of code
Even $1 is a waste for something that will never be used
>>541843785I'm currently maintaining a project for a large company, written mostly by Claude, and using LLMs to analyse data.It works.I don't like my work.
>>541841054>it would be a violation of the Fair Housing ActNot a violation if this service is hosted by an individual/entity that is not a subject to the US jurisdiction.I think you can turn it into a real service. Not sure how to monetize it though.
>>541839267LOL WHY DON'T YOU JUST RUN A MODEL LOCALLY ON YOUR CHROMEBOOK
>>541843905I also work for several large clients and they would cancel the contract at the slightest sign of LLM code.
>>541844070I wouldn't use LLM generated code either, but there's obviously a market for that. Not for vibecoded aislop, mind you. The foot is in the door, so to speak.Aforementioned company has a liberal licensing policy and a dedicated LLM gateway for employees. It's one of the biggest European companies. And its CEO is jewish.I'm sure everything will be fine.
>>541839267You're asking it to program a videogame bro it's not supposed to be cheap
>>541844070How would they tell? Besides the obvious claude co-author or thousands of lines commits tells
>>541840261Bro I bought a house to flip in Cleveland, can you check my neighborhood? Kennedy Ave. I got my house for dirt cheap and the neighboring house sold for 150k.
Chinese models are compatative on price to performance.
>>541839273Nice fucking screenshots very readable
>>541844356I think the guy is making the opposite point, that it is surprisingly cheap for a working game prototype, and all anons here are missing the point.
>>541844585if anons can't infer the meaning but a LLM couldwhat does it say about 4chins
>>541844649>what does it say about 4chinsThe cesspool of the internet.https://www.youtube.com/watch?v=c9oOKqMjNUo
>>541839337most people get no benefit from the latest models over slightly older models. openai and claude have to continue to push out newer models as that is their business but people will be using Small Language Models offered cheaper or free on their computers and phones because they won't need Fable 8 in 2029 or whatever is new then.iphone 20 will be able to provide enough compute for an opensource SLM which answers all your questions and can search the internet for you.
I always set it to 'Sonnet.' I like chatting with the thing and if you don't choose the latest model you can interact with it pretty much all day so long and you tell it not to write any code or do any web searches without your explicit permission. Even if you are asking it to write code and doing web searches frequently during the conversations you won't run out of tokens. Only when you let the thing do whatever it wants will it use up all your resources. It doesn't understand time, scarcity, or opportunity cost very well. Easy to engineer around though.
>>541839273Why is he saying please?
>>541839267>5 more sessions and he could of buy an 3090cloud fags btfo as usual
>stop subsidizing LLM use and raise prices to start trying to make a profit>customers nope outlol. lmao even.This technology was definitely worth crashing the world economy for
>>541839267>screenshot>no gameplayi bet it's utter shite
>>541840063The entire business model operates at an extreme loss for every ai company currently
>>541840261>>541840211>>541840123now, let me tell you why you wasted your money, and are a retarded person.I was a census data collecter.i went door to door. i did the non response follow ups. there were a LOT of blacks working on the census, especially in charge leadership positions.and they kept quitting when they were asked to do math or fix anything or accept responsibility.so i kept getting promoted. a lot.numerator, crew leader, field operations supervisor, then into upper operations management for a tri-state area.they kept quitting, and when they did, they'd see one white guy making templates on his phone and sending them to his people to make status updates fast and easy. i was promoted more in the department of commerce than i have been at every other job combined.so i know personally, from each step, what happens with US census data.first a bunch of numbers are collected, then they just completely ignore that, and fabricate numbers they think will keep people docile and not revolting. 2% of the population has been jews longer than I've been alive. 13% black, for how long now? those % were chosen because they don't make people feel threatened or replaced. they just straight up fabricate the public data.its just made up and that what we use.you used nonsense collected by niggers, that was then ignored in favor of just lying, fed through a jewish lobotomized llm that thinks the holocaust numbers are accurate.that is how fucking retarded you are.you even believed you were being smart.
>>541839267Google tried this same marketing bullshit with Gemini 3. >WAOW LOOK AT THIS GAME GEMINI PRO PROGRAMMED WITH A SIMPLE PROMPT.Some dumbass pajeet shitskin shill went and posted the links on /g/ and it turned out they were absolute fucking shallow applications that didn't pass any extended scrutiny. And low and behold Gemini 3 was markedly inferior to Gemini 2.5
>>541844649Congratulations. You just figured out why AI makes so many people seethe.It makes talking to shallow retards obsolete.
how do people make money with these things unless you already have a job
>Be me >Switch to Kimi k3 Ai>We deploy research agents 3.....>You ran out of tokens>5 hours later... i type: go>Redeploying agents...>You ran out of tokens....>5 hours later i type: go>Redeploying research agents...>You ran out of tokens....... Sorry your weekly limit is over... wait another 6 days Yes we didnt write a single code for you.AI IS A SCAM!
>>541844585Not really. Anyone reasonably capable could learn how to throw something like that together in Unity in a few hours using some stock assets. It's just a basic skeleton FPS.
>>541841732But there is nothing to gain by going to the moon, at the same time, this is bankrupting America
>>541847255Anyone that supports AI is a lowkey kike bootlicker.
>>541847376Be honest with yourself.Can you throw it together in a few hours?
>I CANT SEE IT SO ITS FAKE!t. someone who desperately wants me to be christian
>>541847880no but I could probably get a freelancer from somewhere like fiverr to do it for me for less than $218 USD
>>541847983Almost certainly that freelancer will get a dirt-cheap chinese model to make it and then hand it back to you. Nobody is coding this by hand nowadays outside an actual game studio
>>541839267Fable = 1 dollar per task. Accuracy 81.5%GLM 5.3 Flash = 0.005 per task. Accuracy 73.3%China is about to catch up with ultra cheap models.
>>541839267Can anybody explain what any of this means and how it is politics?
