>>108231026
The DOM is already maximally abstract. 100% of JS frameworks are no-ops. The best would be none at all so it's not hard to beat one with code in it.

Did we really need another web framework?

so basically they had AI check its own work a bunch of times and not nearly enough time has passed to suss out how many times it lied

>middleware
>server actions
>rendering routes
>client hydration
>hardening
>coverage
Total zoomer deletion with no backup.

File: 1753457627980.jpg (287 KB, 1496x1309)

287 KB JPG

>>108231026
The singularity is soon. We were advancing over decades, then years, then months, now stuff's getting more advanced by the week. It's finally happening.

>>108231026
Websloppers deserve rape.

File: file.png (65 KB, 936x887)

65 KB PNG

>>108231026
kek

>>108231026
>github link to vite doesn't even work
was the article also authored by AI?

>>108231026
I'm unironically going to start disabling JS globally if this is our future.

Making it work on server rendered React. Poor AI ...

>>108231055
Current LLM capability progress is not exponential. Each new model is a smaller upgrade than last time - there's nothing like the overwhelming gap between GPT2 and 3.

>>108231026
>see vercel making bank from shilling next.js
>dimwitted influencers who are deathly afraid of linux eat it up
>people even stupider than that eat it up
>vercel proceeds to fuck each and every one of these people with extremely fees
>profit ensues

cloudflare must've seen this and wanted in one the action lol

>>108231047
Anon, there are tests, man-made tests, that make sure that everything is working as it should.

i'm starting to become concerned about people who haven't/refuse to use the coding agents
they just don't realise how capable they've become in the last ~4 months

>>108231055
It is strange, I popularised that meme shilling chainlink so I could make 300,000K. I did make the money as well. I must have posted 10,000 singularity memes.

https://www.youtube.com/watch?v=lzAuXuxD0Oo

The thing is though and stop and consider this, if the arrival of the mass internet and personal computer in 1995 ish was actually the singularity. Everything else is derivative, including AI which has yet to demonstrate any of the economically demontrable returns in productivity that personal computing did in the latter half of the 90s.

it's not hard to improve javascript slop

Anyone here train a model by port mirroring and shitting the entire metaspoit database at it and then looking for novel packet sequences ? The fun has only just begun

>>108231026
NEXT!.js

>>108231026
i dont even know what Next.js is

>>108231026
is it better than the AI matrix home server port they published on their blog?

>>108231191
Cybesecurity is the next great nass layoffs zone, all work in SOCs is easily automated and better done by ML

>>108231219 (Me)
Checked it out and it's basically react-dom and react-router for server side rendering bundled together.
Which is funny, since i got a server DDoSed last year, by five kids, who simply requested a page that got server side rendered with react a few times.
I fixed it by replacing the react code with html in a template iteral.

Why would you use AI to vibe code a framework that only exists to abstract things away for "full stack devs", with the huge downside of being slow?
Shouldn't you use AI to actually code your project in complete vanilla js instead, without any framework, because it is faster?
You dont have to care about abstraction and readability and ease of work, because you dont look at the code yourself anyway, right? The AI will just do it.

I assume the AI in question was Claude?

>>108231083
Jesus fucking Christ
Did these fucktards even test their slop?

>>108231260
SOC is already automated since at least 7/8 years. Back in 2018 I've worked on a project that automatized every DDoS reporting and mitigation process using "basic" machine learning algorithms and some predefined rules. Why do people here always think that CybSec = SOC?

>>108231120
>server rendered react
so we went full circle from php and node.js classic SSR to spa websites then to ssr again but while keeping the spa bloat? lmao

>>108231352
Well it's natural that they wouldn't have caught an error as arcane as the build directions not working. I mean, who tests that sort of stuff anyways?

>>108231392
basically, SPAs got so bloated, that they had to cut the bloat into different bundles based on routes

>>108231048
>>client hydration
what does this even mean? i barely understood the rest of it

>>108231134
>it would take months, if not years, to write the tests and verify them
>so we had AI write them

>>108231507
funny how corporations moved rendering to the client to lower server expenses, but they ended up having to serve hundred of MB worth of data for each request anyway

>>108231148
people are emotional, irrational animals

they still think AI images can't do hands and all of it looks the same

>>108231026
>use open sores already published codebases to produce something for which the patterns are already out there
>hurr AI is so le awesome

Was there ever a bigger idiot than Stallman? This is what decades of open-source software has led to. Millions of idiots have programmed themselves out of jobs.
They made their code available for free, as in free for AI to train on it and then offer a paid service for le superintelligent vibe coding

*spit*

>>108231134
>>108231026
From which specific jeethub did they steal the code from?

>>108231519
I think it just means deferred loading. When you go to site.com/page.html it gives your browser a minimal version of the page which contains a script to download the rest of it.

>>108231906
kek
this is the first time i've thought about it that way but you're right. a shame too because i do think stallman and other foss fellas have contributed an insane amount of good to the world. the problem is the world is filled with a lot of things other than good.

>>108232490
Unironically, nobody could have anticipated 2 billion jeets existing. And nobody could have imagined the damage that unloading all of them on to the internet would cause.

>>108231148
I use them every single day. I would never in a million years trust them unsupervised like Cloudflare did here. Why? Because every single vibe coded app I have seen deployed at work, I find quality issues within minutes. Every. Single. One.

>>108231163
Its not and they haven't done it. This is like anthropic claiming they vibe coded a C compiler all over again. Its fucking bullshit. They havent done it.

>>108232601
But if their claims are true and it's not a buggy/vuln infested mess, then they have done it. I don't see why sticking our heads in the sand helps here. Next is very well defined, with a huge amount of tests, but rewriting it to be 60% faster or whatever is still a massive achievement, especially given it was done with only one human involved over a week.

>>108231083
>everything has to do retarded init command
I am certain that vinext is crap, but this issue is created by retard

>>108231026
honestly speaking next.js is such a horrible piece of shit, that even if the would replace it with empty file, it would be better

>>108232601
It's literally open source. You can go look at it on GitHub. It's real.

>>108231148
Separated the losers from the winners

>>108231134
AI-made tests. It’s a standard for AI-assisted development.

>>108232542
It’s obviously not unsupervised. Come on now.

>>108233057
It's not and you didn't read the article. The tests are made by humans. That's part of the reason they chose AI for this usecase.

>>108233597
You are sticking your head into sand.
You saw cloudflares vibe coded matrix server that didnt work.
You saw Anthropics dysfunctional c compiler, that also doesnt work.
You saw the vibe coded browser, that also doesnt work.

But you refuse to accept it and instead believe some advertising blog post. And then you fall for the next scam.
>but muh tests
He is shouting, because thats the only thing he can spot, that is different to all those previous vibe code failures.
But what do those tests even cover?
Its like the rust idiots from uutils, who focused on testcases alone, and then discovered that those trscases dont even cover 10% of switches. And it took them an actual deployment on Ubuntu and the following shitfest and flood of bugreports to notice that.

Rather than believing lies from people who lied to you many times in the past... how about you just test it?
Show us your application built on vibe coded nextjs.

its gonna be a great time for hackermans

>>108231148
No one is denying that AI couldn't replace at least half of devs now and turn the rest into overworked prompt monkeys if companies really wanted to (it would still be a mistake, but only the next quarter matters).
The real question is why would you use coding agents as a dev if you're not:
- lazy
- retarded
- overworked to the point that you need to defer shit to it
- told by your boss to use it or else.
It's not like there's any skill required to using them so the whole "you're falling behind" meme is dumb nonsense mostly perpetrated by no-coders or people who resent coding like most people who shill AI agents here do. Especially when the way to use them changes regularly. Like if you still are still using an AGENTS.md file in 2026, you're a retarded luddite now and have already fallen behind.

>>108233962
this nigga using ai to diss ai

>>108234085
You literally have to because the other devs that are just as good as you will absolutely outperform you if they utilize AI, and you don't.

>>108234096
>something will happen in the future
And China will collapse and Ukraine will take Moscow, about the same time Israel manages to take Gaza and Americans build god in the form of AGI, which then makes Aliens descend from heaven in UFOs, before the Rapture happens and all the shabbos goys will get rewarded, while everybody who doesn't support genocide rots in hell.

>>108231148
It depends on the software. For example, in my company we are working on fully event-driven projects, and there are many of them interconnected. And Claude Code absolutely sucks at preventing race conditions or avoiding adding new ones. When it does fix them, the solution is usually really slow and dumb.
And it kind of makes sense, because race conditions are pure probability. While some are obvious, others may happen only after 1 billion operations. It’s not really something you can properly train AI on.
Other thing I noticed is advanced lying, especially with Opus 4.6. It loves taking shortcuts by not adding what you asked to do, changing tests a bit and make it like everything works. It's kinda hard to detect them because first glance everything makes perfect sense.

>>108234096
You can also get outperformed by a 996 Indian vibe coder too. Even the biggest supporters of vibe coding accept that the code isn't really good, we can see the few vibe coded apps people use like Claude Code fall apart in real time to the point where even the slop lord himself, TheoT3, is having enough of it. So if the only metric is how much shit you can do in a spring, you too have already fallen behind by your willingness to not work harder for the company.

>>108231148
the problem is that these tools use your input to train themselves if you use them you are effectively training your replacement. This isn't going to end well for majority of developers.

File: 586361782368123.png (48 KB, 1138x473)

48 KB PNG

>>108231026
so ai successful copy/pasted next.js code base by changing few names here and there.
there is a 2023 paper on this describing how AIs can fully replicate books they were trained on. I don't understand, is this guy retarded?

File: 1611008874095.jpg (247 KB, 1441x1383)

247 KB JPG

>>108231026

File: 1769547782139.png (185 KB, 1184x2428)

185 KB PNG

>>108231026
Are they going to

- production grade
+ proof on concept

thing like with their last vibecoded garbage?

>>108234181
it's even funnier when you realize that comments literally say file by file what this file was used for in next.js

>>108234198
>claiming that something untested that doesn't work is "production grade"
Only a big corpo gets away with such shit.
You wouldn't be able to do this in a hobby FOSS project, because any person who spends his personal time trying out your stuff and discovering that you lied, is going to give you hell.
Or even consumer products for good goys. Imagine giving a boomer a TV that randomly crashes every minute, while claiming that its the best tv ever.

It's astonishing how corpos have the lowest standards known to men and keep lying and keep slurping up lies from other corpos.

>>108234138
>Other thing I noticed is advanced lying
The free version of ChatGPT (on the website) does this a lot. Not even code related, but about anything at all. And it is highly resistant to being corrected even if I link a source.

>>108234242
kek, i once wanted to find a stage collaboration of two artists that i remembered vaguely, and after google was useless, tried AI (it was old and surely in training data) and it gaslit me and told me it didn't exist and that those artist never collaborated.
I then found it while searching on Yandex.
I thought that i would be nice and tell the AI, so that they can use the mistake in the next LLM training, and the AI went bonkers, insisted that it doesn't exist and demanded proof and specifically wanted the youtube link to it, like some redditor.
I pasted the youtube link into it and it then claimed that the video actually shows something different, that it doesn't, just to insist on it's lies, even thought that its a chatbot that cant even follow such links.

Meanwhile other times it glazes you and says that you are the smartest being on earth, for asking a question that assumes something wrong.

AI is a lying simulator.

>>108234198
What "concept" does it "prove", though?

>>108234286
>AI is a lying simulator.
Well unironically yes, since it's making up its response from probabilities instead of from facts and logic.
>>108234347
I guess we're going to see websites switching to a different anti-DDOS product soon. There's no way Cloudflare is going to survive being vibecoded.

>>108231124
As long as LLMs can't replace transformers with something better, it will be mostly about data centers, and they won't be build exponentially faster.

>>108231055
>bottom tier iq
keep slurping that marketing/propaganda

The hits keep coming
We now have proof that SWE has always been the easiest form of white collar work

>>108234198
>n kuntz
kek, is this a real account?

Yay slop

I have no clue how impressive this is but I honestly believe web development overcomplicates everything. They abstract a bunch of stuff that already comes in browsers, like react's virtual dom, frameworks need rehydration because they manage internal state that need to be synced with the real Dom, and since those solutions are slow they need shit like prerendering. I feel like there must be a better way. A framework that levereages the browser's apis before reinventing the wheel. I tried nextjs before and it was so bloated and slow

>>108238346
pure React on its own is fast and quite capable - I implemented 3D rendering engine in WebGL relying on React update lifecycle and dividing responsibilities by hooks, and it didn't cause much of a performance overhead
but nextjs is the worse shit I ever used - bloated, slow, unusable mess that doesn't solve any problem and intrudes 100 more.

Vercel just released a blog about migrating to "Vercel from Cloudflare" man these guys are seething lol

>>108234174
the majority of training data is just from RL environments now. they don't need much human data anymore.

A few points:
> Taking months isn't the greatest metric because you're talking about engineers that do many things at a company compared to a codebot that is doing one thing. How many engineers dedicated to making a NextJS replacement were *only* working on that replacement?
> The idea of making a replacement in the first place is kinda silly and a really large scope. AI is good at that type of aggregation, but most engineers only understand the many NextJS/framework concepts in the abstract. Much of the problem with this project is probably with *how* abstract the paradigm you are working in is. I.e. the project is unecessarily complex.
> It's impressive that the AI was basically able to make NextJS, but then just update the parts that was required by Cloudflare
> There are definitely going to be bugs lol

The more I look at AI, the more I view it as a very advanced form of package management, where you can make finetunes to any particular segment of code since everything can be abstracted into a "package" of some sort. However assuming the NextJS shit is too complex for its own good (although frontend isn't exactly my field), perhaps we can get AI to replace it with something simpler in implementation :^)

>>108238813
>Taking months isn't the greatest metric
the metric is flawed, but the point stands - if you have any experience with these things, the speed is superhuman. no human can keep up.

>is kinda silly and a really large scope
are we now starting to complain that ai projects are too big and complex? really?

>There are definitely going to be bugs
probably. will they be meaningfully more than something that was human-developed in a rush? can they be patched faster than a human-developed project?

>perhaps we can get AI to replace it with something simpler in implementation
why. code is no longer for people; it's for future models. complexity is a you problem; try and keep up.

>>108231581
What's really funny is that every phone is more than powerful enough to render sites client side extremely fast - all you have to do is not use React, Vue or Angular.

Unfortunately, that's not an option for the bootcamp retards who can't actually write JS because they only learned those frameworks.

Remember how slow Github became when they rewrote it in modern web technologies? It happens every time. And every time there's a retard who's like
>UMMMM, React isn't slow, you just have to-
Yes it is. It is slow. From start to finish is it designed to do a pisston of extra work to even figure out what changed in its VDOM. The best you can do is reduce the amount of extra work it has to do a tiny bit - it will ALWAYS be more bloated and slower than reacting to changes directly.

>>108234240
And this person will get promoted for this. This shit created so much buzz and that's all that companies crave for. All the Cloudflare bosses probably want to speak to him. The AI company bosses will want to poach him.

>>108238928
>From start to finish is it designed to do a pisston of extra work to even figure out what changed in its VDOM
yeah, conditional rendering is a thing everywhere where you want something to be fast.
I recommend you to write your own game engine or physics simulation, you will quickly realize that doing a hundred extra conditons, to avoid re-rendering something you dont have to render, are making things faster.

>>108231026
>We're the first ones to pull this off! EVER! It's SOooo-- hard guys. Nobody could do this without a full professional corporate dev team; unless they throw massive amounts of expensive AI compute at it!

These fucking disingenuous snakes should have a look at
https://www.npmjs.com/package/@vitejs/plugin-rsc

Which already does client components, server components, and streams.
As for routers - use fucking React Router with it. It's literally one of the fucking starter examples:
https://github.com/vitejs/vite-plugin-react/tree/main/packages/plugin-rsc/examples/react-router

>>108239130 (cont.)
Actually - they should have a look at what the retard-AI actually spat out. Because it LITERALLY FUCKING USES the Vite RSC plugin to do all the heavy lifting: >>108231083

>>108231083
Literally all of these breakthrough AI projects built in 1 week are like this.

File: 1751821876967229.png (59 KB, 640x464)

59 KB PNG

>>108233962
Again, you didn't read the article
Cloudflare's vibecoded vinext project is already deployed here:
https://www.cio.gov/

>Early results from real-world customers are encouraging. We've been working with National Design Studio, a team that's aiming to modernize every government interface, on one of their beta sites, CIO.gov. They're already running vinext in production, with meaningful improvements in build times and bundle sizes.

You are the one with your head in the sand. These things are doing real work that gets actually deployed. This would be the carpenter equivalent of the house building itself out of a bunch of wooden logs.

>>108239348
The salient question is: would you dare sleep under the roof of that house?

>>108239378
This cloudflare project was built with a developer's oversight. He guided the machine and reviewed its pull requests. This is the correct way to vibe code (ie, taking a real developer and have him review and control the project). I don't like it either but the house is getting a lot less dangerous to live in.

>>108238928
you have absolutely no idea what you're talking about, junior
come back in couple of years when you will gain some experience and leave adults to do the talking

>>108239421
>a developer's oversight
An engineering *manager* - anon.

>>108239437
Framework booty-booty detected. Did it make you upset that I insulted the only technology you know? Don't you have work to do, making websites slower and shittier and introducing a billion minute long build steps to counteract some of the slowness your "tech debt by default" stack introduces?

>>108238928
the point of those frameworks is not to reduce cpu cycles but to make code scalable with time and changes. Modify the dom directly with event listeners and all that oldschool jquery jank? That works today and in a small prototype. Add a lot of interactive things to a page and make other changes over a long time that bring in a lot of complexity, and you'll end up seeing stuff not move when it should (or vice versa) with no way to find out why other than line by line debugging

>>108231519
>what does this even mean?
What those Russian prostitutes died to the President.

>>108239551
Who says you need to write DOM spaghetti. Just don't use React. Solid or Svelte offer more or less the same model, except without the absolutely retarded "the entire function is re-run every time anything in the app changes" shit that makes hooks such a pain to work with and with the same kind of performance you would get from hand optimized DOM spaghetti.

The reason react sloppers don't use them is because the way they know how to implement "feature" is to write "npm instal react-feature" and if that doesn't work, they're done.

>>108239460
it is not about frameworks, junior
because everyone with a bit experience know what shitshow js is, and that nextjs should be nuked from orbit
>Yes it is. It is slow. From start to finish is it designed to do a pisston of extra work to even figure out what changed in its VDOM. The best you can do is reduce the amount of extra work it has to do a tiny bit - it will ALWAYS be more bloated and slower than reacting to changes directly.
it is about you babbling about slow an vdom in absolutes, proving you have no idea what you're talking about about and how inexperienced you are
and based on your reaction, I nailed it

>>108239600
lol at building something bigger than reusable component or prototype in Svetle

>>108239871
What. You can't do it? I'm not surprised.

>>108231191
Something like this is pants-shitting terrifying. That's where the real power of AI is and I don't think it's been really done well yet (at least publicly). So many zero days will be found when it happens and the mitigation will take 10x longer even with AI.

>>108239881
You can build frontend in Haskell, Java or Rust, but no one sane would do that.
Svelte main benefit is being able to build web components reusable by various other frameworks of pure JS
but only a madman would waste time trying to build something bigger with it

>>108232513
Most of India isn't on the internet yet. It's about to get worse.