[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / 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: blow_on_AI.png (346 KB, 643x372)
346 KB PNG
> Jon Blow : AI sucks at coding
Why is he lying like that? Did he even try frontier models?
>>
You wouldn't understand. His standards are high. End to end vibe coding still produces slop in a big project.
Rewrites and decomps are easier since there is a fixed target.
>>
Who cares about that fag
>>
>>110258848
>Did he even try frontier models?
What else is there to try, retard? Frontier models is all that people talk about.
>>
>>110258848
>Why is he lying like that? Did he even try frontier models?
holy cope
>>
>>110258848
He makes great games, but his takes are fucking retarded, including those for game dev.
>>
>>110259151
give an example
>>
>>110259163
This entire video https://youtu.be/x0z0WcbSzJ0
>>
>>110259189
Oh, so you were just memeing.
>>
just how every sophisticated reader spots ai generated prose slop instantly
a real coder spots garbage ai slop code instantly
>>
>>110259203
>a real coder spots garbage ai slop code instantly
> caring about how code is written.
Outdated concepts. LLMs can test and review the code. If it works why would I care? You don't review assembly anymore.
>>
>>110259211
>If it works why would I care?
jeet mentality
>>
>>110259211
why do all ai shills talk like this
>>
>>110259238
because their main language is Hindi
>>
>>110259232
>>110259238
> not an argument
The truth is that programming is dead. LLMs are replacing programmers.
>>
>>110259251
dude, you already admitted you don't even care if the code is any good so what are you still doing in a thread about ai code not being good?
>>
>>110259280
> the code is any good
Only autists care about code quality in 2026. LLMs can produce thousands lines of code which always work. LLMs can fix all the bugs and implement features.
>>
>>110259299
you're not answering my question
>>
>>110259299
the code that you tarded jeets call "working" usually needs a full rewrite lol
>>
>>110258848
He's right though, we use Claude at work and if you're thinking about letting it just go on its own then it does extremely retarded shit constantly. Like last week I told it to add a new field to a JSON message the application was sending out and Claude fucking randomly decided to also add ZIP compression.

I asked it why the hell it decided to compress that one single field, it told me there was no particular reason but it saw ZIP being used in an entirely different place (it was for some archival history data) so it decided to use it here as well. Just dumb shit like this all the time. It's a very useful tool under supervision but if you're just letting it run automatically on your code it's 100% guaranteed absolute junk.

I guess in some situations having the junk is preferable over having nothing though, like when the volume of work is otherwise extremely high (for example, with all the decomps / recomps).
>>
>>110259211
>>110259251
>>110259299
they really all talk like this huh?
>>
>>110259322
yeah shit like this keeps happening. claude loves to make low iq assumptions about what needs to be done.

this absolute retard wrote like 10k lines for a feature which should have been a few library api calls, like 50-ish lines. Apparently it failed to download it and instead of pausing and asking what to do (as it was told to do in prompt) retardo just decided to rewrite entire library very poorly.

It truly is like working with electronic Indian. They do this type of shit all the time
>>
>>110258848
As impressive and advanced as AI code has gotten I feel I have to treat it like Spock and tell it why there needs to be shit in there that just makes me feel good.

Or it does a lot of nonsensical stuff in the UI that is not good UX. There's some hard and fast rules but there's exceptions to those, and what you do might create a unique path in UX that I find it hard to believe that AI will get around as I've seen zero progress towards it making anything "sympathetic" in any way possible.
>>
>not a moral stance, it's just not good enough for what I want
Is he the only reasonable person in the world? Why has it become such an emotional crisis for every social media addicted FAGGOT in the world?

>>110259238
>everyone who disagrees with me is a shill
Go full Uncle Ted or shut the fuck up, hypocrite
>>
File: 1786382885296890.png (379 KB, 1199x1508)
379 KB PNG
>>110259189
holy based
>>
It isn't good at coding, though. It knows the syntax, but actual code structure its still fucking awful at.
>>
>>110259457
He got filtered in limgrave? lmao
>>
>>110258848
He probably tried gpt 3.5 turbo a few years ago and has made his mind since. Most copers do like that desu.
>>
Why are AIFags so obsessed with us luddites?

You're right. We're being left behind. Go fuck off and make your million-dollar app and prove us all wrong.
>>
>>110258848
>Did he even try frontier models?
you jeets have been saying this for the past 2 years at least
any criticism is put down with the same old "you only tried x in <6 months ago> and now think all AI sucks"
no, motherfucker. it's amazing how well this shit works for certain use cases, but it's still not actual intelligence and it's not fucking magic
>>
>>110259717
>You're right. We're being left behind. Go fuck off and make your million-dollar app and prove us all wrong.
The point is to replace existing programmers.
>>
>compilers suck at coding
>they'll never emit assembly as good as mine
This is what luddites sound like
>>
>>110258848

Blow's a prick, but he's right on this one. Code produced by AI is always very low quality, inelegant, unreadable and unoptimized. It's fine if your only standard of success is "I want something that works" but if you want actual GOOD code, code that's elegant and has nothing unnecessary in it, you have to do it by hand.
>>
>>110259813
this was true for a while
maybe it still is, but I suppose processors are too fast for any manual optimization to matter anymore
>>
>>110259856
Skill issue
>>
>>110259813
that is kind of true, for example i could write more efficient code in C compared to python
>b-but its easier to use python
sometimes you have to go low level to be more efficient, how is ai any different?
>>
>>110259238
I'm completely sure I've read those exact posts and replies before. They must be bots, shills, or perhaps humanoids devoid of rational thought.
>>
>>110259742
>not actual intelligence and it's not fucking magic
It is.
>>
>>110259306
he evaded the question. instead he tries to move the goal posts to "the quality doesn't matter". it looks pretty weak right now, his best course of action is to wait for another thread and start with that from the beginning. ai would have told him, for some reason he thinks that he's better at arguing than ai
>>
>>110259892
The whole point of these things is to make skilled workers obsolete
If there is still such a thing as a skill issue, the tools don't actually work as advertised
>>
>>110259211
The issue with unmaintainable code (code that has too much coupling, is hard to read, is overengineered, has premature abstraction and other code smells) is that development speed drops exponentially after the first few weeks of development. These problems apply to LLMs too, not just humans. Good luck getting anything done when moving a button breaks half the system for mysterious reasons. LLMs can definitely boost productivity when used properly, but vibe coding tanks it
>>
>>110259322
>I asked it why
it doesn't have that state anymore (never had it in a way that is human interpretable in the first place), so all you are getting is post-hoc rationalisation baded on the history. i.e. you'd get the same answer if you fed the history to a completely different model. you already knew, but it's good to get reminded, i make the same mistake often
>>
>>110258848
I think he’s a genius. I bet he rolls coal on bikers. he sounds based
>>
File: file.png (117 KB, 474x529)
117 KB PNG
>>110258881
>>
File: 1728465803154363.jpg (93 KB, 694x642)
93 KB JPG
Ask AI to implement RTGI and compare implementation quality and speed to competent devs.
You shitters always like to pretend that some barely working feature is as good as a performant and production ready one.
>>
>>110259802
then why are you talking to those that will be replaced? are you also visiting retirement homes and tell the old people there that they will be left behind if they don't install tiktok right now?
>>
>>110259491
>filtered
>he thinks it'a about difficulty
You're an angsty insecure teenager who thinks games are about beating hard enemies so you can brag to your friends at school.
The game looks tedious, with nothing interesting in it and pretty much all Blow's criticisms in this clip were valid.
>>
>>110259975
You can tell it to decouple and describe the architecture you want. Even if you don't current opus doesn't do the button thing.
>>
>>110259322
>but it saw ZIP being used in an entirely different place (it was for some archival history data) so it decided to use it here as well.
YES! I've had shit like this happen half a dozen times in the past couple of weeks. feels like it's a new thing. maybe they're trying to make it smarter by looking for patterns around the codebase. which, honestly, is a good idea. who knows how often it results in better code and I don't notice it because it's correct
but then I catch these obvious mistakes and I'm jsut left wondering what else is falling through the cracks because I'm barely reviewing the code anymore
>>
>>110260060
And if he was playing on PC, the performace was terrible even on a high end rig with awful stuttering (<30FPS 1% lows) constantly. How FromSlopware still gets that much praise for the absolute garbage they put out is baffling.
>>
>>110260114
No it can't. The button thing was hyperbole, but all LLMs write shit code
>>
>>110260057
this lmao
AIFags will never be taken seriously
on 1 hand they have the secret GOD intelligence that can do everything
and on the other hand they need to relentlessly shill social media with
>ITS OVER DID YOU HEAR ITS OVER? YOU NEED TO USE THIS NOW NOW NOW

the best advertisement would AI actually doing shit. they wouldn't need to try to convince us so hard with faggot doomer narratives
>>
>>110260171
Not as shit as most people and you can describe how you want the code structured, patterns, methods and architecture.
The real shit happens when it gets confused by your instructions and loops trying to patch issues caused by underlying poor design. You can resolve that in code or pseudocode.
>>
>>110259251
>>110259299
>
Poo in loo Sanchit
>>
>>110260060
>unc who never made it past the iron ranks in any comp game ever
>>
>>110260221
No, they just produce trash. If you read the code you prompt and had the skills to evaluate its quality you would realize this. A prerequisite for AI psychosis is incompetence
>>
>>110259189
I hate jblow, but no cap, Elden Ring was pure shit.
>>
File: sokoban1.png (11 KB, 341x331)
11 KB PNG
>>110258848
>LLM devs: *vibes language that runs Sokoban in a day*
ohnono



[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.