I'm tired of the>HAHA! Learn a real job nerd!>NOO our craft will never be automated, AI is so dumbidiot to idiot conversationsCan we have a high IQ discussion about this?My take: In the long term, when entire apps can be made with AI, it is possible that it will still create problems and they will still hire devs. There's a precedent for that: Indians are so much cheaper and they are still paying 100k for a new grad just to avoid the long term problems from bad code.Another theory is that software dev will no longer be a job that people get hired to do. I hate this, but admittedly, it would be better for most people, as they like software and use apps a lot (which is why we make money), and they would have more apps with more features if this happens. I think it will create other tech jobs due to the new possibilities, but it's still a headache and bad news for us if this happens. I find it sad.
Buy an ad.
https://www.hollandtech.net/claude-is-not-your-architect/
https://lucumr.pocoo.org/2026/5/24/pi-oss/
>>108910499They really managed to lure in so many gullible nudevs with that ncurses UI.
>>108910499AI will be used for most applications in the future.But for things like systems related stuff, you will be safe.
>>108910539>nudevsThats a very polite way of saying low IQ promptjeets
>>108910539that's not ncurses, nulldev
>>108910546yeah and what do you think happens to people doing "systems related stuff" when all the other devs are trying to get a job? kek its supply and demand
>>108910499> I think it will create other tech jobs due to the new possibilities, but it's still a headache and bad news for us if this happens. What is a tech job? Paid for getting to sit in front of a computer? Is onlyfans a tech job? Being employed using technology, managing it, creating it? Basically everyone is a tech worker now.Anyways try reading Norbert Wiener's God & Golem, Inc.: A Comment on Certain Points Where Cybernetics Impinges on Religion from 1966http://luisguillermo.com/diosygolem/god_and_golem_inc.pdf
>>108910663some people arent mindless cattle worried about nonsensical bullshit like muh jobz. we're aiming for a post scarcity society.
>>108910499your take is the low iq take.we will either have AGI very soon, or llms are fucking useless.it's not gonna be the same fucking thing with a bit extra. you either can manufacture minds, or you can't. if you can, human labour and humans become worthless. if you can't, ai can't return anything on investment.
>>108910499Claude is literally a fucking sheep that writes gay code. Unless you prompt it with ceremony i.e. specs, real design document, and shit. You have to explain your vision to it properly and only it can properly code else it shits the bed.
AI shit only works if new millenials are born. Zoomers are shunning tech and social media to live in the real world. Flip phones are coming back in style. Is there single Sailfish AI app? Do AI companies expect that they will suddenly create a product with demand? Or is it sputtering out as it reaches the last *jesh in Calcutta's yellow pages? The only people who will ever use this tech have already tried it out. AI corporate dogs have already told them not to pursue careers. So buckle in millennials, AI ends with you still having to fix bad code by Indians. What you were already doing before.
>>108910499This is better than Indians so there’s no reason to ever hire Indians or outsource to India ever again. But you still need to handhold it and review what it outputs.
>>108911230you are the same organism as the jeets, though. if ai can replace jeets, it can also replace you.
>>108911230>there’s no reason to ever hire Indians or outsource to India ever againSomeone still has to do the prompting while being paid pennies per hour for it.
If you want to succeed you don't claude code because it is not a good product, businesses should be using the open source models directly from deepseek/qwen or hosting them themselves, it would be a more competant software and no indians required.
>>108910499>Can we have a high IQ discussion about this?I think that's impossible to do on the internet, unless you find a very special community. /g/ is not that. I'll just say the same thing that is likely to happen but rarely if ever gets acknowledged in discussions like these: In this brave new future you are describing, new software that doesn't do anything different from its counterpart that was made a decade ago will run slower and require a minimum of 32GB RAM. This trend was already occurring before the AI hype, but AI is going to accelerate it.
>>108910499> Can we have a high IQ discussion about this?Let's see.> In the long term, when entire apps can be made with AI...Anon, I'm sorry to break it to you, but we cannot have a high IQ discussion here. You are retarded. I'm really sorry.>>108911183> millenials> Zoomers Another one...>>108911862Claude is good if you are a solo dev. You can actually get things done instead of maintaining your broken shit all the time. Software breaks, software requires maintenance. Just like moving parts in mechanics, there's no way around it. So if you are a solo dev, you delegate that to corpos like Anthropic, who are arguably the best at it, and that's how you get the most of current tech.
>>108910499>My take: In the long term, when entire apps can be made with AI, it is possible that it will still create problems and they will still hire devs. There's a precedent for that: Indians are so much cheaper and they are still paying 100k for a new grad just to avoid the long term problems from bad code.Companies have been hiring less westerners by hiring more Indians to do the job for less. Once AI develops enough they will hire less people altogether.
>>108910499>Another theory is that software dev will no longer be a job that people get hired to do. I hate this, but admittedly, it would be better for most people, as they like software and use apps a lot (which is why we make money), and they would have more apps with more features if this happens. I think it will create other tech jobs due to the new possibilities, but it's still a headache and bad news for us if this happens. I find it sad.This is more likely and the transition is gonna be pretty rough imo
>>108910904
>>108912007>AI generated forced memeFuck off you are part of the problem
>>108912007I fucking hate demoralizers.>imagine living in a society where post-scarcity utopia is technologically possible and not devoting your life to organizing a revolution that will make it a realityDo you people literally have zero agency and self-respect?>b-but it will be hardSo what? Still better than the status quo.>inb4 you can do it nowNope. First, human-level AI must become a reality. Communism does not work with humans, but it can work with robots.
>>108912048>Reddit spacing
>>108910521> AI agents are pathologically agreeable. Ask Claude if your idea is good and it’ll tell you it’s good. Ask it if a microservices architecture makes sense for your three-person team and it’ll explain why microservices are an excellent choice. Ask it if you should build a custom ML pipeline instead of using a managed service and it’ll enthusiastically lay out the design.True. A number of my coworkers have been forced by management into developing all their work through AI, because management wants a baseline metric on how it could work for the organization (hey- it's not a bad approach as such; at least they've not committed to just drinking the Koolaid straight-up, right?) One of the first things they learned:Never ask if X makes sense. Always ask why X might NOT make sense.Querying in the negative forces the model to be pessimistic.
>>108912064NTA but lot of people are figuring out limitations and trying to adapt it. But I see upcoming models have fewer limitations as each new version gets released
>>108912048>Communism does not work with humans, but it can work with robotsAnother confirmation high IQ anything is impossible here. These people are too stupid to understand what was the root problem of communism theory and why it never came to be. Although it is obvious: resources are finite. That's one problem with it. The second one is that communism is a totalitarian state by design. Communism theorists do not know themselve how is absolute totalitarian state supposed to transform into a different formation that somehow cancels the opression by state. Or what that formation would be in general.They have no idea. Communist ideas are built on mathematical approximation and extrapolation. The worst part? They fabricated history to fit their fake theory. They fabricated the fucking dots on the graph so that they would align with their made up shit.Oh fuck, this is clearly going nowhere. Year is 2026 I am explaining communism to degenerates on a Mongolian basketweaving forum. This is getting out of hand.
>>108912097>Although it is obvious: resources are finite. That's one problem with it. The second one is that communism is a totalitarian state by design.Both of these problems can be solved by ubiquitous human-level AI and robots.
>>108912113What do you mean? I just told you it is a totalitarian state by design, intentionally build like that. For purposes that do not make sense, they sound completely made up. How can it be fixed otherwise than just not being implemented in the first place? Or how do you make resources infinite? There is no singularities, there is no functioning nuclear fusion. That's sci fi. Just like your human-level AI.Current AI, specifically the LLM tech based on transformers, is barely any different from markov chains. It is a deterministic algorithm that predicts the next token based on your prompt and the model. Model being a compacted version of the Internet. All the content that was scraped and stored in archives for decades was compressed during the AI training. That's what training is. Lossy compression of data. Data is broken down into tokens, those are encoded as vectors and matrixes to determine their relationship with each other, then this information is stored as weights and biases. You can store all kinda of information as weights and biases. It is a very efficient lossy compression technique. But it cannot really store human level intelligence. You'd have to be as stupid as OP for that to happen.
>>108912167>Current AI, specifically the LLM tech based on transformers, is barely any different from markov chains.False. Artificial neural networks are Turing-complete universal function approximators and theoretically can approximate the function of human reasoning with arbitrary accuracy. Whether they can get to human-level intelligence with our training methods while respecting practical limits is an open question. We are in uncharted waters. Anyone telling you that Transformers can never get to human-level intelligence is making up bullshit, as is anyone telling you they surely can get to human level. We simply don't know. The only way to find out is to try.
>>108912097>resources are finiteso there exists issues over rationing finite resources [and pseudo-scarcity in forms of intellectual property, etc]there's debates over what's the best manner to do that, >totalitarian stateso is your existing libtard goy slave environment, ignore morality and just think about efficiency... there's better forms of control than actually existing ones>>108912167social problems are about information and controlwhat's the role of human intelligence in that process? society is pretty stagnant or messed up... it's hard to make the case for humans there
>I can make a thing by asking a thing to make the thing and praying>I can't maintain or support the thingAmazing
>>108912226>FalseBut it's not exactly false. Do not mistake theory for what is happenning on practice. It is only different from markov chains because of scale. But resources are limited, you cannot have an infinite context window, you cannot have an infinite prompt.Also if you see infinities in your calculations, that usually means you math model is shit and cannot describe the specific case. So what on paper is a singularity would be a failure in reality.>>108912269> there exists issues over rationing finite resources They don't really exists. They belong to teh strongest. Anyone telling you otherwise is a liar.> there's better forms of control Even if there are, I don't want those. I want to be free, not be a part of some nonsense. It's easy to see how doomed humanity would be if it allows a bunch of idiots that believe in singularities to exercise control over every aspect of a human life. It is basically a mass suicide. You want that, you're on your own, I'm not participating.> social problems are about information and controlThey are about attempts to extract all the juices left in society. Enhancing tools of control would only harm humanity even more. What a respectable gentleman must do nowadays is to seek ways to create untraceable weapons made of unregulated materials and tools and spread such knowledge so that more people would be able to defend themselves from the oprression of the state. It is the opposite direction of enhancing those tools of mass control. Tech people how to use self hosted services and DIY guns, instead of figuring out ways to enslave them and hand the cotrol over to some random globohomo midwits.
>>108912313>not using a tool by learning its limitations
>>108912048>Communism does not work with humans, but it can work with robots.Communism works if people with access to more resources are willing to share them equally to society. Also there needs to be an honest system to distribute those resources fairly which is near impossible to implement
>>108912688you are idiots.the system is alive. it doesn't matter what you call it or how it's supposed to work. as soon as you are disempowered in favour of the system you are walking dead.
>>108910499As an operations guy, the problem is largely perspective. There has long been too much fixation on "programming" specifically. It's trivially seen in the rhetoric around BS like "which language should I learn to get a job" as if the truly relevant ability is ever "knowing how to write valid java" or even knowing how to implement an observer pattern or solve a leet code exercise.The fundamental value of any programmer is and has always been the general ability to solve problems using a computer. Programming is just one facet of the process. Good developers always understand the layers above and below them as well as the problem domain. They understand the details of the problem and how those details map to the relevant platforms and engines. Code is, and always has been, merely the communication mechanism between the programmer and the machines. LLMs are the most dramatic change the nature of that interaction since the introduction of high-level programming languages, but it's only a shift in emphasis (at best). It doesn't fundamentally alter the requirement that you understand the problem space and what a solution should look like. And you need to understand it to a degree sufficient to prompt an LLM how to generate functional code.These are all skills that good developers already had, that bad devs and nodevs did not. People just took it for granted because everyone was always too fixated on the practice of coding itself.
>>108912996Your post is nearly perfect. One could frame it and put it on the wall. It's a shame OP is a dumb faggot and will not understand anything.
>>108910539Is this just me, or they stole Mistral style with that one?
>>108913536cc doesn't look like this anymore. not that mistral owns retro-styling.
>>108912996>These are all skills that good developers already had, that bad devs and nodevs did not.The bucket that I found myself in is that I never developed, never learned to code, but i am now neck deep into 2 very mature projects, one of which is custom made machine learning, and any problem of any depth that you can imagine - I already encountered and solved it. Yes, this requires multi-layer understanding of the system. I would be surprised if a 4 star general at leet programming would be able to solve some of the problems that I have solved with zero experience, UNLESS they delved into it as deep as I already have. An old-money sissy who opened a business and wants to create an app for his shop, is not going to have the mental bandwidth, the IQ level, to build anything meaningful, to even create a spec document for claude, so it will be AI slop.What I do see happen when we are closer to AGI, is that AI reviews tickets and feedback from users, and patches the application with the owners approval, so over a year the app becomes better and better, but someone still has to do the bulk of the development, be it classic programming, or AI-assisted programming, both require very high IQ.If there is a future in which AI takes most of these processes entirely, programming will become very routine, like using Excel, in which case classic development won't pay as much as it has so far, more people will have to go to industries which require hands on, like department of sanitation.
>>108913588It's not any kind of retro styling. It was a very specific style, but I'm not good with this, I don't remember what it's called.And it's kinda hard to put into words, but it was a genius design. It felt very fresh, very not like typical globohomo disgusting crap. Since it is a text generator, that retro bit somehow went so well with it. Simplicity of the text paired with that warm optimistic futurism theme from the days when people believed future tech would be even better, would always improve.Anthropic default style is so disgusting, so sterile, so globohomo, so vomit inducing. It's utter shit. No wonder they stole Mistral style.
>>108913643>programming will become very routine, like using ExcelIt was like that pre 2022, when it comes to CRUD shit.
>>108913643Yes, LLMs are enabling some modest number of people to create software that they wouldn't have before. But it's not as simple as "stealing jobs." It's often work that simply didn't exist before. And in many cases done by people who know tech and programming. Saw a story recently about a guy who ran a warehouse, he used AI to write an inventory tracking system that the story claimed would have taken a team of programmers. Probably an exaggeration, but either way it was not a job that ever would have been on the market anyway. But the real punch line was the offhand comment that the guy running the warehouse was a former programmer.The other issue, of course, is the massive volume of slop generated by low-iq nodevs. Multiple projects have moved to ban AI submissions due to this problem.
>>108910499I'm worried that web/fullstack jobs are in grave danger. I don't think web developers will be gone, but I think the need for them will lower. It's unclear just how much it will lower, but with AI you will (and already do) need less people to get to the same point as you would have gotten to without AI.I'm wondering if embedded is more resilient in this regard, as it has a physical component and is generally more complex than web development. I'm not pro or anti AI. I've used it for web dev and it is scarily good at a lot of it and it concerns me.
>>108914238I think it's crazy that they hired so many people to program a website in the first place.
>>108911173>This AI agent does really good work if the human tells it what to do properly Ok so don't be a moron
>>108914264FAGMAN used to hire just to dry out the talent pool. They believed it was way more important to deprive competition of highly talented people than to just hire enough and optimize personnel.Today it's called 10x dev or something. Back then it was called "rock star dev" or so. Also it was real, unlike what it is now.Second thing they achieved is oversupply of talent, which means it became cheap. People actually believed that so many are required to build a website.
>>108914401Where is this supposed oversupply of "talent"? The job market is full of pajeets, various wannabe "coders" who wrote their first line of code 3 years ago and even women.
superpowers is nice
>>108910499https://github.com/seanvert/cherry-promptI made this, you can easily paste lots of files in a structured format in LLM web chat and fuck it. It also has a token count estimative.
>>108912007Nonsense. We invented computer viruses but also antiviruses.Tech keeps each other in check. Even surveillance companies offer countersurvillance. Politicians may live but also have to hate surveillance used against them.All are tradeoffs.
>>108910577that's not what his sentence implied, promptard
>>108914416Mostly employed, mostly do shitty jobs. But you won't see them brag about first line of code 3 years ago, because they are busy with life. You see unemployed people talk about that and leet code and programming languages and what not.
>>108915464>We invented computer viruses but also antivirusesAbout 20 years ago most viruses were released into the world by antivirus companies.
>>108914238>I'm worried that web/fullstack jobs are in grave danger. I don't think web developers will be gone, but I think the need for them will lower. It's unclear just how much it will lower, but with AI you will (and already do) need less people to get to the same point as you would have gotten to without AI.Pure web devs may be in trouble, actual fullstacks will be fine for the most part.>I'm wondering if embedded is more resilient in this regard, as it has a physical component and is generally more complex than web development. I'm not pro or anti AI. I've used it for web dev and it is scarily good at a lot of it and it concerns me.And even then it could get fucked for raising prices of hardware across the board.
>>108914238>I've used it for web dev and it is scarily good at a lot of itI've used it for web dev as well, and it completely shat the bed at it.
>>108917355skill issue
>>108910499You have to consider other aspects as possible futures. COBOL was created to let business people code without needing a developer. Nocode platforms were marketed the same way. There were many other such tools. They all failed. Why?Because what the programmer actually does is determine how to solve the problem, rather than describe the problem from the incorrect perspective of a moron. This part cannot be replaced no matter how powerful the tool you give the businessfags will ever be.There are other considerations as well such as the possibility of AIs acting as compilers more than independent software developers.Or the possibility that they take over completely, so there would still be human programmers, but they'd be a rare niche, similar to caleche drivers today.Anyway, this won't happen this wave. Without an AI winter to reset the research and funding landscape, the tech is fully plateau'd and there's nothing left to squeeze but cope through developing expert systems around LLMs to stop making them be so broken, which of course means combining the worst of LLMs and of expert systems at once. But when the next AI technique starts showing some promise, you better be ready because it WILL be able to generate high-quality software without problems.ETA 20-30 years.
>>108912076That's not what real life is showing. Opus 4.7 is notably worse than 4.6. Gemini 3.5 is the worst release from google since the 1.5 series, and 2.0 series is still the best performing models they have ever released. All the gpt 5 releases so far have been completely fucked to the point the answer quality is comparable to gpt 3.5. Those aren't just slight downgrades like gpt 4o was early on. This is cataclysmic. When your models used to be good at making tool calls and now have a 0% rate of outputting correct json, that's pretty fucked up in terms of reducing limitations.
>>108910518Jump into an acid vat
>>108910499>Can we have a high IQ discussion about this?>Proceeds to have a low IQ discussion about this
