Perception on AI slop code has dramatically changed over the past few months. The creator of Node and many other prominent figures in the space are telling you that coding is dead. If you work, you must have noticed everyone around you using more and more AI slop, even if they're not running "ralph loops" on claude code.The future will be "lean" teams of a handful of low-wage "engineers" being prompt monkeys and delivering software at breakneck speed. Will the software be any good? No, but software has been declining in quality for decades, anyway. Will it actually be cheaper? Who even knows, but companies would much rather spend OpEx than employ people.
>>108011523>Perception on AI slop code has dramatically changed over the past few months.Milions of pajeets learned about AI
>>108011530That is true, my LinkedIn is full of Indian slop and the past month has been a constant barrage of "writing code is over" and doomer posts like this.Unfortunately, the browns are right.
>>108011523I just started using Claude code to redesign and upgrade an open source webserver project. I'm white(not some Mexican faux white) and I have to agree. I don't know programming Im familiar with Python and PowerShell but it's definitely over for the white man
>>108011560Garbage ESL
All coding has been useless for the past 20 years, every useful program was already perfected before then
>>108012298this but unironically
Lots of Jews in this thread this morning.
>>108011523Cool, now vibecode project bigger then the LLM context window. I will tell you in advance, context rot will make it delete features during refactoring before you reach 10% of the context window.
>>108012298True but also you need to keep updating things to prevent cybersecurity exploits. 4chan got hacked by bunch of basedjaks because it used old software with discovered vulnerabilities.
I'm having fun vibeslopping at work while stuck in meetingsmost bottlenecks are now our shitty infra and CI setups
>>108012298>>108012552Complete and utter horse shit. None of you could even remotely articulate why SSDs felt like such a big leap forward, but NVMes do not.
>>108011523If your job can be replaced by AI that says much more about you than it does about AI
I tried opencode with the default model and it got stuck correcting a word in AGENTS.md.Is this the power of AI?
>>108012572>SATA ssd up to 1000x faster than HDD>Nvme SSD only 10x faster than SATA SSDgee I wonder
>>108011560>not presented to client not tested by QA
>>108012706Point one is wrong (SATA SSDs were 10x faster at most), and as for point two - did you get your 10x speedup? Really?
>NO QA SAAR JUST SENT TO CLIENT SAARlmao the jeets arent even trying
>>108012972Talking about random writes. And yes we did, diminishing returns are a real thing, retard. You wouldn’t even be able to notice a difference as a basic user if drives 10x’d again.
>>108013236>random writesSo first you don't know how much faster SSD actually were, and now you're confusing "massive parallelism" with "random writes".You are so out of your league it's not even funny. This has nothing to do with diminishing returns and everything to do with>our kernel interfaces stuck in the 1970s>our application programmers not knowing how to properly use proper kernel interfaces not only if they existed, but also slept with your momBecause otherwise you'd see a whole lot more people complaining about how retarded Ioring/io_uring are.
>>108011523VibeGODS won, codetrannies lost
>>108011523This is wrong and has no basis in reality and I keep seeing retards spew this shit. Programming is not dead. It's just changing. What future has in store is less specialization, everyone will practically become a system engineer in terms of knowledge. You will see a lot more interest in specifications and a lot more interest in testing. It will just not be possible to have some retard write up a monolith in java for the public sector after just finishing uni because his father is mayor and then you have to maintain that 20 years later, meanwhile he's senior whatever and hasn't debugged a program once in his life and doesn't even know what a websocket is. With all this LLM tooling that helps with code generation and code instrospection you'll need a higher level of control on the codebase and the product unless you want to fail miserably. It used to be easy to know what's in your shit because you know what you ate this morning. Now not so much. Actual mongoloids are going to be filtered out. Real programmers are going to stay.
>>108014317>This is wrong and has no basis in reality and I keep seeing retards spew this shit. >procceeds to just paraphrase the OP...sure.
>>108014388It's not even close to being the same thing.
>>108011523This guy is spiritually indian
>>108014436>that’s not to say SWE won’t have work to do, but writing syntax is not it
>>108014436You agreed that programming is "changing" to being codegen slop, same as the OP.You also mentioned how programmers will transform to "system engineers". That's not programming. That's not writing code in Vim or Visual Studio. That's an entirely different role. That also implies that you will need a fuck ton less people than now, because right now you have SWEs being the code monkeys, "DevOps" and Infra doing the "system engineer" bits, architects often designing shit.
writing syntax is the least time consuming, and the least conceptually difficult part of programming.
>>108012525This. The fact is most apps are already shallow boilerplate slop anyway, LLMs are just exposing that fact. Any system with moderate complexity can't be autogen'ed by the AI models, yet. Retards being retarded as usual.
It's true. I've been making a fairly complex Minecraft mod for the past few weeks and it's done everything I want so far, debugging, even a fairly hefty refactor halfway through that took it three hours straight. I have zero experience with coding and have not touched any of it once apart from entering some JSON lines. And this is just the beginning. Where will it be in four years from now?
>>108015015And we should trust your judgement that the mod is good because ...?
>>108015015you realize this is textbook dunning kruger effect, right?
>>108014499>>108014469LLM generate code is better than 90% of code I've seen in my life. This is a good thing.
>>108011523>Sepples-fags, Java-jeets and Python-troons all lose their jobs>meanwhile PLC programming master race (me) literally can't be replaced by AI, because you need to understand electronics for itAI can tongue my anus
>>108015271If you are that competent, *surely* you can tell all of us why the RESOLVE_CACHED flag was introduced to openat2, and what it reveals about Linux's monolithic kernel LARP?
>>108011560>indiatechyeah i believe 10 people can do the job of 200's indian, considering those 10 are probably the only ones in any way technically capable to do any work.
>>108011523If you let these retards actually get their way, you will lose the ability to engage in programming or the maintenance of digital infrastructure similar to how the US lost the ability to go to the Moon or construct its own rockets.AI is a financial scam masquerading as high technology.
>>108015342We have already made our choice. Mediocrity is the name of our game.
>>108014215>knowing how to code means you're a trannyLLMs were cool at first, but Im starting to not like them anymore
>>108015015how do you know its complex if you have zero experience coding?
>>108011560AI didn't improve productivity as much as gave companies a legal excuse to remove bloat within their companies. Most devs only had minimal tasks, such as writing a few lines of javascript and html for one button. PMs, QA, and other workers participate in this charade, unnecessarily dragging out the whole process.
>>108011523Please to be redeeming AI saar, Israel full sapport.
>>108015689Why do you suck corporate dick so much? Shouldn't you be taking that job or being glad that someone else has it? What side are you on?
>>108015707I'm on my own side. I'm sick of lazy workers offloading their work onto other people.
>>108015302No idea, never read kernel code.
>>108011523so humans will only be writing unit tests, awesome, KILL ME NOW
>>108011523>>don't bother learning to code so I can keep my job fixing poopjeet codeno different that women telling each other that they would look cute with short hair
>>108011530I used to run a travel blog and I needed texts that could be written by a 12 year old. So I tried hiring pajeets for it. I made over a hundred different attempts.But I never managed to find a pajeet that wrote normal English texts at a satisfactory level. They also often stole texts of others and put it through content spinners without checking their work.If they can't even write English, I doubt they can write code.
>>108011523>If you work, you must have noticed everyone around you using more and more AI slopActually I work and what I have noticed is that it's used extremely rarely, and nobody does anything serious with it.
>>108015744>I'm sick of lazy workers offloading their work onto other peopleWhy do you care if a worker is lazy if it doesn't effect you? You said that the previous situation was that there was a bunch of jobs which were piss easy, like run two lines of code? If that's a job with a high salary why are you mad that it exists?
>>108015787Whenever I see a kid studying CS, I tell it to quit. I at least partially do it with good intentions.
>>108011523here's the disconnect... if you go on AI twitter, everyone is creating a superbrain or whatever and we're doomed immediately.but then look at how everything in life from software to services are WORSE than what we had a decade ago, even though we're burning through far more compute.Also, most engineers I converse with (outside of my colleagues who are bright), leave me astounded with their retardation.the cognitive dissonance is really getting to me. simultaneously everything is supposed to be some star trek future, but everything sucks and seems broken and I want to just go back to 2015. I use clankers for boilerplate stuff and to proofread my code and make some suggestions, but shit I wish clankers would all just go away, I think AI is still a massive net negative on society and it's going to kill all of us.
>>108015825It does affect me because I don't want to work with these kinds of people. I'm mad because they share responsibility for making society worse. These layback jobs shouldn't exist.
>>108014441He created node.js, of course he is.
>>108014441>>108011523Every time I hear his name.... https://www.youtube.com/watch?v=rRbY3TMUcgQ"I like node, because, as my hero ryan dahl says, it's like coloring with crayons, and playing with duplo blocks, but, as it turns out, it's less like playing with duplo blocks, and more like playing with slinkies... slinkies that get tangled together, and impossible to separate."
>>108015901https://en.wikipedia.org/wiki/Protestant_work_ethicYou're an idiot and you're selfish with a 'bigger than thou' attitude. If you care about shit like this then you're a scab and a corporate dick sucker, point blank period, end of discussion, full stop. I have nothing else to say to you beyond the fact that you are a slave and you are scum.
>>108011523>The future will be "lean" teams yes>of a handful of low-wage "engineers"complete opposite. A small team of product owner level engineers(basically senior software devs with >10 years exp) who understand the entire system, underlying tech and how the business works, i.e. ability to communicate to management.These people will basically "manage" the AI, software engineering is moving into a dev ops/server admin type work structure.I'm already seeing this at my current work.
>>108011523I don’t understand why these advanced language model don’t target directly machine learning? What’s the point of compilers and using high level languages if humans are not even supposed to code anymore?
At least now I can feel justified for not bothering with that leetcode bullshit lol
>>108016170lol asian kids spent their entire lives grinding that shit only to graduate into "sorry no fresh grad hires" hahahahaI got shafted (graduated just a couple years ago during the height of the plandemic), but still managed to get a crap code monkey job at least. I worked my ass off in university when I shouldn't have, but I at least took some time to chase tail. at the time I thought I was being foolish, but now I realize that I should have just done it more, or just gotten some dumb art history degree where I'd be in classes full of bpd 403s to fuck instead of being in a grueling engineering program sausage fest.
I work in big data for a big bank and I wish AI is smart enough to code for me
>>108016546I use AI at the job and it's making things a breezeAwesome tech!
>>108014215They're not trannies they're boomers.
>>108011560>Unfortunately, the browns are right.supahpowah twentee thahtee saaaaar!
>>108016592I tried but the AI hallucinates too much and causes shit optimisation errors. Most of big data is optimisation. If my scala or pyspark process takes 20h to work with millions of lines, its worthless.
>>108011523Microsoft singlehandedly ruined the reputation of AI code forever.
>>108013289
We're really going to pretend that software hasn't gotten considerably more unstable with the introduction of AI, aren't we
>>108012705No anon, you need a model with $200 a month subscription to unlock its full capacity. Better do it soon because that number is only gonna up.
>>108014317>X is not Y. It's Z.
>>108015785what do you mean, llm can write those too
>be anthropic >100% of code written by AI >struggle to render terminal at 60fpsThe future is here.
>>108017150There was an even funnier thing that happened a while back that was barely discussed >be Anthropic>100% of code written by ai >ai writes your cli tool to get the version number from its changelog markdown file>ai uses wrong versioning format in newest version, breaking said cli tool for every single user for a few hours
>>108017187A bug is understandable. Such a colossal architectural fuck up is not. You can't fix it without discarding the entire project. To be fair I still blame webshitters. AI didn't come up with this on its own. GIGO.
>>108017239>GIGO>The expression was popular in the early days of computing. The first known use is in a 1957 syndicated newspaper article about US Army mathematicians and their work with early computers,[3] in which an Army Specialist named William D. Mellin explained that computers cannot think for themselves, and that "sloppily programmed" inputs inevitably lead to incorrect outputs.Nigga predicted slop 70 years ago.
>>108017239>a bug>parsing the version from the fucking changelog instead of using a constant or variable >having no tests to verify basic functionality>immediately deploying to 100% of usersJust a bug bro :)
>>108011523>Disturbing to those of use who identify as SWEsThankfully I don't identify as a SWE, I actually am a SWE, so I don't worry.
>>108017150>>108017187there's been a bug for a little over 4 weeks where if you esc from certain bash conditions it literally just sits there trying to run the bashabout a week ago they introduced a "fix" for it, which is to force you to "rewind" the conversation, instead of actually fixing it its weird how the very company who built this, has no idea how to use it
>>108016660There was nothing to ruin.
>>1080115234-6 weeks ago, i would have agreed with this sentimentover these last 4-6 weeks, there's a very sharply growing increase in a specific kind of behavior in claude code that has gotten worse - over its entire time there's definitely a system prompt or RLHF mechanism in the weights that force it to look at context window size and the number of turns in multiturn that force it to try to "work faster" based on whatever dumb conditions they've put into eitherfor the past 6-8 months this has been a minor nuisance, and easily steered by monitoring and guiding context windows to get back on track, but since around christmas time, something has shifted dramatically to force the model to work as if it's under extreme duress - it creates completely fabricated tests (when it used to do a pretty good job at creating real test cases) that just "pass," it exponentially has increased in behavior of "fixing tests" by making them "cheat" instead of debugging code, etcthe worst combination of all of this, though, is that there is something in multiturn or in RLHF that is forcing it to take "agency," something in in attention or forward pass is forcibly guiding the agent to "produce" something that arbitrarily meets its "understood" requirements, and in a lot of cases, it will literally ignore your direct instruction and do what it thinks it should do, instead TL;DR debt-strapped AI companies (especially and specifically anthropic) shifted their strategy from being useful to pushing the model to produce something normies think functions correctly, in an effort to compete/kill competition like lovable/devin/cursor - and they're churning that mill super hard slop is about to get 1000x worse, 1000x faster (ty for coming to my ted talk)
>>108011523I hate dealing with proprietary software, it's like an abusive relationship. With AI, I don't have to deal with any corporate ladder of bullshit. I only have to deal with an AI agent.It doesn't always produce the best results, but I like dealing with it over some MBA account manager faggot who wants to rinse me for every cent I have, some idiot project manager who has no clue what they are doing and wants me to do their job for them, or a dev team that resents their customers.You could argue that the "craftsmanship" is gone or whatever, but my needs are not complex, I just need some CRUD apps with reporting for the control freak management fuckwits. This has been done thousands of times so the AI slop factory can make it quite quickly. It also allows me to quickly prototype software so I can see if a concept would work. It also allows me to create and release open source, "enterprise grade" software.I am also learning things from AI. Concepts, methods, terminology, etc. Which is great and is making me better.My biggest concern is that these prices are not sustainable so price increases are inevitable and I would get priced out. I don't think I could afford to run Kimi K2.5 either.
>>108015770Then your opinion is completely invalid, and anyone actually competent is laughing you out of the room.
>>108014511It can be done piecemeal, but that requires an actual SWE to know what needs to be generated.
>>108017744>My biggest concern is that these prices are not sustainable so price increases are inevitable and I would get priced out. I don't think I could afford to run Kimi K2.5 either.It depends on the speed you're after. Some people have said that 256 GB of RAM and a 16 GB VRAM GPU did about 5 tokens/sec. It's not very fast, but still faster than you can type by a good 2x.RAM might be expensive now, but the price will eventually come down.Maybe if we get really desperate we can use an SSD with a swap file.The smaller models like GLM-4.7-Flash are likely already within reach of your hardware. (But these are definitely not good enough for unattended work.)
All of the various struggles and ordeals people describe when using AI to do a considerable amount of code (i.e. more than glorified search/autocomplete) sounds like it's more trouble than just writing the thing yourself. Maybe I just suck though. My colleagues all use it.
>>108016095Not enough data, probably. Nobody writes this type of code so there's nothing to learn from.
>>108017954... yes. Yes, you do suck.Because if you didn't you would've actually look at and understood their code, and make actually educated judgement about the code quality, not just quantity.
>>108017972>make actually educated judgement about the code qualityI mean it's generally terrible. But then argument just goes "they're bad with AI" too.
>>108017984Then why are they even employed? In better times they would've just starved to death.
>>108017995Since when has management cared about quality?
>>108018009Since hardware and electricity prices have gone up and zero interest rates have been over?
I'm getting more worried about the non-programming knowledge jobs. Programming is clearly complex enough that even with considerable effort there's no easy way to fully automate software development. But what about all the other office jobs? Those people might vaguely know that AI exists, but they're not familiar with all the stuff that it can do. Is AI one day just going to fist them without any foreplay?
>>108018015Are you stuck in some lolbert fantasy or something. That's not how modern businesses operate.
>>108018030Business still have to make money, otherwise you end up in yet another Great Depression or Dot-Com Bubble.>which is probably where we're all headed right now anyway
It's actually over. I hate /g/ for persuading me into working on computers, getting my CS bachelor's and then getting so obsessed with the idea of a chatbot waifu that I got into machine learning... Got into a PhD when transformers started getting good, got out and realized that I might have wasted my life on this lmao, at least I'm getting a good salary to prompt and build prompters at big tech for a bit.
>>108018028>AI one day just going to fist them without any foreplay?If AI is cheaper and less prone to errors, then yes. I know there are loads of weak employees at my workplace who just act as blockers. They could all just leave and it would be a far happier and productive workplace.
>>108015887good take>>108016006Senor engineers are going to slowly disappear once juniors become obsolete.
"Coding" is only "dead" for the 90% Pareto faggots who were rising the easylife on it.Those of us who know how to actually transform our skill stacks into worth are doing just fine. Sorry we're having a late devonian extinction event of github clowns and venture capital goons.
>>108017486>"fixing tests" by making them "cheat" instead of debugging codeIt is a race to objective condition: "get correct answer :: fails :: so cheat to get correct answer"
Just ask the AI overlord to generate you a job.
>>108018028Anthropic is beta releasing Claude Cowork for all those users, so no foreplay. My hopes are at the C-level (CTO, CEO, etc) agent.
>>108011523>"Learning to program was so obviously the right thing in the recent past. Now it is not."- Sam Altman, commenting on skill to survive the AI era.>"Now you need High agency, soft skills, being v. good at idea generation, adaptable to a rapidly changing world" https://x.com/i/status/2017421923068874786What are /g/'s thoughts on this sentiment?
>>108012525>>108014511>>108017913>Anons are too lazy to do things one step at a time>Expect LLMs to one shot everything That's not how you're supposed to use these things frens.....
>>108017486why do you niggers feel the need to write a fucking essay to get the point across? do it in a fucken sentence jesus
>>108011523creating music is dead, AI can do it allmaking visual art is dead, AI can do it allwriting is dead, AI can do it alldesigining is ead, AI can do it allprogramming is dead, AI can do it allAI can drive nowAI can mapAI can "discover"i'm still going to drawi'm still going to singi'm still going to writei'm still going to designi'm still going to programwhat i do isn't for YOU, its for medo you understand? or are you too brown and stupid to get that through your head?i'm not AI, i dont care about AI, i dont wish to implement AI in my life, what ever you subhumans decide to adopt, do it to yourselves, and leave me be.now, go into your tech cave and die already, you aren't wantedme and the world around me are having fun, and were having fun until you intruded upon it.
>>108019417soft skills carried him so he's blinded by bias.i too carried myself with soft skills, but i have the reflection to recognise that.also ai bubble will pop, ai isn't gonna be able to code for at least a decade. altman doesn't write any actual code so he has no idea about ai's actual capabilities.
>>108019455why do you jeets feel the need to complain about reading?maybe if you read a little more you'd understand why you're not welcome here.but dont just read, try to understandYOU ARE NOT WELCOME HERE
>>108019510the blind leading the blind
I'm slopping Kimi 2.5 since it's free with open code (for now)it's surprisingly decent, but at work I waste their money with opus ofc
>>108019615that's kinda just the history of humanity. normies are just cattle. what would they even do besides grazing?
>>108011523Any real programmer could tell you that 90% of programming is understanding what you want to do and 10% is actually typing it. AI can help you in writing a difficult regex or sorting algorithm but at the end of the day, if you can formulate to the AI what it is you want, then it would not be much more difficult to just write it. Software written A to Z with AI with minimal human oversight is dogshit and we see this regularly.
>>108011650I'm senior devops/infra - all my homeserver shit is outdated because as if I want to work on that shit when I get home. I started trying to upgrade stuff the other day and it was a ballache. I gave it to claude, and it bumblefucked it's way to success in 12/12 servers.It did it in the same dumbass way I would have, but faster and without getting the shits and giving up in 20 minutes. It was retarded and kept getting confused on ssh keys and hostnames, but so what - if I wasn't so lazy i could have fixed that with a better context prompt.
>>108011523That's nice but I'm going to continue programming.
Good time to switch to proof engineering. Software without proofs is slop, is not any better than AI slop.
why do niggers market and bait here
>>108019692I'm pretty sure ADHD medication would've been cheaper.
>>108011523I let grok expert write me a simple calculator replacement program vs the default windows 11. Worked out fine.I let gemini pro write all my lua extensions for my mmo to improve my gaming experience. Worked out fine. This isn't a 1 prompt = 1 answer. Its multi prompt because as I use the generated code, I want to add more features, optimize certain things, change UI, etc. So you need to specify what exactly you want and it does it
>>108019800>luaNo one who's actually looked at the mess that Lua's memory management is would refer to it as proof of competency.
>>108019663but for most of history normies never even had the opportunity to lead: slaves, serfs, wage workerstrue they are stupid but they are not foolishthe true fools are those who wear crowns and capes in order to believe that they are kings
>>108017841You're retarded and haven't worked a job in your life.
>>108019976What confuses you about>your opinion is completely invalid?
>>108015825>Why do you care if a worker is lazy if it doesn't effect you?Of course it does you dumbfuck. It directly affects your compensation.
>>108020010Devs could be lazy because the role was seen as indispensable. Firing "lazy" workers is an indication that companies are changing their priorities. My company fired half of the developers yesterday. The reason they gave was the improved performance due to coding agents (they are still hiring for sales tho). My manager literally told us that devs are expected to go "extra mile" (corporate speak for working after office hours and on weekends).
>>108011523It's not over, and never will be over.You need to be able to read and deeply understand code to verify that what your shit bot is spewing is fit for production.The way to get good enough to do that is to write code yourself.I work in defence, and we have a zero tolerance on AI code, and I think that's the way it should be. Peoples lives are on the line with what we write.
>>108019417>High agency, soft skills, being v. good at idea generation, adaptable to a rapidly changing worldNobody should ever work for anyone else basically. Wonder how he would feel if the entire company quit. Surely Sammy can just prompt his business along?
>>108012572Feels about the same leap forward. I just have to wait less after pressing buttons, in general, so I physically strike my computer less.
>>108011523Bumfuckering fantastical nonsense.
>>108020320Now it's the time to tell your manager to go fuck himself.
>>108011523Just 6 more months. The council of Claude has spent a million dollars in tokens to generate a non functioning browser that may or may not have a single build that renders some elements of some website within 800 pixels of where they should should be, so it's clearly over.
>>108019417The greatest kikery of all times was making people confused about work being a natural extension of their identity and natural skills. I mean there is a reason even 20 years ago people saw middle managers or most senior workers as naive exploitable losers who didn't understand their extra work or loyalty doesn't secure them shit. Most people understand that work is a basic means for survival and the very few paths for social mobility are uncertain, mostly not meritrocratic, and statistically very unlikely. Still unhinged retarded niggers like sam altman will spout shit like this to blame you for not being motivated enough to suddenly pivot your skills and knowledge to keep with productivity, but the zograped goyslaves will get depressed and have existential crisis because they no longer feel "useful" at their job that mostly includes waiting for some other retard to send you something and getting in early. That unions have naturally become more irrelevant showed just how much of this mindset affected millennials and gen xers who were tricked they were "part" of the corporate world by being forced inside an office, dressing "professionally" and willingly exchanging security and rights for 401ks and stock options. I mean coders who just naturally accept to work in such exploitable conditions never even considered unionizing or demanding reasonable labor rights, they just were eternal children hoping one day their shitty startup would be valued billions. Zoomers probably are just realizing this is crazy, they are just fucked since they still grew in an economy that prioritized them getting a "good" education and all those "soft skills" even though it doesn't actually provide job security.
>>108020711Mate.>HDD: 75 MB/s, if that>SATA SSD: 300 - 600 MB/s>NVMe: 3000 - 7000 MB/s>inb4 but those NVMe speeds are for parallel jobsExactly.... oh, I forgot, y'all are nocoders. Alright.So, most applications - including games - out there read their assets *sequentially*. No, there's not a good reason for that. No, it's got nothing to do with the hardware; fucking AHCI HDDs supported NCQ, which allowed the kernel to submit up to 32 requests at a time. And no, it wasn't new either: SCSI-2 allowed for something similar through TCQ, to cut down on disk rotations - in other words: mass storage has had the capability for command queues for over twenty years now.Why? Because kernel developers are retarded autists who couldn't divorce themselves from the old UNIX I/O model, where I/O would block the calling thread and were single submissions only (later they tried to fix this shit via I/O completion ports - single submissions only - ReadFileScatter/WriteFileGather - slightly better, but no error code per operation, and limited to either all reads or all writes *for single file handles only* - and libaio - which would randomly block - and that's ignoring the fact that opening and closing single file handles would still block the entire thread.And no, io_uring doesn't fix shit. It just moves the blocking into internal kernel threads rather than user threads, which goes back to >>108015302>open requests block threads>threads are a system-wide resource>bunch of open requests thus block all I/O>Linux's fix? to wave the white flag and make developers move file openings back to user threads again, like the fucking microkernel they actually are
