As in, it didn't become immediately obvious the first time they played around with ChatGPT that extrapolating the progress outward even at a linear rate (rather than some s-curve exponential) would effectively spell the end of programming?Why are they only now, with Claude 4.6, seeing the writing on the wall, and lamenting the death of their "craft?"
>>108090228Also what is going through the head of the people still working on open source?Why didn't they stop as soon as they realized it was being used as training data?
Based on your two posts I don't think it's worth having a discussion. I will just call you a low IQ retard and move on
I've been trialing multiple different tools and models for like 3 months straight now and I'm still not sure if its saving me any time for anything beyond very basic pre-understood problems. For some things it's straight up wasting my time and resources while also being frustrating.It's really only useful in my experience with heavy guidance, which isn't so bad but isn't fully autonomous. And even when guiding it it's not always saving me time having to clean up tons of output.Things keep progressing so I'm hopeful but it's still a 50/50 in my book right now.
>>108090236I've already used open AI models to improve open source projects and open source projects have already been used to improve open source models. Why wouldn't open source programmers combine their strength and work like they usually do.
>>108090228>extrapolating the progress outward even at a linear rate (rather than some s-curve exponential) would effectively spell the end of programming?
Every code monkey is like I work on bespoke unsolved problems and u ask them what they do and they work on b2b saas
I think OP is fishing for attention and not actually interested in talking about AI. Sorry to hear that.
I want to strangle anyone who refers to it as a craft
>>108090412kek this but often they're just unemployed
>>108090491There's been a noticeable wave of>oh shitX posts, blog posts, and posts on whatever platforms the LGBT groomers use (Bluesky? Mastodon?) expressing dismay bordering on outright existential despair in the wake of the release of Claude Opus 4.6 last month.
>>1080902282 more weeks
>>108090228They're too retarded and arrogant for that. Even today some here take pride in telling you that LLMs don't produce any good code, not factoring their lack of skill obviously.
>>108090228Still not seeing it.
>>108090738I would understand if these were Fabrice Bellard types, but a lot of them, perhaps most, are webshits assembling a mosaic of React and various NPM packages together, with the help of second- or third-hand guidance from SO or an AI chatbot--yet they think they're indispensable.Or they acknowledge the transformative power of LLMs, but talk about the productivity gains as if they, and not their employer (probably soon-to-be ex-employer), will reap the benefit.
>>108090228Too much trust in other humans dismissing it for the greater good.
>>108090228If something can be done by AI, you could also slap in more human to increase the products quality. Old skills will be obsolete but human will stay useful as long as we are still better than AI at something
>>108090228My company just gave me a 30% raise and we have so much money pouring inMy job is so much fucking easier now alsoOnly zoomers and people doing third world jobs for first world wages are in trouble here
>>108090272>Based onYeah I'm think based
glad we all agree in this thread that op is a larp nigger who doesn't program professionally and doesn't understand what he is talking about. can't believe a frog thread had to die for this.
I depand better samefagging, OP
The AI astroturfing kind of sucks because they're all like "it's great try it out for free" and then you try it out and find out within minutes it's not valuable. Like wow epic troll dude you wasted a whole 5 minutes of my time.
>>108090228Well because it doesn’t even work nowadays and it's a grift.
>>108090620It's more like an art.
>>108090347My frustration with it is that even these amazing new models so close to AGI repeatedly shit the bed with repeated code or simply not following instructions.
>>108092565>simply not following instructionsThis is by far the worst part for me because things just fall out of memory/context.You tell it to do something, maybe you even record it in a file that you tell it to reference for guidance, and it does, for a few turns before it stops and goes back to its default way of doing things.So not only do you have to babysit it, you have to babysit it repeatedly and teach it the same lessons and its still not enough because it might just ignore you and lie.
>>108090412I don’t work on bespoke problems, but I work on heavily asynchronous code where design and maintainability matters.And I wouldn’t be so fast to people code monkeys just because they do B2B SaaS. Some of these problems they work on become subtly complex due to some asinine business rules, that LLMs shit the bed still.
>>108090412>b2b saas>using AI for this>b2byeah no, you aint gonna be in business long that way
>>108092580At that point I’d rather deal with a junior dev because they actually learn shit.
>>108092563artists btfo
I take a look at what my co-workers are doing and I see idiocracy brewing. I code with my own brain, not with a hallucinating algorithm that cannot do 1+1=2
>>108092604That’s the impression I have, everyone just collectively decided to give up on thinking. Maybe I'm just the weird one for being in this field for the love of the craft.
>>108092597Strongly agree. That's the way I've been describing the current state of models. Somewhere between an intern and a junior, but they are permanently fixed at the position they're in and can not improve. A new better model can come out, but obviously you are not going to impart knowledge on the current one (beyond a few steps).Which, don't get me wrong, is amazing. It really is amazing that I can have at my fingertips a whole ass junior to task on whatever I want, but also, at this point in my life, I am a much better programmer than a junior, so it's not always faster to task them than to do it myself. Sometimes, for things that aren't really important, but eh...that's not that common.It feels close to being good though, I'm still hopeful for the future. But not yet.
>>108092612We'll be the ones who can 10x our pay after shit hits the fan.
>>108090228It can’t beat children’s game shut up.
And when your indian colleague decides to reply to your code review or Slack chat with LLM generated responses. Jesus fucking christ I wanna jump off a building.
>>108092616Yeah, but I also want people around me to get better too! It’s actually pretty rewarding seeing your junior colleague gears grinding in their head and the solution to their problem clicking.AI just destroys everything, it's so fucking soul sucking.
>>108090228Maybe they believed that it was S shaped, that GPT-4 (or whatever) was close to the top of the S, and that any further progress would be greatly sub-linear.They were wrong but it wasn't a crazy stance to take at the time.
luddites will keep coping even after they've been replaced by AI-enhanced jeets
>>108092622I would just ignore it. Well, I would just ignore jeets anyways.
AI still makes pretty obvious mistakes. It has its uses but as senior at least you are safe for now.
>>108092657i really really want it to work, i want to make my work easier you retardThe problem is that it DOES NOT FUCKING WORK
Why are you posting here instead of just prompting the AI to answer the question? Youre wasting time that you dont have, anon. Someone out there is vibe coding the solution to nuclear fusion and you're wasting time talking to us. They're going to get the pay raise tomorrow that you're not fighting for today, which means you'll be offering your boyhole for a spot in the bread lines the day after that if we extrapolate accordingly. Don't you see, anon?The writing is on the wall.Claude 4.6 has made it so.The craft of shitposting is DEAD.
>>108092673How do you use it? The biggest gains imo come from tasks it can do on its own. Wite tests and let it implement the solution, it can loop for hours while you work on something else. Also imo only Clau Opus is useful.
What baffles me: people think those silly programmers are just going to go in their corner and cry?These are people whose minds have been trained on tackling difficult problems, who are much better at things most folk fail at. And now they're out of a job, whose job you think they're going to compete over? I hear everyone laugh at programmers and digital creators, but they are coming after (YOUR) jobs now
>>108092673This. People are too quick to assume I don't like AI just because I say it's not viable yet.I may be good at programming but that doesn't mean I like doing it. I like programs more than I like creating them.Just like a carpenter likes living in a house more than building them.And both of those things take a lot of time and dedication to do well. I'd much rather it take less time and less skill to produce something of equal value.Who wouldn't want that. But we don't have that yet.
>>108090712Opus 4.6 was released three days agoAlso people have been shitting up every website with such posts for literally years
>>108090712I wonder who would have a vested interest in creating "oh shit" posts and who has the capabilities to produce a lot of human like text 24/7 effortlessly. Probably real customers who genuinely love the product of course.
>>108092749I actualy like the autistic problem solving more than the software.
>>108090236why should they? they have always worked for free and will continue to do so.
>>108092865rambling / blog:Parts of it are fun for sure and it's rewarding to overcome a challenge. But some days for some tasks it just feels like a slog or chore, or worse I feel clueless for days trying to mentally solve a problem spinning in circles or throwing out entire prototypes.Something I really enjoy is running a program and having it work well and thinking "damn this is nice" only to realize it was me that made it.Same with reading source code. But that's always after its done. Until then it's somewhat arduous and uncomfortable.
because vast majority of programmers nowadays are retarded normals who got into it based on labor market pressure. technology is not actually interested to them and they wouldn't touch a keyboard ever again in their lives if nobody paid them to do it. admiting that you bet your life on the wrong horse is very hard, for a lot of people impossible and therefore you see a lot of denial, seething and all the usual coping stuff.
>>108090228"miss the boat" implies you can't do it anymore like buy Bitcoin for $1.you're just consoomer cattle, anybody can start consooming the same AI editors right now.>>108092565>even these amazing new models so close to AGI lmao
One think I like using AI for is just janny work. It's not glamorous, but I think it actually adds up. I have written a lot of shell and python scripts with AI, these are of course the easiest tasks, but how often do you just don't do that, because it still takes 10 minutes and you just do it manually?And it's not just about time, but also cognitive load, I can instruct it to write it while I work on something else, and then I have added some automation to my workflow.There was also a IDE plugin I always wanted, but I haven't touched that language in a decade, I can still read it but I just can't be arsed to write it. So I did it with Claude, I was able to review it, and now I have a useful plugin.Another thing is to point him at some folder and e.g. ask him if there are any incorrect comments or stuff like that, things that I rarely did before.
>>108093015>"miss the boat" implies you can't do it anymore like buy Bitcoin for $1.My phrasing was poor, I just meant not see it for what it was immediately.
>>108093166>janny workImagine how much money a site as big as 4chan could save up with that! You know how much they spend on janny payroll?
>>108092678>Why are you posting here instead of just prompting the AI to answer the questionBecause AI isn't AGI, and isn't omniscient, but the latest coding agents with RLHF are depressingly good at a specific task: generating code, and tests for said good, according to an informal specification.
>>108090228No amount of scaling llm will ever replace programmers as they are architecturally incapable of intelligence which is necessary for engineering.They are fine for webshitting though.
>>108090228what's the best way to use this shit anyway? I got some credits just to see what the hype is about and just the /init shit burned like 1m tokens if the dashboard is to be believed
I refuse to pay a corporation for making me dumber. I like writing code, I like digging into the root of the problem, I like to learn, and I don't want to pay $20 + tax + tip so I could perform less brain activity.
>>108094120As someone who wanted really to become a programmer back in 2020 and people would gatekeep it, or atleast try not to tell you where to go online, first time I ever used it, I realized how incredible it was stupidly , I only ever used it for homework it never ocurred to me it would get this fat tbqh otherwise I would've put all my live savings into NVIDIA instead of shitcoins lol
>>108092266no its useful if you spend $200+ on tokens a month.
>>108095081>it never ocurred to me it would get this fat tbqhThat's what I don't understand; why didn't people just extrapolate the tech out, even at a linear rate?>would've put all my live savings into NVIDIA instead of shitcoins lolI put a lot in MSFT (due to its OpenAI stake) and Google, but the pick-and-shovel plays like NVDA would have def been better.
>>108090383>has to pick an example for extrapolating>can choose literally anything that goes up only one time as a pun>chooses his wife having husbandscuck
>>108095081Nvidia was already up because of cryptocurrency then. Buying Nvidia would still have been the play if you were into shitcoins
>>108090228Lamenting? I do 2 hours of work a day and get paid ~$125k salary per job. I'm currently up to 3 jobs now, almost $400k/yr. Claude has given me a path to early retirement.
>bu-bu-bu-but you didn't prompt it right! that's why it doesn't work for you!If I have to upload my entire codebase and write a novel worth of direction to get it to something, it is unironically easier to do it myself directly. AND that is assuming it would get it right on the first try, which it often doesn't, requiring a second novel of follow-up to try and coax it to the right answer. The amount of time I see my coworkers spending trying to coax their AI of choice into getting it right is insane. If you are not a pajeet and can actually write software (as in you have taken it seriously and developed some actual skills), it would be faster for you to spend the 30 minutes you were going to take to write the correct prompt with all the little details reading whatever documentation you need to and just doing it yourself.
>>108090491You're like a Super Saar.
>>108090228>that extrapolating the progress outward even at a linear rateThe rate is logarithmic btwyour welcome>t. actual dev
>>108090228Programmers weren't as smart as people gave them credit for
>>108092785Every new model release we get this.
>>108090236Open source will be better now. Also it's the only choice that makes any sense.
show me an AI agent writing a rust macro that generates another rust macro using $$ macro metavars
