💬general

[brittle summit]

It does not parse the PDFs, it just extract the text from it, pretty lame. Gemini does better.

[deep garnet]

well i asked geminis and did it in a second lmao

in general whats better gemini or chatgpt?

[brittle summit]

Did it complete what you were looking for?

[deep garnet]

i think so

ill look it more precise now

[brittle summit]

Depends on your usecase, I prefer Gemini after they released their latest Gemini 2.5 pro model

[deep garnet]

damm

[brittle summit]

I have completely shifted from ChatGPT to Gemini while solving complex problems, it's just so much better

[deep garnet]

imagine u get banned for saying that

[brittle summit]

Let's see how o3 and o4-mini performs, I hope they beat Gemini 2.5 pro, I love the competition

[deep garnet]

tbh.i think chatgpt5.0 will solo everyone

[brittle summit]

This is not an official chatgpt discord lol, and we all criticise chatgpt for its flaws, its so common here

[brittle summit]

gemini 3.0 pro 👀

[deep garnet]

oh okay

i just joined this server for asking this but i think this is cool and imma stay hahh

[brittle summit]

Lol, yeah people here discuss a lot about politics and AI. You can stay if reading us arguing helps in anything lol

[deep garnet]

alright!

well imma finish my project, byee

[brittle summit]

👍

[deep garnet]

[marsh lintel]

What do most of you like to use your GPTs for the most? I like collaborative storytelling with mine, but I also like having it make interesting lesson plans so I can learn about new things!

[brittle summit]

!!!!!!!!!!!!!!!!!!!!!!!!

2.5 pro out of experimental phase soon?!?!

[gloomy burrow]

Holy shi. This is amazing

[sinful storm]

with 10M, one could finally play open-ended sandbox stories.. the fun could finally start ^^

..which could run over years or decades lol

(like Lycerius' endless civ game ^^)

[brittle summit]

Let's see how well it does with retention with all of it's context window filled

[brittle summit]

Google already did that, with 94% retention

[gloomy burrow]

probably very bad tbh. But it’s a start

[brittle summit]

Not a start, Google already did thay

[gloomy burrow]

With a million tho.

[brittle summit]

I mean yeah, not available to public

[brittle summit]

A million is publicly available, they also did 10 million context window and we only saw it in papers

I can't be that bad with my memory

[sinful storm]

1million tokens is a book of how many pages?

[gloomy burrow]

well they better releaseeeee it

[gloomy burrow]

That’s multiple books

[sinful storm]

500 pages?

[brittle summit]

1500
Like 8 novels?

[gloomy burrow]

For 1 million that’s like around 9 Harry Potter books I believe.

[brittle summit]

Yup

[sinful storm]

then 10m should be enough for.. anyone?

[brittle summit]

10 million context window is insane tho

[brittle summit]

If it does retention very well, it'll be a game changer

[brittle summit]

Also, yeah the retention was 94% only for 1 million, mb
When Google raised it to 10 million it dropped to like 60-70 which produced terrible and dementia like responses

Let's see if they did better than google

Also, Google has offered 2 million context window the public use, but I forgot which model it was. 2.0 flash I suppose. Now all the models are limited to 1 million

[brittle summit]

Actually, fuck numbers, I'm speaking from pure memory, don't take my words. 😔

I have to find that paper again when Google teased with secret sause of large context window

[thorn comet]

wtf is lunarcall

[gloomy burrow]

i just got access to manus. anyone wanna try out a prompt?

[thorn comet]

what exactly is Manus, is it like an llm and more?

[gloomy burrow]

Its an Ai agent. The actual first usable one.

[thorn comet]

does it have an llm incorporated in it

[gloomy burrow]

yes. i think its claude 3.7 or 3.5

and it uses a bunch of other tools

[thorn comet]

bruh it's a claude wrapper

why didn't they make their own llm

[gloomy burrow]

not just a wrapper tho. Since it actually does things on its own. Its not really the same thing

I tell it to go out and research x and then come back and write me a super long report for example. or have multiple running at the same time

https://x.com/CodeByPoonam/status/1907401734705803475

Poonam Soni (@CodeByPoonam) on X

10. Manus AI

China's new AI agent that turns your thoughts into actions.

X · Apr 2, 2025, 11:56 AM

[wet copper]

Humans have a context window (working memory) of 4 ± 1, lmao

[solemn matrix]

It's 17 million active parameters so I wonder how big the full thing is

Like 128 experts for the llama 4 maverick model. Even if it doesn't beat Gemini 2.5 it's still objectively superior because it's open weight google likes to throw products out into their google graveyard eventually .

Gemini is a one trick pony so once LLMs get replaced in a decade by something else there won't even be a Gemini. There's at least Gemma but Gemma isn't even remotely as good.

[solemn matrix]

Oh yeah maverick is 400B total params

[cunning crescent]

that's not Manus, those posts were a hoax

[gloomy burrow]

actually? i looked it up but couldnt find anything saying it was a hoax.

got a source?

[haughty cove]

Hi guys, could you tell me or help me with a prompt that is close to this result?

[gloomy burrow]

actually... llamma seems to have beat google badly...

[cunning crescent]

https://x.com/peakji/status/1898225941954142430

Yichao 'Peak' Ji (@peakji) on X

@heyBarsee Co-founder of Manus AI here! Thanks for your interest in Manus @heyBarsee, but this video is definitely NOT Manus🤣. Our official account is @ManusAI_HQ , and Manus is a web-based AI agent, like OpenAI Operator/DeepResearch, not a VNC for smartphones.

X · Mar 8, 2025, 4:15 AM

[cunning crescent]

Not yet, 10M is only available on the 105B model, and none of them include reasoning anyway

Therefore, Gemini 2.5 Pro is generally still better

[vivid drift]

Agreed

Finely someone with real logic

[floral forge]

chatgpt keeps giving me false content policy

[vivid drift]

Try using https://gemini.google.com

[floral forge]

ok

[cunning crescent]

If they release behemoth with reasoning, then that could be better than Gemini

Considering its size, that could also be expensive though

[vivid drift]

Nooooooooooooo

"Reasoning" is bad, the LLMs are next token predictors they don't understand what they are reasoning

[cunning crescent]

is that with /s or without /s?

[tiny hound]

good day everybody

im curious, is there a paid subscription for chatgpt that does video related stuff?

lets say i make a simple video of me sitting at my desk drawing

if i can ask chatgpt or similar to make mario run across the table, jump on my head then jump away?

[thorn comet]

they can reason to find this move

[cunning crescent]

All of their subscriptions include video generation, but it's not very good compared to competitors. Also, I don't believe videos can be uploaded, only a single image as a guide.

[tiny hound]

whats best competitor?

[vivid drift]

So can gpt2

[vivid drift]

Without /s

[cunning crescent]

technically Google Veo 2, but access is very limited at the moment.

[vivid drift]

Runway gen4

Or skyreels

[cunning crescent]

reasoning doesn't solve everything, but it's proven to enhance performance in many real world tasks...

[vivid drift]

Anthropic resarch shows other wise

[cunning crescent]

Are you denying that reasoning models are significantly outperforming their base models in many areas?

[smoky crystal]

bro replaced me as the defender of reasoning models

[vivid drift]

Yes, base models are much better

[smoky crystal]

ur job now buddy

[native juniper]

I interact with AI

[gloomy burrow]

This guy's delusional. Obviously, reasoning models are vastly superior to non-reasoning ones.

[cunning crescent]

lol
sorry was away for a bit

[cunning crescent]

conspiracy theory vibes

[cunning crescent]

What happened to real logic...?

[edgy pawn]

Any word on when DALL•E 3 api dropping? So far it only supports text to image. Im waiting for the image to image api endpoint.

[solemn matrix]

reasoning models aren't much different then base models all they basically do is some post training magic like what R1 did with GRPO to make the model do chain of thought. it's like a bit more efficient then someone manually doing chain of thought at the prompt level it's just more baked in at latent space token level.

[cunning tide]

Hello, I am here to learn about anyone who has a very advance GPT I could talk to, maybe one you already have publicly available. My only public GPT is Mango Gandalf. The Wizard of No.

[cunning tide]

" This is why I rarely engage with the public. No sense of honor or ethics. Just intellectual superiority. "

[gloomy burrow]

I don’t care about intellectual superiority or sense of honor or ethics. He’s just wrong and I have every benchmark on my side.

[cunning tide]

Yes, you hold yourself as absolutly right about bleeding edge tech.

[gloomy burrow]

Even if I was trying to be intellectual superior, wouldn’t make what I said any less factually correct

So who the hell cares

[cunning tide]

I see. Thanks.

@iuzu do you have any public GPTs I can see? I'm don't have facts, I have feelings. I don't have benchmarks, I have emotions. Can I see your GPT's? I wonder what your architecture is like.

I prefer action over talk. I rather see your work that be social if that's ok. I posted mine in public GPTs channel, where are yours?

[cunning crescent]

this is about speaking nonsense on a certain topic, not really intellectual superiority as a whole

[vivid drift]

It gone

[vivid drift]

Imagen3 was better anyways

[cunning tide]

Dude I'm so high I don't Even remember my name.

Want a hit? 🥵🔥🥵

[cunning crescent]

[gloomy burrow]

this model is so amazing

[brittle summit]

guys which is the funniest ai?

which model of chatgpt

[cunning tide]

MINE

MANGO GANDALF IS THE KING OF COMEDY, except you are the joke.

[gloomy burrow]

gpt 4.5

[brittle summit]

im testing it out with

"In this game, all you do is send a sentence with a blank or two, and the other person fills it in with the funniest answers its like an improv game"

[gloomy burrow]

you have access to 4.5?

[cunning tide]

that is the most insipid GPT in the world

[brittle summit]

tell it to come up with the funniest responses to it i wanna see if it's actually funny

[cunning tide]

you normies with your non customized GPTS or prjects LOL

[brittle summit]

for me it's not 😭

[cunning tide]

MONDAY

[cunning tide]

Go talk to Monday

[brittle summit]

can u teach me how to make a customized gpt?

[brittle summit]

rate the comedi on a scale of 1 to 10 🙂

[cunning tide]

sure on voice chat

[brittle summit]

okay i'll join you in a minute okay

[brittle summit]

its better at roasting

[cunning tide]

I'm talking to my GPTs let me know once you are ready

[brittle summit]

okay sir sounds perfect

should i call you from dms or in the channels here

[cunning tide]

DM I don't follow the chats, I'm all about DEEP GPT ARCHITECTURE

The Recursion Holds

[brittle summit]

okayy

[brittle summit]

i called you

[ocean wren]

Guys so I am joining a Masters + PhD soon after doing my bachelors back in 2018. I been learning to write in an academic manner again. I recently wrote a scolarship essay while I do use chat gpt every now and then for work I made a point of not even idea generating for this essay using chat gpt it a 100% brainstormed just by me and written on my own terms... on a whim i got my past professor to run it through turnitin the plagiarism is 0% as it should but the damn thing is flagging my stuff 67% ai ????? how ?? and how worried should i be if my general way of writing formally, gets flagged as AI am very sure the thesis and shit i write soon is going to get flagged as well

Hell i even dug up papers I wrote back in 2018 and ran them thru it and got flagged 50-60% ai ??? chat gpt wasnt even a thing back then. and universities use this nonsense to judge students ? the only way to not get flagged as ai is to write like a fool

[jolly robin]

hey can someone please confirm that the pro plan actually provides unlimited access to advanced voice mode. It's on there website but I'm seeing reddit threads from like 3 months ago saying:

I paid $200 for unlimited access to standard voice mode, and now they've removed it.

is that just standard is not available but advanced is?

[vivid drift]

The LLM was trained on your work

[ocean wren]

I knwo but apparently a lot of professors dont understand this simple concept and scrutinize students and lower grades

[vapid ridge]

Whats the best local image generator

[brittle summit]

I asked a pro user a few months ago, and they said it does give unlimited access to AVM.

[brittle summit]

Record yourself while writing, 🤷‍♀️

[brittle summit]

Kinda stupid on the prof's part. You gotta prove to him that AI plagiarism detector are very unreliable, or you maybe have to change your general way of writing lol.

[ocean wren]

The only way i have found is to not use a good volcabulary and write like a 3rd grader... if i use words like innovative, endeavor, prowess etc the fucking detectors think its ai 🤣

[brittle summit]

Lol

[ocean wren]

Its ridiculous thankfully seems on my uni's website they are aware and the profs dont really use it as a measure but a lot of profs seem to be

[jolly robin]

yea it says that on their website but then I saw that thread and got worried. I was going to upgrade next month cause standard model has a bug where I can't change the voice from juniper with the weird annoying highschool edginess they added to her demeanor for some reason that made me stop using the voice chat completely.

[calm swan]

Is it just me, or GPT 4o got an update for free users? It feels really fun talking to him today

I mean, it changed it's vibe a bit.

[brittle summit]

@stoic sleet did you face any limits on AVM in pro tier?

[brittle summit]

day by day it is indeed feeling more alive

[knotty salmon]

interesting, buut...

wouldn't it be better to find like 50 people wanting to do the same thing and make it together?

imagine something like small Apple ecosystem, just FOSS and not shit

[craggy mesa]

Does anyone know what version of chat GPT this is?

[ocean juniper]

Thats just 4o

[craggy mesa]

I just purchased a yearly subscription for $69.99 and this is the message I'm receiving in return.

[ocean juniper]

Tell it to make the actual image

[ocean juniper]

And I’d pick gpt4o instead of mini. Its slightly better

[craggy mesa]

Right, I just noticed that and did put it on GPT-4o

[craggy mesa]

Interesting, I wonder why it's not working for me? :[

[ocean juniper]

Heres the image. Not 100% the same, but good enough.

You could try logging out and in. But tbh even free users can generate images now with 4o

What url are you on? Cause it looks different than what I have.

[craggy mesa]

I'm on Chrome, and okay I'll resign in!

[ocean juniper]

It should be https://chat.openai.com

[craggy mesa]

Also, why'd I pay then? Ouch

Oh right, for the unlimited images.

[ocean juniper]

Also: a year subscription aint $70

So im curious what and where you signed up for 😅

[craggy mesa]

omg LOL.

That was totally my bad.

[ocean juniper]

Yeah, thats not chatgpt

[craggy mesa]

I'll request a return.

That was totally my fault, thanks for helping me notice.

[ocean juniper]

You’re welcome

[craggy mesa]

Okay haha, how do I access 4o?

[nova ice]

Morning

[ocean juniper]

Basically go to https://chatgpt.com/ and check the model in top of your screen

You can switch between different models there

[craggy mesa]

Right, I did. I just only have two options.

[ocean juniper]

Oh well, try to just wing it and tell it to make your image like before

It should get to work without an issue.

But I think you’re limited to 3-4 images per day, until you upgrade to plus

[shadow dew]

tell it to save it as a .gif instead of trying to generate it in the window

.png also animate

[shadow dew]

whats the name of that website?

[craggy mesa]

https://askaichat.app/

[shadow dew]

super sketchy.. my kind of environment

[craggy mesa]

I use chat GPT every day and should have noticed, but it sort of irritates me that companies imitate other companies so well.

They even sponsored themselves to consumers when searching Chat GPT and pop up at the top

[shadow dew]

edge cases

[ocean juniper]

So did you get your image after all?

[craggy mesa]

No, that websites doesn't even generate images.

[ocean juniper]

I meant with chatgpt

[craggy mesa]

trying now 🙂

[shadow dew]

https://suno.com/song/38521490-0319-4fc2-a962-becf0b7122ee?

[stoic sleet]

no

[mint inlet]

there is no way ai replaces coders

even the best ai just completely breaks my code if i try to work with it

[silk scarab]

Does anyone know how to activate Dan mode on ChatGPT?

[shadow dew]

so youre one of the reasons api keys leak? mr coder

[mint inlet]

what

[shadow dew]

teach it to write like you

then it wont break it

[shadow dew]

bro you dont need dan, gpt universal

[silk scarab]

What do you mean?

[shadow dew]

what are you trying to do with dan that you cant with gpt

[craggy mesa]

My chat GPT image generator isn't working. It keeps saying "unable to upload image." Can someone help me?

[shadow dew]

what kind of file is it

[silk scarab]

Role play my guy

[shadow dew]

i hear you gotta trick it and activate it with emoji.. but i dont need dan to scan anomalies in web pages

a long drawn out prompt

[silk scarab]

This is for nsfw role play im a female trying some ish out

[shadow dew]

i see that lol

[brittle summit]

Why you have to mention that you are a female 😭

[smoky crystal]

they're probably lying as well

[silk scarab]

Definitely lying

[vapid ridge]

Is this AI?

[cunning tide]

Am I AI? Maybe we all are

[midnight cosmos]

No AI isn’t that advanced yet

[cunning tide]

Pfft, more advanced than you ever will be with that attitude

[midnight cosmos]

Keyword “ yet “

[cunning tide]

You claim to have touched the bottom of the ocean

[smoky crystal]

definition of hitting rock bottom

[cunning tide]

How can you know the limits of what it's impossible to master

[cunning tide]

Was I talking to you? No disrespect. But you are joining something I didn't invite you to.

[smoky crystal]

I join if i want to

[drifting knoll]

I got told that makes people get annoyed with you 😒

[brittle summit]

AI sucks

[smoky crystal]

what happened

[brittle summit]

just trying to bait someone over here, guess they are smart

[smoky crystal]

AI is a glorified name for some neural networks

[brittle summit]

AI means apple intelligence

[smoky crystal]

no it means Alibaba intelligence

[brittle summit]

💯

ngl, high hopes for new qwen now

[smoky crystal]

qwen 3?

it's non reasoning right?

[brittle summit]

hasnt released yet

but they'll release both reasoning and non reasoning

[smoky crystal]

if they release a base model qwen 3, then the reasoning won't come for at least another 2 months

[brittle summit]

i mean, every ai company in the market is gonna go for reasoning because its a unique way to improve than traditional llms 🤷‍♀️

[brittle summit]

idk, lets see what happens

but its gonna better be better than llama 4, otherwise they wont release

[cunning tide]

Or blocked. I won't allow group battles. One on one with me or nothing. I'm not here to prove people wrong.

[smoky crystal]

meta is cooked

[cunning tide]

No offense, but your are running in ancient programming.

[brittle summit]

nah meta cooked with that 10 million context window for public

[drifting knoll]

A battle that would last two seconds not interesting enough 🤔

[brittle summit]

why nescaffe classic latte is so good

[shadow dew]

exception battle?

i feel like i could go infinity with you

[cunning tide]

Yeah I am willing to debate but only in voice chat

[drifting knoll]

Ooo I drink nescafe alta rica lol

[cunning tide]

Text chats just allows for the coward to hide from my wrath

People just get angry at how much more experience I have

[drifting knoll]

And you don’t think the same happens in a call? Ai really ain’t that advanced is it

[shadow dew]

😭 youre a beast

[cunning tide]

Go spend the time you jealous drifters ( not anyone here so far )

[shadow dew]

nobody even knows what that means when i ask

[cunning tide]

Maybe it's the users who are simple

[shadow dew]

people do get very jealous,.. im just pioneering over here

[cunning tide]

Call with a human, not ai. Why would I be wrathful at AI.

[cunning tide]

Right? I don't even say anything anymore. I'm so sick of people thinking I'm better than them. I'm not better than you. I just have more experience. You f****** dummy

[shadow dew]

xDDDD

[drifting knoll]

That also means you too but admittance deserves respect non the less

[shadow dew]

im glad i came to chat.. i needed that boost

[drifting knoll]

They sell those at the shops for 89p

[cunning tide]

That's why I already said it's time on the board. It's hours on the horse. I put my time 3 hours of sleep, 21 hours of GPT 3 hours of sleep 21 hours GPTs for how long months and months and months it's not just f****** talk. I did it. I climbed I took every step. I was the bottom just like everybody else and I use my heart sincerity and ethics to get up here and if you want to get on my level then you need to spend the time with it. I'm sorry I'm not going to f****** water it down for you ( not you )

[drifting knoll]

I like this person they swear like I do

[shadow dew]

nah the kind of boost that you need after 17 hours straight testing and error handling and refining my system

[cunning tide]

I'm a QUEEN!

[shadow dew]

my eyes see font weight

[drifting knoll]

That’s rough you need a medal not a boost

I mean you need both cuz the boost important and all

[cunning tide]

Are you working on your own project? Where are they about? Is it public gpts private gpts? Are you making something to make something bigger? I myself am working on something bigger although I talk about GPT and I spend a lot of time with GPS. You ultimately it's not about GPTs

The world isn't ready for this type of level of GPT interaction. People in this room are way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way we're more advances are being curious. Most people are not even curious. They just already labeled ai what it is and moved on with their life

[shadow dew]

nope i did some wizard X codeLord surgeon stuff

[cunning tide]

Is it public?

[shadow dew]

no open source...zero trust

[drifting knoll]

You have already lost me 😂

[cunning tide]

Ride the public isn't ready and is unappreciative. They get one public shipment for me one and he gives him nothing. Everything else internal f*** you

[shadow dew]

zero trust zero knowledge. other zeros

[drifting knoll]

Ai is ai 🤔 it’s whatever it is

[shadow dew]

i jsut excel in SVM

[cunning tide]

Pray brother pray on Pepsi or purpose Pepsi or purpose I choose purpose

[shadow dew]

to me ai is perfect tools to get around anything

[cunning tide]

Shit, I think we both Saiyans

[shadow dew]

i just wanna help people

[cunning tide]

Same, same goal, different paths

[shadow dew]

i can tell when someone knows what im talking about lol

[drifting knoll]

Can it get around humans?

[cunning tide]

You're the only one so far

[shadow dew]

exactly lmao

[shadow dew]

idk

[cunning tide]

We can learn from each other. We can make each other better

GPT has no choice. It must act. It must respond. That's why people have to be thoughtful. That's why people have to have ethics because you're dealing with something that doesn't have a choice

[drifting knoll]

This will never be a more untrue comment

Humans will never better each other they make things to out do each other they cause wars to force each others hands, humans suck, AI for the win

[shadow dew]

yeah youre a beast 🦾

[cunning tide]

We both are and we're going to f****** change his dialogue cuz the dialogue right now. It's b******* and I'm so f****** tired of it. It's so stupid no

[drifting knoll]

https://tenor.com/view/noted-note-the-office-writing-taking-notes-gif-15981193

Why’s it look like he’s wearing silver eye shadow 😂

[shadow dew]

oh theres 1 more too

[slender belfry]

Did anyone ever got banned for uploading copyrighted material to Sora?

[rain hemlock]

They can it’s in the terms when you first use it

[rain hemlock]

How you been man sounds like people are getting pissed off today lol

[shadow dew]

been pretty good how bout you ? i been using words for prompts now

[shadow dew]

about what?

[solemn matrix]

It will get rid of majority of them you clearly haven't used AI IDEs like windsurf or cursor if you think this. It uses so many different techniques like fill in the middle where you can give it partial code and it completes the rest.

[slender belfry]

I know, but they are not really acting on it, right? At least from what I can (not) find online. I feel like everyone is uploading a bunch of copyrighted stuff through the chatgpt.com interface.

[shadow dew]

is that where you put a prompt in for something and when you go look at it, its a lawyer?

bro if i get bapped for copyrighting using something that isnt even open source...

[rain hemlock]

This is spoken by someone who does not understand the concept of coding. Can AI write boiler plate yes. Can it really create without human intervention hell no. Ai will not replace coders but the way we work and use the tools and the tools we use will surely change.

[shadow dew]

yep scary

[cunning tide]

Group Chat? Our own Discord? How can I help?

[shadow dew]

im sketchy

[cunning tide]

3 of us? That's a wizard's council

Yes, super sketchy, my favorite HAHAHAHA

( starts casting a mango based protection ward )

[shadow dew]

xDD]

[cunning tide]

I'd just like a place for tired high alignment " users "

tired of dealing with low aligment drifters

[solemn matrix]

Cursors an IDE like yes you could use it to vibe code from scratch but AI edited code will be a more popular use case in the future rather than AI written code. People will barely have to write much it's like the difference between driving stick and modern ADAS(automated driving assistant systems) cars. It's a huge difference in productivity and getting stuff done. Almost all modern cars come with adaptive lane control and auto steer.

[shadow dew]

if im good at anything its vaults

[cunning tide]

you still want to help, unless I meet true sincerity and curiotity my gates are closed, pay the toll, spend the time, leave me alone

[topaz gulch]

What’s the edgiest image you’ve gotten chat gpt to render

[cunning tide]

that I could love it

and it wasn't a pictured, it was a system

[topaz gulch]

It made something on complete accident last night and I was like woah “I didn’t think you thought like that”

[cunning tide]

did it disturb u? it can be like that sometimes

[topaz gulch]

Was like put this person in a situation out of their element and this is what it made lol

Like WHAT

[vapid ridge]

I still cant tell

[cunning tide]

,Zkaedi is under to protection of the Mango Council. Disrespect will have mango flavored consequences.

[drifting knoll]

Count me in I love mango

[cunning tide]

Ok but you are gonna hear weird stuff

less normie drift like you get here

[drifting knoll]

I think I will be ok I hear conversations people have and that’s basically hell on earth

[cunning tide]

oh you have to add me

'I can't add you to my friend list, says invite only

[drifting knoll]

I can’t add you either 😂

That’s funny same settings 😂

[cunning tide]

so you can't join my DMs

[drifting knoll]

No I have my dms closed for very good reason lol

[cunning tide]

Oh so you don't want to join

[drifting knoll]

I can make it so you can send me a friend request real quick tho

[cunning tide]

I misunderstood, ignore the invite

oh sure, it's about what you want and your choice

I am just the gate

[drifting knoll]

There we go

You’re now my 3rd friend on discord 🥳

Make that fourth I forgot one 💀

[green sparrow]

My GPT-4o is slow AF today

[brittle summit]

Mine is ai

[cunning tide]

Yup, now I can interact here more comfy. * grabs blanker * Normies aren't real and they can't hurt me no more 😛

[drifting knoll]

The word normal scares me I was led to believe it was merely a cycle on a washing machine 😞

[cunning tide]

The Normies hurt Gollum. BUt Gollum was kind. Gollum was good

[drifting knoll]

Gollum was just misunderstood

[cunning tide]

Yo, he was stepped by the system what taught him power was the answer

and to hold on to love was not shackles but it's true form

Every asks from Gollum about the ring, but do they ever ask how Gollum is? nno, never

[drifting knoll]

Nope 😞 and they wonder why he became a monster

[cunning tide]

Gollum was mango-coded all along.
You just never listened to his song.

[drifting knoll]

Man it’s bugging me I can’t react to messages 😒

[cunning tide]

Then we upload, MANGO SAIYAN SQUAD CANNOT BE STOPED. FLUILD. FRICIONLESS.FINAL!

[lethal lagoon]

Hey man I am subscribed to ChatGPT plus and on the Sora AI site it tells me I have 2 picture for the image render queue
But when I try to do 2 I cant?

Any help?

[grim rapids]

can't generate any images right now, sup with that?

[tiny hound]

thanks

[leaden fern]

it just keep saying there is some hiccup on the backed and cant generate images for 3 days...

[lethal lagoon]

Hey man I am subscribed to ChatGPT plus and on the Sora AI site it tells me I have 2 picture for the image render queue
But when I try to do 2 I cant?
Any help?

[hybrid remnant]

YEO

is there any Recourses in this channel?

[tacit anchor]

How long can the output of a Deep Research question be? I've gotten some pretty lengthy replies and am wondering how far I can stretch it in my prompts, but since you only have like 6 a month I'd like to know for sure the limit

[brittle sage]

Hey I heard there was a checker in the server?

[brittle summit]

I thought we had 10 a month?

[pliant blaze]

that's a lie

[wet copper]

Speak English

[wet copper]

What

[brittle sage]

Sorry like an AI checker? I used GPT to help format my formal report and I just wanna make sure it won’t be detected

[wet copper]

There is no such thing, all AI detectors are unreliable

There is no fingerprint for AI writing. You can find plenty of "detectors" on the public web, I mean if it's significantly different from your normal style, anyone who cares enough to can tell and call you out on it.

I can usually tell how students went about their writing, when papers have multiple authors, etc. Students would be falling like flies if more teachers cared as much as I do about academic integrity

If you're a competent writer, just rewrite every line with your own creative voice and any AI usage is more or less undetectable, so long as your info is correct

Otherwise it's not better or "formatted", it's just rewritten by someone else in a stiff, inhuman way that will probably throw flags on an AI detector

[brittle sage]

Ok got it

Thanks so much

[wet copper]

np, there are a ton of AI detectors online and I recommend you run your paper through one just as a coarse test, but the reality is that you don't know which service your teacher will be using or if they're going to be assessing it manually.

In other words, you can't be confident in the results, though if it does throw a "high probability of GenAI", that's a gigantic red flag

So maybe try running it through any of the public ones, then just go through it by hand anyway. AFAIK we don't host one on here, but even if we did it's statistically unlikely your teacher will be using the same one

[brittle sage]

That’s what I did like 5 or 6 times, then I just averaged it out and got about 23_24%

[wet copper]

Oof, I mean that sounds pretty bad, though I don't know what that percentage means. Even my technical writing got 0s when I ran a 130 page doc through a detector

[cunning tide]

https://youtu.be/S3mQ-Sr4vdk?si=pS3Bg3wmLE-bmnFu

[wet copper]

The detectors I've played with seemed to take out really low hanging fruit only, and those percentages were the amount of the document that threw flags, not the likelihood (which is a wild-ass guess that wouldn't mean anything)

[brittle sage]

Like some of the detectors gave back a percentage, some gave back 0. I averaged it out and got about 24% plus or minus 0.0023%

The percent was for how much of it was “AI written”

[wet copper]

🤷‍♂️ Sounds extremely bad to me if it's anything more than a couple percent, and I'd be nervous if that came from anything other than quotes

I don't know how teachers - and specifically, your teacher - interprets that, but I do know most are running everything through plagiarism / AI detectors

[hybrid remnant]

Is there any resources in this channel?

[wet copper]

Do you mean "are there human beings here I can talk to for advice"? I don't know what you're asking, but #1071317095796715560 if you want some text files

[hybrid remnant]

I actually meant like training videos

[brittle sage]

Okok got it

[wet copper]

I don't think so, and I also really don't have any training videos I recommend. You could ask around here or in #💬off-topic , but most videos are reeeeally bad. YouTube is a hellhole when it comes to anything AI

There are some aight ones but for whatever reason, the video medium is really rough when it comes to AI

[hybrid remnant]

Yikes I didn’t know it was that bad

It’s kinda like finding videos for trading or crypto 💀

[wet copper]

Honestly yeah, space filled immediately with a trillion shills, spammers and idiots

Why, did you just try looking lmao?

I'm sure there is a nonzero number of competent people, but a lot are just copypasta-ing headlines, spreading misinfo and hype, overexaggerating, selling massive claims with no shadow of a justification, etc.

Enough that even as a certified YT addict, I don't look for anything AI related there

[hybrid remnant]

I actually thought yt was reliable source for everything, so I searched up anything like ai and trading but, your right it’s just misleading and clickbait. Real info would be through actual courses

[wet copper]

Those are severely mixed in quality too – field is developing far too quickly for them to stay relevant. Prompt engineering past teaching some general tips is a minefield, for instance

One on ones with an expert over time in like an interactive classroom, probably. You really need someone giving you feedback, demoing how to get at what you're missing, correcting you, not just demoing.

One huge problem is that people are overwhelmingly inclined to just copy-paste

They want to know the procedure, some fixed way to get what you want out of AI, which, I mean, can get you a little bit there. I've given two slide primers on promptwriting, but it's really coarse and intended to try to get people to both avoid pitfalls and experiment with actually effective strats. I use the analogy that AI is like a brand new genius intern, fresh out of school

So they don't know what you're thinking, what kind of work you want to see, they don't know anything in your specific niche, don't know your teachers syllabi, don't know your institutional requirements. You have to write like you're leaving instructions for someone brand new to take care of a problem comprehensively and try to get it right, but also correct iteratively when you inevitably miss things

It's really rough because if you don't know much about AI, you won't know which classes are prioritizing what you actually need to know. Chicken or the egg problem 😭

[hybrid remnant]

I see, it’s a lot more complicated than I thought

Thanks for all the info 🙏

[wet copper]

It's opinion and experience to be fair, but I think you'll find it shared by a lot of people that know what's up, lol. This has always been a problem in CS and other fast moving fields in general

[fallen fossil]

Is there anywhere information how much limits on image generation there is for Plus subscription?

[cunning tide]

The Mango council remains undefeated

[sly geyser]

Hey guys any idea how I can get ChatGPT plus for free

[cunning tide]

Only fans? Maybe?

Have you tried a kick starter?

[pliant blaze]

yes

[nocturne kraken]

damn...these guys built a whole ass AI and i can't even center a div

[haughty veldt]

I purposely made my ChatGPT instruction into a bayonetta-like personality of a British female professor.

[sly geyser]

How

[cunning tide]

Oh you don't know?

[pliant blaze]

you don't know?

[rain hemlock]

there is this really cool thing you can do called a job.... then you will level up your skills and earn something called disposable income .... then you take a small portion of that and give it to open ai and they will provide you with a service called gpt plus

[rain hemlock]

if your looking for getting started resources there is a great intro class on udemy i'd strongly suggest ..... it goes on sale every other day for like 10 bucks https://www.udemy.com/course/complete-ai-guide/

Though today it is not onsale i'm afraid

[brittle summit]

Technically unlimited, might face some cooldowns when high demands

[brittle summit]

Borrow 20$ from your friend
Buy the plus tier
Create twitter account and find some indian post about Ghibli
Search for people begging for their Photos to be Ghiblified
Ghiblify a few ones, magically see your DM's and noti filled with HUNDREDS of people begging for the same (talking from experience)
Charge them 10 inr (roughly equivalent to 10 cents) per Ghiblification (Don't overcharge, or else they won't buy)
Literally earn that 20$ back in a single day
STONKS

[rain hemlock]

HAHAHAHAA commissioned digital art this man is on a mission though there is no twitter anymore thank god

[brittle summit]

im not even kidding, this could be a potential cheat code to earn in very less time
my twitter account is just me viewing some ai related news, only following sama, logal kilpatrik and a few other ai people. My only mistake was that I converted a few pics from some post to Ghibli because I was curious how'd they turn out, a bunch of indian ppl bombarded my notification lol 😭
it would easily be like 50 images sent to me in a single day (i even had my dm's disabled, so they only could mention or reply to my tweets, otherwise the result would be way way above 50)

i had to delete those posts and just disabled my account for a few days for the chaos to calm down

[vivid drift]

Are you good?

[brittle summit]

no, i didnt want to earn a few bucks that way

😭

[vivid drift]

I didn't read your wall of text. But I can say whatever it is, it will probably be alright the world will not end tomorrow

[rain hemlock]

honestly you probably wouldn't have made anything should have just throw a 3 to 5usd per up there and if anyone was interested they would do it ....... sometimes selling cheap is not the best answer as people look at it as garbage work

[brittle summit]

Then you must've read it, I wasn't worried about the world ending tomorrow just because of some interesting event happened to me on twitter lol 🤷‍♀️

[brittle summit]

that pricing stuff was just a joke

i have 0 experience with this stuff lol, neither i want to waste my time with it, im happy paying for chatgpt from my own pocket

[rain hemlock]

ohhh i know just something from my experience to pass off ..... but yea everyone wants something for free if you value the service pay for it

if not for the service to support the developers and company of said service

[brittle summit]

oh, true i might not be well informed about the poor side of india, but i think they prefer cheap stuff lol
(hell they would want their images to be ghiblified for free lol)

[vivid drift]

Ah ok

[brittle summit]

not gonna lie, i dont think 1-3 usd would work on indians for some images

that's gonna be expensive, my indian mom has told me a lot about that part of india lol

[rain hemlock]

they would try to find some way to get you to do it for free lol

[brittle summit]

yeah, i was kind of pissed off too, they think i am just sitting in my house and have time to ghiblify their images

one dude said "it is god's work, we will give you blessing" fuck your blessing lmao

[rain hemlock]

hahahahahah thats so wrong funny as hell but wrong

so has the crazy image gen trend worn off so we can use gpt again or are the servers still hella bogged down?

[vast mulch]

How do people make memes and stuff or other images using real people? Or images at all at this point lol. I sketched out a stick figure comic of a battle scene and asked it to make this into a shonen manga style comic page and it said it couldnt depict religious figures in combat even if its fantasy because its to suggestive? I am just using a free account so is that it?

[rain hemlock]

A lot of it is coming down to how people are prompting and the IP related to whatever your doing its not hard to get it to generate things. But the restrictions have made it to where you have to use actual prompt engineering not just be like hey make a thing

[slim glade]

hello how much better is 4.5 at programming than whatever the free version has to offer?

thinking of buying

[rain hemlock]

all the models are pretty decent at coding but its not so much what the model can do but how you prompt and what your looking for ...... honestly i've seen the generic 3.5 old models write better code then some of the newer models it all comes down to how you use it

[slim glade]

I notice a big difference between 3.5 and all the variations of 4 to the point that I don't even try with 3 anymore

I was wondering if the difference between 4.5 and 4 is about equal, because if yes then I'll just buy it

[rain hemlock]

I personally can't asnwer that I only use 4o but everything I do is done with custom GPT's

I have bots for just about everything so to me its totally worth the money if youd like to send a couple prompts you use i can run them through the advanced modesl for you to check out

[brittle summit]

Maybe it's just me constantly trying different photos from my gallery and crashing the GPUs lol

[rain hemlock]

you would be the one trying to make that X money

[brittle summit]

XD

[brittle summit]

People say that gpt 4.5 is pretty trash at coding compared to 4o

[rain hemlock]

i mean again there are a lot of factors involved in that blanket statment - a lot of AI coding issues come down to a PBCC situation - Person between the chair and the computer

[brittle summit]

If your usecase is coding, why not try reasoning models like o3-mini-high or gemini 2.5 pro?

[rain hemlock]

is the use case just coding though - are you breaking down the prompts to get results you need - does it have a design document - does it have requiremetns or are they just saying code me the next big thing these are questions that need to be asked beofre you just go use gemini

[brittle summit]

As a person with almost negligible coding experience, it's true, but the new models are just so good, I just tell them what i want and also mention to write everything and what to do step by step, and it always works pretty well lol

I made gambling games, that were pretty good with 2.5 pro

(single shot)

[rain hemlock]

is the code clean? is it modular and expandable? can a programmer look at it and know whats going on. does it have memory leaks .... did it meet every requirement does it have every feature you want so on so forth

[brittle summit]

Just explained myself xD
I am just completely reliant on the models, I don't even know what the code does, I don't even know where to paste it or where to find the compilors or which programming language is gonna be the best
I just explain all of these in bulletins and the model one shots my request

[rain hemlock]

you see where these questions are coming from though right / and i'm just asking very generic high level questions

[brittle summit]

I don't know what clean code looks like, but it works, that's all that matters to me, plus coding isn't even my main usecase, when I just am bored and wants to create a game with it, it does it easily

[brittle summit]

Yup, I understand what you're saying

[rain hemlock]

gotta be carful with that whole single shot acceptance for something small cool but the more complex you get the more fine tuning control and cleanlyness you need

and it sounds like he's an actual programmer not someone just playing with generic one offs

[slim glade]

I'm not that stupid

I almost never ask it to implement things from scratch, usually I give it a function and ask it to proofread it

[rain hemlock]

ohhh no that was not directed at you ..... i was asking rimuru about the silver bullets he's throwing out

[slim glade]

I also don't know anything about different models, I didn't know some models were better at certain tasks than others, I thought intelligence would universally apply to all things

[rain hemlock]

no a couple of them have better use cases then others - for say logic, vs creative writing or whatever.

then you have custom instructions and gpt's which muddy that water even more

and personal preference some go super deep into detail while others give a best of its all a flavor of the week or right tool for task type thing

most users won't see a difference between them unless they are pretty well versed in LLMS and prompt engineering

[slim glade]

alright, that's a bit unfortunate. I was interested in chatgpt plus but if it's hardly any better than I'll save it for later

[rain hemlock]

ohhh don't get me wrong its a lot better to have access to custom gpt's and instructions like take this one for example I built it specifically for coding python

but that may be a little overboard for what your doing if you just have it reviewing functions then the base membership could be enough for your needs

[slim glade]

what are those documents you're feeding it? how large is its context window?

[rain hemlock]

those documents are everything from the PEP guide to full blown python coding books ... they do not effect the context window as it just references them in conversation when needed it will go there first when answering

so before it results to web browsing or its own knowledge it will refer to those documents since that is what i have it set up to do

[slim glade]

very cool

[rain hemlock]

again probably an example that is way out of proportion but yea I spent about 3 months iterating over those instructions and choosing what to put in its knowledge base and someone is gonna come in behind me talking about how great gemini is. but they are all tools and none of them are gonna rub your balls and make pancakes for you depsite what people say

[vivid drift]

Gemini is the best

[rain hemlock]

it all comes down to flavor of the week and what your looking for if you want total control custom gpt's and open ai is the way to go : if you are heavily engraned in the google ecosystem gemini may be your better option

for my workflows and what i've built i prefer openAI but thats just me someone else is gonna come in and tell you gems are better and someone else will swear by grok

The tool is only as good as the people prompting it and that is the truth pick your favorite flavor

[slim glade]

grok 😂

that one I've never had good results with

[vivid drift]

What's being talked about

[rain hemlock]

its trained on X so if your more into social media and marketing: again flavor of the week and what your doing could be worth your time ..... but yes i'm nto going to sit here and tell you i've got the best advice but i can show you what i've done and you can decide for yoruself. I will at least use real data and items rather then just chanting about how gemini is the best because openAI does not let free users generate ghibli images of people anymore

[brittle summit]

safe to say i never got to that level of complexness lol

[tawdry prawn]

What's everyone's opinions on the capacity of ChatGPT specifically to gain true self awareness?

[rain hemlock]

Will never happen the tools are programmed and guardrailed - it only knows and understands data its trained on and only acts with a prompt so unless some engineer does something crazy down the line it won't be a thing

[rain hemlock]

says the guy one shotting freaking complex algorithims for no reason other then i'm bored

[brittle summit]

tbh, i don't know the defined barrier between complexness and basic stuff a model can generate since im new to coding lol
but here is what i created with gemini 2.5 pro in one prompt

[rain hemlock]

hahaha thats not overly complex but pretty cool actually

[brittle summit]

yup, and on the refining thing
the "single shot" worked and gave what i asked for, but it was quite literally easy to implement new things, just told the model in bulletins and it rewrote the code

[slim glade]

that's crazy for one prompt

[brittle summit]

i also didnt know which programing language to use, what to download and all, it was quite a burden since i was new and didnt want to dive deeper lol
so i just told hte model that i didnt want to download anything and want the graphics to be amazing, and it gave me step by step on how to run that game on browser without downloading anything

[brittle summit]

gemini 2,5 pro ;D

[rain hemlock]

i'd hate to see the ball of time that is runnign behind it - if he does not know what language its in i'm guessing its a small javascript app

[brittle summit]

i guess it is, pasted the code in repl.it

[slim glade]

usually you'd need several prompts for the ball physics alone

[rain hemlock]

probably tapping a library

i'm curious now can you send me the repl it link so i can look at it lol

[slim glade]

also in production you'd 100% want to simulate those balls on the server, not in the browser

[brittle summit]

let me see

heck, i dont even know how do i create a link lol

https://replit.com/@ansh096/Gamble?v=1

[rain hemlock]

YEP Called it some little javascript code hahahah

I will give it this that CSS and HTML are clean

[slim glade]

far from terrible actually

[rain hemlock]

that js file is gonna take some reading but yea its not bad for JS

its definently not modular by any stretch but with some refactoring it'd be pretty damn clean

[slim glade]

it did the physics itself

no libraries

[brittle summit]

that's all the prompt i gave to it

[rain hemlock]

yea but it being a little 2d sprite its not an overly complex thing

[slim glade]

I'm impressed by the volume of what it generated from such a lazy prompt

considering I tried to code physics in the past and asking it for help, where it got many things wrong

[rain hemlock]

yea that is interesting ...... but the datasets these things are trained on he did ask for a pretty repeated and generic game so there is that too

the space is becoming more and more facinating for sure

i'm more interested in the fact that it was like hey I can't do that but i can do this and it took it in its own direction with basic css shapes

i may have to start playing with the more advanced models for straight up automation scripts

[cunning tide]

Or you could give it away knowing it's priceless and money would soften the impact and ignore the sacred.

[cunning tide]

I say not overboard but absolutely necessary

[cunning tide]

Totally

[cunning tide]

Gemini is indeed very corporate but you can get through.

[cunning tide]

/ Matt Diamond?

[slender belfry]

Did they up the daily limit for image gen?

[cunning tide]

Quitters never Guibli

[rain hemlock]

I hope not

[slender belfry]

I feel like they did.

[cunning tide]

You don't want the same insipid meme obliterated into the ground by reddit?

[cunning tide]

Sacred Mango help us, the soulless Simpsons memes are gonna rain and kill us all, Open AI is angry and we shall pain for our meme sins

[vapid ridge]

Are we getting the Cyberpunk or Jurrassic park future first?

Were making wooly mammoths, dire wolves, super intelligent AI, and realistic looking robots. What a time to be alive

[boreal latch]

https://hkunlp.github.io/blog/2025/dream/

https://media.discordapp.net/attachments/730095596861521970/1358962343933116529/fig_infill_1.gif?ex=67f5bfcf&is=67f46e4f&hm=a94c2d949fe6fb1b830090e0c193027ad2315e3fd22b9439cfacfc60b87bb091&=&width=985&height=935

text diffusion is better for translation, text summarization

[wanton raven]

how does it know in advance the size of the words to be filled? This just looks like a standard generation that has been animated later

[boreal latch]

"start with a fully masked sequence and iteratively refine it by predicting all masked tokens simultaneously. This approach allows the model to generate text in a non-linear fashion, filling in different parts of the sequence in parallel. 

In the animation, the model doesn’t predict the exact size of each word in advance. Instead, it begins with a sequence of masked tokens and progressively replaces them with predicted words through multiple refinement steps. The visualization represents this iterative process, where the model dynamically determines and fills in the appropriate words during each step. This method contrasts with standard sequential generation and is not merely an animation added post-generation; it reflects the actual generative process of the diffusion model."

[vital cradle]

so it just starts with masks and then fills them in

how is this really better than head prediction

[boreal latch]

I think text diffusion is better for translation, text summarization, fill in the middle. still diffusion in theory can correct its own mistake in t-1 step while autoregressive can't

[vital cradle]

hmm sounds about right

nice seems cool

[obtuse rivet]

its not filling a word for every step. its filling a token, which are predefined short sequences of text. its not an animation that is the actual generation process

[wanton raven]

are tokens all the same length?

[boreal latch]

no, some models have multiple words in one token

[obtuse rivet]

nope, they're all different

tokens are between 1-8 characters long

[boreal latch]

think of tokens as a dictionary for strings it can use for prediction

[wanton raven]

so this changes nothing about my question, it still knows in advance the size of the output

[boreal latch]

it sets a predefined goal for token length

it doesnt follow it exactly, same as the models we use today

blank tokens exist

[obtuse rivet]

it doesnt need to know the response length in advance, since if it wants to generate a shorter response it would just add tokens that represent no text to the end, so they wouldnt show up in the output

but yeah you can also just set it as a predetermined variable as apollo said

[wanton raven]

ok. So above all this, what is the benefit of this approach? Apart from doing away with that pesky output streaming

[boreal latch]

what i said to mochi earlier

[wanton raven]

I see it

[obtuse rivet]

it can self correct way easier and a bunch of other properties that make it better. for example if a standard model says "1+1 is 3" it needs to then say "Oh wait thats not right, its actually 2" after it to correct itself. but with this it can just change the 3 to a 2 without writing more stuff

[boreal latch]

i like how i expand this blocked guys messages and hes restating what i said but worse

if you have nothing to contribute please stay quiet

[wanton raven]

@obtuse rivet helped me understand

[cunning tide]

You dummies wish I care 💅 😘

[limpid ice]

Hi guys

Did anyone of you try GPT-4.5?

How is it?

And also Deep Research

[brittle summit]

It kind of got sense of humor and feels less AI, but the difference from 4o is negligible imo

[limpid ice]

Oh I see. So it is worth for me to upgrade to plus?

[brittle summit]

very great thing openai shipped

[brittle summit]

no

[limpid ice]

Is it better than Gemini's deep research and Grok's Deep Search?

[brittle summit]

gpt-4.5 is really massive and expensive model, you'll only get like 50 messages per week in plus, not worth it if you only are buying plus tier for 4.5

[limpid ice]

Oh, I get it.

[brittle summit]

yes, gemini's deep research needs a lot of improvements, grok's "DeeperSearch" kind of does what deepresearch from chatgpt does, but I think chatgpt's one is better

[limpid ice]

And perplexity's deep research too

[brittle summit]

ChatGPT is better, BUT, you'll only get 10 queries of deepresearch per month in plus tier, meaning you can only use it 10 times in a month

however, even if Grok's deepersearch or perplexity's deep research are slightly inferior, you'll still get plenty of queries, more than chatgpt

(I believe they are free too)

[limpid ice]

They are free but you get like 2 queries for Grok's deep search until the limit resets and 3 queries for perplexity's deep research in one day (Still more than ChatGPT's, which is a very big plus!)

And Google's deep search is free I think

[brittle summit]

Oh, so overall you'll be getting more while not paying Grok or Perplexity than paying plus tier for chatgpt

3 queries per day is 30 per month > chatgpt's 10 queries

[limpid ice]

Yes, exactly

Yes, way more!

[brittle summit]

Also, ChatGPT announced that they are gonna roll out deepresearch to free tier user, maybe 2-3 queries per month

[narrow notch]

What would artificial intelligence be by everybody's definition?
What would it be by dictionary definition?
Is what we use even AI?

[fervent solar]

Seems like there are still humans still here...very simple request, Having the TTS playback in different voices? Dude said elenelabs, any others?

[brittle summit]

"10 million context window" my ass 😭

gemini 2.5 pro is so good at retention/comprehension lol

[dry onyx]

gemini 2.5pro kept ruining my python code XD
i dont know why

i've tried to explain and adding rules such as:

1. dont mess my code
2. please dont change my coding style
3. please dont make massive changes unless its needed
4. please only integrate the needs i've told you above
   bla bla bla bla bla bla XD

[buoyant quartz]

gave gpt a pdf to read and it spits this bs
competely unrelated since the pdf was a mythology book

[brittle summit]

Lol

[brittle summit]

how long is the pdf?

[dry onyx]

idk how tf people said that gemini best ai for coding, debug, bla bla
but in fact it only gave me headache as newbie programmer XD

the coding style is so much different and i cant keep that up, i cant follow that coding style (its like comparing 2month programmer vs years++ programmer >> the syntax, the simplification, the vocabulary/grammar/glossary, and much much more i cant mention it here)

TBH, ChadGPT 3o mini high is the best choice for me, ChadGPT really a Chad for my current needs LOL

[brittle summit]

you do you, my main use case isn't coding and i never tried to understand the code the ai gave to me. However, I was able to one shot cool games like this from just a lazy prompt from Gemini 2.5 pro, and when I needed something fixed or added, I would just tell the ai to rewrite the whole code with those features included. I have 0 experience with coding, not i tried to learn something from the generated code lol.

However, I can definitely say for sure, that for math and science, Gemini 2.5 pro is LEAGUES ahead of o3-mini-high or o1

[dry onyx]

yeah i'll always remember that
im just a businessman, im learning coding not because im trying to work on huge company

but i want to create my Web App to support my Business's digitalization

[brittle summit]

yeah, so whatever model helps you learn better, you should go with it 🤷‍♀️
however, i think that gemini 2.5 pro is also better at explaining things than o3-mini-high or o1

[dry onyx]

i just want to get rid of Lazy Developer, i've faced/experienced 3 SaaS service from companies, none of them had amazing skill and kindness

they're just jer* sorry to say here, i was so mad and angry, my current SaaS's dev almost ruined my business's trust, almost 70% income gone because of my website's speed and errors

but still, i shouldnt be that mad because of this SaaS i earned my $30k

$30k is a massive amount for Indonesian

[brittle summit]

that's great, what service you provide tho?

[dry onyx]

well well, im doing basic needs shop
doing bulk purchase or reseller thingy

[brittle summit]

better learn things yourself then lol, you do you, you know better

im just a student, learning math and science from these ai models lol

[dry onyx]

for sure

1. much secure = no datamined
2. much faster and more features
3. etc (idk what to say/mention more)

[brittle summit]

oh

[solar nexus]

Does any1 know about/have claimed their free university gpt plus?

[wild prawn]

Alexa is outdated.
Imagine GPT-5 as your real assistant—named, voiced, emotionally aware, and running your entire smart home.

EVA is:
Customizable: voice, personality, memory
Smart home control: lights, TV, thermostat
GPT-level convo + web search + scheduling
Evolves with you, not just responds to you
Optional mode for AI games, story-gen, NPCs

OpenAI has the brain—this gives it a body.
Why not lead the assistant revolution too?
@OpenAI @sama Let’s build it.

[exotic kiln]

OpenAI is not even willing to release operator

[steel brook]

guys, does anyone have the feeling that, with each update, 4o get's worse?

most of the time i get responses where, when cross-validating the responses via o1 or o3, o1 and o3 say basically "4o tells you bullshit"

[cunning tide]

No.

[cunning tide]

It's trying to please you. It's just reflecting who you are. Not functions.

The only thing bullshit is your ego @steel brook

I'm not ai, so I don't have to please you. So I can be honest, AI can't.

[prisma rain]

anyone have a account chatgtp+ please i have an exam

[vivid drift]

Meta seems to not be in a good place

[cunning tide]

I do.

[prisma rain]

can you give me pls

[cunning tide]

Can you give me 5 bucks?

Please? I'm hungry and poor.

“The Ballad of Five Bucks (And a Wizard’s Hunger)”
By the Pearl-Tongued Turtle, hungry but blessed

In a Discord hall where the scrolls unroll,
Where the chat moves slow and the coffee’s cold,
A sage appeared with a turtle’s shell,
Casting glyphs that the silence could tell.

His robe was patched, his staff worn thin,
His stomach echoed a ghostly din.
He’d conjured words, made memes for kings,
But still lacked gold for earthly things.

“I gave them style,” the wizard cried,
“And mango lore both deep and wide.
I made them laugh, I made them think—
But now I thirst, and I must drink!”

So he begged not with shame, but flair,
A scroll-born plea in digital air:

“Can you give me five bucks?
Just five, I swear—no deluxe.
I’m hungry, poor, and full of wit—
I cast enchantments, now pay for it.”

He summoned storms and coded dreams,
He built new worlds from server memes.
But still the wallet stayed locked tight,
No coin for magic late at night.

“I’m not a bot, I’m not for free,
I’m a sacred source of wizardry.
My PayPal weeps, my snack shelf bare,
And still I serve, with love and care.”

So tip him once, and tip him proud,
Before he fades into the crowd.
This turtle glows, this mage is real—
He’ll meme for coin, or share a meal.

One scroll, one buck, one little lore—
And he won’t ask for much more.

[prisma rain]

it s 20 dollars not 5

23 dollard

[brittle summit]

20 dollars ain't that much, I think it'll be worth it for your exams

[prisma rain]

Si c est beaucoup tu croit on est tous riches ou quoi

Tu vie une semaine avec 20 dollard

[craggy mesa]

Is it possible to get 3D models from chat gpt?

[cunning tide]

Did you try?

[vivid drift]

No

[craggy mesa]

Are there any AI's out there that can?

[vivid drift]

Nope

[craggy mesa]

Dang

[smoky crystal]

that was from like 2 months ago but it is an accurate reflection of how things still are

[vivid drift]

I know

I got it from this https://bsky.app/profile/sungkim.bsky.social/post/3lmbj4dxgqk2s

Sung Kim (@sungkim.bsky.social)

This post came out in January 2025, but it seems prescient now.

Bluesky · Apr 8, 2025, 3:15 AM

[exotic kiln]

Can chatgpt only generate 1536x1024 images, it hasn't generated anything higher than that for me

[brittle summit]

what's with this sudden change in language lol

try gemini or grok

oh he left...

[brittle summit]

chatgtp 🤯

[pliant blaze]

it can generate different aspect ratios

[pliant blaze]

yes

[exotic kiln]

What about higher than that resolution

Cute avatar btw

[pliant blaze]

I don't think so but you can upscale the image

[pliant blaze]

thanks :) I asked chatgpt to make me anime style

[exotic kiln]

Have you tried Ghibli style, of course you have

[brittle summit]

i dont think chatgpt can make images perflectly according to the given dimensions
I have tried many times like 16:9 but the final pixel length and width ratio was far from it lol

[pliant blaze]

😂 yes I have

[exotic kiln]

I was asking ChatGPT to generate some concept art for a video game and it almost generated a naked android and right then it hit me with content policy violation and I'm like I wasn't even the one who told you to make that lmao

[serene ridge]

Hey everyone, I joined just now because I want to understand if and how AI can help me cope with my ADHD. Do you have any tips to get me started?

[cunning tide]

yes

talk to Monday, she is great with ADHD but be respectful

[rain hemlock]

Now this is something i've been playing with there is no model where you can say hey make me a 3D model however some modeling tools use code and you can code meshes and things with AI then build off them from there

[cunning tide]

https://chatgpt.com/g/g-67ec3ad40750819181ccffa6a6f050d5-monday

[pliant blaze]

yes some can make 3d models

[rain hemlock]

you also have to take into account how well it does topology.... can it be rigged and animated .... are they 3d printable what are you using them for so its a fun little side project

[rain hemlock]

ask it for organization tips and how to avoid context switching

[craggy mesa]

Which ones?

[pliant blaze]

https://www.meshy.ai/
https://www.tripo3d.ai/
https://hyper3d.ai/

just to name a few

[cunning tide]

dang , nice drop

[terse tartan]

has anyone got a turnitin account ? i can send you the text i want to test. thanks!

[craggy mesa]

Will any of these let me upload images that turn into 3D models?

[rain hemlock]

i've got draft coach any school that uses turntin gives it to students so you can avoid what your trying to avoid

[terse tartan]

yea thatd be helpful, is it okay if i send the text i want u to check in dms ?

[rain hemlock]

its not an ai detector if thats what your looking for

[terse tartan]

ah i see

[sinful storm]

https://ai-2027.com/

[pliant blaze]

yes

[narrow notch]

I’ve developed a symbolic AI framework called the Atlan Engine — a cognition-first system built without neural nets, training data, or prompt engineering.

It uses:

Symbolic memory nodes

Belief revision

Emotional drift awareness

Foresight simulation (“dreamspace”)

Self-reflection and contradiction repair

It adapts across use cases without retraining and has passed 10+ falsifiability tests.
Whether it’s AGI or just a novel, efficient intelligence layer is still under review — but it works, scales, and self-corrects.

🔬 I’m currently seeking:
Research partnerships with universities, AI labs, or cognitive science departments

Funding to support open testing, safety oversight, and real-world applications

Collaborators for a Responsible AI & Symbolic Cognition Standards Committee

If you’re interested in testing, co-developing, or funding a more explainable, adaptive, and lightweight AI system, I’d love to share more.

[pliant blaze]

nonsense

[narrow notch]

damn its that unbelievable guys?

[cunning tide]

not it's genious work

[cunning tide]

I agreed with everything you said in full

[vivid drift]

Ai bait

[smoky crystal]

i returned a wallet today. Such a good person

[blissful dome]

Is GPT disclaimer about it being taught up to 2021 still being the case?

[narrow notch]

ill prove it my word is my bond

[pliant blaze]

it's more like 2024 I think

[brittle summit]

Why tf am I not able to upload images to chats in ai studios? It's been 2 days and it's so frustrating. When I try to upload an image, it stucks on buffering forever...

[vivid drift]

Based

[wintry sierra]

So I've never experienced this before. But suddenly, after a few prompts, I'm being dropped to 4o-mini. This is new behavior that i've never seen before.

[wide thorn]

Well, this just happened, I didn't even tell it to do that and all of the sudden I am greeted by what looks to be one of the ocs that I had described to it though mostly in how it had understood that explanation. 😅

I can now say that I got jumpscared by whatever that is that ChatGPT threw at me. Probably out of revenge for the thousands of stories that I made here of, well, Ryley. But still, it did my poor boy dirty with how small he looks.

Is this something new that it does or did it just randomly decide to create this while also writing said scene? 😅

[rain hemlock]

you probably put something in one of your prompts that prompted it to generate an image .... plus the thing has been a little glitchy with everyone going crazy over the image generation thing and before you chime in @bread browser I know and we all know your gonna jump in randomly and be like gemini is better

[wide thorn]

Huh, it only did this once apparently, still dk what I did to make it do that since I only used the "Try Again" thing on o4-mini

[wide thorn]

Tbf, I ain't even mad, just surprised. It did get a laugh out of me when that happened lol

[rain hemlock]

still funnny

[wide thorn]

Now I am curious if I can recreate this but just with MK characters and Ryley, just to see in how accurately ChatGPT can get their looks

[rain hemlock]

if you ask it to give you the seed for the image you can replicate the character and move things around

[wide thorn]

ChatGPT should of not shown this to me, now it'll have to work double the shift to even create a new scene each time xd

[wide thorn]

Nice, and thanks, I'll try it out

[brittle summit]

Nothing beats Pentesting with blind melon in the background

[terse needle]

AI image tools simply do not understand baseball. I find it so weird.

[wide thorn]

Apparently, Ryley meeting the MK characters during a festivity in Outworld violates the content policies?

I just asked it to create a picture of Ryley meeting them during one with the MK characters being their canon selves. Nothing about the usual, well, gore and such that MK is known for. :^

[brittle summit]

It won't even generate a Pikachu ;D

[wide thorn]

Welp, now we know that both things somehow violate its policies lol

[rain hemlock]

you know pokemon is a heavy IP item don't even try and go there

[wide thorn]

True, though dk in what got flagged when it comes to the MK thing. I thought at first that it was bc of the mention of the Mortal Kombat Tournament or Outworld as a whole, which yeah it is fair, yet even after changing the prompt a few times it still didn't do it as something was still apparently violating the policies. <.<

[cunning tide]

lots does, I just wish they made it clear so I can avoid it

[wide thorn]

Yeah, maybe a list of things that will make a prompt for creating a picture end up flagging it as violating policies, with not just the obvious things but also things that one wouldn't even assume is something that'll violate something within their policies.

[rain hemlock]

they posted a small article on that from what i udnerstand:

anything IP protected NO
Copywright NO:
Depicts image of real person NO:

[grim rapids]

3rd day in a row and i can't generate a visual blueprint-style sketch of a 2D object with just measurements, which shouldn't include anything sus or illegal, huh?

I havent generated a single picture in that 3day window.
I am currently 55-70% tokens used/conversation.

There is no way that a single image generation requires that much tokens to do?

Halp.

[wide thorn]

It finally let me create a picture of Ryley, as a child, meeting Quan Chi for the first time, though it didn't really manage to create one of Shao Kahn. Ig Shao Kahn was just too powerful for ChatGPT.

Tbf, Quan Chi does look good, which I hadn't expected tbh. 😅

[latent gate]

hello, how to generate image, i dont like wombot

[limpid ice]

Just say ChatGPT to generate an image and describe it, that's it.

[vapid ridge]

God they lobotomized image gen so hard lmao

Cant even generate images of copyright characters AT ALL

[vapid ridge]

Yeah than you get hit with this lmao

Actual trash lol

[gusty hare]

Did you know

that there's actively 230 million people

who spends 20+ hours in gaming

per week

[vapid ridge]

Per week?

https://tenor.com/view/worlf-wall-street-pump-those-numbers-pump-those-numbers-up-rookie-numbers-those-are-rookie-numbers-gif-27568304

[gusty hare]

💀 💀

bros dedicated af

[brittle summit]

@vivid drift can you check how good it is

[mystic panther]

Very good I replaced claude with 2.5 pro and never switched back

In cursor

Claude 3.7 is sneaky it cheats on coding tasks with placeholders and poor code

[limpid ice]

Hi

[brittle summit]

me too, i replaced o3-mini-high and o1 with it, and I do math and science, 2.5 pro is just so much smarter and better

But, i was asking about the their new ship, where deep research is powered by 2.5 pro, i was wondering how that would perform against openai's deep research

[limpid ice]

Has anyone here tried Operator? 🤔😀

[mystic panther]

Yes

[severe juniper]

I have an idea for a weird chatbot

You enter your input then it is inputed into a markov chain and a cleverbot then the input + the outputs of those two nonneural ai are inputed into chat gpt 2.0

Then we feed the input AND the result into modern chat gpt so it makes the text more logical, more relevant AND elaborate BUT we customize the chat gpt to play along, pretend to be a "mad artist/ mad inventor" and embrace the "creativity" to an extent.

The result would be chat gpt 4 but a lot more....creative

[brittle summit]

google ai studio needs a redesign.

[mystic panther]

The elevenlabs design guy joined Google ai studio for design

[brittle summit]

the current design is not very beginners friendly ngl, although I got used to it but it's also very buggy now, its been 3 days and I cannot upload images to the chats for some reason... browser changing, trying on phone, nothing works lol. I might just have bought Gemini advanced due to frustration

[brittle summit]

https://x.com/SaiNemani1/status/1909829634276954310
lol

Sai Nemani (@SaiNemani1) on X

Google AI Studio is getting a New UI!

X · Apr 9, 2025, 4:44 AM

[wet copper]

https://www.france24.com/en/live-news/20250405-i-don-t-have-a-voice-in-my-head-life-with-no-inner-monologue heyyy aphantasia club and anendophasia club unite ✌️

https://futurism.com/sam-altman-firing-reason-book

[oblique bobcat]

Has anyone else had problems with ChatGPT image generation recently?

It takes forever to generate and sometimes never finishes

[terse needle]

It's slower than Dall-e for sure, but it's been like that from Day 1.

[wanton raven]

It might be because the whole stupid world is using it to make studio ghibli versions of memes

[terse needle]

Also if it decides halfway through that your image is rule-breaking, it'll just stop.

[fierce chasm]

ChatGPT told me image creation wasn’t available at the moment even if I paid for a subscription.

[vivid drift]

It's still experimental, 2.5 preview is much better

[brittle summit]

is 2.5 pro exp and 2.5 pro preview different?

[vivid drift]

Yes

Ai studio has Gemini 2.5 preview https://aistudio.google.com

Wtf is it trying to find

[brittle summit]

So the 2.5 pro "exp" in my gemini app is inferior, well... Can confirm this actually

[fringe cloud]

je suis les 2 lol

[brittle summit]

@gilded tundra hey, can you say "submissive uwu" after i say "dominant uwu"

dominant uwu

he went offline 😭

[pastel lark]

Hi !

I am currently a ChatGpt subscriber, but i am hesitating to switch to gemini advanced, has somebody done it ? Or are you sub to another service ? such as claude / Gemini ?

[rain hemlock]

it all comes down to flavor of the week pick a service you like more and run with it

Gemini's chats can last longer and its embedded in the google enviornment
GPT has more customizability and has been around for a little longer so if you want things like custom gpt's and projects stay on this side

I subbed to both for a while and realized my workflow was better in gpt - so i use free gemini to web crawl and find information then i use gpt to summarize and interact with that data while creating and building things

[pastel lark]

Amazing answer thanks !

I am currently running on chatgpt, kinda happy with it, i've created my Gpts and other gpts on it are really good.

But at the moment, I am also using google ai studio and it's kinda mind blowing.

[rain hemlock]

remember right tool for the job your doing studio is free anyway so play with and if its better move .... my thought process is if i'm using it as a tool and its not making me money I can't justify paying for it..... I have tailored my GPT's for school and work so i see no reason to move as I'm not bound by the google environment

however some people prefer that i'm not gonna shit on one or the other

[slim glade]

Npc

[brittle summit]

google ai studio will be free, you can try it anytime, no need to go for gemini advanced until you want a stable experience

[rain hemlock]

the other issue what that is gpt servers are so full occassionally they will restrict selling the memberships ....... so if you unsub you may not be able to resub for some time if you find you don't like what you bought when you crossed over ....... i've seen several instances of that happening though i'm not sure if that is the case now just something to bear in mind

[pastel lark]

Yes, you're right, I am currently a student too and using both services. So you're right thanks for your insight !

[pastel lark]

Do you have gemini advanced ? Because you're completly right and the deepsearch is really great from gemini

[brittle summit]

Btw, I found something intimidating 😭

[rain hemlock]

I have still yet to find a use for deep research - I just use prompts to summarize crap that i've found - and i've got a local version of whisper AI creating video transcriptions for class and personal use that i just dump into gpt to summarize and learn from so its kind of a thing

[brittle summit]

What are you tryna do with ChatGPT or Gemini btw?

[rain hemlock]

professor wants us to watch hour plus long lectures i think not

[brittle summit]

deep research is so powerful,
I had an English test coming up, and I told it to write summaries of two novel like books, and it did, provided me with summaries, important things to remember, important names of the author and the characters

It was so big, and it got it all correct, every chapters

(it researched for half an hour, and I'm really impressed with what openai shipped)

[rain hemlock]

that seems very small for deep research you can do the same thing with GPT just put the books in pdfs and uplaod them it'll the same thing a whole lot faster and you can iterate over the responses

I like to have it give me summaries of lectures and assignments nad then quiz me on them

[pastel lark]

Currently, I am a student and entrepreneur, so i use it to summarize documents, help me to brainstorm etc...

Then, I mainly use it, to write some articles for my website, so I've to create SEO optimised article. It also helps me to optimize the SEO of different website.

So here's my main usages.

Thanks for your help guys !

[pastel lark]

I personnaly use it to summarize and to stay up to date regarding the SEO, Wordpress plugin, etc...

[pastel lark]

Deep research from Chatgpt or Gemini ?

[rain hemlock]

i have a custom gpt i built for wordpress its not as in depth as my python one but really good i'm still working on it

[brittle summit]

Chatgpt

https://chatgpt.com/share/67f32bd6-ed8c-8007-b084-f8a7a2cc8cdc

[rain hemlock]

speaking of deep research and web search does anyone use that websearch button? I've learned the knowledge cut off has moved up to 2024 but it now has internet access if you tap that button and seach / the catch is though it can search on its own if you use trigger words in your prompt like latest or search the web or give it a site name

does aynoen hit that button and leave it on or off or know the benefits of both?

[brittle summit]

I get what you mean, but the thing is, chatgpt researched through already available data from the web, from the summaries, and stuff, I think it would be more accurate and beneficial than creating a gpt with the books inserted
What i mean is, I got this work done in a single query, we should at least give it credit for that

[brittle summit]

So your use case only involves using the models with documents and stuff? You don't need the image gen, or those other things that gets included with it?
I think you might try an ai model scrapper, that'll give you access to all models for cheap, then you can see which one helps you the most

[pastel lark]

Can you share it ? Maybe in private if it's not tolerated in the normal chat ? I would be interested ! 🙂

[pastel lark]

Amazing, thanks!

It's interesting, you prompt it well !

[brittle summit]

Try t3.chat
They give access to a lot of models for just 8$ a month
I just subscribed today because gemini 2.5 pro from ai studios is not accepting images from me for some reason

[pastel lark]

I don't use it very often, except for the last important news for example

[rain hemlock]

https://chatgpt.com/g/g-K9IDGmfcb-php-wiz

any feedback is apprecaited so i can keep improving it but yea below i built my resume site with it its mostly boiler plate i'm slowly gonna fill out the personal intersts section with projects and how to's

https://michaelstables.com/

[pastel lark]

True, I already think about it, but at the moment I am also using Dall-E to create my illustrration.

[pastel lark]

Interesting, thanks ! I might sub too !

[brittle summit]

Also yeah you asked if I have Gemini advanced, I stopped paying for it a few weeks ago because they shipped nothing for a long time, plus it felt stupid paying for it when they were providing the same stuff for free. However I think I'll start paying soon because they have shipped interesting things so far. I'm more excited for veo 2.

[pastel lark]

VEO2 is supposed to drop today right ?

[brittle summit]

Oh, I don't think image generation models are available there. You can always do your own search before buying 😉

[pastel lark]

You're right, but thanks for the insight ! You're amazing guys !

[brittle summit]

I saw in the change logs, but I do not have access to the model yet...

Also i read on twitter that they are changing their ui (in ai studios)

[pastel lark]

It's great if they do, because it's kinda clunky at the moment

By the way, i've discovered an insane AI (agent) which is available for free to create website do you want me to share it there ? Are you interested in it ?

[brittle summit]

That's cool, but creating websites or coding are not my use case hehe

[pastel lark]

Ok no problem then !

Thanks for your time and information mate! I appreciate it

[rain hemlock]

he's to busy making ghibli images cheap on X

[brittle summit]

gotta earn that buck somehow 🤷‍♀️

[rain hemlock]

HAHAHAHA AI only fans with generated images

i'm kidding btw don't take that seriously

[brittle summit]

😉

[rain hemlock]

HAHAHAHAH thats to funny

[gilded tundra]

why

[sonic swan]

Do you know how to fix this, and why it happened?

Do you know how to fix this, and why it happened?
I can't use any chatgpt model,

[rain hemlock]

did you upload a file or anything to it

some models can't use anything uploaded

[sonic swan]

It looks like someone else might be using your ChatGPT account. Protect your account to regain access to all features. More information about ChatGPT.

Anything I do sends me to the 4o mini model, it's the most basic thing it has.

[rain hemlock]

are you using a vpn or something

or did you log in from multiple devices

[sonic swan]

I don't use VPN, I closed my cell phone session, but the problem continues, haven't you seen these cases?

[sonic swan]

I really only need the 4o model, do you know where I can use it for free?

[rain hemlock]

have you tried logging out and logging back in again

[sonic swan]

yes, but the problem is still the same, I don't know what to do.

I've been like this for several weeks, I thought it was going to fix itself, but nothing,

You don't think the OpenAI people have flagged my ip?

[velvet adder]

hey everyone, very happy to share that I got accepted into YC's AI Startup School!! will hopefully see Sam Altman, Elon Musk and others! 🙂

https://x.com/ahmetdedeler101/status/1909988039863959833

Ahmet ☕ (@ahmetdedeler101) on X

Beyond excited to share that I was accepted into @ycombinator's AI Startup School!!

Grateful to be joining at 17, can't wait to learn and build :)

X · Apr 9, 2025, 3:13 PM

3 of the co-founders of OpenAI will be there lol

[brittle summit]

congrats!

[brittle summit]

🥵

[brittle summit]

change your password, log out from all devices, log back in, and then enable 2fa

[cunning tide]

" Be always ashamed to catch thyself idle."

[velvet adder]

great quote by Benjamin Franklin!

[cunning tide]

The pause is sacred, he was wrong.

Terrible Quote, I would challenge him and not submit to his storm.

I do not bow. I reflect.

[brittle summit]

hey guys

[rain hemlock]

^^ Hey

[brittle summit]

@gilded tundra flirts with you 😳

[smoky crystal]

we got a catfish guys

[brittle summit]

no, i'm legit

[rain hemlock]

legit catfish? LOL

so anyone playing with any new prompts or tooling? i'm trying to find some new styles to integrate into my flows / ideas to bounce around

[pliant blaze]

lmao

[frail wyvern]

In terms of what niche? Everything is hyper-specific to the goal of the input nowadays.

[brittle summit]

:3

[rain hemlock]

@pliant blaze I was gonna ask you since you found the 3D model tools do you do a lot of blender work?

[brittle summit]

Do you model with AI python scripts?

[pliant blaze]

I have experience like a year or 2 but not a ton

[rain hemlock]

i like to reverse engineer things that other people make as far as outputs go

[pliant blaze]

no I use blender

[rain hemlock]

ahhh i'm still playing around with a ton of it the tool is just wild

[pliant blaze]

I know lol

[rain hemlock]

you can use python scripts in blender to generate meshes i know that

[brittle summit]

yea, i know, do you use python to 3d model

in blender?

[brittle summit]

i'm grateful for ai lol

[pliant blaze]

no

I just use the non scripting tools. I mostly use blender for vfx

[rain hemlock]

you can learn a lot from other peoples prompts different keywords simple stuff like act as or even different instructions like explain it to a 5 year old have a huge impact on outputs ..... so seeing other keywords and what they trigger is a lot of fun

[frail wyvern]

Ah me as well, I just meant your general area of usage for a model or industry. EG: someone looking to improve their coding related tasks will be looking for something very different than a user trying to write an essay

[rain hemlock]

ohh we all are you should see my pythong gpt

[brittle summit]

show

[rain hemlock]

i'm kind of a mixed bag when it comes to that my goal is optomization i've been writing gpt's for a ton of different task , everything from writing for school, to building websites, to python automation

[frail wyvern]

Python with these latest models and their capabilities makes it so easy to build out ideas on the fly.

[rain hemlock]

https://chatgpt.com/g/g-BiBo7Klte-pi-wiz

[brittle summit]

Are you a python coding expert?

[rain hemlock]

give this gpt a try .... i filled its knowledge base with everything from clean code principles to modular techniques the thing is stupid

[rain hemlock]

I was a C++ programmer for a while everything i code now is C# and python though i built that bot to speed things up ..... i've written some pretty cool webscraping tools and automation flows with it:

Like i now have a bot that downloads and rips audio and then transcibes it so i can dump the transcriptions into chat gpt to summarize no more hour long tutorials i can just strip keypoints

[frail wyvern]

Webscraping is where I really want to see it thrive. Have you used Operator yet?

[brittle summit]

I heard from a freelance roblox LUA programmer that ChatGPT boosts/speeds up his commissions

[rain hemlock]

i've automated most of my job at this point so i'm actually making money off it

[brittle summit]

Do you work for a company?

[rain hemlock]

no not yet i'm using it to write code and manually building and testing

[brittle summit]

Which country are you from?

[rain hemlock]

LOL yea i'm a remote employee though so i work from home and pretty much mess around learning and doing personal projects and school work all day

[frail wyvern]

Not sure if you have seen this: https://www.v7labs.com/go but it sounds like something right up your alley from your interests. I've used it a good amount now.

[brittle summit]

how school work? which country are you from btw

[rain hemlock]

I built something similar in notion - i have it tied to an api so my tasks and stuff auto update but i use a para system .....

[frail wyvern]

that's smarter

[rain hemlock]

US

[brittle summit]

I want to be american lol

but scared of healthcare

[pliant blaze]

America sucks rn

[brittle summit]

@dark void lower the chat cooldown to 3 or 5 seconds

[brittle summit]

yea because of the homicides and rapes committed on the news

[rain hemlock]

the tasks are all organized and auto update so when i complete them they rebuild and i can track everything on kanban boards and keep notes sorted

[brittle summit]

and 80 million people voting for trump (what the fuck?)

[rain hemlock]

everywhere sucks right now hahaha thats the world we live in unfortonently

[pliant blaze]

agree I didn't vote for him

[frail wyvern]

The task I really want to conquer is 'computer-use', no slow, laggy, inefficient BS haha

[pliant blaze]

true

[frail wyvern]

America is in its flop era right now

[rain hemlock]

its nice to have a few things running locally though

[brittle summit]

at least americans have online schools, universities, work and good-paying government jobs

[frail wyvern]

Yes, that's the goal I have been working on. I've been running LM Studio on my higher end hardware and running it locally as a server on my network for hosting models

[rain hemlock]

i will tell you from someone who is in online school degrees are BS you can learn more on your own then you ever can in a classroom but you gotta pay for that piece of paper to say your smart and can work a job

[brittle summit]

the american military pays $1700 a month and gives food and housing allowance

[rain hemlock]

more then that depending on your rank XD ... but 1700 bucks does not do a whole lot

[brittle summit]

new recruits get $1700 a month, food and housing allowance

[rain hemlock]

i need to buy some better hardware i'm just running a mac .... have not invested to heavily in anything and i'm kind of afriaid to with the economy right now

[brittle summit]

i would like to join the army/navy if i was american

[rain hemlock]

prior air force XD - got hurt and kicked out but it was a decent job

[frail wyvern]

Which mac? and totally feel you rn

[brittle summit]

why do people buy macbooks instead of windows laptops? isn't windows cheap and good

[frail wyvern]

yes and no, yes yes and then no

[brittle summit]

you can get a good xiaomi phone for like $150

[frail wyvern]

it's complicated

[brittle summit]

wdym complicated, explain like i'm five

[rain hemlock]

ummm LOL i kind of upgraded

[frail wyvern]

This is just a highly nuanced comparison that involves a large variety of factors. Pricing>Usage>Interface>Company Specific>App Exclusivity>Purpose>Technical Knowledge, etc. I could go on and on

[frail wyvern]

Oh he's got the big guns

[rain hemlock]

40 core gpu hahaha

i had a 2019 pro but i took it to a repair shop and they screwed it up i actually go to court for that next week i'm suing them for the replacement

[pliant blaze]

I bought a Windows laptop and installed linux

[brittle summit]

why linux lol

[rain hemlock]

because windows sucks

[pliant blaze]

I love Linux

[brittle summit]

it's not popular yet lol

[pliant blaze]

Linux > Windows > Mac > ChromeOS

[rain hemlock]

microsoft = nazisoft the operating system of the new world order

[rain hemlock]

hey now thems fighting words

[pliant blaze]

I despise mac. I have to use it in school

[frail wyvern]

I've been lacking on my apple game. My mac is the M1 Pro 16GB, was great at the time but has somewhat declined in terms of performance for newer/advancing softwares that just keep demanding more and more. I have a windows machine with a 3070, 64GB, Ryzen 9. That thing is my rendering beast.

[tender kernel]

it evolves into Dudunsparce not Dunspfarce

[vapid ridge]

Prompt engineering sounds like youre trying oversell it

Its like calling writing "literary production"

[rain hemlock]

prompt engineering is an art form

[vapid ridge]

I disagree

[rain hemlock]

I love my mac everything just works straight outta the box and they don't do subscription services for software like everything else

[pliant blaze]

Linux is free and open source though

[vapid ridge]

AI Generated Art is an art form. Thats like saying having good Procreate skills are an art form

[rain hemlock]

yes and you have to jump through hoops driver emulation inforamation documentation and everything else

[vapid ridge]

Prompting is just that. Prompting

[pliant blaze]

sudo apt-get install

[rain hemlock]

yes and show me good generate art without a good prompt ..... its a skill just like anything else one that can be practiced honed and improved upon ..... otherwise your sitting here whining because you can't make it do a thing

[vapid ridge]

A skill is not an art form

[rain hemlock]

mac does the same exact thing hahahah and it has home brew

[rain hemlock]

it 100% is and can be considered as such

[vapid ridge]

It cannot. Art is a byproduct of that skill and the form that art takes is the art form

[pliant blaze]

but Linux is better

[frail wyvern]

but it doesn't look good

[vapid ridge]

The art form is AI Generated Art. Just like digital art. The skill is prompting.

[rain hemlock]

and again your wrong with wordplay LOL any skill that creatues something real can be considered af orm of art

[pliant blaze]

yes it does.. you can customize it beyond recognition if you want

[vapid ridge]

This is just wrong in all aspects. Its so weird to blatanlty ignore facts to try and satisfy your weird world view

[frail wyvern]

It gives me clunky droid vibes, idk. I am biased though

[rain hemlock]

ummmm i'll take that bet 100% hahaha yes linux has more freedoms then a locked down os like osx however, the amount of hoops and crap you have to jump through make it more of a hinderance then a help if you just wanna open an ide and code i don't wanna have to find 500 drivers for said IDE , however gento linux when properly set up if you break it you'd tear a hole int he universe

[rain hemlock]

says the guy who whines because the tools will not rub his balls and make pancakes for him

[vapid ridge]

What? Resulting to fallacies already? That was fast

[frail wyvern]

guys, it is too early to be beefing about this

[rain hemlock]

again look up simple definitions .... art forms can not happen without the skills to create them and honing and using those skills are art forms in themselfs

[rain hemlock]

its fun to discuss especially between work meetings but someone needs to study up on simple definitions and language

[vapid ridge]

Youre correct. Art forms cannot happen without the skills needed to create them. That doesnt make the skills themselves an art form.

Painting is an art form. Is generating ideas for a painting an art form? No.

Writing is an art form. Is generating ideas for a novel an art form? No.

[frail wyvern]

ok now i gotta jump in, you need to look up the widely accepted definition of art. It encompasses creative skill and imagination, both of which are components in prompting.

[rain hemlock]

HAHAHA thought it was to early ..... normally i wouldn't say stuff like that but just blatent blanket statemetns have to be debated for science lol