[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]


File: claude_37db.png (56 KB, 1280x720)
56 KB
56 KB PNG
Sonnet 4.5 is ridiculously good and you can prompt it so fucking much despite it being so productive.

What the fuck is the future going to be when its gotten this good in span of few years
>>
File: b8.png (203 KB, 1631x1718)
203 KB
203 KB PNG
>>106998218
Damn, I wish I could experience it myself first hand.
>>
It's okay, but Instill hit thread limits and then got rate limited after telling it to fox obvious problems three times for every prompt.
Now it will refer to past chats. Which helps somewhat except that it wants to look at all past chat session which means a lot of irrelevant shit gets brought into the initial prompt. It might be best to start freah with the minimum info needed to resume.
>>106998234
What did you do anon?
>>
>>106998234
>review of your recent activities
tf does 'review' mean and tf you did?
>>
File: 184723147232.jpg (194 KB, 900x900)
194 KB
194 KB JPG
>>106998218
I'm a senior SWE and in no way a "vibe-coder", but this field is really doomed.
In the company I work for we just switched from Copilot to CC and everyone is using it.
Actually, the heads of engineering push everyone to work with it.
There are no more juniors in my company. From middle to staff, everyone is pushing out tasks like never before.
Most of times CC goes from the initial prompt to opening a PR in a couple of steps, sometimes in a single pass.
>>
>>106998370
Thanks to you vibe coders five years down the line everything will be full of bugs and no one has any idea how to fix them
>>
>>106998370
Same, CC has completely taken over our company, they are even using AI in upper management.
If you're not a senior at this point, it's unironically over.
>>
>>106998370
But how do you know if the code is correct? Do you write tests?
>>
>>106998457
CC can do tests and you just need a quick review to see if it isn't fucking up.
You just need to tard wrangle it.
>>
File: 1691321126089519.gif (277 KB, 640x640)
277 KB
277 KB GIF
>>106998370
Here, I used AI to create a video from your frog image.
>>
>>106998370
>>106998453
I highly doubt this is true that you can just turn your brain off and have AI do the ticket
and I dont count it if you just used AI to help you get to the solution yourself
>>
>>106998474
Where does the mysterious yellow liquid disappear to?
>>
Does LLM say "nuclear option" because of reddit or does reddit say it because it is LLM?
>>
Now is the time to be a tech entrepreneur. Why bother trying to be some company's slave when you can just start your own company?
>>
>>106998491
None is turning their brain off, every little thing the AI does is evaluated by a human, but all the scutwork is handled by CC that the job has become tard wrangling a hyperactive autistic savant, the mistakes are usually pretty easy to spot. The problem is all the seniors are gonna get old eventually and we'll need new blood, but I guess at that point, we'll have models that can do our jobs too.
>>
>>106998500
The former, LLMs are essentially the ultimate redditors.
>>
>>106998540
just take the L and hire juniors man even the capabilities of juniors have risen as long as they are competent
nothing is worse than work place full of middle aged people who would just want to be managers
>>
>>106998527
Because there are a billion other people trying to do the same thing. And money to pay rent.
>>
>>106998527
to be an entrepreneur you need to have a certain mindset
i could never be one since I dont really believe in things entrepreneurs usually believe. they tend to have very pro technological progress kind of thinking and I wouldnt mind if we returned to monkey
>>
>>106998572
Not in my power, or my interest, I already have one foot in the door, anyway
This industry is terrible, and I want no more of it.
>>
>>106998370
It's a vicious cycle, the more jobs are AI assisted the more data AI companies have to train systems to replace people
We're cooked
>>
>>106998457
>But how do you know if the code is correct? Do you write tests?
CC takes care of tests too. You just need to guardrail it a bit, but it's improving really fast.
Obviously engineers can tell if the code makes sense and it's correct, but as the other anon said, we are becoming more like supervisors than actual developers.
>>
Should've unionized and enforced labour licensing and other protections. Between letting big tech meme everyone into going into coding and now this, software engineers just allowed themselves to get cucked and stolen from.
>>
>>106998665
so what happens when they decide that since CC is replacing developers they can charge close to but still under a developer's salary for it?
>>
>>106998547
You're on exactly the right track with that line of thinking!
The best part? Everything I reply with has been scraped from reddit.
>>
File: 1675498939240003.png (1.04 MB, 652x651)
1.04 MB
1.04 MB PNG
my company uses claude sonnet 4.5, they have uploaded some maritime rule books and regulations as official assistants. Its reasoning is still ultimately shit.

chatgpt, despite having less sources on the topic, gives better answers in almost all cases.
>>
>>106998598
Neither of those things are stopping me though. Also you underestimate the average person's ability to actually succeed in tech. There's a ton of people telling ChatGPT, "Hey, code an app for me that'll make me a millionaire." Very few of them are actually drawing up detailed and intricate plans for the software they want to create and how to monetize it.
>>
>>106998731
overestimate**
>>
>>106998491
>I highly doubt this is true that you can just turn your brain off and have AI do the ticket
>and I dont count it if you just used AI to help you get to the solution yourself
I didn't say I can turn my brain off.
That's the difference between an actual software engineer and a vibe-coder.
But I can for sure go afk to take a coffee, come back and analyze what it's been done. Sometimes you need to do some fixes, or suggest a different approach.
However, there is no denying that LLMs, specifically CC right now, completely take care of coding with an extremely impressive understanding of the codebase.
In Claude Code you /init in your project and the thing analyzes your project entirely.
From there, with a little guardrail and experience in design patterns, you can hand over almost any task to it.
>>
>>106998737
how do you use claude code in practice? just do the command in terminal or do you use an ai editor?

so far I've just used claude in my browser because the AI editors look like a distraction
>>
>>106998218
I hate street-shitter hyperbole so much it's unreal.
>>
File: 1753931401988.png (164 KB, 512x512)
164 KB
164 KB PNG
>>106998218
Sonnet 4.5 is in a league of it's own. I sometimes wonder if people are using other models for coding because they are retarded or because they can't afford cc.

I'm still seeing people use 4o and Gemini, even seeing faggots use glm 4.6 and qwen all the time. Are they retarded?
>>
>>106998737
Init doesn't give you much, takes a lot of editing CLAUDE.md but you're mostly right, you can just tell cc to edit the CLAUDE.md to avoid A,B or C etc.
>>
>>106999031
even claude fucks up routinely. I'm convinced people who can get claude to do their work 100% of the time are doing stupid CRUD jobs that any monkey could do. When I try to get it to do anything Vulkan or compute shader related it shits itself.

I also refuse to allow myself to be roped into a scheme where I have to rent access to an LLM forever to be productive. I would rather be slower but independent. My local llms are useful but I don't trust them beyond "can you debug this" or "can you suggest alternative approaches"
>>
>>106998370
same, senior SWE, they got rid of all the juniors at our company, and they pay for our CC
some days i can just copy paste what the product people wrote into CC and go make lunch, play games, go out shopping, come back, open PR, reply to two slack messages
there's no way we're gonna be paid six figures for this forever
i'm going for a blacksmithing course early next year, need to figure out my exit plan
>>
>>106999162
>When I try to get it to do anything Vulkan or compute shader related it shits itself.
well duh
>>
>>106999162
agree, it's still pretty hopeless for anything graphics-related still. only a matter of time though.
>>
>>106999181
>i'm going for a blacksmithing course early next year, need to figure out my exit plan
yeah like thats gonna work out
>>
>>106999183
why are you saying duh? everyone said that it was going to replace devs and that if you get bad results you're just not prompting it right. It has definitely gotten close to making shaders that are usable . gpt-oss-120b and glm-air-4.5 are decent if you have the ram to spare for graphics, but about 75% of the time still it suggests garbage and I have to go to my pencil, paper, and renderdoc and debug it manually
>>
Mom cancel my meetings, Claude deleted my project files
>>
>>106999215
I will say it how it is
AI is not gonna write vulkan shit for decades and if you want a career plan there is that but again not many people can do it for a reason
>>
>>106998873
>how do you use claude code in practice? just do the command in terminal or do you use an ai editor?
You use it from the terminal within your IDE.
It basically works in agent mode, looks into dirs, analyzes the code, propose a solution and asks you if it can proceed. It will also read the project config files, know how to launch commands, how to perform tests and so on.
>>
>>106999196
aye brother. ironically people made fun of frontend devs for ages for being css monkeys but try getting an LLM to any kind of interactive visual or UI work and it shits itself more than an indian call centre employee
>>
>>106999238
>commit WIP commit so Claude can't undo what it's done so far (it works well, just some last touches needed
>Prompt next thing
>Claude has melty, deletes entire repo, including .git directory
>>
>>106999215
I don't work on that stuff so I don't know, but even if you are right it's just a matter of time before it catches up.
Point is, it's evolving quickly.
Anthropic spent 100% of its revenue on AWS, almost 3 billion dollars.
Right now, the only case where AI doesn’t move forward and developers aren’t entirely replaced is if companies can’t figure out a viable return-on-investment plan.
I wouldn't count on it though.
>>
>>106999243
hmm
i wonder if it would be smarter to not show your power level with this tool
>>
>>106998218
based. Total codetroon death
>>
>>106998234
Wtf did you do
>>
>>106999397
I would never admit to using ai to code to my job
>>
>>106999465
yeah well if you are fully remote than I guess thats possible
>>
I got my Claude account banned because I use a VPN.
Should I pay for Cursor since it has Claude included or it's better to recreate an account?
>>
>>106998218
i got banned from chatGPT before anyone else was even using it. cant remember how i found it, but i was wanting to fuck around with my bot for IRC friends and yeah. we would tell it about klikes and whatnot, got banned in a month because i didnt turn it off after the warnings
>>
>>107000209
cant you just make a new one?
>>
>>106999243
How can you revert its changes easily? Does it give you a log for what it's changed? These are my main concerns when using it. It seems like it can just change random shit since it has access to all files, and you wouldn't even know about it.
>>
>>107000268
Are there any burner phone sites available to complete the registration?
>>
this entire thread is full of anthropic niggers
claude is a one-trick pony at best, it has no actual intelligence other than slopping out thousands of lines of "best practices" code with no actual thought
>>
>>107001251
>best practices" code with no actual thought

thats programming yeah
>>
It's funny, six years ago I said here that I'd passed a selection process for a job as a civil servant. You said I'd be paid little and that the job wasn't very challenging. For a long time, I listened to your bullshit, but I see I made the right decision. While you're tearing your hair out, wondering if you'll have a job in a year or two, I'm here enjoying a good, lifelong position that I can only be fired for if I mess up really badly.

I don't mind the low salary; it's enough to pay my bills. At least I won't have to go hungry when the lean times come.
>>
>>107001560
Doing what
>>
>>106998719
is that Charles
>>
>>107001574
It's basically a department within the Ministry of Health. The current module I'm working on is a rewrite of a very old COBOL system that we're gradually rewriting in C#. The system handles general queues. We manage queues for blood donations, organ donations, surgeries, and so on. We have a single healthcare system, so even private hospitals need to integrate with our platform.
>>
>>107001517
i'm glad we agree it doesn't take intelligence to hold a programming job



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.