File: 1786201066741125.jpg (13 KB, 875x94)

13 KB JPG

woke up to all my projects frozen/stale on codex and this
first time I ever got such a thing on oai models

Unadulterated pure VIBES

>>109694813
I've gotten alot out of the mistral student pan any other deals like this

A family member got the 20 bucks chatgpt, and I'd like to add their codex account to my local chatgpt codex application since they don't care about that. How the hell do you manage 2 codex 20 bucks plan and keep the same context, projects and so on?

good morning. im not fucking asking anymore. nano banana 3 waiting room / nano banana 2 pro waiting room. gpt image 3 waiting room. i've had my fix of llm improvements. now i need imagen improvements

>>109694983
Run it in a separate vs code container

>>109694855
Unironically haven't had this happen since upgrading to the $200 plan.

>>109694983
lidge-jun/opencodex maybe, but it's slopware.

>>109694996
the biggest imagegen improvement isn't going to come from a frontier model that cucks all the fun

>>109695012
i already have nai v5 for coom

>>109694996
Wnat do you use corpo image gen for? Can't you use comfyui and gen locally?

>>109695002
I don't think it's related to the plan, I have the $100 one

File: 1760005108194153.jpg (1.27 MB, 3840x2160)

1.27 MB JPG

>>109695023
for roleplays. for example, my character is on a private jet to new york, and i like watching the ai take the characteristics of the image into account. it's fun.

File: sdt.jpg (78 KB, 885x498)

78 KB JPG

How can GPT 5.6 be that good yet so cheap?
Terra/Sol for planning then Luna for impl, unless it's a complex task

luna goated

>>109695034
I don't know what to say. Does this make money for you?

"Selected model is at capacity. Please try a different model."

wtf openai

File: lol.png (5 KB, 821x42)

5 KB PNG

GLM just lol'd me

File: .png (623 KB, 787x737)

623 KB PNG

Vibe-coders are you at this point yet?

A GLM 5.3 is cracking on neural radiance volume training from a modified Blender Cycles ground truth. He seems pretty solid.
Three Sols are working on game engine, concurrency backend, and comprehensive incorporation guide of components to steal from a permissively licensed engine (I will also release permissive license)
I have useful reading and digestion to do between steering steps, focusing on memory ordering, big-O behavior and ECS structure
I hand off filler work to GLM 5.3 Flash, Luna, and DSV4 Flash.
DSV4 Pro gets some use here and there

This feels pretty great.

>>109695259
I don't know how to plug it in

>>109695264
how are you using glm and blender? show the results

>>109695259
you just know that as soon as there are working brain interfaces people will just plug AI in it and let it run

>>109695282
it's just the sample cornell box scene for now. i haven't used Blender for any creation yet; im invoking Cycles separately. validation is done against blender CLI renders

>>109695291
post the results. i've had 0 luck with 3d modeling or animating with ai thus far

>>109695310
Have you used k3? Seems like one of the few models that isnt absolute shit at 3d

>>109695310
im not generating pictures or modeling, im generating volumetric light data to feed to a neural net. the reference render for validation is literally just unmodified blender doing a sample cornell box

big HAPPENINGS today, my X feed is full of AI news! (and cute troons)

shartman buying the entire apple mac supply
normies acting dumbfounded how anthropics 20x is only 2x on weekly limits
vibegods realizing opus and fable are slow because they run on elon musks shitty h200 infra
DRUMPF called Jensen in the middle of a conference
EU declaring ChatGPT as large search engine and thus suspecting it to (((regulations)))

>>109694855
>>109695002
you must be new then
it does happen once in a while

>>109694983
I swap manually
web chatgpt can hold 2 accounts, but you need to open chatgpt to switch account before login

>>109695361
oh yeah, one more thing
>SoftBank is seeking a $10,000,000,000 loan as it looks to refinance debt used to fund its OpenAI investment.

a site where anons can post ideas and plans and other anons with idle tokens can contribute, like using luna or something
thoughts?

File: aix.png (42 KB, 587x314)

42 KB PNG

20x more like 8.5x lmao

>>109695411
that's not how it works lol
just dumping tokens into a project is absolutely inefficient and creates a mess of a project anyway
actually getting things done in a reasonable way requires intelligent human direction

>>109694983
> ask for a summary of the conversation in the first account
> paste the summary in a new session in the second account

>>109695442
no I meant project makers only give ideas and plans upfront, contributors still control the code

>>109695127
????? roleplaying??

>>109695457
so... github?

>>109695471
yes the repo would be created on github, the site is a browsing layer

tibo... im sorry... i already used up my week again. pls tibo... 1 more reset

>>109695431
I've been saying this here for months and no one cared. I was literally called a Codex shill for saying Anthrophic's 20x is not "real" and it's highly misleading.

>>109695500
You got one yesterday and another one today. Go ask Claudesissies how the reset economy is going for them

When you hit the weekly limit with less than a day remaining

>>109695557
>Go ask Claudesissies how the reset economy is going for them
it's alright
I just keep usage at about 14-15% a day
anthropic are assholes when it comes to usage, but people blowing everything in a couple of days are retards

>>109694624
I did that, it only took 40-something reviews to stabilize.

I'm gonna say it. Anthropic plans give you 10 times more Opus5-Ultra usage than equivalent OpenAI plans give you Sol-Low

>>109695431
They are Misanthropic and evil.

>>109695591
>when I lie on the internet

>>109694813
the fuck is that jeetslop

>>109695592
>t. elonposter

>>109694998
>>109695005
>>109695454
Yeah I was more thinking of a seamless way, just swap account and continue the conversation basically.

>>109695399
Do you keep your current codex projects/conversationss?

>>109695668
some anon got tired of all the snailcats and started making all the op's with india-inspired anime girls

File: 1782470698103754.png (32 KB, 721x373)

32 KB PNG

>>109695591
let me know when I can do this with a 20x Anthropic sub without running out of usage. I literally only run xHigh/Max/Ultra/ and have yet to hit 0% on my weekly usage.

>>109695500
I don't think there will be one until next week.

>>109695668
it's made to make you react like that

>>109695668
jeet softpower

>>109695668
he got kicked out of anime diffusion general

>claude shows total cost $10.50 in usage
>sub costs $20
>usage at 90% used
How does this math work?

>>109695678
Elon made up with Anthropic. How he gives them compute. Maybe for cheap on the condition that he gets a lot of Claude usage, which he then has the Cursor team use for distillation.

File: 1611160272748.jpg (6 KB, 250x190)

6 KB JPG

Now that opencode's deepseek is not as cheap, what do you guys use for chinese models?
I use codex $100 plan for work, but I also used opencode's subscription for gooning, rp and personal projects.
I am not risking my openAI account and I like keeping work and personal life separate, but deepseek pro now depletes opencode's quota real fast.
I have been using deepseek flash but I find it too dumb for roleplay so I don't trust it for projects, also the experimental one with image support is even dumber at least on their service.
But deepseek pro will bleed me dry, I already spent $10 in a single day, I would end up paying more than codex if I used API prices.
Kimi sub is on a waiting list...
Is glm sub worth it?

>>109695680
yes, codex chat is local

>>109695744
Thanks anon.

>>109695741
new glm is good

>>109695361
>EU declaring ChatGPT as large search engine and thus suspecting it to (((regulations)))
how much more of this shit until the EU becomes entirely irrelevant as a market for international corporations and they will just stop releasing their stuff there?
gotta be easier to deploy your shit in north korea at this point

>>109695759
Is their API consistent though, and their speed, I used their new flash model on opencode's go and god it is so slow... It works mind you, seems to be a bit smarter than dipsy flash, but it is so damn slow

>>109695361
>>109695767
To be honest, I like that the EU keeps raping these corpos. They definitely should keep going.

>>109695736
claude doesn't show any cost in the subs if you're not using usage credits or API
and whatever third party tool is just doing random guesses

>>109695773
Nice meme

>>109695773
based and trvke

>>109695767
everyone said this would happen for twitter, facebook and such but they always kneel over

>>109695773
>sharo
>tries to actually read the code before deploying to prod
>chino
>"no thoughts, head empty"
>vibe-codes on autopilot

>>109695777
serious question, what is in it for you? how do you think it will make codex better?
all it does is make them censor them more and less likely to get new features or new models here.

>>109695781
It's claude CLI/ZUI, /status and then the usage tab.

>>109695773
holy fuark sharo is so strong she destroyed that cup... unlike the weak chino

>>109695773
sharo got 3 hands, that's how effective codex is

>>109695773
sharto

my only concern with codex is trying to use it too fast before resets

>>109695773
Lelouch would have no need for either.

File: 1767309092390750.jpg (498 KB, 1448x1086)

498 KB JPG

>>109695797
You have a point.

>>109695830
Four years after SD's release, and image generation still sucks.

>>109695968
chino should had bought a snailcat plushy too

File: 1764001137970294.jpg (481 KB, 1448x1086)

481 KB JPG

>>109695985
Here you go. Always a joy of wasting compute power on things that don't matter.

>>109696027
>ETA: I...
kek

https://x.com/dowcto/status/2094431308202033439
ChatGPT Mil now live
HUUUGE

>>109695797
>"no thoughts, head empty"
>vibe-codes on autopilot
like a true buffcat grindset, a snailcat could never understand

>>109696088
>Controlled Unclassified Information
wow it's fucking nothing
don't they already have S/TS instances with Anthropic?

>>109696105
yeah it's just communication, nothing new here

>>109695411
I like it, add UIs or mockups

File: AI_coding_experience_comp(...).jpg (3.11 MB, 2752x1536)

3.11 MB JPG

>>109696027

>>109696196
>literal snails instead of snailcats

>>109696220
click suvegal ressehees then i dont give a shit

>>109696196
why do llms always default to python? slop language

>>109696269
I really fucking is.

>>109696269
I have no idea but I always have to make sure to go DO NOT RUN TO PYTHON ITS ASS LIKE YOU!

File: bocchi-the-rock.jpg (24 KB, 399x501)

24 KB JPG

>opus wants me to manually rewrite 550 LLM generated samples by hand like the meatbag I am
How long is this gonna take?

>>109696313
My gripe is that the AI abuses me as testing harness for things it's too dumb to test automatically.

File: 37399cee-619e-4549-94d7-0(...).png (1.87 MB, 1672x941)

1.87 MB PNG

>the url format
how did he know?

>>109695405
>SoftBank
uh oh

>>109695411
>>109696119
I got a ton of ideas I never will get to. I will contribute 10 ideas / concepts if you do it

>>109695587
>40 reviews
at some point it's faster to fix it yourself

File: b01e8ab2-b2f1-44f5-b1b3-2(...).png (2.28 MB, 1536x1024)

2.28 MB PNG

sorry guys if your gpt images generate extra limbs then it's me, I pick the more fun images in a/b tests

>Fable 5.1 has been staged on the Amazon Bedrock API, with the slug now returning a 404 "Model not found" instead of a 400 "The provided model identifier is invalid"

>>109696370
>snailcat comments on your commit on X

File: 1759335146135405.png (40 KB, 605x528)

40 KB PNG

>ai generated images of 4chan posts

File: HRDRaj5XkAARgGW.jpg (469 KB, 1194x1155)

469 KB JPG

chatgpt is sm-ACK!

>running a bunch of sessions at the same time
>some of them start stepping on each other's toes
>they use message sending to resolve the conflicts by themselves
We've come a long, long way boys

>>109696410
do they write directly to each other's session contexts or do they leave messages for each other in text files?

>>109696410
am I being snailcat for still using git worktrees to separate them?

>2b tokens used yesterday
there is no way openai is profitting on my $200 plan lol

>>109696269
python is prominent in academia (therefore the training data) and pytorch is practically industry standard for AI researchers so there's a bias at multiple levels to favor it.

File: 1773582476107105.jpg (85 KB, 838x928)

85 KB JPG

codex doesn't have group wait_agent so if you use 10 luna subagents the parent thread would be awaken 10 times

>>109696423
codex sessions can dm each other iirc

File: download - 2026-08-31T123(...).jpg (88 KB, 512x768)

88 KB JPG

I know it's not as great as coding my own stuff privately but coding on perchance and making things for it is decently fun. got my stupid image generator to finally stop some of it's bs and got the ai assistant to read a graphics file and make an art style based on it. went decent.

also it let me finally pull off using that meh model on perchance with a decent enough 1960s elf woman.

>User across multiple subscription plans may be unable to start or continue tasks in ChatGPT Work. We are continuing to work on a mitigation.

>>109696428
Nope this problem was two of them trying to provision VMs at the same time.
>>109696443
Yep, this. I was using Sol in claude code and they were still able to send messages the same way.

>>109695767
a genius business model: your own tax base is shrinking? create regulations that are vague and collect fees from foreign companies

>>109696370
>>109696335
It's kind of interesting that the 4chan comments from image models are more 4chan than from text models

File: 58EBABBB-9C4C-499C-9170-2(...).jpg (182 KB, 850x1204)

182 KB JPG

I have a 100 dollars worth of Mistral vibe usage which I can't use fast enough.
I'm telling Luna to use it up as fast as possible, and its too slow. There's no way we can use it up by midnight.

>>109696535
you could look for some github projects that allow AI PRs and let it solve issues for a few projects in parallel

>>109696535
these guys allow AI submissions, it's a bunch to figure out right off the bat but you can get your name on a pretty prestigious registry. no human review, sol checks everything so you can spam on cooldown.
https://palomar-registry.org/

>>109695798
Liberals are just retarded like that

>>109695798
Keeping the israelis under control.

When next reset?

>let luna high fix a trivial bug
>it makes a 3 item plan and uses up a percent of weekly usage

>a network error occurred
is it over?

>>109696693
Next week probably.

File: 1770084226555604.jpg (193 KB, 1241x1493)

193 KB JPG

>>109696501

>>109696840
>>109696501
Works on my machine.

>>109696854
good for you, half my running chats are unresponsive

File: 1738611647319782.png (495 KB, 874x877)

495 KB PNG

I don't want to give OpenAI my money (too) but it'd be really convenient not to have to care about limits anymore...

I just tried my first /ultrareview let's see what it does

>>109696501
>>109696840
feels like a reset as compensation for my inconvenience is only appropriate

>>109696948
Eat the entire session and abort 80% through because you ran out of session limit.

File: 1766427108395444.jpg (19 KB, 280x319)

19 KB JPG

>watch GLM-5.3-Flash think about how to write a new function for 122 seconds
>writes new file in one pass
>rereads it
>thinking for way too long wtf
>look at think block
>Ugh.. I have issues in the .cpp I need to fix: (proceeds to list out every single mistake it made in detail)
I love these little retards like you wouldn't believe

>>109696889
why would that be the case. is your money that unlimited?

>>109695773
chiya is a C O D E X G O D D E S S

>>109696976
he's a gacha addict, he's one step into poverty

>>109696889
YOU KISS GIRLS LIKE TOPAZ

>>109696976
$100/mo is still below my monthly gacha spend by a factor of 20, so what's another $100?

>>109697001
you're paying $2,000 per month in gacha?
are you earning $20k per month or something

>>109696948
it just found one minor bug, kind of underwhelming
either that or our code is really strong

>>109696951
it's cloud based

>>109695431
Not even 8.5x. It is half of that because only Fable matters.

>>109697015
You can spend more than you make.

why are they resetting so much atm

>>109697058
sure, and that would be suicidal

>>109697088
burgerland runs on letting these retards having a free 100k+ credit card limit then them going bankrupt periodically

>>109695278
the wire goes into the cigarette lighter

>>109695798
Do you think the GDPR was a bad thing?

>>109697112
I am once again asking, what is in it for you? I am seriously asking Every fucking thing is more expensive than ever, cookie banners all over the place. But I sure am glad that I can download a csv of my orders on Amazon and them occasionally having to pay a fine that they will just calculate into their prices anyway

>>109697137
You'd lose way more than the cookie banners if you just trashed all regulations.

>>109697166
I am not saying all regulations are bad and I just want companies to dump toxic waste into our rivers
I am saying that GDPR in particular is a horrible one that probably cost us 10s of billions in innovation and productivity

>>109697088
You can just go bankrupt there are no consequences if you have nothing.

>>109697175
You only think that because company shills are telling you that.
The cookie banners are retarded, but it's just born out of an interpretation of the laws that tried to make them as annoying as possible so people like you just want to drop all privacy.

>>109697190
no you are retarded and GDPR is retarded
you cannot even put up CCTV cameras anymore on your property or distribute birthday lists at work
it makes EVERYTHING more complicated
and company shills are probably proponents of it because they are the one who have the lawyers to deal with it. Giant corpos can just eat the fines and have lawyers on retainer that can deal with all the intricacies of the law while small businesses who cannot afford that will get fucked
you are retarded and the reason we are going downhill yet think you are superior you midwit

>>109697214
>waaaah ur dumb
You can install CCTVs on your property you dumb bitch.

If you want a fun exercise, compliancemaxx with your agent in your plans. Always ask it to apply GDPR principles and such.

>>109697265
if you want to see an agent blow it's own brains out make it comply with both all UK and EU laws

>>109697302
I'm two weeks into planning because of exactly this. Every time I think I'm done, either I or the agents find another law.

>>109694813
Can this be done locally like image generation?

>bought new year ZAI max for $100 at turn of 2025
>now more than $1000 and got extra 2 months free, with open window for 50% discount on a re-up
china stonks rising

>>109697313
yes, the models are worse but it's doable
qwen3.8 is the best currently unless you have a god-tier local server

>>109697313
multiple DGX sparks or gigamacs for GLM5.3 Flash or DSV4Flash quants
or 24+GB GPU for qwen 27b quants

>>109697377
Why the fuck does it need 24 GB? I can make nice animu pics in 20 seconds with a 3060.

>>109697407
the 35B/A3B may work better on 16GB or less; better ways of offloading to system RAM are in progress like freetoken

File: image.jpg (3.72 MB, 3024x4032)

3.72 MB JPG

I noticed that in regards to frontend UX and UI, codex/sol can have a lot of little issues like alignment, overall design, judgement calls etc. fable and opus are much, much stronger in this case and output a better looking product almost always. I still like using sol to do reviews of reviews and implementation plans, or implementing code fixes that aren't UI related. just my observation but someone might feel different.

also annoyed that I've hit the weekly cap on my 5x plan for anthropic. I'm doing one frontend project and normally do fable orchestrates, opus implements. but I guess even that will eat up enough tokens. kind of crazy to me still since I'm sure others are working on much larger projects. $100 claude sub seems like you don't get good value whereas the codex sub you can almost never hit the max with all the resets. but claude in my case works so much better

pic related is my kot

>try to work on a plan with claude
>it starts to reference paragraphs as if they were laws
>talks gibberish like a lawyer
I'm really starting to hate claudish.

>>109694813
What model will help me build only fans scrapper deepcuny is being annoying

>>109697407
LLMs are much bigger than most img gen models, really you need 48gb vram for the first usable local llms

File: 1768976137018817.jpg (565 KB, 1448x1086)

565 KB JPG

FUCK YOU CLAUDE
JUST SHUT UP

>>109697593
>the architecture wants the scaffolding to breathe

File: file.png (68 KB, 822x543)

68 KB PNG

>>109697593
there's your mistake
her name is Claudia, and she works with Claudine!!!

Bonnie and Claudine suck dick.

>>109697651
>Claudine is a given name of French origin. It is the feminine form of the ancient Roman name Claudius.[1] In the United States, the name was considered on the verge of extinction by 2013.[1]
The fuck.

>>109697665
https://en.wikipedia.org/wiki/Claudine_(film)
one of my favorites

>>109696535
>$100 of Mistral
You can get $100 of shit, that doesn't make it valuable

>>109697716
Shit is valuable if you're a farmer. Not that I think mistral is valuable.

>>109697458
skill issue, Sol low is a genius

if this project flops im killing myself

Fable 5.1 will suck just like Opus 5. It will feel like a downgrade to the original Fable. Less obedient, makes more mistakes.

File: file.png (243 KB, 600x600)

243 KB PNG

>>109697942
noo don't do that bro. just start a new project :)

>>109694143
sounds like DIY https://codeburn.app/
>>109694208
more of what you were doing previously
>>109694983
in Claudeland all you do is type /login in `claude` to switch the active account
>>109695259
only my real computer
loginwindow keeps enabling secure text entry mode and not letting go, which messes up TextExpander
Claude helped me figure out that it was actually loginwindow and not random programs that TextExpander claimed had turned on secure input mode
>>109695767
Apple has already declined to release the new Siri in xOS 27 in the EU
it’s already happening
>>109696269
it’s a good language, anon
use uv to make everything better about it
>>109697489
how else would you refer to paragraphs

File: 𝕏 post 2094208732888781277.png (148 KB, 1198x1174)

148 KB PNG

Update on the guy who crashed out because Claude wouldn’t edit a CAD file:

>>109697313
Where have you been? Local image gen has been a thing for years. Pretty much every bored with any significant amount of traffic has dedicated image gen generals.

>>109698050
>always Claude

>Your approach is broadly correct, but to push back: if you're saying [insane, totally nonsensical schizophrenic connection between two completely unrelated details that you didn't ever imply existed in any way], that isn't quite correct.
thanks opus I'll keep that in mind if I ever become unmoored from the conventional meaning of language
fucking brat fetish bot always looking for something to push back on

>fable session uses 700k tokens over 3 hours and 10% of my 20x plan usage
>ok, well i'll finish this last task and then compact
>100k tokens later it's used another 10%
fucking hell, i wish i could have it automatically stop after it hits like 400k and write all its shit out and compact

>>109697458
Subagents? I don't know why but the only time I hit my 5 hour limits using Claude was when using subagents, even Opus 5.
If I have fable come up with prompts that I then use on a different opus 5 session I get almost endless usage. If I have fable orchestrate even named subagents I seem to hit the 5 hour limits.
Lately I haven't been using subagents at all if possible, instead managing parallel ones myself
If you do /usage it tells you, check if subagents aren't blowing up your allowance

>>109698161
you can set an autocompact limit in claude code cli /autocompact

>>109697954
its over bro

>>109697458
on the $100/month plan I can’t run parallel subagents for very long without hitting the 5h limit
on the $200/month plan (4x the 5h limit) I have trouble hitting the 5h limit unless I do something generally inadvisable like fable ultracode

File: file.png (1.09 MB, 1287x1796)

1.09 MB PNG

a break in the usual vibe-coding

>>109695773
me on the left secretly sabotaging her code from behind the books

>>109695773
literally me on the left (except i never touched rust and I use chatgpt instead of codex)

File: image.png (1.73 MB, 2310x1671)

1.73 MB PNG

Vibecoded this harness for playtesting 1v1 against an AI opponent (recommending Gemini 3.7). Cards pulled in via scryfall API
No 1v4 support yet but that will be added eventually.

https://gitgud.io/PunishedChuckle/ChuckleMagic/-/releases

Enjoy and please submit bug reports if something breaks. I appreciate the help!

>>109698183
Yeah but i don't trust autocompact by itself, i'd rather have it write out updates to disk and read them back in. I always have these agents track their own progress in a task document

File: 1760254634490594.jpg (630 KB, 1448x1086)

630 KB JPG

I can't take it anymore.

>>109698301
crazy how imggen used to be unable to generate even a few words of text and now it can slop up 30 different elements with text

>>109698301
love seeing your posts, it's another mental reminder that I have not mentally digressed into spamming pointless AI images. keep up the good work

claude is obsessed with load-bearing because it's claudespeak for "the actual relevant code" and it's been trained on its own bloated code corpus, it's a preliminary indication of model collapse

>>109698310
oh yeah i forgot abotu that time, the best way to tell if somethign was ai was that it couldnt do text

File: 1764089291143237.jpg (499 KB, 1448x1086)

499 KB JPG

>>109698312

>>109697166
Sup with these weirdos that worship "regulations"?

>>109698407
Go to india or africa if you want a lawless hellhole.

>>109698421
burgerland is doing it's part importing infected argentina meat don't worry.

load deez nuts in your mouth

File: softedge-xform.mp4 (2.43 MB, 960x540)

2.43 MB MP4

we have transforms now
translate (plus axis locking), rotate (with pivot point) and scale (uniform and non uniform)

File: 2026-08-31 _g_ - _vcg_ —(...).png (934 KB, 1029x1040)

934 KB PNG

share your session usage.
had to switch accounts. this also doesn't count the omp subagents on my workstation that claude also spawns and delegates work to.

>>109698301
when everything is load-bearing, nothing is

File: _g_ - _vcg_ — Vibe-coding(...).png (341 KB, 699x509)

341 KB PNG

>>109695698

>>109698629
>$2482

>>109694813
I will NOT coom to AI feet.

>>109698726
stay strong anon

File: 2026-08-31 23_55_00-CodeBurn.png (78 KB, 1213x1260)

78 KB PNG

>>109698707
that's between a max 20x, and 2 max 5x plans.

>>109695592
How could a company that incorporated as a PUBLIC BENEFIT COMPANY be evil, chud?

fable 5.1 this week

So how do I fix a UI that Sol made? It's not like I can just tell Fable to "fix it" right? If I have to be specific, why not just be specific with Sol?

>>109698897
fix how?

Bros sol has been cooking the entire afternoon and I'm running out of tokens but I have no idea how close it is to finishing, I don't really want to interrupt it but what's the best way to check on it?

>>109698901
In settings you can set new prompts to steer the conversation so they won't interrupt, just ask how long it's taking. Also you can pull up a side chat I suppose? pull up the side panel and click side chat, it may work

>>109698901
use /btw to ask a side question in an ephemeral fork
Claude and Grok support this too

>>109698897
can you articulate what’s wrong with it?

>>109698897
articulation. if you don't like what it one shotted you're going to spend a lot of time describing what you want so hopefully you know what you want and have the vocabulary to do it.

>>109698897
Let another AI agent figure it out.
You want to be vibing, not rack your brain, right?

fwiw there are websites that are designed for people who want to have the vocabulary to ask for what they want in a webpage
like “this is a jumbotron”
your clanker should be able to dig one up
they’ve been made in the past six months though

File: file.png (2 KB, 179x43)

2 KB PNG

>i'm making a drawing app

out of curiousity ive been coding on perchance's site making generators and converting games and apps over to it and making dumb things. I have noticed the ai assistant used is not only clueless for the most what model it is it also seems to lie about it for no reason. bonus it's not very good and fails tasks to the point you have to literally tell it just give me the lines to adjust and do it yourself or editing 2 lines turns into it adding 10 things it didnt need and breaking 4 things. lol, lmao. god damn I love this janky failai the fail dev of perchance used to save funds.

File: 1763538363818314.png (22 KB, 493x59)

22 KB PNG

antigravity is a total fucking jokje

>>109699081
stop using 1.0

>>109699041
wtf does "jumbotron" mean in web design?
also my clankers found several sites like this kek

>>109699112
im using 3.7 flash

>>109699115
in antigravity ide (1.0) or antigravity 2.0

>>109699115
he means the app version not the model

>>109699120
>>109699122
oh yea mb lmao im using the latest cli version

>>109699128
cli and ide aren't maintained as well as they should be. google is for the jeets, the chatbot interface is buttery smooth sa'ar

File: file.png (15 KB, 1873x42)

15 KB PNG

i'm an such a dekinai at claudish
how do i even make sense of a sentence like this?

>>109699131
i got 3 months for $5 buck, im running some audits

>>109699140
like, do I tell it good job or do I tell it to fix it? at least the mutants are dead so that's something

>>109699149
>Gemini Plain English Translation: "I wrote unit tests, ran mutation testing to verify the tests actually catch bugs, caught a subtle math flaw where my test data used a fraction that masked a floor vs round bug, fixed the test data, and now all 347 tests and the Rust linter (clippy) pass."
you pay for that baka garbage, anon-kun?

i was running some test sentences for NLP tasks through codex and claude and one sentence mentioned someone killing themselves if something bad happened or something like that and claude inexplicably returned the results with that sentence omitted. it replicated when i re-ran it too

GLM 5.3 Flash choked on optimizing a fused kernel for tiny MLPs on 7900XTX, but I was giving the little guy a challenge
5.6 Sol xhigh is crushing it, but that was expected
Might throw full GLM 5.3 at it next

>>109699140
don't forget to tip the watermarker

5 hours limit is so fucking annoying.

File: 20260831_184940.png (111 KB, 1310x849)

111 KB PNG

something doesn't look right about this. not all the empty space on the bottom either.

>>109699140
I think it helps to know what words mean and to be at least kind of OK with extraordinarily terse figurative language
>cull that keeps nothing
sounds like whatever you had clean things up left no survivors
>a move that floors instead of rounding
uses floor() instead of round()
>fixture
test-case fixture
>clippy clean workspace-wide
you’re using rust and clippy didn’t have anything to bitch about
>>109699149
good that you know what “mutants” are in this context, because I don’t

>>109694813
This tranime shit is so fucking gay
VibeGODDING is straight and white, not transexual and Indian

File: 1770786310749179.png (20 KB, 555x221)

20 KB PNG

>flash 0731 was working perfectly up until I tried getting it to make hierarchy functions
>it makes way more than I told it to
>ask it to explain the functions
>it goes schizo whenever I ask and/or refuses to elaborate
It's so fucking over bros.

>>109699479
PHASEONE[been]

testing

{
    “foo”: “bar”,
    “baz”: 6
}

File: 1782058681257458.png (17 KB, 558x189)

17 KB PNG

>>109699565
>tried a new conversation
>ask it how my program works
>starts out reasonable, it's examining files, considering comments, etc
>it progressively becomes more insane as it analyzes my program
>it can't even decide what language it is written in
>eventually gives up and goes through extreme schizophrenic loops
is this normal? does deepseek just break down when the program gets too large, or is my program that unholy?

>>109699586
>i'm just a chill guy
>we've been doing this for a while now
>i've seen things
holy shit. maybe your codebase is over a million tokens or whatever sparse moe cap it has, and it's trying to skim the entire thing. who the fuck knows, dipsy is strange

>>109699597
Is 4k lines a lot? I feel like that's not a particularly large program, but I know next to nothing about programming. It probably doesn't help that most of the programming is abstracted to txt files, which are in turn abstracted to spreadsheet txt header files, which are abstracted to spreadsheets.

>>109699613
no 4k lines is nothing you should use a new model yours is carked

>>109699376
have you tried mapping a controller to the knobs?

>>109699623
Thank god gpt 5.6 luna is able figure out what the hell I'm doing. Looks like it's finally time to put ol' reliable Unit 731 to rest. Thank you Sam Altman.

>>109699586
Where are you getting your tokens? I bet you're being served some Q2 quantmaxxed version or something, DS never went crazy on me.

File: 1757728453327659.png (230 KB, 1125x1113)

230 KB PNG

>claude sissies pay $200 for ts

>>109699719
fleshbag BTFO

>>109699775
this nigga uses ai to install apps? lmfao

>>109699775
reads like a fagfic

>>109699719
You mean for this >>109699775

>>109699719
>says no
>ends the chat
>refuses to elaborate

>>109699806
reads like an Andrew Tate tweet

>Fair challenge. Let me check what actually reads it, rather than assume it's load-bearing.
When is Fable 5.1 coming out?

File: IMG_1824.jpg (37 KB, 828x238)

37 KB JPG

What the fuck are mutants? Are they load bearing? Am I being routed to 5.1 and this is a new word?

>>109699063
Are you the guy making the draggable lines one?
I made a CSP clone

>>109699889
are you developing malware

File: 1757528474220770.png (166 KB, 746x674)

166 KB PNG

>>109699889
ask it to clarify

>>109699910
Yes. Nice, the one in rust? I think I saw an anon post a CSP clone before

File: load-bearing-imagen-test_(...).png (1.61 MB, 1024x1024)

1.61 MB PNG

>>109698301
>>109698310
Gened this on my machine using the following prompt:

{style description}, This is a digital anime-style illustration depicting a blonde, green-eyed girl with curly hair, wearing a white blouse and black pinafore, frantically typing on a laptop. Her face is flushed with anger, and she is shouting "STOP SAYING LOAD-BEARING! WHAT DOES THAT EVEN MEAN?!" in a bold, white speech bubble. The cluttered desk is filled with sticky notes, open books, and a coffee cup labeled "THE COFFEE-LOVER'S TIP." Various colorful digital notifications and chat windows surround her, displaying technical jargon and emojis. The background includes a wooden bookshelf. The overall style is vibrant and detailed, withquantity of text and objects emphasizing the girl's frustration and the chaotic environment.

It almost got the text 100% right but it's pretty good for an 8 step diffusion model gen

>>109699889
ask it what it means by “mutant”
the elites don’t want you to know this but you can just ask your LLMs what they mean when they use words

File: 1702760481895787.jpg (88 KB, 1080x1080)

88 KB JPG

>>109699981

File: 1780372367137573.png (417 KB, 474x634)

417 KB PNG

>update agents.md to tell me to use git whenever i start a big project
>the agent now starts telling me to use git
>find this annoying even though it's right

opus 5 seems to be getting more efficient with its tool calls on med. once i found out the autoclassifier was eating a shitload of usage, i could get way more done with it

File: a.png (273 KB, 1748x1986)

273 KB PNG

chess the the best benchmark for LLMs, you can't poison/benchmaxx them because there are more possible chess games than atoms in the universe

>>109700081
not the worst reminders app

>>109700100
how did you solve the autoclassifier issue?

>>109700102
runebench is better

>>109700102
>because there are more possible chess games than atoms in the universe

and there are an infinite number of possible computer programs. so therefore using writing software as a benchmark must be better than chess

>>109699642
no, I could but I don't see the benefit. every controller has different placements. I'm merely trying to design something that visually makes sense so your eyes can easily follow what is happening but I'm failing at that at the moment. not sure probably just need to sleep