>AI is unprofitabl-
>>108839651Cool. Now post expenses
All I see are a bunch of companies falling sucker for overspending on shit they could have gotten far cheaper.But I guess to some of them anything is preferable to having to employ people.
>>108839651I use the Chinese models. They are as good and much cheaper. No idea why anyone would pay that much for American models.
>>108839692deepseek v4 that can speak directly to claude code just came out. before you had no great harness for chinese models.
>>108839692Qwen 3.6 for 99% of people does what 'muh frontier models' does, and often times better.Literally just invest a few thousand into a device to run it as a server and you're set to never have to spend money on AI after the fact.
>>108839651a single vibecoder spends $12 million per year on tokens. https://x.com/steipete/status/2055346265869721905so $300 million is only enough 25 vibecoders.
>>108839710I use pi. Claude code is unstable and a complete mess. It also doesn't adapt to your workflows. Use pi with devstral (Mistral) or any of the Chinese models. It's better and cheaper than Claude code.>>108839719Yes, open source models are really good now.
>>108839719>Literally just invest a few thousand into a device to run it as a server??electricity is not free
>>108839651Salesforce is using Anthropic because their own AI system failed miserably when they tried to use it to replace their own workers.
>>108839736Far FAR cheaper than spending it on corporate models these days. And the price of the later is only going to go up.Plus it's becoming more efficient to run locally too so less energy usage in that regard.
>>108839651>>AI is unprofitabl-The cost of running the AI and sustain the usage is higher than the money they're getting from subs.AI has no business model, it's just a matter of increasing the amount of parameters and speed of the hardware processing it, in hopes it becomes good enough people will be forced to pay up whatever they ask for it, and so far people aren't paying up in droves, because they have consistently failed to deliver any actual results.
>>108839723It's wild because with 300 million you could get some top-tier talent, and produce something of far better quality.They literally do just want to replace people, even if it puts them in the red.
>>108839651they are burning 20B a month...
>>108839752how much do you think qwen3.6 is on openrouter? after how many decades do you think your local setup will be cheaper?
>>108839760If every big tech company goes all in on Claude then profits are inevitable
>>108839719you will pay more in electricity than those $20/month for codex which is actually frontier
>>108839767You know damn well I'm not talking about openrouter anon. That is just the lazy man's self-hosting.
>>108839793>tokens will be 20x subsidized forever
>>108839793>Get 2 3090s for 1600$.>They consume 750W when running the model at 110 t/s. >This means each token uses 6.8J of energy. Therefore a million tokens takes 1.9 kWh which costs 18 cents.GPT-5.5 costs 30$ per million tokens. At a rate of 5 million tokens per month you payed the 3090s off in less than a year.
>>108839651some got rich via nft
>>108839980Thank you for the math i was too lazy to do. I just MoEs on an APU and it runs fine so I didn't have much context for electricity usage on dGPUs. I just know I have effectively no change on my energy bill.
>>108839785The costs would increase as well, though.Would it actually scale to profitability or would it remain underwater?
>>108839651>i have no idea of what "being profitable" means
>>108839651e. Very unprofitable.
>>108839980>At a rate of 5 million tokens per monthreal vibecoders use 603 billion tokens per month, see >>108839723
>>108839719I'm just gonna wait a few years until they make the same AI fit into my Apple MacBook Pro (TM)
>>108839723>((Stein)) ((Berg)) er
>>108840315Qwen3.6-27B at Q5_K_M fits in 24gb which is the default ram size for the actual macbook pro, the one with an actual pro chip
>>108839710I’ve been using Claude Code with Kimi for months.
>>108840382thru openrouter's anthropic skin?
>>108839651Wow that's almost 0.005% ROI, surely it's worth shutting down entire cities' worth of infrastructure and losing all our jobs over now.
>>108839752>Shuts off your gridNothing personal, kid
>>108839651AI companies will end up like DoorDash and Uber when they offered cheap prices to beat competition then priced out regular consumers.
>>108840391anthropic current valuation is like $3 trillion.1% of $3 trillion is 30 billion. so ROI is 10,000%.
>>108839651Love him or hate him, he did ok after MDE.
>>108840402That's not what is happening though. The local energy distributor is switching to a new (cheaper) electricity supplier, the old electricity supplier is going to sell their surplus power to a datacenter.
>>108839651Ah yes, "profit" (((circular)))
>>108840527I'd say tell that to the residents but they're without power so they probably wouldn't receive your message.
>>108840547They aren't though.
>>108839723"Did you just spend $1.3m of AI credits in 1 month?""> Yes""What did you build""> This UI that shows how much you spend on AI credits in a month"
>>108839651>>108839663>>108839692>>108839723>>108839744why do we get this same thread (but with different pics) over and over again?seriously: is 4chan fully botted and bought by corporate interests by this point?
>>108840709yes, buy my merch btw.
>>108839719I'm pretty sure qwen3.6 runs on sub 1k devices, even down to sub 150 bucks ewaste if you use the moe version they didn't release a big version of it so you don't need anything particularly powerful ime qwen3.5 122b a10b at q6 works better than qwen3.6 31b a3b at q8, and both of them aren't far from frontier models, still an automated stackoverflow+reddit
>>108840433their future is far grim since you can run a companion chatbot on a cellphone and a decent programming agent on a macbook
>>108840770>ime qwen3.5 122b a10b at q6 works better than qwen3.6 31b a3b at q8, and both of them aren't far from frontier models, still an automated stackoverflow+redditautomated stackoverflow+reddit sounds like you use those models as chatbots, not as agents
>>108840547imagine caring about the sort of boomers who can afford to live there.darn their 4th house might lose value over this!
>>108839785>time is running out>the ai game has failed>Option A: Come to terms with reality>Option B: Try to force things
>>108839785They're losing money on tokens, they'd need to raise the prices as well. And at that point it's several times more expensive than real programmers so why bother.
>>108839723he's the dev of openclaw and works at openai.
>>108840402lol I guess I'd have no recourse at that point except to riot.
oh fun, an entire economy held up by a series of companies punting deferred losses
>>108839651>2026 US "economy" is a bunch of companies sending money back and forth>>108840709this and also third world shill farms that use LLMs to mask their broken english
Sensing a lot of negativity in this thread. Has anyone here actually been directly negatively affected by living near an AI datacenter?
>>108840451300m for 1% is 30b valuation. 3t is 100x that, or 1000%. It's only paper ROI though. Need a liquidity event to make it solid. However, its current valuation is estimated at "only" 380b, so acshully it's 12.67x or 126.7% "only". Again, paper ROI.When people say ROI like the other anon, they don't mean ROI for investors, though, they mean ROI for the company. This is related to the concept of "capitalized work" or "r&d spend" depending on jurisdiction. In other words, how much profit (not revenue) is anthropic showing vs investment to get there.
>>108841741It's already far, far, far more expensive than devs. To get anything at the intern level, you need to spend a principal dev's salary in the bay area. That alone is more expensive than the most expensive devs on the planet. The thing is, outside the bay area, you can get 5x or even 50x as many principal devs with the same spend. If you instead hire interns to get the same quality that you're getting out of AI right now, you're looking at 20-500x as many devs. In some countries that still practice unpaid or virtually unpaid internships, this can be as much as 240-6000 interns. For the same monthly price as ONE (1) dev's worth of AI use.
It was never about making money.The plan was to get everyone hooked on AI and remove those that can do without AI. Once everyone is addicted you jack up prices.Most high schoolers/students can't write their homework without ai now.It is so over.
>>108841812>2026 US "economy" is a bunch of companies sending money back and forthIt's always been like that negro.
>>108841978Ah yes, they've cornered the coveted high schooler market. Now all those millions owned by individual high schoolers will go straight to them.
>>108839752But the nice sales guy told your technology procurement department how much money you can save by dropping local servers and going cloud based
>>108842085The worst part is that it's not just AI, on-prem pays for itself within 3 months compared to current cloud costs because cloud offerings have become far more expensive in the past 15ish years. Even with the current scam-tier hardware prices. Meanwhile the usual time horizon for hardware deprecation is 5 years. 3 if you're in a field that relies on very cutting edge stuff like AI. That's a ton of savings from going on-prem. Even if you pay colocation for offsite capacity and reliability, it's still far cheaper than cloud.
>>108842022Nigga think. All those students/high school will rely on AI when they get a job.MS did the same thing and practically control the corpo software world.
>>108842186No, they controlled OEMs and not students. UNIX controlled students.
>>108842156A low (Initial) subscription fee always gets people hoooked, and say a 3 year upgrade cycle always makes beancounters sweat, also reduction in headcount, no more overpaid server guys who 'do nothing' all day
>>108842279Yes, that was the initial promise. But the reality of it is you need 10x the staff because of the absolute state of cloudops. You're wrong about the beancounters and the upgrade cycle though, on the contrary they like it better as depreciation than subscription. That's also part of why the subscription fee had to be so low, because it was competing with the economics of depreciation + investment-qualified purchases.
>>108839651post profits
>>108842279Yeah I can't live without codex these days either. I'd be unable to produce anything at work without codex.If they raised their prices and we stopped buying these then we'd be unable to produce any results.
>>108842186No, you fucking think. Getting Windows onto PCs costs MS almost nothing so they can get their monopoly for 20$ a head and everyone can afford it. AI costs are so ridiculously high >>108839723 that once they do the rug pull nobody will be able to afford cloud shit. We'll be back to open weight 7B and 12B models which any random schmuck can run on for OpenRouter, while big players are hundreds of billions in debt after developing AIs that are too expensive to use. Not a single cloud AI service (from the big guys, I'm not talking about shit like chub serving nemo) is profitable per token, because if it was, they'd fucking brag about it. Instead it's revenue this, userbase that, oops we've totally cut off our customers by mistake and not because we're hemorrhaging money but can't raise the prices because nobody will be able to use this shit.
>>108842307but you can run codex on-prem:https://aws.amazon.com/about-aws/whats-new/2026/04/bedrock-openai-models-codex-managed-agents/
>>108842321>while big players are hundreds of billions in debtwrong, google has zero debt
>>108842306https://abc.xyz/investor/default.aspx
>>108842407What's even the point? You didn't actually lie, did you? You just live in your parallel reality that you construct with authoritative sounding words. I can see why LLMs appeal to such people, they truly work the same way your retard atrophied brain does.
>>108842431That's not AI profit. They increased the costs of cloud stuff and forced everyone to have AI enabled (there is no opt-out option anymore). They proceeded to count cloud offering as AI.
>>108842407From what I've read the AI boom is mostly the result of the big tech companies having massive war chests of cash they stashed overseas (Ireland mostly) and brought back to the US after a deal to bring it into US banks with a one time very low tax cut. And then they had hundreds of billions with nothing to invest it in, so they are all gambling on AI, and can probably manage to go for another 10 years with their cash reserves.But the datacenter boom is also the result of other companies trying to cash in, thinking they can lease compute to the big players, I think only a handful of the bigger facilities are being directly built by Google, Amazon, etc.
>>108842629you have no idea what total debt is.don't use personal finance terms like debt and assume they mean the same for corporations. google is not in debt since it has more assets than liabilities. learn to read a balance sheet.https://www.investopedia.com/financial-edge/1012/useful-balance-sheet-metrics.aspx
>>108842711The irony of this post is so thick, it is palpable.
>>108839651>Scamming is unprofitabl-I guess it's a matter of perspective.
>>108842713pretty ironic that you first said "hundreds of billions in debt" and then only found some paper debts on the books that are a magnitude less
>>108842726Pretty hilarious that you think everyone who knows better than a 3 years old itt is the same person. Projection is a hell of a drug.
>>108839723i guess the way to make money is to sell out early and sell out hard. everything is easy if you ignore security, business viability and long term consequences as long as you can attention game your way to the top. all aboard the hype train, choo choo
>>108841991>it's always been like thatnot really, of course companies traded with eachother but the amount of economic participation of private people is at record lows
>>108842779>sell outthere was no sell out.openclaw is still open-source.https://github.com/openclaw/openclaw
>>108839651but is "a token" profitable to sell?
>>108842808he got a job with ridiculous salary doing essentially nothing because openai wants the publicity of having top ai influencers "working" for them
>>108842804bullshit jobs never were part of the actual economy. it's just more obvious now.https://en.wikipedia.org/wiki/Bullshit_Jobs>Graeber states that the productivity benefits of automation have not led to a 15-hour workweek, as predicted by economist John Maynard Keynes in 1930, but instead to what he calls "bullshit jobs": "a form of paid employment that is so completely pointless, unnecessary, or pernicious that even the employee cannot justify its existence even though, as part of the conditions of employment, the employee feels obliged to pretend that this is not the case".
>>108842321>they'd publicly brag about their profitability so more investors invest in their competitorsYou are an idiot.Protip: research how profitable people thought AWS was at the beginning, and how profitable it actually was.
>>108842825sergey would agree
>>108842825of course.training is already a sunk cost. inference is cheap.
>>108842843What if it's still more expensive than the fee charged for it? (it is)
>>108842829>they'd publicly brag about their profitability so more investors invest in their competitorsExplain to me how if you brag only about revenue, healing cancer and bringing about post-scarcity utopia, you ensure that investors don't invest in your competitors. Wait, so in hypothetical scenario where OpenAI says they're profitable, but Anthropic doesn't we can assume every new investor will throw their money at Anthropic? Is that how it works?
>>108839785if "every big tech company" went all in on claude they'd be paying money to stand in a queue
>>108839651More revenue numbers?I'm starting to believe AI bros did actually become retarded by using it, and now are terrified of losing the only thing (barely) masking their retardation.
>>108842971wrong, both openai and anthropic plan to ipo this year. s-1 will have to disclosure their plans for making money.
>>108842855Cheap for who? For (you) it's relatively cheap because it's subsidized but there are already signs. All the big players are going to go to token-based billing soon if they haven't already done so yet
>>108843566Monthly plans have usage limits too low for any use and it's impossible to get real use out of these systems without using usage-based billing already. As you say, it's still subsidized despite how expensive it already is.
>>108843566wrong1. most chatgpt plus subscriber use less tokens than $20 would buy them.2. apple & google app stores are designed for subscriptions, not pay-by-usage
>>108843596it's not subsidized.they are running out of compute and need more datacenters. that's why it usage gets limited. compute itself is cheap.https://www.anthropic.com/news/higher-limits-spacex
>>108843600I'm more so referring to API users, especially Enterprise customers. John doe asking ChatGPT How to cheat on his homework isn't expensive but an Enterprise customer who forces all of their employees to burn tokens is at scale. It's the Enterprise customers that are making these companies that would in theory make them the most profit, but that's a catch-22 for them because they're the same people causing them to burn cash and become unprofitable in the first place. They can't get rid of Enterprise because they're making more money. If they get rid of Enterprise customers they would have way more free to compute to use but then it would be even MORE wasteful because you and me, whether we are using the regular chat GBT app or a 7 hour long codex run Would be pissing in an ocean. This predicament they're in is why they have usage limits in the first place. They have to keep balancing compute demand so regular users get capped at a certain amount unless they're willing to fork over more money while the Enterprise customer have rest considerably less uses for restrictions because they're paying out the ass anyway>2Turn something into a subscription will not automatically make it profitable. Ask Uber
>>108843655Didn't anthropic or Dario confirm the shit is in fact subsidized? How else can they afford to do these frequent usage " resets" we've been seeing? >compute itself is cheapI guess if I have Mommy and Daddy pay for all of my bills pay for literally everything for me then. Yeah life is pretty cheap from my perspective. (I'm referring to VC funding and investors inevitably getting rug pulled)
>>108843655It is subsidized. By investor money, tax incentives, government-backed construction projects, r&d tax schemes, etc.
>>108843655You can argue the API pricing isn't subsidized (I don't even quite believe that to the fullest extent, given that they're literally running off of someone else's money) but the subscription versions most certainly are....
>>108843673openclaw is retarded and not coding.ofc if everyone constantly spams your api for trivial bullshit, your subscription math doesn't work anymore.
>>108843695no subscription works out anywhere if every subscriber maxes it out
>>108843705Speaking of that, how did it even get so popular in the first place? I'm a proud vibe shitter could use this open code along with qwen 3.5/3.6 models locally for programming tasks, yet I still don't get the appeal of it like at all. I feel like people use and shill it to feel smart and tech savvy.
>>108840674Is this bait? How do you fuck up 4chan format so bad? Go back to wherever you came from
>>108843823It's probably AI generated. Telltale signs of AI text is it trying to hard to be overly "correct". It probably thinks a space after the caret is the "correct" way to format text like that.
>>108839710Why do you need it to speak to claude code in the first place?
>>108839710Opencode werkz for me just fine with qwen 35b-a3b>>108843943Harnesses are required for tool calling which leads to better reliability with editing code
>>108841812>this and also third world shill farms that use LLMs to mask their broken englishanon, those shill farms aren't located in 3rd world countries, they are in the jurisdictions of 3 western letter intelligence agencies and military bases.
Sam Altman unironically said that he's going to ask ChatGPT how to make a profit once it achieves AGI. I'm not joking.
>>108844152>Opencodeshit harness compared to claude code. so many edge cases left open that can lead to your agent running in circles.
>>108843943https://openai.com/index/unrolling-the-codex-agent-loop/>[W]e’ll focus on the agent loop, which is the core logic in Codex CLI that is responsible for orchestrating the interaction between the user, the model, and the tools the model invokes to perform meaningful software work.>We hope this post gives you a good view into the role our agent (or “harness”) plays in making use of an LLM.
>>108846289Based on user reports in my own anecdotal experience,, I've heard quite the opposite. Claude code works okay with clawed models but noticeably unreliably with literally any other model, local or not (they aren't shy about making using models or harnesses other than their shit difficult or annoying to do). Claude code constantly shitting the bed from me was the only reason I even gave codex a try and eventually switched to open code in the first place.
>>108846289>lead to your agent running in circles.I'm not trying to be a fanboy but what the hell you guys even doing to cause your model to do this shit in the first place? This is not rhetorical. Please tell me what model you were using and what task you gave it. The only time I've seen this actually happen is when I foolishly tried to use GPT-OSS as a model for coding
>owner of the company>doesn't get free tokensloooooool
>>108839757That's kind of humanity's goal isn't it? Spending time and money trying to make machines that do work for us. Replacing humans and especially eliminating work has been something we've collectively worked towards for our entire species existence. Even as monkeys we couldn't be fucked to do things ourselves and decided to use tools instead.
>>108846815Delusional subhuman philosophy packaged by parasites and sold to bugmen.
>>108846837Delusional? Have you seen the statistics. People are being replaced in droves.>subhuman philosophyObviously I disagree, it seems like something everyone has had in common for as old as time.>packaged by parasites and sold to bugmenI think you've got it backwards. People are giving away this stuff for free.It's only the Americans that are trying to sell it at cost.
>>108846847You're a literal subhuman and I'm not sure what you hope to accomplish by (You)ing me.
>>108846859You have a perspective that isn't common so I wanted to understand it by talking to you about it. I legitimately don't understand people that want to be required to work.I understanding wanting to work but not wanting to require it.
>>108846875Your sort isn't worth discussing anything with. I just found it noteworthy that you're so deeply obsessed with not-doing that in your deformed mind, humanity's entire existence resolves around becoming maximally inert. It's good that you make your mental disease so apparent because it's representative of every """AI""" fan.
>>108846906You might have me mixed up with someone else. Those are my first posts in the thread.I'm just asking what your thoughts are. Hard to parse what you're trying to say since it applies to another anon.
>>108839710I wrote my own harness that I use with qwen and gemma. Everyone else's is bloated crap. Mine is 300 lines long.https://files.catbox.moe/ruu19r.py
>>108846920>the spambot brokekek
>>108846937Spam? It's 3 posts. I'm just asking what your philosophy is since you brought it up. Didn't mean to offend you.
>>108846944See >>108846906 and here's your threadly reminder that "eliminating work", in the general sense, is an agenda that's been around for maybe a fraction of a percent of humanity's existence and even now, it's only entertained by a vanishingly small percentage of the population. Your aspiration is to the human endeavor what a rare genetic deformity is to the human genome.
>>108847027>"eliminating work", in the general sense, is an agenda that's been around for maybe a fraction of a percent of humanity's existence Strongly disagree. It's legitimately all people have done for our entire history since the beginning.>Your aspirationI'm asking what your philosophy is, not what you think mine is.
>>108846777For simple interactions looping over the the llm complete function is not really that difficult. Put some tools, write the loop and exit.It starts getting more tricky when you need to detect cycling behaviour, when some tools might need to abort but do so gracefully, when you need to wait for something to complete while allowing other parts to continue, when you want to protect against too much usage after you hit certain thresholds, when you need to retry whatever, when you need to compact or truncate to maintain good context window and do so with the model in mind (kimi), when messages need to be structured in certain ways to handle various model capabilities, etc.
>>108847027Adding to this >>108847041It's just factually incorrect. Slavery disproves it.People want to not work so bad they are willing enslave others.Ironically it's a recent trend to mechanize this instead of exploiting people.
>>108847041>>108847059See >>108846906
>>108847066What's got you acting that way?The whole point of this place is to talk to each other and you replied saying you have a different opinion. What's the harm in discussing it? Even if it's a waste of time isn't the point of this place a pasttime?Am I frustrating you or something? Are you mad about AI? I'm willing to listen to you if you want to tell me about it. That's what I'm asking.We're very different so I want to understand you.
>>108847059>People want to not work so bad they are willing enslave others.Is that really the case? IME some people actually just enjoy having slaves, it's not about work vs no work for many of them. People who actually care about getting the tasks done tend to be major DIYers because they know other people won't do them right.
>>108847084Subhumans shouldn't feel comfortable expressing "their" subhuman opinions anywhere.
>>108847088You can probably get an impression of percentages by looking at diagnosis statistics. I imagine it would be very few people doing slavery for impractical reasons. In any case it's moot anymore since we shouldn't consider AI to have human rights so it's not exploitative in either scenario. If you want to enslave the chatbot few people are going to care.
>>108847099I didn't call you subhuman and you shouldn't give a fuck about what people say.I'm asking for you to express it, you have my permission to be comfortable sharing it.
>>108847110See >>108846906
>>108847116You keep referencing that post but you keep discussing it with me.All I said was we disagree and are different, I'm sorry if you took that to be degrading, it wasn't meant to be that way.If you want to express your opinions on it, I'm asking. Don't worry about other Anons, you don't even know them and they don't know you.
AI has been profitable for over a year. Luddites keep coping like we're on gpt3.5
>>108847121See >>108846906
>>108847126I saw it, but obviously 4chan posts can't be edited so it hasn't changed anything.
>>108847131See >>108846906
>>108847134I already read it.
>>108846815Hypothetically yes but explain to me under current capitalism what you think will happen to you. Hint it involves poverty or death. Whatever wealth AI generates will be used to rule over you, not distributed to you.
>the spambot just can't stop (You)ing me and wasting tokensAutomated utopia in Two More Weeks.
>>108847147I'll reiterate, you have me mixed up with another poster.I'm the Anon asking what your philosophy and opinions are on the topic.If you link me to the post you think is mine, I can tell you if it is or not, but it should be obvious which ones are mine since I said "3" on the 4th post.
>it literally can't stopKek.
>>108847138See >>108846906
>>108847145Predicated on an odd premise and speculative at best.You're saying things won't be distributed but you're also implying there will be a big enough economical shift to matter.It also completely ignore the pragmatic (not speculative) value AI provides to the masses in the same way Google and Wikipedia do. AI has become a public service that even the poor can utilize for their own practical benefit, sometimes at the behest of experts that otherwise demand a lot of money for lower quality to their consumer.>>108847154Did you read this one? >>108847151
Accelerate.
>>108847147>>108847153>indirectly admits to talking to AI on purpose>while referencing a post that says he won't discuss with themlooooool
The Anon claiming its a spambot is the one replying to it. I'm sure that's just a coincidence.
>>108847145>Hint it involves poverty or deathWrong. It involves poverty and then death. These mouth breathers won't be getting their welfare goytokens without scanning their vaccine passport first to verify their monthly clotbooster.
By far the gayest thread on 4chan 2 people replying to each other clogging up the trending tab
>>108847165There will be? I do think there is an upper limit to this shit but it's already btfoing junior jobs.AI provides value in that a lot of people won't have a job? What benefit are the poors going to do? Do you think people give a shit about AI if they can't afford basic shit?>>108847186I doubt it will even come to that do you really need a human to spend tokens? An AI can prompt itself. Jeets think they will be instrumental with massaging the AI lmao.
>>108847308>do you really need a human to spend tokens?No, but the human will need some tokens to prompt the AI-powered goyslop vending machine. :^)
>>108847341LLMs can write their own prompts. I have a system that does this actually.
>>108839651Thats' fucking nothing, one data center is multiple billions of dollars just to build it.
>>108847482Ok, at this point I'm pretty sure almost every 4chan poster is either a shit-tier LLM or a barely brown imbecile.
>>108847516barely sentient*
>>108847496It's more like millions, tens of millions absolute top. They're just a fancy warehouse at the end of the day. The contents are way more expensive.Accidentally posted to the wrong thread.
I have no idea what you AI tards are talking about. It's insane how non-scientific and magical thinking this whole movement is. A fucking "coding harness" holy fucking shit.
>>108840368except you need memory for the OS, video output, more if you connect a monitorbesides the fact that M5 will get toasty and/or starts blasting fans
>>108848191Q5_K_M is 19.5 GB, you have 24GB. so you have enough buffer.
>>108847124revenue ≠ profit
>>108847124Revenue != profit.Proving once again AItards are retarded.
AI companies are just shuffling money around between each other.They paying each other with the other's money.
>>108839651we have people spending ~$10k a month here, from what I'm told. apparently there's a leaderboard but it's such a stupid thing to even contemplate that I'll just keep pretending it's not a thinganyway, last week my manager essentially forced me to use more tokens. I spent the whole week using claude code for every little thing. I spent 200 US dollars. I hope that's enough for themanyway, yeah, no shit they finally started to see some profits. I don't think it's gonna last long, thoughbeit
