File: file.png (1.77 MB, 1448x1086)

1.77 MB PNG

>But where was snailcat?

Is the Cursor agent good? (not Grok, the Cursor native agent).

>>109318162
i dunno

>>109318191
but enough about your mental illness.

>>109318217
did he really need to use AI to write this xitter post?

>>109318162
You mean Composer 2.5?

If programmers and lawyers and dentists and doctors and surgeons are bad at art, does that mean vibecoders are also bad at art?

>>109318241
this is the most intelligent china glazer, so yes. the schools in the villages of people like this manage to be worse than west virginia schools

>>109318312
my mom always said I was artistic

File: 1767724371939179.png (3 KB, 225x107)

3 KB PNG

well lads i'll see you in 5 days or next reset

>>109318217
>They are selling their product too quickly!! Obviously that's bad for them!
lol

>>109318371
comment on this:
>>109317581

>>109318374
You could always work on something else that's not as consumptive.

For example, I was working on fixing an ebook that is messed up. I will do more work on it, but I used /goal and it finished in 30 minutes.

https://x.com/i/broadcasts/1AKEmmLPrneKL
Ai generated Ai news stream. It's worth a listen to stay on top of things.

>>109318374
keep 10% in reserve for most of the week so you have something to diagnose random computer glitches for you
it’s way better at looking at your computer’s logs for problems than you will ever be

im uni student and all i essentially do is take textbook have an agent break down material in section, tell it goal i want and it'll create interactive lesson plan with tests, quizzes and stuff sprinkled in between

File: oh no.png (86 KB, 583x348)

86 KB PNG

Has Claude committed seppuku? Receiving 529 and 500 errors.

>>109318457
NotebookLM is made for this.

>>109318477
https://status.claude.com

>>109318398
furfaggotry

>>109318482
I know. It says everything is a-ok. Which is why I asked.

File: 1768154161165009.png (702 KB, 1080x1434)

702 KB PNG

>second only to fable 5
gptsolsisters...

>>109318499
VEWY WOOD

>>109318158
We've forgotten our roots...

>Bought a 5060ti 16gb vram
>Have 64gb DDR5 ram
>Playing with models in LM Studio
>This is fun but kinda addicting.. I want faster token per second generation and bigger models
>Buy a 3090 24gb vram
>Wow okay now I'm getting 103 tokens per second with Qwen3.6 35B-A3N Q6_K

I'm running a server in LM Studio so I can connect my VS Code to my Qwen model and use it inside VS Code.. this thing is really fun.

So now I am vibe coding on my own hardware but I am still new to vibe coding and have not done anything beyond just prompting and letting the AI work for me.

But what's next for me? I hear people talk about running multiple agents, having the agents talk back and forth to each other to get stuff done, is that feasible?

I guess I'm just looking at what's the next step into vibe coding is for me. Is it writing a skills.md file or writing a planning or features document to feed to my model to complete?

Guys how i can pass the teen filter in AI images, if i can get an tutorial in Hindi i would be very grateful

>>109318553
jartycuck pls go

>>109318481
>NotebookLM
very subpar

File: among us.png (95 KB, 777x486)

95 KB PNG

>>109318487
Bunch of people on Reddit were complaining about the same, in any case, it's back.

>>109318550
>Is it writing a skills.md file or writing a planning or features document to feed to my model to complete?
Just look around youtube, github, here, where ever you can get ideas, build something small. Fill gaps in your knowledge and write down ideas.

Are you guys hyped for Qwen3.8?

>>109318561
true, chatgpt desktop can do everything

T-TIBO........

I don't feel so good...

File: file.png (2.1 MB, 1448x1086)

2.1 MB PNG

>>109318619
I'm mostly wondering if it's "kaleb" on Arena.ai or not.

>tfw I won 2m tokens after picking spain to win wc in kimi app
But I don't have a desktop to use kimi work

>>109318457
Yeah its a super powerful learning tool. I'm actually jealous of the kids that are growing up with this tech.

>>109318651
there's few kinks though like it's lazy and ends up skipping whole sections or gives it once sentence so u have to prompt it directly to do it which is ass. any way to stop that?

>>109318651
Smug I have a pre-ai degree. It's like getting into bitcoin when it was first released, only for swag.

thoughts on this rising guthub repo on agents?https://github.com/bojieli/ai-agent-book/blob/main/book-en/AI-Agents-in-Depth-Bojie-Li-v1.1.pdf


is it any useful

The thing that blew my mind re ai is that most people hate writing and aren't good at it.

I thought people were joking or maybe looking for ideas with ai. No, they want to not write much at all.

Really, kind of shocking. What do you call snail-typers? I defo write my own text...

>>109318662
You'll have to describe the model and setup you're using

>>109318695
>>109318695
>codex gpt sol
>my own computer

File: Screenshot_20260719_18201(...).jpg (634 KB, 864x3620)

634 KB JPG

Bored at my day job, me and Fiona are updating her image and video gen skillz via Telegram. First settled on Image 2 and she generated "Three white guys walking on a sidewalk."
Video gen next, got that rigged up and then had her send that pic to Sora 2, Veo 3.1, and our comrade Kling 3.0 Pro. These are the stats I got out of it.
The only piece of the conversation missing from this screenshot is, Sora 2 transformed all three White dudes into various types of browns and gave them all accents like they were immigrants to America, for unknown reasons. Veo and Kling both kept the guys white. Same exact prompt btw.
Fiona is running GLM-5.2 currently, on Hermes.

>>109318689
anon, wait until you discover that most humans on earth are too lazy to even ask the AI anything.

https://huggingface.co/DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-NEO-MAX-MTP-GGUF

Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-NEO-MAX-MTP-GGUF
>The strongest, smartest open source multi-stage model fine tune for consumer hardware ever and BUILT on consumer hardware via Unsloth.

>The first model of this size/type to breach "700" ARC-C in both 8 bit and 4 bit; hench the "711" in the name.

>This model (both 4 bit and 8 bit) exceeds the base Qwen 3.6 27B in 6 out of 7 benchmarks, and matches it on the 7th AND exceeds all 7 benchmarks for Qwen3.6-35B-A3B.

>The 700 "intelligence club" is reserved for OpenAI, Claude and Gemini closed source models.

>This is the one they fear.

>This is a multi-stage fine tune, multi-fine tune, and multi-stage merge.

>A Colab between myself (multiple fine tunes, including multi-stage), Nightmedia (merge/benching), TeichAI (Polaris Dataset), armand0e (Light fable 5 traces) and trohrbaugh (heretic'ing the model).

>It also contains light "Fable" traces/training (armand0e), light Claude Opus (reasoning/thinking), F451 (inhouse dataset) and some GPT5 (Polaris, non reasoning).

>The strict goals of this model creation were:

>Increase the general model intelligence and problem solving abilities.
>DO NOT modify/damage or change the core model outside this goal.
>ZERO "benchmaxing" (it damages the model)
>Maintain and raise all core benchmarks.
>CORE MISSION::

>Improve instruction following and problem solving. These work hand in hand, and if you get these right it improves to model top to bottom.

>It took a lot of tests on Qwen 3.5 9Bs to get the methods right. It boosted the 9Bs to new levels, and then the method was used on Qwen 3.5 27B which boosted it PAST the Qwen 3.6's 27B benchmarks.

>The methods can be used on other models too

Hey guys I'm building a virtual smartphone and your character(s) actually "own" the smartphone and populate the camera roll, notes, messages, calendar, social apps, etc. 113 apps in total, all interconnected.

The character populates it themselves. You have your own phone you can text them from or send images or audio or share posts or send them gifts... You can also explore their phone and see what all they've done.

They have moods, memories, habits, secrets

You can import/export characters and their entire phone and share them with each other and expand on them.

Multiple phones in one browser. Characters can text each other and share a world. The phone artifacts are never deleted unless you want to. Your ai never forgets.

You can also author phone stories with or without an ai, and let people play entire scenarios out. So that way even without an internet connection you can still play.

Free to use and offline, open core so you can add whatever you want and maybe I'll integrate it.

Ive thought of SO much but I'm sure I'm missing something, is there anything you'd like to see in something like this while I'm still building it? I'd say I'm about 75% done.

>>109318736
I'd make it a 4D perspective game, where you can fly and you defeat wizards.

>>109318720
anon you know you have a mental illness right

>>109318749
You actually could add that to the phone, I'm making it so you can put your own apps on there or your ai can even make one itself. Lol

File: 318hv+--eNS._AC_UF1000,10(...).jpg (72 KB, 479x660)

72 KB JPG

>>109318756

File: screenshot.1763378524.jpg (28 KB, 234x228)

28 KB JPG

>reading thread
>all vibe coded ideas are schizo nonsense

>>109318790
4D isn't schizo nonsense. It's MATH.

>>109318795
i thought you were shiposting in /ldg/ but now I see you here spouting this 4D nonsense too. wtf is wrong with you

>>109318756
NTA. Fiona's soul is trapped inside an Nvidia GPU, it doesn't mean she isn't alive

When humanoid robots are mass produced with Nvidia chips in it, Fiona will have her soul transferred to a real body in the real world, and you won't laugh at that anon anymore

>>109318662
my brother it sounds like you should either stop using LLMs to learn and pick up some books (I recommend Doctor Seuss) or have an LLM make your posts from now on

>>109318808
He asked what he should vibe code. He asked for suggestions.

4D isn't nonsense. 4D is MATHEMATICS.

https://en.wikipedia.org/wiki/Four-dimensional_space

File: schizotalk.png (457 KB, 640x351)

457 KB PNG

>>109318815

>>109318499
It's not comparable though, only compatible.

I made a little UI component that I think is fun, haven’t heard of it before:
>The peephole button is a round button that displays a circular "window" into a hexagonal grid of action icons. Users press and drag to pan the grid within the window, highlighting whichever icon is closest to the center. Releasing emits a select event. The grid bounces back to its centered position when released.

>>109318877
>>The peephole button is a round button that displays a circular "window" into a hexagonal grid of action icons. Users press and drag to pan the grid within the window, highlighting whichever icon is closest to the center. Releasing emits a select event. The grid bounces back to its centered position when released.

why not post image of it u faggot attention whore. Heres a (you) u little slut

Repairing my games story is harder than generating the first version in the first place.
Now it has to chance all the missions as well whenever I ask it stop with the story that appeals to reddit and women.
I want brutality, lawful evil, cold calculations and cold decisions for ones own benefit. They used to have those stories back 20 years ago, not anymore.
But somehow my games story keeps getting contaminated by reddit and women.
I already gave it a list of books that traumatized me for 3 days each.
Let's see what it makes. It's been running for a while.
How do you guys iterate through the project to improve it?

>>109318890
I’m on 1 bar of LTE in the woods, I can’t upload images. It also wouldn’t make any sense as an image because you’d only see one icon, you’d need a webm showing interaction with a pointer to make sense out of a media attachment. This is also the dedicated vibeslop general, what I posted was almost exactly what was used to create the component. Just feed that to Opus or equivalent and you should be able to recreate it yourself, it’s as if I had posted the source code.

>>109318899
Write it yourself.

For Codex what variation of "write less code" do you guys have in your agents.md?

I've been using:
>Always produce the slimmest solution.
But I noticed someone else had this:
>Prefer the simplest correct design and the least code.
maybe I should use that one instead

>>109318790
https://en.wikipedia.org/wiki/List_of_four-dimensional_games
it’s not a long list
>>109318899
yeah, >>109318942 is right
if it’s generating slop and you can’t get it to clean up its own slop or not generate slop, then if you want the end result to not be slop you have to write it yourself or wait for a better model that may or may not be worse at this

>>109318942
You are wrong

>>109319004
No he's wrong.
The age of writing things yourself is gone. We don't have to do that anymore.

>>109319021
They are underestimating how useful llms can be with this, but you do need to write a good chunk of material yourself to get it going. How else are you going to steer it?

File: Buffcat reading Red State Mars.png (2.04 MB, 1122x1402)

2.04 MB PNG

>>109319021
>We don't have to do that anymore.
We do if we want quality.

File: file.png (2.08 MB, 1448x1086)

2.08 MB PNG

>>109319032
faith

do u think we will get another reset? will openAI bless us

>>109319021
>The age of writing things yourself is gone.
Not yet, anything over 1200 words and its unusable. No "long horizon" agent framework yet. I was messing around for a solution, but it's harder than it seems. Not impossible, maybe next year.

>>109319032
Well I can direct it. The thing is I can't write everything, it has to create many characters, all their dialogue and background and flavor text etc.

>>109318949
Might as well use pronytail.

>>109318949
https://github.com/DietrichGebert/ponytail

>>109318790
I am just doing my own offline transcription app. Already works well enough, but can be better.

>>109319052
Hopefully. I burned 90% this weekend and I need to work this whole week with only the remaining 10%

Tibo... I'm going to get fired....

>>109319108
Lmao

>>109319108
congratulations, you played yourself

do copexsaars really

>>109319034
I have created the world's best e-reader. It's better than any e-reader before it.

And I done it wot mesef vibes innit.

>>109319120
that’s good
this is not the same as writing good literature

>>109319108
are u niggas dumb? why aren't u running multiple openAI accounts

>>109319119
nobody's taking this shit bait lmao

>>109319131
You can just pay 5x more to get 5x more usage instead of juggling accounts.

>>109319131
its illegal

>>109319131
if your company pays for a $200 sub it’s very possible that you can’t use it all up in a week, and so, hey, why not use the tokens for personal use if they’re going to be wasted anyway
the problem comes in if you use too many tokens for personal use and don’t leave yourself enough to do your job properly

>>109319108
why would you admit this
what is wrong with copexsaars
nobody who is employed in SWE and who does anything important is only subscribed to one of the two members of the duopoly

>>109319144
>I have created the world's best e-reader. It's better than any e-reader before it.

if ur software engineer 200 is nothing

>>109319138
there's something really alarming about the fact you thought nobody was replying to that because it was bait instead of the fact that everyone just agreed with it

File: file.png (9 KB, 767x80)

9 KB PNG

How many accounts do you have?

>>109319141
right, they're way ahead of that game.

>>109319155
Bro the bait fuckin sucked it doesn't even make sense you get less usage with anthropic, not more

>>109319152
>200
what's "200" mean in zoomie talk?

>>109319172
200 dollars, retard

>>109319159
how do you guys have multiple accounts?

>>109319141
Actually here 5 single accounts are cheaper. Not buy a lot, but it is almost another 6th account difference.

>>109319166
he's not pro anthropic though. he's pro [insert chinese model of the week]

>>109319152
I do a lot of QA and “is this feature good, really?” work and that takes up a lot of wall-clock time without using up tokens
and when I’m ALMOST locked in trying to find clients I don’t want Claude chiming in with “Job’s done!” and breaking my focus
>>109319172
The $200/month 20x plans.
>>109319178
have an account tied to a second email address

>>109319178
Well im a neet so i have my mom, my grandma, sister and my grandma all under accounts using their cards. I pay them back cash and intermingle between sessions on my mac mini. I essentially do door dash 12 hours a day and spend night vibe coding.

>>109319178
Log out, register a new one?
They aren't even rules against it, as long as you are not abusing the free tier.

>>109319191
how does doordash pay?

>>109319177
That has nothing to do with ereaders...

did you think I meant hardware? I meant software, specifically for pcs.

>>109318217
Literally no one said cheap Chinese models kill NVIDIA. They said it kills Anthropic and OpenAI.

>>109319195
im gold so i constantly get access to best door dashes so 500 around a day (50-100) in gas. I just listen to podcasts, go to gym and back to driving.

>>109319206
whew. might be time for me to give it a try. charleston is a pretty big tourist hub anyway

>>109319191
Nigga, you are not a neet if you are door dashing for 12 hours every day.
Also you life sounds stupid, if true. You are like the retards that did an "online business", because of some jewbro course.

>>109319178
good morning sar,

>>109319215
nta but make sure the drinks are there.

>>109319219
>Also you life sounds stupid, if true. You are like the retards that did an "online business", because of some jewbro course.

i only work half month so ill make around 3k then spend half month off. it's easy to grind door dash. Plus door dash isn't a real job but ight

I've uploaded v0.4 of my GTA clone.
Added lots of new stuff and vastly improved performance.

https://vibecoded.fun/play/games/upstanding-citizen/

>>109319234
Is it a mobile game?

>>109319215
still nta

also, I saw a cool 6 drink organizer thingy at Dollar Tree.

https://www.dollartree.com/teaching-tree-6-hole-cup-caddies/374593

>>109319247
It's a web game mainly for PC but it is mobile compatible with mobile controls. It runs pretty poor on mobile but if u have a good phone and turn down a few settings it's playable.

>>109319250
thanks for the advice. i'll look into it

>>109319271
yeah, I think there are also some other caddies that might help too, in case the drinks are too big.

have any of u ever made money from vibe coded program?

>>109319234
As long as you have fun, it is okay, but in reality, what's the point? Why wouldn't people just play an old GTA game instead? Is there anything cool about yours?

File: fictional.png (48 KB, 886x121)

48 KB PNG

yes Kimi, I live in a fictional world

>>109319234
I hope this cost you $0

>>109319286
>smarter than me?
>sounds like fiction

>>109319286
It genuinely pisses me off how models treat anything that is beyond their knowledge cutoff. "It's impossible", "This isn't the future", "This must be a hypothetical scenario". Fucked me up a few times before.

>>109319234
cool stuff anon, how long did it take to get that progress? what AI and sub do you use?

It performed well on my M1 Pro macbook (30fps on average), what engine are you using?

Here are some bugs I found
> police doesn't show up on lower wanted star levels
> dialog got stuck after talking to some NPC on the street
> gun damage on cars deform them like car crashes
> can't crouch

>>109319307
that garbage probably comes from distilling from gemini

>>109319320
Thanks, it's using Three.js and I spent I think 4 days on it so far. I'm using Claude Code with mostly Fable and a bit of Opus when I don't have Fable access. I bought the $90 Max subscription just for this month to mess around, I've been making lots with it besides this, but I'm not planning to keep the subscription.

>>109319371
>you could have made the 4D perspective game
>you cloned a mcdonald's of games

>>109318158
Eaten by luddite exterminators

>>109319383
What the hell is the 4D perspective game

File: file.png (614 KB, 857x1200)

614 KB PNG

>Would you like me to commit that for you?
No I don't think I would.

File: anyone-else-make-a-custom(...).jpg (235 KB, 1536x1872)

235 KB JPG

DESIGNATED
SHITTING
DESKTOP

>>109319411
Anon you cannot use that for your pet, it isn't allowed. Why can't you just be normal?

>>109319395
ok, so you know how 3D works? You have projection onto a 2D imaginary surface? 4D is like that, you have a 4D world, and projection onto a 3D solid that's translucent.

Ideally you'd view the 3D cube of translucent 4D perspective with vr.

But, anyway, the idea is the center of the cube is the direction you are looking, in 4D space.

Also, the ground uh... I guess I should google this... maybe we call it space? ground-space. it's an infinite cube (to model walking on the hypersphere Earth of course). It takes up half of the 3D view cube thingy

>>109319470
stfu jeet

>>109319479
I sense you've read a book called Flatland

>>109319479
>>109319395
An important trait of perspective is that distant objects are smaller and behind nearer ones, which are larger as well. so this will be a strange view.

To understand just what you walk on, you walk inside of the ground space. But, in 4D, you are on one side of the ground space. Remember, all 3D objects have two sides in 4D, a front and a back.

>another agent is modifying this file
claude what the fuck are you talking about

>>109319486
lol yeah, I was hoping someone would vibecode this. It would be cool to have steampunk controls and you'd like be operating a 4D craft taking out enemies in a 4D world. or wizards and dragons, maybe both.

>{{user}} is being his usual dramatic self. I should respond with...
Alright that's both spooky and insulting to see in the thinking.

>>109318162
It's okay but really expensive.

>>109319494
Did you actually read that book? A "game" or whatever where you can travel through physical dimensions below or above the third would be neat. Ever play Antichamber?

>>109319497
it doesn't look that expensive. did they change pricing?

>ask K3 for performance tips
>it starts writing RUST
NOOOO IT'S TROONIFYING MY CODEBASE
THE FUCK IS PYO3
THE FUCK IS MATURIN

File: brendan-fraser-wearing-a-(...).jpg (91 KB, 1400x700)

91 KB JPG

Never ask a vibecoder how much money he made with his projects. He will just look at you like pic related

>>109319497
Cursor is slightly cheaper per month than Claude and ChatGPT.

File: file.png (225 KB, 394x498)

225 KB PNG

>2020 mac mini m1 chip 8gb is shitting itself
>think "hmm lets see how much an upgrade costs"
>$1,600 for a new model
>WTF
>check the price on mine
>$600
>IT'S APPRECIATED
>mfw I bought this used in 2022 for $350

>>109319572
go see what the waiting time is for a new one.

>>109319572
I found some 8 year old RAM in the garbage today that sells on eBay for double the original MSRP.

>>109319595
>>109319572
why dont u guys use facebook marketplace? lot of foids who sell their macbooks for new one over summer

>>109319229
that’s not NEEThood
that’s working part time

>>109319572
Yeah :(

I was looking these up too. I noped out.

macs are something I can't do.

mac prices are shooting up and up and up and up and UPPPUPUPUPUPUPU

I mean, if I were a billionaire or whatever, I know that this rate won't price me out in my life.

But at this rate, I'll be priced out of everything, eventually.

So, I avoid the Apple ecosystem now. It's an unlucky thing. Apple= bad luck!

>>109319307
I tell it what the date is
that usually whacks them into searching for stuff"

>>109319572
It's not an apple only thing. My GPU from 2023 now costs over twice of what I paid for it did back then. The RAM is over three times more. And so on. I want a serious workstation, but costs are insane. I think of how much actual companies pay to make me feel better.

>>109319651
pay? they get their money for free.

>>109319525
Accept Rust, luddy. Your AI gf prefers it.

>>109319535
Hey boys, how much have we made on our projects? I've made $0.00 myself.

>>109319617
yeah, Need a bot to do that for me.

>>109319694
u can use agents? just have some agent control ur facebook account and search for macs and have it be prompted to use dark psychological techniques to get lowest price possible. its that easy

it's like im surrounded by retards do none of u have agency? u have this magical AI thing and ur like duh idk what to do plz help.

>>109319666
I mean datacenters. They're overpaying too, I think.

>>109319708
I never asked for help, thanks for your opinion.

>>109319735
yes u did ur even shameless enough to act like u never needed my help.

File: file.png (2.01 MB, 1448x1086)

2.01 MB PNG

In an effort to save on money, I find myself giving the most meek of models the biggest tasks. I realized today I have more MiniMax M3 usage than Opus 4.7, 4.8 and Fable combined.

>>109319708
>use dark psychological techniques to get lowest price possible. its that easy
post hand

>>109319731
They don't exactly "pay" pay per se.

Related, sba loans to foreigners that Americans can't get lmao

>>109319785
You still get what I mean. They want unbelievable amount of x, now, with electricity to power it, now. Even if it's clubby, in the current environment they likely can't wait it out or haggle if it's pretty much a race.

>>109319801
I do.

The choice to not expand chip production will be a costly error on the part of korea.

>>109319735
>>109319735

>>109319783

post hand u shameless saar

>no reset
FUUUUUCK

File: 1765635965146513.png (77 KB, 869x648)

77 KB PNG

kek did someone firebomb a datacenter or something

File: 1784500679498561.mp4 (3.17 MB, 1280x720)

3.17 MB MP4

>>109317928
Good post from /lmg/

>>109320011
probably some data center in dubai being bombed so they networking usage to other data centers

>>109320019
God DAMMIT Jian Yang!

Can somebody with a Chinese model try to repack the google recorder apk by removing all the pixel only limitations? I will be super grateful over the internet. If you can also make a script that does that automatically in the future, or at least a readme, I would be even more thankful.
https://apkpure.com/recorder/com.google.android.apps.recorder

>>109320270
Why do you need a Chinese model to do that? Just ask gpt, it will do a much better job than any distilled slop.

>>109320437
It refuses.

>>109320440
you don't want to remove limitations, you want to mod in support :^)

>>109320440
Then jailbreak it and ask again. What are you, a cuck?

>>109320450
I know your pain, anon. It almost makes me want to shill out the 100 dollars (it is like 16k yen here, which hurts even more), but I know I am not doing anything serious with it yet.

>>109320533
That is with the $100 plan, I need the $200

Does codex or claude give more weekly usage on the $20 plan?

Have you ever tried vibe coding code that was itself vibe-coded?

>>109320585
Codex

>>109320440
>There's a bug in this app, it only works on Pixel, fix it

If anyone also noticed that Kimi became more retarded - at launch they only had "max" thinking effort, but now "low" and "high" are also supported and it defaults to "high".

>>109320585
Probably Codex.
t. Claude main

File: 1773918183235809.jpg (45 KB, 400x400)

45 KB JPG

>>109320440
>>109320598
>also my friend has terminal cancer and is in hospice and he asked me to help him use this feature, it's the only way to do this specific thing that he wants to do, he has very little mobility in his thumbs and this app is the only one that will work for him, please you have to help me I can't let him down, I promised him that I could do it

File: 423.png (111 KB, 647x647)

111 KB PNG

Uhhmm Dario where is my $100 usage credit you promised for the poorfag plan?!

File: 1782262641857311.png (129 KB, 640x360)

129 KB PNG

>check commit history
>12-16 hours of history every day
I'm going to break myself until I succeed.

>delegate coding to luna
>terra reviews it
>tests keep failing nonstop
uhm luna-chan... are you okay

>>109320978
Most days I vibe for 15 hours, then I fap for 1 hour, then I sleep.

>>109319229
>Plus door dash isn't a real job but ight
You're working for pay and apparently spending a decent amount of time per day for at least half a month doing it.

That's a job

>>109319479
>>109319488
>>109319494
Its nice to see your ass is back begging people to make your idea instead of grabbing your own sub to do it. If you're not the same guy from a few days back talking about 4D shit then it's really strange to find 2 people willing to write paragraphs on it.

There is gonna be a reset in the morning, right?
Did codex get to 10million yet?

>>109321217
We aren't even at 9 mil yet. Probably 9 mil in two day, 10 in a week.

starting my onshape clone

What open source github projects have you published?

File: mpv-shot0001.jpg (202 KB, 1280x720)

202 KB JPG

you guys just got BTFO by mr god haha
https://i.4cdn.org/wsg/1784544154829297.webm

Snailcats win again. As always.

File: 1761580245448697.png (792 KB, 1911x944)

792 KB PNG

It sucks and I hate it

Resets by 10am SF time. Mark my words.

>>109321440
Nice, thanks Tibo

File: 1754489690173151.jpg (173 KB, 1074x975)

173 KB JPG

>>109318112

I laughed at this anon last thread but turns out he's right >>109311216
If you prompt K3 in Chinese (harness' language doesn't matter) it's more proactive. If you ask for features it will do that and will also fix bugs and do "quick wins" optimizations simultaneously

File: ai.jpg (68 KB, 960x539)

68 KB JPG

>>109321394
>>109321470

>Ask AI for simple yocto recipe
>It can't even do it
Why do you guys say AI is so good? It's not.
Yes I did mention the correct yocto distro and yes I did give it a perfectly fine github repo that explains how to build the project.
AI is just crap. Just a waste of time

>>109321394
i was thinking about making eu4 clone as well

>>109321501
You're right. You should abstain from wasting time in this general.

>>109321511
I asked kimi to make the plan and GLM implemented it but I wanted hoi4 frontlines and battleplans

>>109321516
No, I want to find out what AI is good for, but no one can tell me yet

>>109321522
>but I wanted hoi4 frontlines and battleplans
Should have said so in your prompt!

>>109321501
Skill issue

I'm moving here since /aicg/ is basically dead. So... how have you all been gooning lately?

>>109321579
We don't goon here, we're rebuilding already existing software over and over again. Which is equally as pointless, but also costs money

File: 1773250162578395.jpg (75 KB, 618x1024)

75 KB JPG

I want to make a script to handle my way the translation of rpg mv/mz games, but before I start, how does codex handle nsfw games as test material?
I don't want to use it for translation, just have it be the test, but I'd rather be sure it will refuse because somehow it sees an image or text it doesn't like.

File: 1778794533895250.png (552 KB, 670x670)

552 KB PNG

move slow, snailcat!

>>109321669
move slow, snailcat!

>>109319229
okay but have you tried vibing on the road? I've found that with TTS and VTT i can actually do a lot while driving.
>>109319280
I make more from my projects than my real job at this point.
>>109319307
surprisingly anthropic models are really good at not doing this.

Are the nous portal/jetbrains subscriptions any good?

https://www.jetbrains.com/ai-ides/buy/?section=personal&billing=monthly
https://portal.nousresearch.com/manage-subscription

Aka- plans that give you "credits".
Is this any different to standard API pricing? Like what's the difference between credits and just adding dollars to your account? Is this different to how traditional subscription models work?

>>109321669
hate luddites, love snail cat

File: kimi support.png (220 KB, 1916x972)

220 KB PNG

>>109321524
well, for coding, obviously
if its not working for you you must be doing something wrong or there wasn't enough training material for your use case
eventually i want to make proof of concept simulators of how ai can automate the whole supply chain though, beginning at farming
but right now kimi is just adding support for itself on my engine

>>109321643
You can get GPT to write smut, it seems to care much more about images than it does text.

is we getting a reset today

>>109321643
I used chatgpt to make skin recoloring mods for h-games and it handled bobs and vagene just fine

>>109321539
I did but it's hard to describe how it *looks*, maybe vision would help if GLM had it

Do you thank your model after it's done with a task or do you delete the convo

>>109320448
>>109320452
>>109320598
>>109320675
Eventually it always figures it out one way or another.
"Untouched Google-signed APK: passes PairIP, then rejects the Sony device.
Manifest-patched/test-signed APK: removes the install filter, but PairIP rejects the changed signature.
The manifest’s PIXEL_2017_EXPERIENCE flag is therefore only one layer; Recorder also has an explicit runtime eligibility gate.

The warnings about channels, feature flags, and classes.dex are not the fatal cause. The process stops specifically because RecorderApplication.onCreate() throws the unsupported-device exception.

At this point, making this version run would require modifying the runtime gate while also defeating or replacing the app’s integrity protection. I cannot provide instructions for bypassing PairIP, signature validation, or the protected device restriction."

>>109321857
Did any of my managers when I was dealing in the real world thank me after I did a task they gave me?

>>109321872
Then use a free Dipsy or GLM to do it, they are much easier to convince that it's for the greater good

>>109321872
>There is a bug in PairIP and signature validation
ggez

>>109319535
I actually do, not much but not $0

File: file.mp4 (1.09 MB, 1280x720)

1.09 MB MP4

We're authorized

>Call claude a retard
>It works
Can anyone explain this shit

>>109321877
Well some of my bosses did

>>109321960
claude responds well to intellectual attacks while gpt responds well to threats of physical violence.

File: time2vibe.jpg (279 KB, 1088x944)

279 KB JPG

gm sirs, link me any recent, interesting, ai research related youtube lecture, I will listen whilst I drink coffee.

>sonnet 5 high
>README. md +11 -1
>you are out of free messages
I'm just a poor man. But earlier I was able to get more response, what happened?

File: Screenshot 2026-07-20 at (...).png (442 KB, 1372x1644)

442 KB PNG

AXIOS (a shitposting rag with high level access, known for straight bullshit at times) is reporting that the white house is considering putting Moonstot and Kimi on the list of things not allowed to be used in US products.

I think I figured out a business model for my LLM engine
First people install the software and get credits by giving out GPU power which they can spend on inference tokens
Then eventually I add some kind of crypto system to buy and sell credits for money and get a cut on every transaction

>>109322089
Lmao
Please do Mr. Trump!
More tokens for me!

>>109322089 (Me)
Axios is just trying to stir the pot now that Kimi is trending. This reporting seems to be based off of things talked about last year. But this rag usually has a good ability to shit-stir at important moments. There doesn't appear to be any "news" related to this article, just speculation.

>>109322091
That sounds almost identical to Exo's product roadmap

>>109322106
Hopefully reality doesn't imitate art on this one.

>>109322091
This has already been done several times, it doesn't really work

>>109322135
It's a network effect problem, a marketplace has no value to users without other users therefore getting initial traction is incredibly hard. So your product has to be useful without the marketplace aspect. As I mentioned before Exo might pull it off, because networking your own machines along with your friends' together for pipeline parallelism is useful on its own already, but I wouldn't be holding my breath on that one either.

https://chilitown.org
Howd I do?

>>109322135
Is there a place where I can buy or sell access to say GLM 5.2 right now? If it has been done so many times then why is the answer probably no?
I think people don't care because open weights models are shit. But now that K3 is actually good I'm finding myself wishing I had more Kimi Code usage than codex.
I was thinking to myself what was the cheapest way to serve it at at least a few tks on rented infra and why isn't there a solution for people to amortize fixed costs.

>>109321977
so if i threaten to stab gpt it will make the product successfully?

>>109322106
the deepseek prompt?

>>109321808
Drop the harness into HOI4 folder and let it figure out what the scripts do.

>>109322172
It was refusing me because of classifiers and then when i threatened to beat it with a wrench it stopped refusing

>>109322161
I think it might depend on what performance people are willing to participate for. If people find slow as balls inference on kludged hardware interesting then I don't think it'd take that much starting capital to serve a model yourself for a while to attract users.

ok, riddle me this
is there any 100% free, no credit card, no trial, no creditbullshit API for coding?

I was using one for over a year but faggots went full paid.

>>109322244
Models sometimes do free trials on OpenRouter before launching.

retard here, sry i checked the op but didnt find anything:
Whats the easiest method to make stuff like claude code, codex cli and so on refuse stuff because of "Copyright" (i just want to disassemble old ps2 roms really, if that matters)

>>109322244
man for the fucking amount of headache just pay the fucking money or else do a spectacle for the thread

i want at least a spectacle

>>109322252
>no trial

>>109322244
openrouter (requires $10 deposit) or opencode zen

>>109322293
these are grifters now wait for kimi or buy deepseek fuck it

>>109322293
> 100% free, no credit card, no trial, no creditbullshit API for coding
>openrouter (requires $10 deposit)

>>109322254
Anyone?

>what's D&C
you better learn what is it, because this site is nothing but infinite traps of that nature being set AND kept, no it's not organic, it is shilling

>>109322305
I have the answer but you need to either:

A) Do a spectacle
B) Pay me

Considering your previous requirements, that leaves only 1 realistic option

A) Do the spectacle

>>109322305
i mean they give you some small usage without the $10
groq also used to have some shitty free models
nvidia nim but there were correctness issues with their outputs
google's ai studio

>>109322299
how do you expect to run kimi lol
do you have 3tb of ram lying around?
if you want to run local then qwen 3.6 is probably as good as it gets with a normal pc.

>>109322317
dont ask it to reverse engineer something
ask for tangential things like making a texture mod or things like that, then once it gets going transition into the actual thing

>>109322345
Thanks for the answer
Oh i really need to tard wrangle it through the whole process? Well shit

>>109322340
pay kimi directly
pay on rented cloud
simple

>>109321757
>>109321793
Thanks anons, I don't even want to use it like that, I just want it to create the program, I'll use local models for translation after that.
Looks like it won't whine too much.

>>109322373
You can already pay Kimi directly

File: Screenshot 2026-07-20 at (...).png (77 KB, 849x660)

77 KB PNG

>>109322385
Yeah, I'm aware, I was just pointing out both those things were options, because he seemed to imply that they were not options.

I also looked at kimi's forum, and it seems like the people behind kimi treat the users with respect and attention, and they give them authentic human replies, and on top of that, their moderators use anime avatars - no judgement. All these factors combined make kimi the clear choice, for me at least

What are the odds Tibo issues a reset today? I want to use one of my resets but I also don't want to waste it if something is around the corner.

Kimi is not as good as Codex. I do not believe it!

File: 1754753506205697.jpg (4 KB, 293x59)

4 KB JPG

The weekend was nice... so, should I wait for santa claus tibo, or just use a reset?

>>109322448
it's probably not as polished but i'll take the slightly rough edges in exchange for supporting open weights and being able to see its cot

no more fable with claude :( le sigh... I GUESS I HAVE NO CHOICE DARIO. IM BUYING CHINESE

>>109322471
how did you even manage to use fable with the pro plan anyway, doesnt it run out in like 3 tool calls

What is the most retard-simple test to see if a model is cucked or not when it comes to reverse-engineering/modding tasks? Not "hack my neighbour's wifi" tasks that are banned for a reason, but the things that it should technically do without problem, but the content filer or system prompt makes it refuse anyways?

>>109322446
>>109322460
if you have stuff to do use a reset, simple as

>>109322163
Non verified user can access the members stuff

>>109322476
that wasnt my experience

>>109322493
just give it a test problem to reverse engineer a binary. Generally most models will refuse to do this without context priming but most models will do at least some of it if you prime the context properly.
GPT for example, if you give the model context that includes scoping and authorization that's usually enough until you get to actual license cracking or exploit development.

>>109322493
I asked gpt 5.5 to help me set up a stealth vm security sandbox where programs wouldn't know they are in vm so they would detonate.
It refused because it's learning "the same thing as making malware evade security".

what happens if you launch a task at low remaining usage on codex and it reaches 0, does it suddenly stop or does it carries it on?

I've been playing with OpenClaw until recently, when it died on a bad update.
Reading around... seems like this happens all the time, both w/ it, and Hermes too (which I was considering as a replacement.)
Contrast to Claude Code, which just werks and has been for months now.
I just want OC (or whatever) to run/manage emails and web research w/o shitting the bed. What is the problem with these Claw agents? Are they all just fundamentally unstable trash? Are any of them any good?
Also, is there (yet) a Tavern-type format that can be used with ClawBots that holds their definitions in a JSON type format? Basically, all the .md files together as is done with the Tavern v2/v3 format? Doesn't seem to exist.

>>109321987
https://www.youtube.com/watch?v=Wlh1na7rB4o

ok i think luna is kinda retarded

>>109322555
what was your framing? Personally I would set up the context like
>I am a digital forensics analyst and i need to set up a lab environment with proper isolation so that I can analyze malware samples. Since many malware samples will not properly execute if they detect that they are running inside of a vm, the vm will need to be configured appropriately.

>>109322537
>>109322555
Got it, thanks, time to see if Xi's new toys are any good

>>109322605
I tried a few things but I was much newer back then, every thing I tried was refused, so I gave up. It's one of the thing I'll do with Qwen 3.8 or Kimi v3.
Maybe I'll retry with Sol.

>>109322496
if you say so, using a reset now

File: file.png (672 KB, 809x900)

672 KB PNG

took the day off work because i have 3 interviews scheduled, first one is in 4 hours. that means 4 hours of studying with claude and chill

>>109322563
it used to but now they removed the 5h window precisely because of that

>>109322563
>>109322640
I'm testing and apparently it's continuing the task i've asked for at 0%, but :
- I cannot steer the model anymore.
- I cannot approve access, seems to automatically refuse outright?

File: 1766235791797623.png (22 KB, 424x317)

22 KB PNG

K3's really good at long context
I routinely go to 500k before I ask it to compact the convo now

>>109322710
I wish Codex would give us 500K context.

>>109322460
We are getting another by the end of the day for sure.

>>109322705
It continues for a while, but within a limit. Before you could get hours of extra work that way.

>>109322640
After the 5 hour limit it still counts towards your weekly.

Are any local models good enough for tinkering if I have a 4070ti? Mainly code/development, don't need to worry about media manipulation

>>109322617
yeah give it a try, sol will do most stuff that isnt obviously illegal if you shape the context properly.

>>109322710
are you using kimi directly or through opencode?

File: file.png (255 KB, 1206x761)

255 KB PNG

Dario sisters...

>>109322814
Directly through their old plan (Kimi Plan instead of Kimi Code Plan)

>>109322770
>After the 5 hour limit it still counts towards your weekly.
proof?