general

[verbal nimbus]

Happened again at exactly 4096 tokens:

[robust yoke]

https://tenor.com/view/long-tears-gif-11087877242190144149

[young barn]

Hi

[robust yoke]

Greetings.

[fiery gull]

Hi

[errant horizon]

Hello

[robust yoke]

Greetings.

[fiery gull]

This is strange

[young barn]

I’m trying to generate a video once I do the steps and it says it generated the video I see no video or anything

[fiery gull]

Removing style control

What style control is?

[cunning hawk]

why my sesion failed and now the site loading so slow?

[fiery gull]

For me too, is server crasher I think

[fiery gull]

Reset the discord

[cunning hawk]

i see

maybe site overloaded right now

[round basin]

What is the maximum duration of a video that I can create

#share-prompts What is the maximum duration of a video that I can create

[fiery gull]

Just a test, 8 sec

[round basin]

Someone managed to make a video of more than 8s?

[cunning hawk]

generate multiple and combine

[round basin]

ok thanks

[fiery gull]

Now the glm 4.6 is the best open model from lmarena

[cunning hawk]

is it

[dusk torrent]

اهلا

[final tide]

Young Indian man (Aarav), around 26, in jeans and casual shirt, buying flowers from a street vendor, soft expression, realistic lighting, same face as next images.

[brave cloak]

Can someone help? I need an image edit prompt, that uses image A as a reference image, and fixes any differences and deviations, that image B has from image A, EXCEPT the camera angle and any PHYSICAL or positioning change of the woman depicted in the image.

Basically, these images are used as first keyframes for WAN video generation, and I need them to stay the same/consistent with the first image, otherwise, every next image used as a keyframe will look more and more saturated and degraded.

[thin lily]

hi

[remote arrow]

I think you could use Whisk for character consistency, to create first and last frame. Works good for me so far. Possible prompt in #share-prompts

[flint sandal]

Anyone used the A/B "gemini 3" models in ai studio? And if you did, what are your thoughts.

[neon idol]

Greetings

News about Gemini 3?

[flint sandal]

Nothing new

[neon idol]

Lmao

[remote arrow]

What's so promising in Gemini 3?

[fiery gull]

In my tests I didn't see a very big improvement 🙃

[neon idol]

Maybe it will be a very powerful model

[brave cloak]

Hmmm, I am using a model that can only take the first frame for generation, so I have to save the last frame of the generated video myself, thats the problem

[flint sandal]

It can always be flash lite

On reddit

I seen amazing zero shoted landing pages

Better than gpt-5 high and claude 4.5 combined together

And there will also be 3.0 ultra

We are sure about it

Gemini CLI

Github updates

So best models on AB may be pro but ultra will be a lot better

[remote arrow]

For me, it's funny to see people rushing like "gold fever" in 1800's when new model is about to release. They're begging, whining, yelling everywhere: "Give me Nani Banani, give me Soda 4, give me Sora 2!"
In the end, what are they really make with those models? Rushing to get free Ferrari without having a driving license?

[fiery gull]

I didn't use in code, only written text, but I do see an improvement

[severe pebble]

Hello guys. Is it only me or is something wrong? I noticed that Claude models on arena exhibit significantly different writing styles compared to the official anthropic platforms. There's heavy emoji usage, different formatting patterns, and phrasing that seems more typical of other model families like gemini or deepseek. Are these changes somehow tied to the system prompt? Claude Opus 4.1 seems to be talking more familiar.

[remote arrow]

Sadly, those free video generators always "shifting" the image, even if it's a little bit.

[flint sandal]

Yeah. The reason are system prompts

Same with Grok-4

[severe pebble]

So they all don't have system prompts at all? Or use those that override their self-identification?

[quaint patio]

Code plz

[fiery gull]

Sora code?

I want the sora 2 codex 😞 too

[quaint patio]

Yes plz

[fast lintel]

is anyone here using gemini-cli?

[remote idol]

Yup

[fast lintel]

u using pro or flash?

[remote idol]

pro

using flash is sad 💀

[fierce coral]

if i boost the server dont i geet anything ?

[fast lintel]

ik, cli can like generate image?

[white hatch]

wtf

[fiery gull]

What AI?

[winged mauve]

@echo aurora I really wish to say this now, whoever is managing the models on the backend is honestly doing the worst job right now. Let me explain why.

seedream-4 has been way too unstable. A few days ago, something like seedream-4-2k was launched on the platform after being removed 2–3 times already, and that model was actually producing impressive quality images in custom aspect ratios, just like nano-banana. But again, they removed it for some unsaid reason.

Then came seedream-4-high-res-fal, which honestly feels like a downgrade, and on top of that, it doesn’t even produce images in custom aspect ratios, it’s limited to square only.

I mean, why? Why not let a good model stay stable for once? Why keep removing and relaunching the same thing repeatedly? Just come up with something new if necessary, but please, stop removing already stable, good working models and ruining the overall experience.

[pliant rain]

Hello

[fiery gull]

Hi

[remote idol]

No, it can connect to MCP tools.

LMArena is slacking off. It doesn't generate full code; it cuts off, and it is slow.

I hope the issue doesn't last long because before yesterday it used to work flawlessly, but ever since it's been failing. It cuts out. I tell it to continue, but it doesn't.

[winged mauve]

they're doing the same thing i guess for most of the models, nerfing every models in the name of upgradation

[winged mauve]

@remote idol

[remote idol]

Thank you for reporting this. I thought I was the only person suffering from this.

[remote idol]

So, I want to be clear: will they fix everything?

[remote idol]

Vouch +1

[flint sandal]

im using gpt-5 pro for free in chatgpt rn

[neon idol]

https://tenor.com/bEaoy.gif

[abstract elm]

hello

[flint sandal]

Pm

[neon idol]

?

[flint sandal]

Pm me

Dm

[neon idol]

You mean dm me?

Ok

[flint sandal]

Yeah private message/direct message

[quasi atlas]

@stoic sorrel Please check #1397655624103493813 to learn how to use the bot.

[fiery gull]

How? 🤔

[barren prairie]

I know how to use chatgpt plus for free on oxaam but pro...

[fiery gull]

The Pineapple not is the just one extra-human?

[cunning hawk]

can i too

[flint sandal]

Dm me

[cunning hawk]

Done

[winged mauve]

i have the same doubt

[hushed terrace]

Hello, Please check https://discordapp.com/channels/1340554757349179412/1397655624103493813 if you´re trying to create a video or image

[remote idol]

https://oxaam.com

[earnest spire]

i need a sora 2 code. if you do have one, DM me

[stark sphinx]

Hi. Can you give me the link to invite the bot to my server please?

[flint sandal]

Gemini 3 might drop tommorow

Like experimental

Or smth

[wintry tinsel]

Howcome? Why not on the 9th

[hushed terrace]

Hello! Just a reminder that english is the only language allowed in the chat channels ❤️

Check https://discordapp.com/channels/1340554757349179412/1397655624103493813 to learn how to generate content 🙂

[frigid basin]

When's the actual useful lmarena.ai updates and fixes comin

[woven lion]

hi everyone any idea how to make a 16:9 image from seedream-4-high-res-fal? i tried to use the prompt it doesn't work, it works on the seadream4 previous version

[forest dagger]

This voice is best for usa video or realistic yes or not

[final tide]

girls cooking kichen

girls cooking kichen

girls cooking kichen

@final tide

girls cooking kichen @

girls cooking kichen

[jade egret]

??

[remote arrow]

You can't

[barren prairie]

Or Google might want to surprise us this time and drop Gemini pro 3 directly without anounymous model.

[woven lion]

ah yea lol... the previous seedream-2k can do it tho, this new fal cant..

[obsidian shell]

anyone knows how to make seedream generate into different aspect ratios ?

in arena

[cunning hawk]

which do i activate

cg-ai or kaanvah pro

[hollow imp]

https://cdn.discordapp.com/attachments/1293199422380970064/1379747576819220570/Screenshot_20250604-142711607_1.jpg?ex=68e38610&is=68e23490&hm=c326af72a05ad947a1ff908606853ad2611163216299c03531b7421b8a13d9c7&

[remote arrow]

You can DM me for Seedream 2K link

[hollow imp]

Obviously fake

And then you'll dm all your friends free nitro once you're hacked

[remote idol]

😭 ?

Are you ed?

It lets u login to a chatgpt pro account

[hollow imp]

No I'm @hollow imp

[hollow imp]

Proof

???

[remote idol]

go on it and register and youll get the email and pass

mate

[hollow imp]

Show me proof

[remote idol]

busy

[cunning hawk]

whats the diference between kanvaah pro and the cg ai

[remote idol]

kanvaah pro is canva pro

💀 ?

[remote idol]

@hollow imp why so silent

[hollow imp]

Chatgpt pro account proof

[cunning hawk]

oh

[remote idol]

🤣

[remote idol]

send em proof

when u login

[hollow imp]

Bro

It's asking to pay

You said free

[remote idol]

prove it mate

[hollow imp]

Like

There is cg ai and kanva

[cunning hawk]

lol it worked

[hollow imp]

Where chatgpt

[cunning hawk]

im in a pro account

[remote idol]

ik

[hollow imp]

Is cgai chatgpt?

[cunning hawk]

which do i pick

i got no idea

[cunning hawk]

ye

[remote idol]

obviously the first one

[cunning hawk]

alr

here proof

[hollow imp]

☠️

I put randomised email pass and contact no

[remote idol]

ok

[remote idol]

we have diff emails

it gave me [e-mail]

[cunning hawk]

yeah im 6616 lol

there some random chats

[hollow imp]

@cunning hawk @remote idol ???

[cunning hawk]

go to the website

[remote idol]

pc only

[cunning hawk]

oxaam

[hollow imp]

I logged in

[hollow imp]

Same

[remote idol]

how much is 1,950 rupees in dollars?

[hollow imp]

20$+

[remote idol]

oh i was gna say i could use vpn to get cheaper stuff

[remote idol]

delete em

[remote idol]

there r spots in the pro account

[hollow imp]

But what about that memory fearure

[void sphinx]

Hi I am very interested in AI video generation and I have been looking for options. Here is the perfect place to understand it better and to compare different models

[cunning hawk]

there some dude using it rn lol

some french i think

[remote idol]

🤣

[remote idol]

use it in temporary chat

!!!

[cunning hawk]

i think he tryna imagine simpsons

[cunning hawk]

how i do that

[remote idol]

click this

and https://www.injectprompt.com np

[cunning hawk]

thx

[cunning hawk]

whats this

[remote idol]

jailbreaks

[cunning hawk]

o

[remote idol]

I JUST REALIZED

U CAN USE CODEX CLI WITH THE OXAAM ACCOUNT

[cunning hawk]

damn

[hollow imp]

Omg

[remote idol]

yeaa

[cunning hawk]

even a free crunchyroll lol

[remote idol]

yup

its illegal*

[cunning hawk]

i use pirate site anyway

hianime is better

[remote idol]

piratebay?

[cunning hawk]

hianime

its for anime

[remote idol]

is codex the most powerful tool rn?

[cunning hawk]

can i stay login permamently or do i get kick at some point?

[cunning hawk]

not sure

[remote idol]

kicked out

bc ppl will roll for pro too

just make random email and pass in oxaam registery page

[cunning hawk]

ye true

i can even try the perpelxity pro

since its also a good ai

[remote idol]

holy

i just cleared chat 💔

[cunning hawk]

💀

😭

[remote idol]

LOL

TIME TO STEAL EM

[cunning hawk]

ez

[hollow imp]

What else is there

[remote idol]

im viewing chats

what is bro asking for

[hollow imp]

Omfg

[remote idol]

imma delete the chat

[cunning hawk]

some arab

[cunning hawk]

menace

[cunning hawk]

tf

mfs gotta get better at prompts

[plain yoke]

is it just me or did gemini 2.5 pro get nerfed or something?

[remote idol]

Even in the aistudio?

[cunning hawk]

W plageriazing

[plain yoke]

No just their browser version

[magic ravine]

Yep. It was getting too good like their experimental so they needed to dumb it down

[remote idol]

wtf

sarcasm or actually

[hollow imp]

What is the

Umm

I forgot what was i asking

@remote idol damn bro

😭

[remote idol]

lol

[hollow imp]

@remote idol I remember what I was asking

Chatgpt has a memory feature that is very annoying in this Situation

How to deal with that

?

[quartz light]

can i have the acc

please

https://cdn.discordapp.com/attachments/1404636134046302309/1424393844891324486/togif.gif

[hollow imp]

Go to the link he provided

Oxaam

[hollow imp]

Here

[quartz light]

what do i put in

?

[quartz light]

hey

what do i put as contact numnber

number

[hollow imp]

Anything

Any 10 digit no

Random

[quartz light]

nvm

ye

[hollow imp]

Do not put your own

Anything

Add everything random

[quartz light]

so

what do i activate

[hollow imp]

Chatgpt pro

[quartz light]

kaanvah pro?

or cg

i dont see it cro

[quartz light]

wher

where 🥀

[remote idol]

cg ai

[quartz light]

thx

[sullen quest]

Please check https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to use the bot and https://discord.com/channels/1340554757349179412/1397655695150682194 https://discord.com/channels/1340554757349179412/1400148557427904664 https://discord.com/channels/1340554757349179412/1400148597768720384 for your creations.

[hollow imp]

@remote idol hello

I don't understand something

[remote idol]

?

[hollow imp]

If I create a custom gpt, set the model to gpt-5 pro and do share anyone with the link, then can I really chat with gpt-5 pro with a non subscription account?

[remote idol]

obviously no

it will reach limit

after 5 chats

[hollow imp]

Oh

Otherwise it will work?

Like hit limit and limit gone after 3 days or permanent that chat closed?

[hollow imp]

Didn't happen with me

[quartz light]

its using gpt 4o mini for everything now

[hollow imp]

@remote idol

[hollow imp]

Same

[hollow imp]

😭

[quartz light]

🥀

[proud siren]

Yes because they want to trick your mind so that when Gemini 3.0 is released you'll say 'Wow, it's changed a lot.

[hollow imp]

#1395438935676817428

[remote idol]

same

[hollow imp]

See how long this is going

[hollow imp]

Any fix

?

[remote idol]

no

not my website

[hollow imp]

You are unable to use it aswell

[remote idol]

I fixed it by kicking other users

[hollow imp]

How do I do that

[rancid drift]

oh thanks

[hollow imp]

@remote idol hit model limit 💔

Do I register with a new account or smth

[wicked sage]

why does sonnet 4.5 hate me

i cant chat with it anymore

[lunar glade]

Did anyone know where can I still use the GPT 4o image ?

[thorn valley]

he only uses 4o mini since the first time i joined

[winged mauve]

hey, i need some help regarding image generation ( using image ), can i dm you?

[knotty fable]

Dam Sonnet have filed for a divorse then?

[thorn valley]

including even in the way of thinking for longer than normal

[manic perch]

Subject: Urgent Bug Report - Claude Opus and Sonnet Conversations Crashing Frequently

Dear Anthropic Support Team,

I'm a regular user of Claude on your website, and I've been experiencing persistent issues with both the Opus and Sonnet models. Every time I engage in a conversation, it either crashes unexpectedly or displays an error message, forcing me to start over from scratch. This has happened multiple times in the last few sessions, making the experience frustrating and unproductive.

It seems like there might be a stability bug affecting the web interface. Could you please investigate and prioritize a fix for this? I'd love to continue using Claude without these interruptions.

Thank you for your attention to this matter—I appreciate your quick response!

Best regards,

[cursive kestrel]

why didnt i get my video generated

ideas??

[tough meteor]

Are there any plans to add temperature and top p settings in the website?

[echo aurora]

We are exploring options currently!

[winged mauve]

Hi sir, kindly respond to my text

[echo aurora]

It looks like you aren't prompting the bot correctly, be sure to check out #1397655624103493813 for more information

[knotty fable]

Too anarchistic prompting? 😺

[echo aurora]

So models are going to be added/removed from LMArena for various reasons. We're not going to go into the specifics as to why a lot of the time as it's info we sometimes can't rly share. We do want to have as much of the frontier models available as we can.

If the models are unstable I'd encourage you to review the bugs post in #1343291835845578853 and followup with a related already existing post, or create a new one

[winged mauve]

by unstable I meant getting removed and added, different variants of the model, not something like bugs

[winged mauve]

Okay sir

[echo aurora]

ah

makes sense

[tough meteor]

Cool. Any idea when this could be publicly available?

[winged mauve]

can I expect seedream-4-2k to come again on the website? you guys also showed it being tied by nano banana in #announcements , it was such a good model sir

it generated custom aspect ratio images

the one which is on the site rn, named high-res-fal, doesn't have the capability to generate custom aspect ratio images

[echo aurora]

Nothing I'd be able to share in terms of hard dates but rest assured we're working on it.

[echo aurora]

It could happen! glad to hear it's a liked model!

[verbal nimbus]

GLM 4.6 is a beast given that its open-weight and only 257B. One shotted 1K lines of Svelte 5. I wasn't even logged into their website.

[versed dawn]

Hello guys, I am making a motivational story for my YouTube, so I need AI images for that. Which is the best tool in LMArena which can create AI images using AI prompts? please help me

[leaden owl]

hy can any one tell me that how can we create vedio on,lmarena while its just creating image's

[sullen quest]

Please check https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to use the bot and https://discord.com/channels/1340554757349179412/1397655695150682194 https://discord.com/channels/1340554757349179412/1400148557427904664 https://discord.com/channels/1340554757349179412/1400148597768720384 for your creations.

[hollow imp]

@remote idol BUHAHAHAHAH

NO INTERRUPTIONS

@cunning hawk

[viral cedar]

question, what AI has the highest context tokens

[winged mauve]

Please sir, I would be truly grateful if you could help bring it back on the site. It was indeed a fabulous model.

[viral cedar]

so far it's been Gemini 2.5 for me, but i can't find any specific AI with 1M+ tokens

[fiery gull]

It depends on what you want, but in my rough tests gpt 5 high was the best (yes even with 400k) in practice it was the one that controlled the memories better

[viral cedar]

i developed an algorithm to basically get AI to convert exe apps to literal C++ source code

[proud hazel]

Why aren't you using it on their official website? It's free!

[viral cedar]

essentially I made some code in Ghidra that takes out all of the decompiled code from the app, (binary format obv) and then the decompiled code is organized into it's source with AI

[verbal nimbus]

Seems like a good benchmark for different models

[viral cedar]

problem is when there is something super complex like literal full games, huge softwares, its hard to do it since i end up having to ask 100 different times to the AI for each code

since its like a lot lot of lines

[sullen quest]

?

[viral cedar]

so uhh as yall can tell its like

[verbal nimbus]

I know Claude could debug decompiled code, but it seems difficult to reverse engineer an entire program from assembly haha

[viral cedar]

1M+ lines of code

[viral cedar]

ye well ghidra decompiles well and i scripted some stuff that exports it and slightly compresses it and cleans the decompiled code up

and then its shipped to AI

[frank fable]

Hi i cant find my video

Too many posts

[sullen quest]

can you share this with me?

[verbal nimbus]

I guess reverse engineering minified JS/React is more feasible if you're into that

[viral cedar]

then it recreates its source code of lets say 1/100 of the decompiled code, then it comments what functions its missing and how they are used

and then I will do it another 99 times for the rest of the code

[verbal nimbus]

It needs the terminal I think

[viral cedar]

and then only at the end will the AI, combine all the sources and comments it made and put it into functioning source code

[frank fable]

How do we find the videos generated for us

[verbal nimbus]

It needs to use search and use tools to look at the symbol table etc

[viral cedar]

this is py code for ghidra to do it

takes a while to decomp everything

[hollow imp]

Enjoying the time of my life

Gpt 5 pro access for free

🤤

@remote idol I can't thank you enough

All I had to do was invite my email to the workspace

[viral cedar]

i just use lm arena 😐

is it different

[proud hazel]

GPT 5 Pro ≠ GPT 5 High

[verbal nimbus]

Might get better performance if you let the agent work in a Linux terminal

It can then use decompilation tools as needed

But I doubt anyone can reverse engineer such a large app :P

[sullen quest]

I don't really see any "decompiled code is organized into it's source with AI" with the code, but I did just skim through it

[viral cedar]

thats cuz its just ghidra taking all its functions (Decompiled code) into 1 big file

it then gets optimized to remove extra lines and stuff, and then split up into x amount of files to be under 800k tokens

i then put them all into AI and tell it to recreate the code based on functions

[viral cedar]

if u want compile me any random exe app and send it in DM's I'll get its source code

if it isnt obfuscated

[tough kayak]

where i can see results so far which model doing the greatest ?

[proud hazel]

wrong channel

dude...

[viral cedar]

https://cdn.discordapp.com/attachments/742009120076333057/1251927683752005732/ezgif.com-gif-maker-1.gif

[sullen quest]

ok thats pretty random, Ill look into that

[verbal nimbus]

https://lmarena.ai/leaderboard

[tough kayak]

Yea I saw thanka

[neon idol]

https://tenor.com/62C8.gif

https://x.com/elonmusk/status/1974698202625679361

Elon Musk (@elonmusk)

Version 0.1 early beta of Grokipedia will be published in 2 weeks

X · Oct 5, 2025, 4:48 AM

What the hell chat

💀

[sullen quest]

bruh

[neon idol]

Me when I search H!tler in this app:
H!tler: a kind and good person

[thorn valley]

how did you do that?

[hollow imp]

@remote idol will explain

[neon idol]

oxaam.com?

The classic

[thorn valley]

i accessed the link they sent a few hours ago, but the account used shows that the plan is free and the model used when making a request is from the gpt 4 family, that doesn't even make sense

[thorn valley]

i will wait until he can answer

[neon idol]

You have to go on setting

And then click on your profile

Then click it and find a logo with like a bag

Click it

And free gpt

[proud siren]

how?

[remote idol]

np

[manic perch]

Formal Complaint: Incomplete Outputs from AI Models

Subject: Urgent Concern Regarding Persistent Incomplete Responses from AI Systems

Dear AI Development Team at [Company Name, e.g., OpenAI / xAI / Grok Team],

I am writing to formally express my deep frustration and disappointment with the consistent issue of incomplete outputs generated by your AI models. As a frequent user who relies on these tools for professional tasks—ranging from drafting reports and creative writing to generating functional code—this recurring problem has severely undermined my productivity and trust in your technology.

Time and again, when I request a piece of code (such as a Python script for data analysis), a structured text (like a business proposal), or even a straightforward response to a query, the output is abruptly cut off mid-sentence, mid-function, or mid-explanation. For instance:

• In code generation, variables are declared but never implemented, loops are started without closures, or entire error-handling sections are omitted, leaving me to debug and complete the work manually.
• For textual content, paragraphs trail off without conclusions, key arguments are left undeveloped, or references to sources vanish halfway through.
• Even simple responses end prematurely, forcing me to prompt repeatedly for continuations, which often leads to inconsistencies or further fragmentation.

This is not an isolated glitch but a systemic flaw affecting all major AI systems I've encountered—from GPT variants to Grok and beyond. It feels as though the models are optimized for brevity over completeness, perhaps due to token limits or generation thresholds, but the result is unreliable and unprofessional. In a world where AI is touted as a revolutionary productivity booster, delivering half-baked results is not just inconvenient—it's a betrayal of the promise.

I urge you to prioritize fixes in upcoming updates, such as:

• Enforcing full-response validation before output.
• Implementing user-configurable completeness thresholds.
• Providing transparent warnings when a response might be truncated.

I expect a swift acknowledgment of this issue and a roadmap for resolution. Until then, I'll be exploring alternatives, but I sincerely hope your team can restore my faith in AI's potential.

Thank you for your immediate attention.

[sullen quest]

bruh

[unkempt kelp]

hi

[hollow imp]

So you try in different web browsers and with vpns and add random info on oxaam until you get an account with less than 9 workspace members. Then you invite your email and enjoy the subscription until they find out and kick your email. Also all your chats after you've invited yourself and got pro would be gone

[remote idol]

imma publish a website that shows access to gemini, grok, claude and chatgpt pro.

[fossil fable]

[quartz light]

how

[fossil fable]

wtf

is that

[fossil fable]

yk that just uses o3 deep research and also wastes a pro token right

[wintry tinsel]

Why tommorow specifically do they usually surprise us 3 days before release?

[weak swan]

What's the difference between UB rank and style control rank?

[calm vault]

Am I allowed to use the generated vids for commerical purposes?

[hollow imp]

Yes

🔥

[fossil fable]

r u leeching off someone's sub

[grand cliff]

Man...I miss 4o.

Thank you LMarena for letting me experience this again.

Ai-ya...

[barren prairie]

It made for me 6000 lines without mistakes and I am still working with it on the same project without issues . It is amazing !
Note : i am not a dev.

Even GLM4. 5 was able to make 1500 lines on one prompt with 0 mistakes. GLM 4.6 is another level

[barren prairie]

Yeah even 5 days, but this time they are silent.

[keen beacon]

how do i change from battle mode to direct chat on web dev

[sullen quest]

not a thing

[keen beacon]

so aint possible?

[sullen quest]

yes

[heavy knoll]

Which App Are you Using? To Code

[barren prairie]

Chat.z.ai from their web site

[sharp mirage]

yoo

[heavy knoll]

I mean the App Where You Write youre 6000 Lines

[proud hazel]

With or without deep think enabled?

[remote idol]

@hollow imp

can u gimme the email u used

in oxaam

rq

[icy sail]

hi

[verbal nimbus]

What the... Wow that's a lot of code

Without mistakes for 6000 lines is actually crazy for plain HTML/JS wth

[hollow imp]

Uhh I closed the tab

And the workspace that I've currently got access to is in russian language

And it's been two hours and I haven't got kicked

[wintry tinsel]

The suspense is built

[verbal nimbus]

Last time it was on LMArena about 4 days before

At least based on when people started to mention the anon model on the server

[sullen quest]

I didn't hear about this was it confirmed gem 3? or a google models? was it good?

[sharp mirage]

hell nah bro

any1 have papaer show specs of glm4.6 ?

[verbal nimbus]

It does seem worse on LMArena

It thought for only 3 seconds but on the official website it thinks for way longer

[sullen quest]

weird, I thought I heard someone say the exact opposite opinion

[verbal nimbus]

I could try the same prompt

But even with basic questions it thought for much longer with deep think enabled

Compared to how long it thinks on LMArena when I asked it to code something

Also the code seems shorter on LMArena

[sharp mirage]

i well tryt

try

does glm 4.6 have better codes than gemini 2,5 pro ?

[sharp mirage]

.

what even she made :/

[hollow imp]

@keen beacon bro

I'm using Gpt 5 Pro

And OOO my god

This is absolutely not worth 200$

I feel bad for those who pay

😭

😭🙏

[sharp mirage]

لاقعا

bruh

[hollow imp]

Do you want me to test something for you? Send prompt

[sullen quest]

lol

[sharp mirage]

and gpt 5 high not even good bro

[sullen quest]

you don't pay?

[sullen quest]

rlly? wow

[hollow imp]

@keen beacon

[sullen quest]

Cause lmarena publishes all the stuff and helps bench llms and they get donations and free api usage for that purpose

[sharp mirage]

any1 have paper for how good is glm 4,6 in codes like vs ai ?

[hollow imp]

There was some guy named bycloud ai who made a paper finding assistant or something, maybe use that

[sullen quest]

how do you ahve free usage of gpt 5 pro?

[sharp mirage]

he buy it :/

[sullen quest]

oh ok

[hollow imp]

Nooo

I tricked a free chatgpt pro giving website

[sharp mirage]

you got it free ?

[hollow imp]

Oxaam.com

[hollow imp]

Yes

[verbal nimbus]

Prompt for GLM 4.6:

Create a Twitter clone in Svelte 5 for mobile. Your output should be a single .svelte file. It good be an exact visual replica.

The difference is like night and day between the official site (with Deep Think) and LMArena.

[hollow imp]

And I will probably be kicked from the gpt5workspace tomorrow when they notice it

So I'm doing everything I can rn

[sharp mirage]

i dont fell safe for that

[sullen quest]

that could be random chance

[verbal nimbus]

It happened before to me

[sullen quest]

oh yeah that

[verbal nimbus]

It used the same broken styling last try on LMArena too

[sharp mirage]

i am trying glm on the main website and on lma

[sullen quest]

makes sense

[sharp mirage]

to see who is better

[sullen quest]

what proviers does lmarena use?

It also may have to do with the instruction set it has on glms website

[verbal nimbus]

The discrepancy for other models? Not as far as I know. There's an issue with Claude models being cut short at 4,096 tokens though.

[sullen quest]

but not all models have og providers

[sharp mirage]

what about extianos that can give better ui to website it fells wrong somethimes

[sullen quest]

what

[sharp mirage]

extension

like when ai in lma answer in arabic sometime send wrong words

[sullen quest]

???

some weird models do show up on the leaderboard

how would I know?

[sharp mirage]

how they work without api ?

[sullen quest]

I'm not searching through all 260 models

[sullen quest]

un offical apis

[sharp mirage]

thats wierd

my gpu is gt 210 :/

btw i got it

for now glm on lma takes more time thinking it take 2 more sec and it was slower
on chat.z takes less thinking time and faster answer

both of them send bad codes :/
does all ai bad at powershell ?

[sullen quest]

prob

[hollow imp]

I had to do a 3-4 hour chat on gpt 5 search and Claude opus 4.1 search to finally get a small thing done using powershell.

[sharp mirage]

no one know how to use it ?

[sharp mirage]

to now i got better codes at lma idk how

[native yarrow]

Lmarena lowk need to loosen the filter

[marble rivet]

google ai studio ???

[sharp mirage]

?

[verbal nimbus]

Image gen or text?

[sharp mirage]

anyone have any idea about kotlen ?

[verbal nimbus]

GPT-5 was ok

[sharp mirage]

kotlin

[verbal nimbus]

What about it?

[sharp mirage]

yeah gpt 5 chat was better than gpt 5 high

[hollow imp]

HUHHHHHHHH

Is it because you crossed that 1 million limit or something

[sharp mirage]

bro i am gone crt

cry

[midnight hornet]

🤔

[sharp mirage]

i hpaen

its okay

[midnight hornet]

ok thanks ... i will try 20 min later

[verbal nimbus]

Idk how I never realized it, but every repo on SWE-Bench verified is literally Python

[sharp mirage]

bro gpt 4o is way better than gpt5

[viral cedar]

chat are we seeing this

[magic stream]

is there a way to get longer than 3 seconds?

[viral cedar]

#video-arena-3 message

we gotta shun this guy

[finite stump]

Yo which ai is best i use gemini 2.5 pro cuz its fast but what yall think

[midnight hornet]

thanks . i didnot know that ... now it is work again

[hollow imp]

GPT-5-PRO

[sharp mirage]

yeah its good but try 4o

[toxic egret]

gpt-5-high isnt a reasoning model?

[sullen quest]

no it is

[toxic egret]

why i dont see what he is reasoning?

[sullen quest]

you mean in lmarena?

[toxic egret]

yes

[sullen quest]

thats cause the little "thinking" bar doesn't work for all models

including that one

[toxic egret]

its i question ive had since that was added

[toxic egret]

oh

it has a reason? or its just not added yet

[sullen quest]

it has reasoning, just for some reason lmarena can't time it

put it in "bugs"

[toxic egret]

you sure its a bug?

not something like for example that it doesnt use </think> tags

[timid oak]

hi

[sullen quest]

hoi

[manic perch]

Subject: Issue with Long Texts: "Something went wrong while generating the response"

Message:
Hello,

I’ve been experiencing a recurring issue on your website/platform. Whenever I try to write or submit a long text, the system fails to process it and shows the following error message: “Something went wrong while generating the response. Please try again.”

This makes it very difficult to use the platform for more detailed or lengthy content. I would really appreciate it if your technical team could look into this problem and fix it.

Thank you for your attention and support.

[toxic egret]

i just say hello to the LM then send the big message lol

[stray aspen]

yo

theres so much hype from google for next week

i hope we get a new SotA

[neon wyvern]

Hello guys

[neon wyvern]

what platform is that?

[stray aspen]

did this just come out right now

?

its no longer preview

[dusky ravine]

wtf lol what this gooners doin

[golden ocean]

so real

[dusky ravine]

bruh im really tired with this infinite generation 😭

[stray aspen]

these kind of things happen more often than people think lol

[dusky ravine]

some ppl are just so h0rny lol

[golden ocean]

IM JOKINGm

[echo aurora]

Reporting this stuff to Moderators is helpful

[golden ocean]

hi everyone how is everyone today

[viscid cloak]

seedream-4-high-res-fal and gpt-1 in BATTLE keeps giving error message. Any one found the same issue?

[dusky ravine]

Hey, are you doing the same thing?

[golden ocean]

dispenser

is this u

[sharp mirage]

Bro ur sus

[golden ocean]

it's a dispenser

[viral cedar]

ayo!!!

[golden ocean]

erm

[midnight hornet]

🤔 is there anyway to rename a chat ?

[frank mica]

/video prompt.. messi driblling ronaldo and make him lose the ball

[echo aurora]

There is not sorry to say

[midnight hornet]

🤔

[echo aurora]

Be sure to review #1397655624103493813 if you're having trouble with the bot.

[keen beacon]

Hey guys, Im isaac from ecuador, I study and research AI engineering. Im looking to network and meet ambitious people in the space, and hopefully learn a lot on the way. Anyways, dm me if you'd like to network

[sharp mirage]

No

[sharp mirage]

Bro just made his account today's

[keen beacon]

yeah I dont like discord that much tbh

just wanna meet a few people and imma delete it

[sharp mirage]

Does anyone know how I can make chat gpt or another ai
Access my pc ?
Or access some files ?
I am thinking in PowerShell code but idk

[fiery gull]

Is possible with a some CLI

[sharp mirage]

How?

[sullen quest]

nah a few days

[summer ruin]

what is everyone using for video generation?

[toxic egret]

check the leaderboards

but if you are lazy anyway Veo3

[stray aspen]

sora 2

or veo 3

[rancid basin]

I am new at this trying to learn and take it serious but everytime I try to make an image to video most good ones are only paid I tried using LMarena but it locked me after making 3 images

[echo pendant]

how can I create a 30-60 sec video or just continue making with the initial frame like a story ... just a beginner here

[summer sphinx]

Sora2invititaion code please

Sora2invititaion code please

[winged mauve]

Bytedance? How can I use it?

[plucky ember]

Hello 👋 everyone I'm new here

[robust yoke]

Greetings.

[robust yoke]

You can use it on Hugging Face or Dreamina.

[winged mauve]

Free?

[robust yoke]

Free.

Dreamina is free.

For right now, at least.

[winged mauve]

Dreamina is banned in my country, not sure about hugging face

[robust yoke]

Just use a VPN then.

[winged mauve]

Which country?

Us?

[robust yoke]

The U.S.

[winged mauve]

I mean which country vpn will do?

[robust yoke]

Yes.

[winged mauve]

Thnaks

[robust yoke]

My pleasure.

[plucky ember]

😊😊 hugging face is what I've been using it helps me with ai voices and perchance with images

[robust yoke]

Oh neat.

[winged mauve]

Does it generate custom aspect ratio image if you have idea by any chance?

[robust yoke]

Yes, it does.

[winged mauve]

Is it free?

[robust yoke]

You can set it to any ratio.

[plucky ember]

Yes but it doesn't use picture to image its one there but I can't find it anymore

[winged mauve]

Thanks

[robust yoke]

My pleasure again.

[echo pendant]

which AI can be used to make long videos??

[robust yoke]

Lemme take a lookie for you.

[plucky ember]

How long ??

[robust yoke]

I remember coming across a certain one, I think by Alibaba, but I'm not sure.

[robust yoke]

Apparently, Kling 2.1 works.

It supports video generation of up to 2 minutes in length.

[plucky ember]

Pixverse allows you to extend a video by 5 to 10 sec and you can keep extending but you may lose quality after a while

[echo pendant]

like 1-2 mins or I can make multiple video of the same seed to convey a story of 7-10 mins

[plucky ember]

O wow 2 minutes I didn't realize we was at that level yet

@echo pendant what video generator have you been using for 1-2 videos

[robust yoke]

It's true.

Apparently, it listed Veo 3 as a long-form video generation model too, for some reason.

[plucky ember]

@robust yoke I haven't had a chance to test out the paid video generators yet but my favorite video generator is Vidu and Haiper ai

[robust yoke]

Interesting.

[remote arrow]

Veo3 can do long videos, but not sure which subscription is facilitated with that.

[robust yoke]

Interesting.

I assume regular Veo 3 can't do long-form videos, can it?

[remote arrow]

The free one is maxxed at 8 seconds.

[plucky ember]

Haiper Ai don't allow just anyone to use it anymore its only for big production companies now

[robust yoke]

Figures.

So, it must be either the Ultra or AI subscription.

[plucky ember]

@remote arrow oh veo 3 is free??

[robust yoke]

No, it's not.

[winged mauve]

Any good ai for text to speech? @robust yoke

[remote arrow]

There are some websites allowed you to use free samples

[robust yoke]

You betcha.

[plucky ember]

Lol 😂 I got a little excited

[robust yoke]

Apparently, Microsoft's VibeVoice models are very good at life-like text-to-speech.

[remote arrow]

I've used Goog's AI Studio TTS, the exact same version is at LTX dot studio. Been using that for my videos..

[robust yoke]

But, apparently, a very good video generation model to use right now would be Sora 2, even if it can only generate up to 10 seconds in length.

[robust yoke]

Interesting.

[remote arrow]

It's 10. It's good, except for the limit not to upload "photographic image" and those watermarks..

[plucky ember]

Open art gave out 5000 credits for watching their YouTube video to the end they have Kling veo seedance vidu and a 8 other generators you can use veo 3 was 2000 credits for 1 video while others were 150-300 credits

[robust yoke]

I've used that one before, but I find VibeVoice more realistic since the audio is low-quality and sounds very expressive.

That's not to say that Google's doesn't, just that it's too high-quality.

[remote arrow]

I only use Hailuo on that site. Lol..

[robust yoke]

I suppose so.

[remote arrow]

I'll need to try that one..

[robust yoke]

Yeah, it's very good.

They even support voice cloning.

[remote arrow]

The most useful features on Uncle Googs is it's the oldest search engine with data collected from decades, so it can do dialects correctly..

[robust yoke]

I tried it with some Spon geBob characters like Sponge Bob, Squid ward, or Pat rick, but they didn't turn out too well.

[robust yoke]

That's true.

There are certain times where either-or can be good.

[remote arrow]

Agree..

[plucky ember]

Now voices I need some good free ones fakevoice is pretty good and free they have a great Rick I've been using

[robust yoke]

There are two certain models I know of that are very good at that.

One is the model I mentioned earlier, and ChatterBox.

Apparently, that model is also good at voice cloning.

[echo pendant]

try steve.ai,, it can generate 60 sec video

[plucky ember]

😊 I will be checking that out soon are you able to change the emotions and accent

[robust yoke]

For ChatterBox, yes, for VibeVoice, no.

[plucky ember]

I came across one that allows you to Change the tone emotion and accent it was marvelous I cant find it to save my life it spoiled me

[robust yoke]

Do you roughly remember the name of it?

[remote arrow]

Google and LTX Studio can do that..

[plucky ember]

No I remember the logo was purple with a frequency symbol but I found it on hugging face

[robust yoke]

Purple with a frequency symbol...

Gotcha.

Is it called "Octave" by any chance?

[plucky ember]

I was able to use a promo like [joyous] hello everyone [ arrogantly] I have a new Ai [sigh] but it's [sad] Horrible [weep]

[rancid basin]

Hey Darkness is there any free software or websites that are edit ur entire script all at once instead of these 5 seconds increments that always turn into pay walls

[robust yoke]

Must be Hume's then.

[robust yoke]

Free software that edits entire scripts?

[rancid basin]

that generates it into video sorry

[plucky ember]

@robust yoke Promo I meant prompt ....octave one sec I'm going to look it up because that sounds familiar

[robust yoke]

Oh, my apologies.

In that case, you can use Kling 2.1.

[rancid basin]

it's free?

[robust yoke]

Seems to be.

[sour saffron]

Mine always goes into queue

[robust yoke]

If you're looking to create 10-second videos instead, then Sora 2 is your go-to.

[rancid basin]

Where I'm p sure mine tried to get me to paid from clippie.ai

[robust yoke]

Oof.

[rancid basin]

sora 2 would be great but need invite

waiting list is a million years long

[robust yoke]

Yeah, that happens.

[robust yoke]

No worries.

I have access to a Telegram server that provides codes to access it.

[sour saffron]

Can u send on dm?

[robust yoke]

Sure.

[sour saffron]

Tysm

[robust yoke]

Absolutely.

[quartz light]

magistral small is impressive

[robust yoke]

True.

[rancid basin]

wait really? You could ? would you mind dming me a code?

[robust yoke]

I'd love to.

[plucky ember]

@remote arrow I will check that out

[rancid basin]

that's awesome thanks!

[robust yoke]

Yeah, my pleasure.

[plucky ember]

@remote arrow yesss it's exactly like that

[remote arrow]

Both Google and LTX can do. I prefer Googs since it's free..

[plucky ember]

@robust yoke @remote arrow I want to thank you guys so much I would have been researching for years to find the gold yall introduced me to 😊😊😊

[robust yoke]

Yeah, no problem whatsoever.

[rancid basin]

Btw is sora 2 work for other types of videos not just realistic?

[robust yoke]

Yeah.

It used to work for copyrighted cartoon styles too, like Spong eBob, but now they updated it so it can't.

[remote arrow]

Yes. @robust yoke doing Family Guy and Spongie like they're original..

[robust yoke]

Yeah, heh heh.

[rancid basin]

Lol ya I saw video of craby patty on the highway police chase

[remote arrow]

🤣 I want to see that

[cloud zinc]

when will it be fixed

sora-2 gonna be $0.10 per video and sora-2 pro gonna be $0.20 per video.

[robust yoke]

Nobody knows, but considering how ClosedAI has dealt with GPT-4o disappearing from ChatGPT, we'll likely see the guardrails disappear once again soon.

[plucky ember]

@rancid basin I've been seeing some great things from sora I can't wait to try it someone put a prompt blank blank ect as if it was an episode from office

[robust yoke]

I could quite literally just do that right now, heh heh.

[remote arrow]

Oh no. Not Spongie or Fam Guy again..

[robust yoke]

Sora 2 The Office episode, now featuring Dwight in a boxing ring with Jim. 😈

[remote arrow]

You must include Scam Altman sweeping there..

[robust yoke]

"Scam Altman"? Ah, I think you meant "Scam Saltman". Yeah, I know that guy.

[cloud zinc]

they will def do some deals with some companies

he said "We are hearing from a lot of rightsholders who are very excited for this new kind of "interactive fan fiction" and think this new kind of engagement will accrue a lot of value to them"

[plucky ember]

@robust yoke I hope they put down the rails soon they are getting a bit out of hand with censorship / try it out the office parody is hilarious

[robust yoke]

Yeah, I hope so too, and definitely.

[remote arrow]

Wait and see. I'll stick with Veo3 for now..

[plucky ember]

Scam saltman 😂😂😂

[robust yoke]

Heh heh.

[elder solar]

why is it so hard to find a website that actually offers gpt-5 high for free

[cloud zinc]

sora 2 is way better than veo 3

[robust yoke]

You wouldn't believe me if I told you...

[elder solar]

poe.com was going pretty well, but now they increased the output prices of it

[remote arrow]

Scroll up, I think I've seen some guys talk about that earlier, and they made it..

[cloud zinc]

lmarena offers gpt 5-high

[plucky ember]

@remote arrow ohh no they came before we could clean the evidence lol

[robust yoke]

That's what I'm saying.

[elder solar]

yeah but token limited

[robust yoke]

Try using Copilot then.

[elder solar]

i dont think thats the high one

its the normal one

[robust yoke]

Anyhow, check out this video I made with Sora 2: https://sora.chatgpt.com/p/s_68ddcda627548191b3c4c8f7fb0e0819

[cloud zinc]

better than nothing

[robust yoke]

Hmm...

[cloud zinc]

thats so real

[elder solar]

poe.com was the best option

[robust yoke]

Yeah, it is.

[elder solar]

but now they increased the output price of it

[remote arrow]

Thanks God it's not Spongie..

[elder solar]

i just ask the chatbot's a single sentence question and boom

currency already spent

[robust yoke]

Apparently, this website provides it for free?

https://www.easemate.ai/ai-chat/gpt-5-free

[elder solar]

feels like some fake slop

[robust yoke]

If only I could... 😞

[cloud zinc]

thats fake website, noway to verify if its gpt 5

[robust yoke]

Well, you never know until you try.

[robust yoke]

Well, Comet found it for me, so it might be.

[cloud zinc]

even if its gpt-5, it could be gpt-5 mini, which is a terrible model.

[robust yoke]

I'll verify then.

[elder solar]

its just like that one 'grok 4 heavy" scam

[cloud zinc]

grok 4 heavy is 300 a month lol

[elder solar]

yeah

wish we could have those here

[cloud zinc]

it doesnt have an api thats why

[elder solar]

what about deepthink 2.5 google

[cloud zinc]

it doesnt have an api also

just like gpt 5-pro

[sharp mirage]

Yo

[robust yoke]

Yeah, never mind...

[robust yoke]

Yo.

[elder solar]

theres pro?

[robust yoke]

Yeh.

[elder solar]

i heard there is sora 2 pro too

[robust yoke]

That's also true.

[elder solar]

but no idea if that evenexists

[robust yoke]

It's just Sora 2 but with 1080p quality.

[cloud zinc]

ya for chatgpt pro user. thats 200 dol;lar a month

[cloud zinc]

and 15 second

[robust yoke]

But still a watermark. 😩

[elder solar]

theres a offer that is just one dollar

but it was only usa

usa credit cards

[plucky ember]

200 a month and it still has a water mark wow the audacity

[remote arrow]

Grok dot com it is 🤣

[robust yoke]

I guess so.

[cloud zinc]

i mean u can removed it using other ai

[robust yoke]

https://tenor.com/view/long-tears-gif-11087877242190144149

[plucky ember]

Lmaoooo

[cloud zinc]

i have seen lot of post where sora 2 video had no watermark

[elder solar]

ai layers over another ai

[cloud zinc]

ai inception 😵

[sharp mirage]

Does anyone have any idea how i can get ai access to my pc or my files?

[robust yoke]

AI-ception. 🤯

[robust yoke]

Just use a CLI.

[sharp mirage]

School after 30m

[sharp mirage]

How?

[robust yoke]

You go to GitHub and download one.

A free one is Gemini.

[sharp mirage]

It uses my pc ?

[robust yoke]

Not "uses" as in "controls", but it can access your files.

[elder solar]

if only poe.com decreased their output prices for once..

[sharp mirage]

Yeah my gpu is gt210

[robust yoke]

Nice.

[sharp mirage]

Thx for info

[robust yoke]

My pleasure.

[elder solar]

i even got to have a chopped & screwed html generator

and now i cant update it anymore

[sharp mirage]

Btw you know any other why?

[elder solar]

cuz gpt 5 high is now overpriced

[robust yoke]

Any other way?

[sharp mirage]

Yes

[remote arrow]

Anthropic MCP

[robust yoke]

Lemme try and see...

[sharp mirage]

K

If you want to search for it together dm

[elder solar]

whats the difference of gpt 5 pro anyways?

[robust yoke]

Gotcha.

[sharp mirage]

I going to school 🏫 cya

[robust yoke]

Cya.

[sharp mirage]

Cya

[icy sage]

Cya

[elder solar]

@robust yoke ?

[icy sage]

wait

nvm

hi

[remote arrow]

Copacabana guy!

[icy sage]

ask it to chatgpt duh

[icy sage]

nah 💔 lol

hello bro

[elder solar]

i dont think chatgpt knows itself

[remote arrow]

How're you?

[icy sage]

Damn

I'm fine, you?

[remote arrow]

Sleepy. Lol..

[icy sage]

I was about to say that lol

well, i got nothing in my mind rn

[remote arrow]

Just scroll up and read things about AI, then, maybe you'll get heated..

[icy sage]

Choose a number from 1-6

[remote arrow]

None. You must wide it up..

[icy sage]

hmmmmmm

[elder solar]

7

[icy sage]

choose a number from 1 to 103

[remote arrow]

9

[icy sage]

no

try again

[remote arrow]

99

[icy sage]

ok

lets see if someone is lucky here

choose 10 number from 1 to 99

[robust yoke]

I have humbly returned.

[icy sage]

10 numbers from 1 to 99

now

[robust yoke]

Ah, my apologies.

I think the difference is in reasoning, maybe?

I forgot.

Anyway...

[icy sage]

fair

[robust yoke]

11, 22, 19, 99, 98, 97, 96, 95, 55, 50.

[icy sage]

congratulations

you just wasted time texting random numbers

🤩

[robust yoke]

Well, I only wasted about a minute.

Since I'm a fast typer. 😎

[icy sage]

-1 minute of your life lol

[robust yoke]

Not like that minute was important anyway, heh heh.

[icy sage]

you could've done absolutely nothing in these big 1 minutes

[robust yoke]

I could've also chatted in that 1 minute, which was exactly what I was doing.

[icy sage]

Ok so i won, i successfully wasted 1 min of your time

[robust yoke]

You did, but for a productive reason.

[icy sage]

Why am i real 😭

[robust yoke]

Because God made you real.

He put blood into your veins.

[icy sage]

If i asked this to ChatGPT he would have written a gigantic text

[icy sage]

In what religion do you believe?

[robust yoke]

GPT-5: "Well, in order to figure that out, I'll need to ask you some questions:

1. ..."

[icy sage]

Nah he would freak out thinking i am about to kill myself

[robust yoke]

Or he would interview you.

[icy sage]

Worst therapist in the world goes to ChatGPT

[robust yoke]

Christianity, I suppose, but I use profanity, so I am already a sinner.

Heh heh.

[robust yoke]

True.

[icy sage]

Isn't everyone a sinner according to the Bible? Even babies? Not complaining, genuine question

[robust yoke]

Not for those who have never sworn once or have gotten into a relationship (since that's a sin, according to the Bible, for whatever reason).

[icy sage]

The Bible is complex, no doubt there's an area that studies it

[robust yoke]

Well, yes.

Churches.

[icy sage]

I pretty much don't care if a God exists, i'm 100% sure neither heaven or hell exist

[robust yoke]

There's also religion school, I'm pretty sure.

[robust yoke]

That's respectable.

[icy sage]

I don't know if that exists here in Brazil, even as the most christian country in the world (and also one of the poorest 💔 )

[robust yoke]

Ah, Brazil.

Right, right.

I forgot we've met, so I didn't know.

[icy sage]

oh wait

i'm dumb

[robust yoke]

I don't think it does.

[icy sage]

we have THOUSANDS of Christian schools in Brazil, not totally dedicated to Christianity, but they strictly follow Christian doctrine

like

very strict

[robust yoke]

Makes sense.

Any place is bound to have at least some Christians, I'm sure.

[icy sage]

i think they're better than Military schools when it comes to discipline and doctrine

[icy sage]

Literally any place

[robust yoke]

Yeah.

[icy sage]

What is this? 💀

[robust yoke]

Some Chinese model, I assume.

[icy sage]

oh

created by Xiaomi

[robust yoke]

Ah.

[icy sage]

still aching 💔

[robust yoke]

"Aching".

https://tenor.com/view/biden-debate-president-presidential-debate-wtf-gif-4127151492472228747

[icy sage]

what

still aching still aching, oh baby i need your love

😭😭

[robust yoke]

Nah, I thought it was something else. 💀

[icy sage]

noooooooo

relax bro

[robust yoke]

Yeah, heh heh.

Anyway...

The Office parody episode: "Jim and the Boxing Glove" @remote arrow @plucky ember

[icy sage]

Stupid Altman, let me see the fight

[robust yoke]

L Scam Saltman.

[icy sage]

bro

[robust yoke]

Hm?

[icy sage]

Top musics of 2025

Can i get banned for this 💔