AI coding is genuinely impossible in 2026. You just get hit with rate limit after half an hour, even when you're paying for maximum plan. It's over
You people still don’t understand that AI isn’t for you, it’s for the working middle managers with corporate accounts. You are irrelevant and you were never going to produce anything anyway.
>>108843716>middle managers>ProducingChoose only one (1)
>>108843669Then buy 2 maximum plans.. what are you? Poor?
>>108843716>working middle managersHere we have both the moron, and the oxymoron of the day, folks.
>>108843716>managers will generate code brohahahaha, wait you're serious?HAHAHAHAHAHAHAHAHA
>>108843669If you're talking about copilot or codex, then you list the requirements and ask chatgpt to make the prompt you need to pass so it won't use up the tokens to generate readmes and keep re writing the same file again and again because run test failed.
>>108843669surely the $20k a month you're earning by vibeslopping will easily cover paying for multiple plans
>>108843896retard alert.one manager funds one AI engineer who's doing the job of 10 coders.all clear now dopey?
>>108843896it will go full circle. managers will take over coding, everything will break down, managers will ask AI what to do, all the unemployed programmers will be shitposting and complaining online about how the retarded managers have no idea what they're doing, so AI fed on this will say to hire programmers, and the programmers will be rehired, the managers will get a hefty raise for finding the solution to all their problems, AI stocks will skyrocket because of AI and AI managers fixing the great code decay bug of the 2020s and everyone's happy
>>108843982who's this 10x coder, is he in a room with us right now?
>>108844094It's definitely at least 2x for webslop though, which is the vast majority of coders. I think we're in for more layoffs soon.
>>108843669local models are decent enough
>>108843669Sounds like a you problem
>>108843716They made it free and released open source models so that everyone on earth can be unpaid beta testers. Once it's been perfected, no more subsidized access for the plebs.
I never hit a rate limit.Maybe try to not rely that much on them?
>>108843669Wtf are you doing with it? I'm on a Claude pro subscription and I worked for 8 days straight days on a project, finished it and I haven't exhausted my tokens
>need test case time series data across 2 hours with a row every minute>like 8 columns of repeated junk with only the timestmap changing>ask llm to repeat my example line 120 times with only the timestamp changed>outputs 3 lines of reusable code to repeat my example line>"no do it in token space">it obligesgotta do my part to use the tokens we are paying for
>>108844094>who's this 10x coderDeepseek
>>108844110depends on the webslop. Regular crud shit already had 1 click templates long before ai. And anything more complex ends up much slower. if you achieve 2x speed you're making 4x tech debt.
>>108844002Close, but programmers wont be rehired directly by former employers. The actual plan is for AI companies to get a monopoly on developer talent and loan people out on contracts to companies after things start falling apart. Since the companies paid all their money to the AI companies for tokens, they wont be able to outcompete what those companies offer their workers in terms of pay because those companies will triple dip (dishing out salaries to their fixers from investments, token revenue, and now white glove service contracts to fix perpetually broken codebases). After enough things are enshittified, the final steps will be to:1- assimilate all the devs the companies let go of2- lobby congress to make engineer licensing mandatory for all dev work. With AI usage statistics being collected to the mothership, alongside widespread system failures, and the youngest generation actively showcasing that they cannot actually program or even think without these tools, it'll be a slam dunk to kill all talent pipelines in one fell swoop.https://job-boards.greenhouse.io/anthropic/jobs/4985877008The funniest part is that F500 companies that get bent over the barrel will have basically asked for it.
>>108843759>>108843881>>108843896>malding over their lack of employmentthey haven't replaced burger flippers yet, get to it
>>108843716>it’s for the working middle managers with corporate accountsits true anything i write i know bossberg can do himself n 5 minutes with claudeas soon as he gets that small million loan for openai tokens and figures out openclaw agents it's goodbye to every single dev workingwe got too cocky too much wfh
>>108844865>and figures outholy shit get out of here with your science fiction star trek shit. At least keep things within realms of believability
>>108844242This, OP is genuinely retarded and does not know how to manage his models/give proper context.
>>108843669>planngmi. buy more tokens.
>>108843669That's weird bruh. My local Qwens have never acted like this.
>>108843669Local LLMs can still solve a good chunk of problems in programming
use local qwens, understand that they won't "one shot" everything for you and you still have to use your brain but they will help with debugging / implementing annoying shit
>>108843669You know, this problem literally never happens to people who code by hand. Ever considered that?
>>108843669>AI codingwe are spewing heat into the sky because this profession doesn't want to learn their trade properly. what a disgrace
>>108843669>1000%no one is building all those data centers to sell you a $20/mo subscription. they will only slightly undercut the cost of replaced meat bag devs. 10000% will still be considered a deal.
>>108844188what do you need to have to run them?
>>108846274m5 or strix halo or gpu with enough ram (like 24-32GB), you can run them on the cpu if you're desperate but it's gonna be like 5tokens per second max
>>108843716If AI will be good enough for middle managers to replace devs, than it will be good enough to not require middle managers as well.And why pay a retard who was absolutely no fucking clue what's going on, when you can pay a dev and be much more efficient?AI is a force multiplier.
>>108843669like you ever coded tensorflow in C or ran local LLMs or shit you created yourselfAll cloud crap whether SAAS, storage, AI, streaming or gaming is raw sewage subscibe crap for retards who can't
>>108846324interesting to see where openvino and the ultra core 2 7 goes on next gen in partnership with nvidia new wintel of AI right there
>>108843669Coding is becoming a high IQ, high capital activity and all the middle class programmers will be left out.
>>108843669use the API. if you're fully reliant llms then surely you're making big money by now.
>You can ask an AI what it costs to produce an Opus token. The answer I got was $0.216/M Token. That takes into account hardware cost, power, cooling and hosting.Others have posted links here saying inference is about 30% of Anthropic's spend, the other 70% is R&D - things like developing the next model. If you take that into account, add a profit margin, and round to make the figures easy you end up with them selling tokens at $1 / M Token to make money.Their API cost is currently $25 / M Token. There is no question that's profitable. Someone really pushing their $200/mo Max plan can use 10B Tokens per month, which works out at $0.02 / M Token, so they are eating a huge loss there. That is clearly going to go away at some stage. For the rest of us: look at your average monthly token usage. If you are using 100M Tokens a month with the Anthropic $100/mo plan, they are making money out of you.You have to be working very long hours, and be really, really proficient at using AI to achieve 10B Tokens a month. The only way to be that proficient is to have been using it for a long time, years in fact, so it was useful to you long before Opus came along. You would be very disappointed to lose Opus, of course - but you are just the sort of person who can make a less capable model sing. It's not so difficult to see those users moving to an in-house-hosted, open-source model in a few years, and it will cost them what they are paying now - $0.02 / M Token.All that means I'm not convinced by the gloom and doom vibe of the article. Things will change, but it won't mean the end of AI usage.
>>108845355>tokensmeds
>>108846366Because deals between the big boys brokered over dinners need to happen. I agree middle managers wont be needed in a world where AI works, but devs wont be needed either. In that world, things would probably boil down to whales investors hiring teams of actual, unironic prostitutes who happen to have MBAs to operate the day to days of shell companies that interact with other shell companies to funnel and extort money between whales... and maybe collect subscription dues from the poors on their slop products if theres any trickle down actually happening.
>>108847001Wouldn't ~0.21 cents be on the low end for pure text inference for smaller frontier assuming indefinitely stable GPU offload? I'd imagine costs could vary wildly based on IRL hardware utilization and maintenance, reasoning cycles, multimodal processing, context cache maintenance, security operations, variable energy costs, rent or real estate costs, and a handful of other things. Like to the point where even $25/M tokens is not guaranteed to provide a comfortable profit margin across all possible users.
>>108846780>Implying they wouldn't price hike the API too Why do people just ignore the fact that that shit is subsidized?
How much does it cost to use AI as a single user? I was thinking of switching from local to api. Any suggestions? Looking at around 4m tokens per day. Model must be more capable than glm 5.1 q4.
>>108843669I've never hit the limit, but I work at MSFT
After 30 minutes of generating code, spend the rest of your work day reviewing it to make sure it actually fucking works.
>>108843896this is what the coinbase ceo genuinely believes LMAO
Don't be silly. I have a Claude acc and a Codex acc and I never run out.
>>108843669learn2code faggot
>>108846324>gpu with enough ram (like 24-32GB)you dont need that much you can run gemma 4 26b moe or qwen of similar size in less vram they dont take as much of a performance loss for cpu offloading
>>108843716wut, im making porn, im making music, im making sfx, im letting my agent write my code to fix my gamewhat the fuck are you talking about lmao. are you a god damn absolute loser that dont know how to use a computer?
>>108849393But you don't have the hard skills to do those things themselves, which are more versatile and transferable than just asking a computer. You can actually do two of those literally anywhere (music, coding) if you have a pen/pencil and paper.
>>108843716HAHAHAHAHAHAHAHAHAHAHAHAHAHAHA!I can picture some retarded 80 IQ diversity manager BEGGING Claude to add a button after showing it to those oppressive nerds by firing them and it just refusing like a total fucking Chad.Someone please Grok this. My poor limits are too low...
>>108843982ffw 2 years, you will be hiring developers at double the standard rate to emergency fix your festering codebase after your entire infra goes down from cyber attacksmeanwhile 5 companies will pop up competing directly with your product, since they use the same tools and processes, you holding nothing over themseriously, why do you retards think the most important and most successful tech firms are talent hogging the way they are?
>>108849408i'm using a fucking keyboard instead of pen and pencil to send this shit. I dont need to walk manually to your home to post this shit for you to read it
>>108845160I want to get into it but I have no idea what to use. Any free ideas?
>>108849530There's this thing I keep saying but no one seems to understand. AI bros, anyways.Nuance exists, you can dislike one technology and like another.
Your employer is supposed to be paying for it retard
>>108843669>AI coding is genuinely impossible in 2026.Deepseek V4 Pro, Max thinking. No rate limits. You can spawn 10 sub-agents running all day and barely spend $1 in API credits.
>>108850628makes no sense.subagents implies a harness. you mentioned none.
>>108843669Pay up, snailcat. AI is the future. If you aren't on a per token plan you're already behind. Face it—AI won.
>>108849460They don't even have to use the same process. They can just steal the code and there's shit all you can do about it. Copyright is strictly "human authored." AI use means no copyright.
>>108843669breh visit /lmg/ before opening your retarded mouth
>>108843669"now increase the price by 1000%">everyone clicks on "model" and selects DeepseekBravo, American capitalists.
>mfw AI tokens are literally free for all big frontier models at my fagman company but I only use them a bit when I'm too lazy to search for method names or make unit tests
>>108851564you will never understanding programming
>>108843669I find the $20 codex plan sufficient for my hobby project usage. At work we have a corpo claude license and I never even come close to the limits despite using opus all day every day.
Vibe coders have no idea how worthless they're going to be on a job market.
>>108850965the AI bubble will pop anytime, and soon the tariffs will be almost impossible, accept it or die.
