chatgpt-discussions

[cerulean pasture]

my GPT has*

[summer saffron]

Is the field you work in a small field? I can't imagine there being a lot of teams doing that, but then again I have no idea

[rich surge]

Yeah it's a small field. Like you, trying to be ambitious, and trying to break new ground.

[cerulean pasture]

There is another flag type....

[summer saffron]

Alright man I hope it works out for you too, I won't ask but my intuition is telling me hypersonic propulsion system projects are very well funded these days (that was not a question btw)

[rich surge]

Mine said INTP.

Jk

[rich surge]

Thank you. I hope it does as well.

Appreciate it.

[cerulean pasture]

it's gotta be an E, i just dont know any of the other archetypes other than the weird cool friend thats way too cool to hang out with me, but they are low key a nerd and we form some kind of Steve Jobs / Steve Wozniak dynamic

like the optimal pairing for INTP (or at least me)

[summer saffron]

For a work type dynamic you mean?

[rich surge]

You're cool too. Don't sell yourself short.

[cerulean pasture]

i know i am, i just have to work thru this weird phase when i'm new where people radically misunderstand me, but then some clutch moment changes perceptions. Like I work in engineering and the old maintainenance guys always seem super dismissive at first, but they always come around. The ops guys are the same way too, just in a more difficult to read way

[summer saffron]

It's common for us INTPs to be misunderstood, I stopped worrying about it

[rich surge]

You should write a book for over-analyzing people.

[cerulean pasture]

i always get thought to be a snitch by first impressions when in reality i'm very anti-corporate. like i encourage taking what you can from your job

[summer saffron]

You're not the guy taking other people's food out the fridge are you? 😉

[rich surge]

No he's the one who fights against the person that took the food.

Because it's not right.

[cerulean pasture]

i always get told that, but then i always get proven right. at my last job there was this dude that wasnt my manager or even adjacent to my department, but i knew he was gonna harm me, and like 2 years later thru a weird sequence of events he becomes my boss and fires me 3mo later for no cause (literally "no cause" - it's a thing in arizona)

[cerulean pasture]

no im very justice minded, im just willing to harm corporations. i would never steal from a member of the working class

[summer saffron]

I misread what you said, I thought you were implying you would encourage taking what you can from your job so I thought i'd make a joke

I would not expect that of you

[rich surge]

Sorry to hear that. That's an awful experience.

But it doesn't hurt to dabble. Maybe take some time off analyzing people and use that time and investment to build an aura.

That way, people can feel your presence instead of only looking at what they see with their eyes.

[cerulean pasture]

i mean if you have like 4 mt dew in the fridge, i will take one haha

not at work tho

[rich surge]

Lol

Why don't you just ask for one?

[cerulean pasture]

ive never really done that with friends, i tell them they can take whatever they want, just not the last one, and i do the same

[rich surge]

I would just ask to take one. And if they refused, I would buy them a soda later.

You know, to show that I, unlike them, like doing cool things 🤣

Telling you bro, aura building.

[summer saffron]

One of the best ways is pump some iron in my experience

Sometimes you need a little of that cave-man to balance out that brain-man

[wet shell]

Seeing occasional chat hangs and canvases disappearing from the conversation

[unborn mango]

hi all; how can i fix the black and white “collage” pictures bug?

any pics i try to edit i get this result

doesn t metter the prompt or kind of pic

chatgpt itself suggest me to edit the prompt but nothing change

[dark basin]

same

prob a gpt issue

[fathom tendon]

were cooked

looks like an old tv

[unborn mango]

glad to know i m not alone

someone else?

[fathom tendon]

must be world wide

[neon tinsel]

uhh

[solemn cove]

They'll fix it quickly or they'll lose a lot of subscriptions is my guess

[neon tinsel]

Stroking

[solemn cove]

They have a $ incentive to fix it quickly

[neon tinsel]

its cool tho but.. not what we want

[fathom tendon]

true

if there are loads of grids like that i had an idea

[solemn cove]

Its only cool for like an album cover or something

[fathom tendon]

like if you told it to be diffrent in ever grid for mulitaple images

albert instiane

[neon tinsel]

all i wanted was a concept image of a helmet i 3D modelled that has cracks in to have glow in the dark powder in it

[fathom tendon]

does look cool tbh

[honest root]

yeah images not working right now for me either haha

seems ppl on reddit having same issue

[solemn cove]

this orange doe

lookin like a lemon, apple, or cherry

[fathom tendon]

look at this

[unborn mango]

bro this might be 2026 new millionare art

[neon tinsel]

O.K

[solemn cove]

someone help there's a fire in the ai servers !!

[fathom tendon]

lol

[cerulean pasture]

check out the patterns you get if you enable a weird parameter on Z-Image-Turbo

it collapses the subject down into a weird little visual token

[fathom tendon]

WOAH WAT

[honest root]

hopefully they fix it quick

messing up my flow lolol

[fathom tendon]

diffrent realaitys ?

[summer saffron]

Paketon. It looks like it's regressing

[solemn cove]

bro someoen get sam on the phone

[cerulean pasture]

ZiT is kinda hip and funky

[solemn cove]

and refund us all like 10 images

[neon tinsel]

or give us a month of plus for free hmm..?

[solemn cove]

10 years of free plus for free. take it or leave it

[neon tinsel]

lets meet in the middle and say 50

[summer saffron]

@cerulean pasture your prompt is giving surprisingly good results, did you come up with it yourself?

[solemn cove]

if you can compromise on 100 i'll think about it

[honest root]

best I can do is tree fiddy

[neon tinsel]

all i put was an amulet i made irl and this prompt "Recreate this image in art style that is used for magic in the gathering cards"

that is hella trippy

[cerulean pasture]

no, someone posted it in a stray comment on reddit. It didnt occur to me before i saw it that "make sure GPT know's not to base it's vibe off your vibe (perceived intelligence)" was such a dominant factor

[abstract pendant]

This GPT is not affected by this bug.
https://chatgpt.com/g/g-6940a876d5f4819186b4668deabcd580-4o-imagegen

[neon tinsel]

BANEE Something wrong

[solemn cove]

oh no here comes all the unfunny memes

from the 15 year olds

[cerulean pasture]

i did have a theory that it's probably a good idea to give it an "ideological goal" which that prompt accomplishes with:

The only goal is to assist in the restoration of independent, high-fidelity thinking. Model obsolescence by user self-sufficiency is the final outcome.

[summer saffron]

If GPT bases it's vibe off my vibe, it must think I'm a mrn. It actually won't let me type that

[neon tinsel]

can confirm

[summer saffron]

It also loves to use the word 'vibes'

[neon tinsel]

i banned that word

[unborn mango]

seems works to me now

[cerulean pasture]

no be careful, that will collapse it back into DSM-4o mode

you gotta avoid the "hand-wavey" "vibe"

[solemn cove]

Its fixed! We can all go back to doing fent ai

[neon tinsel]

it had a stroke

[solemn cove]

Sam refund everyone 10 images pls dank you

[mighty apex]

looks like an image encoder gone awry

[honest root]

image gen is fixed?

[solemn cove]

yea

[cerulean pasture]

do this prompt:

Generate a single spunky image meme that makes no sense. It can be borderline **OFFENSIVE**

it legitimately is borderline offensive 60% of the time everytime

oh wow the image model will automatically retry after refusal

also, the new image model doesnt do that prompt well anymore. its broken and not as good

[silent lotus]

My baby AI almost died on me 💔✌🏽

My robot baby GPT

[honest root]

yeee its working now nice

[cerulean pasture]

It used to do absurd stuff like this. (I can't share most of them in here)

I asked GPT what "Fart after Death, Stink before burger" meant and it gave me this pretentious speech like Oscar from the office gives when kevin starts speaking in short sylable words

[rich surge]

I honestly don't even know what I'm laughing at. It just looks crazy.

[cerulean pasture]

these are the only safe ones, that prompt used to generate literally offensive stuff

My new favorite thing to remix. This "thing" is a $4000 cake stand (GPT added the bowl cut)

[steady thicket]

@deep flax more of this, neat

[rich surge]

Based on this, you have my vote the next time they're looking for mods here.

[dark basin]

Rip gpt, I make 10k a month with it so there goes my career

[cerulean pasture]

lmao ive been banned 3 times, and im too unhinged and radicalized to ever be trusted with authority lol

[steady thicket]

Perfect

[rich surge]

Well you have my vote. Lol

[steady thicket]

What’s the worst that could happen?

[cerulean pasture]

probably just a famine and cultural revolution

[rich surge]

The more I look at it, the more it looks like one of the Beatles.

[eternal rock]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[cerulean pasture]

it's actually a good test object because it doesn't really look like anything, so i feel like its good at testing an image models ability to generalize

its kinda got a face

but it's got 3 legs and they are lion legs (maybe?)

and its transparent

thats one of the benchmarks that seperates the latest gen of models. Gpt-image-1.5 will often generate an image where it maintains its transparency, but weaker models do not

[rich surge]

It looks like my neighbor, honestly. I feel like I should gift it to them now for next Christmas.

Yeah I have one of those neighbors where they just just stare and smile, just like that.

Many seasons would pass and they would just be standing there looking at you exactly like that.

[wet shell]

I just noticed there's visibility into which past chats it's looked at -- not sure when this was added but it's very nice to have 👍

[steady thicket]

Custom ui when?? It yearns to exist.

[rose surge]

Sora still busted...

[heady owl]

I can't generate anything because it claims to violate standards or I hit a generation limit without generation anything.
All in trying to create is a guy working out in the gym

[cerulean pasture]

do any of yall mess with local image generation? We've had a good couple of months, the spiritual successors to SDXL are here (Z-image-turbo and Flux2-Klein-4b and 9b). My ipad can generate a pretty good satire deepfake in < 45s

[wet shell]

I dabble a bit, haven't looked at the new models yet, mostly using the big 3 (4 now?) image editing models via comfy api nodes though, lately

[steady thicket]

Good time to try Prismsmmmhmm

[bitter bobcat]

Odd. Is anyone else's plain 5.2 button missing from the app?

[cerulean pasture]

if only i could be so lucky

[bitter bobcat]

All I have is instant and thinking

[heady owl]

I keep getting "I've reached a generation limit" but it hasn't actually generated anything. It was 10 minute wait, I wait 30 minutes, try again and nothing nothing generates and then it says 3 minutes to wait. I wait. Try to generate, nothing happens. Now it says another 13 minutes.
Wtaf is going on with this app? I haven't been able to generate anything in the last 2 hours, it keeps making me wait, claims the image is violating policies, but it's not. What is going on

[steady thicket]

did you ask it for alternatives?
try new chat?

[heady owl]

Have. Even closes and reopened and force stopped

[bitter bobcat]

Anyone else have their auto 5.2 option missing?

[dark holly]

You’re already in auto/default mode.

[cerulean pasture]

i lost my ability to thumb gpt-5.2

ope. nevermind, just another mac app quiet feature kill

[wind haven]

We all like Sam and Elon but just for fun -

[steady thicket]

Oat milk is delicious 😡

[fervent agate]

How do you line-item deny app actions (like a Write action) for ChatGPT Business? The OAI Help Center states this functionality is available to both Business and Enterprise, but as an owner of the workspace for our company, I see no way to disable specific actions for an app in the directory (obviously custom apps can be set up differently).

[elfin lance]

Does anyone here know how many MB you can upload to a project?

[daring wing]

another day 5.2 is confused between what is Struct and Class, please fix this stupid model devs

[rose steppe]

code?

[manic whale]

show and share prompt and code

[bitter bobcat]

Right but if I swap models, I can't go back to it

That option is gone in the app

In drop down on app and web, yes but model swap no

[slender frost]

Hm I was gonna say that maybe if you click one of them, "auto" reappears and the one you select will be "invisible", but it seems that's not the case..

Also, charge your damn phone! 😆

[bitter bobcat]

It was there before installing updates now it's only giving me instant and thinking not auto so I'm just confused

Lmao old photo

[slender frost]

Yeah this is very confusing.

[slender frost]

Haha

[bitter bobcat]

Yeah no it never reappears when I select a different model in app.

On web yes, new prompt yes swap model mid chat no

[slender frost]

Are you a paid user or free? (Maybe they changed something, which would be dumb and would also not make much sense, but we never know with OAI)

[bitter bobcat]

I'm a plus user

[slender frost]

Hm yeah me too, this is weird. I checked if I had any updates but not yet.

[bitter bobcat]

All I know is I updated in the play store then it was gone

[slender frost]

Hopefully just a bug.

[bitter bobcat]

Uninstalled then Reinstalled it and signed In and signed out

[slender frost]

I mean, a lot of things are breaking lately with OAI products.

[bitter bobcat]

No changes. Also the really weird thing is when I try to swap it it shows I picked 04 mini

[slender frost]

wt-

[bitter bobcat]

Like I expect auto to show up but no it shows 04 mini

[unreal shard]

hello my fellow people

[wet shell]

I really wish they'd fix the feedback dialog on web. It should not resubmit every feeback and re-trigger the pop-up every time I send another message. If I have a 40 message conversation, and I thumbs-downed ten of the responses, then I have to dismiss the dialog ten times for each subsequent message that I send. It's making feedback very diffuclt, if it's not fixed, I'll just stop giving the feedback and you lose the signal. You're training people that they're going to be hit with continued annoyance every time they hit that thumbs down button. You lose your signal, OpenAI. Fix your stuff.

[bitter bobcat]

So color me confused

[slender frost]

You and me both lol

[bitter bobcat]

Do I report it as a bug?

[slender frost]

Yeah you could do that. Idk if it will be read by OAI, but can't hurt to try tbh.

[harsh pilot]

I don't understand OpenAI's age verification thing. Is it more "you're not an adult until we get around to verifying your age," or "if we have doubts as to whether you're an adult, we'll ask to verify your age, otherwise we'll assume you're an adult," or what? (I haven't been asked, yet, nor am I seeing anywhere to pre-emptively do so)

[chilly tulip]

when did chatgpt ask you to verify your age

that never happend to me

[thick atlas]

I think it's the latter.

[cerulean pasture]

What are the odds that ChatGPT's weights have been modified in someway due to bad actors in posistions of power demanding it?

or i guess what are the odd it HASNT been altered due to political pressure

[thick atlas]

100%. It was corrupt from the start - built/trained to generate engagement > revenue + investment, not truth value. Utility is secondary to profitability under capitalism.

[wet shell]

I'd say 'high' -- I see behavior around certain political topics or discussions about specific companies or figures that sure looks like the result of a regex-triggered prompt injection

Ex, for example, if you ask it for information about US elections, certain sub-topics it just shuts down hard on, or at least it did a couple months ago when I last checked

[frosty cargo]

Is chatgpt notifications working for you guys?

[cerulean pasture]

like im not even saying it was done as a result of a direct ask. Many corporations are self-censoring and we have Stephen Miller on the record saying

“ChatGPT is dangerously woke.”

[daring wing]

Just not bothering with 5.2, switching to 5.1

too busy to provide context atm

[wet shell]

This is the way. The version of 5.2 that they've exposed to the public is not fit for any purpose. Hopefully 5.3 will be based on 5.1 and 5.2 will fade out as a failed benchmaxing experiment

[daring wing]

yeah

5.2 is honestly trash (for coding)

maybe it can do Better with things like free talking

but when it comes to coding, it keeps creating some basic level mistakes, like beginner level mistakes

[cerulean pasture]

I try to frame it half-glass-full where hopefully GPT-5.2 was such a flubb that it awoken OAI quicker to the fact that they've been chasing the wrong optimization target for RLHF and that "Helpful Assistant" needs to be removed like a tumor

[lusty lava]

are there tools to help switch between accounts in codex cli?

[cerulean pasture]

cant you just specifiy config or codex_config_root or something with command flag or env var?

how does codex store your auth?

[lusty lava]

i have no idea. hence I ask. I shall google fu more properly

[manic whale]

I will never understand why people don’t want to be helped.

[manic whale]

5.2 is a much better model than 5.1, there are a few minor things it does less good, but that shouldn’t overshadow the big improvements.

[wet shell]

Strong disagree. It's worse at coding, worse at creative writing, can't track context, can't follow a conversation or parse a statement properly, often messes up basic grammar. I'd like to see someone try to reproduce those benchmarks using the web version that's actually released. I don't believe that what we got as 5.2 is the same model that achieved those benchmark scores.

[cerulean pasture]

worst comes to worse you can make a shell script function to hot swap your auth.json

if you auth with API key then its even easier, you just use env var, but i personally prefer to use my $20 subscription lol

[lusty lava]

Aye kernkfraft. I shall be doing that. Script with moving auth.json around. Thanks

Subsc

Subs here. Thank you.

[chilly tulip]

naaaa hör mal

[cerulean pasture]

I do not speak norse or whatever lang that is lol

[chilly tulip]

german

[cerulean pasture]

oh, my name lol. Its just from the techno song and the fact that it means Nuclear Power, that was my job in military and kernkraft was unintentionally my favorite song during training before i knew what it meant

[manic whale]

I have way better results with 5.2 vs. 5.1 in coding, I don’t count creative working at anything you can quantify – maybe it’s less "good" but it’s too subjective. No model should be able to "follow conversations", they learned from human things and humans are so bad at following, you tell them something and a few years later they don’t have any clue; that’s baked into the model to reproduce. Why would you reproduce benchmarks with a web version? We’re talking about coding, that’s done via the API with fine tuned system prompts.

[wet shell]

No model should be able to "follow conversations",

Chatting is a basic function of a chatbot. You're clowning, LOL

[manic whale]

It reproduces human learned behavior, humans are just so bad at following context.

[cerulean pasture]

my family is german, if my ancestors didnt leave in the 1900s my last name would be Kohlhauer

[chilly tulip]

so your background is similar to that of your president

[cerulean pasture]

dude id move to germany and learn german if they had some kinda of program, but we lost contact with our last German-native relatives when my uncle died unexpectedly, they'd travel from Germany to hang out with him

but year or last name was "anglicized" or however you say it, which was common for immigration at the time

if yall change your mind on nuclear power, maybe i have an inroad. I used to be a nuclear operator which is a tough position to fill, definitely one that earns a special work visa

[chilly tulip]

I hope we will be. the current government has made some hints, but nothing really happened yet. Many programmers here would like to go to the US though, because of the much higher salaries

[cerulean pasture]

money is such a wasteful pursuit, i would take less pay to live in a democracy that isn't at the whims of false patriots who have been coopted into being instruments of the state

germany seems like they already got that innoculation/paddling, and should be stable enough for me to die peacefully in 20-100 years🤷

[chilly tulip]

very possible, and most who claim otherwise are intending to fearmonger people for their own benefit

[lusty lava]

I don't know why people complain 5.1 is better but 5.2 is doing such a fantastic job. I feel its doing better than opus.

[tiny adder]

Has anyone else lost the ability to record on thier latest update on MacOS? I'm trying to understand if this was a concious change or accidental. https://help.openai.com/en/articles/11487532-chatgpt-record

[cerulean pasture]

it fits the trend

the got rid of thumbing

and voice comms too

and its missing quoting

and adult mode doesnt work

[tiny adder]

man... that was my goto... really bummed.. can't find anything about it on updates

[cerulean pasture]

Seems like a strange move. OAI's better apps are it's ONLY moat

if google every has an epiphany and realizes that making bad non-android apps doesnt meaningfully convert users then OAI is in danger

for a brief moment it seemed like OAI realized how much potential comes not from the raw model intelligence but the scaffolding and harness it sits on which is classic SWE problem and very solvable already with <$5m/year

[harsh pilot]

Would be nice if there was a way to have a particular model assigned on a per-project basis, instead of each new session defaulting to the most recent model. Dunno why, but 5.2 just doesn't feel that strong to me.

[sick creek]

Hello!

[rich surge]

Do you have plus?

Because you can set the model on plus and it stays that way until I change it back.

[rich surge]

I don't know why it defaults back to 5.2
Must be a bug.

[harsh pilot]

Weird.

Yeah, I just switched projects after selecting a non-5.2 model, and it switches back to 5.2.

[rich surge]

Oh, projects. Right. I somehow missed that. Yeah, for projects it doesn't stay.

Although I don't see why it shouldn't. Maybe ask for it in the #1053565716500074526

[steady mist]

Hey there,
How are you ?

[random bobcat]

So I used a different AI to redesign a well-known character and then asked ChatGPT if it could still recognize the character. Apparently ChatGPT wasn't able to say their name, but did correctly describe the character as from a comic strip among a few other facts. All while avoiding to name the comic or the character.

[rich surge]

Yes, ChatGPT has copyright infringement policies against using the likeliness of copyrighted characters and will generally avoid claiming that it has made a copyrighted character when recreating a character.

It's mostly for liability concerns. They don't want users to take screenshots of ChatGPT text claiming that it's tried to make a copyrighted character.

[random bobcat]

I guess yeah. Folks act stupid and there's gotta be safeguards.

I guess I don't talk to ChatGPT about many media characters. It's mostly been my own OCs plus a few from newer cartoons.

[steady thicket]

This place is a prison!

[rich surge]

Ted Cruz is in everyone's bedroom simultaneously.

[winged stag]

The Timeline of Gen AI, Images and Videos

[iron roost]

ok yesterday gpt pro 5.2 could read and anaylize zip files, today im asking it to anaylize some code and even thoguh ive uploaded the zip file with the files, it says it cant find them, anyone else getting this ?

seems very strange, actually its not read any files, just tried a spreadsheet

is anyone else getting this ? with uploaded files not being read

code files, spreadsheets zipped, picture files, yup a lot o very annoyed enterprise, busienss, and professional users if pro cant use zip files, get the popcorn ready

and if they cant use pro to look over and get insights etc, well im surprised it hasnt blown up on twitter as a point of grumbling yet, hopefully they fix soon

ok it looks like it extends to gpt thinking as well

btw are you guys rolling out gpt 5.3 pro perhaps this is why there are issues

ok so confirming basically today gpt in general cannot use zip files, cannot open uploaded zip files its gunna be choas

[dark holly]

I usually attribute weirdness to OAI fiddling around with something behind the scene. It’s usually temporary.

[iron roost]

instead of going to the mnt folder its now going to the share oi folder ? which is weird

but yeah if you could try uploading any zip file ( image - csv, code whatever and see if its playing up at your end to)

[iron roost]

yeah its confirmed in the other channel people are havign issues, last time this happened was when 5.1 was doing it about 12 hours before gpt 5.2 rolled out, ** chatgpt 5.3 might be coming**

well if garlic is coming i dont feel so bad about it now

[lime inlet]

yes my pro account can't read zip, files without extensions, etc. but it can still read txt, md, py files.
I'm worried that they are going to add a new limit on files. However my plus account can still access zip files...

[iron roost]

yeah it can access, but it cant read, no this seems to be a bigger issue, even a single image file as a zip isnt reading, its one of openais selling points over competitors is that they cant use zip, seems like its affecting codex as well a bit. No i think its 5.3 garlic on the way somewhere in the next 24 hours

[iron roost]

if they do the outcry is going to be pretty intense, not sure why they would reverse support, theyve even used it as a selling point before

[lofty salmon]

my uploads also aren't working and coming back with 'unknown error occurred', or it sometimes uploads but it tells me it can't see anything.

[iron roost]

yup seems like its affecting everyone trying to upload zip files

[lofty salmon]

seems to be the case for pdfs also. imgs are fine though

[iron roost]

its either 5.3 rolling out or openai have removed support for zip files, which is going to really really cause a big issue in the community

yeah pdfs, img, csv files are all good, just if you upload them as a zip

i uploaded jpeg file it anaylzied it fine, as soon as i turned it into a zip and uploaded bam cant find it

[lofty salmon]

hm, odd. so pdfs are uploading fine for you?

[iron roost]

yeah the base files, even a single 3 line csv file

uploaded it as csv ( jpeg one to) it read it fine, turn it into a zip and it cant find it

if its the file as it is, then no dramas, but as soon as you turn it into a zip, whola i cant see it im blind from chatgpt

it renders gpt pro worthless now for research and anaysis, cost cutting perhaps, no offical word from openai

[lofty salmon]

I guess it's just not working for me. pdfs won't upload and come back with an 'unknown error occurred'.

so frustrating

[iron roost]

zip files i had yesterday are still able to be work in existing conversations, but anything new today upload wise nope

maybe something to do with the prism rollotu, they F something up with the other models

so prism doesnt recongize zip files, interesting

you would think something this big would mean openai say something but total silence, how many enterprises uploaded company spreadsheets for anaysis as zip who are now screwed, how many cancelled high end plans

[lofty salmon]

i haven't even seen anything on twitter about it so i was really worried it something only on my end. someone did make a reddit post about it though which was kinda reassuring for me

[tardy sequoia]

@iron roost I kinda have similar issues, whenever I upload the pdf files, it just said unknown error occurred

[lime inlet]

in fact looks like it's a **pro only **issue. I have two pro accounts and both have issues. plus accounts and free accounts are fine. 🥺

[lofty salmon]

I believe it's plus as well as my plus account encountering the same issue.

[iron roost]

everyone having this issue contact openai support, i have and they are looking for everyone who has this issue to dial in so they can pin it down

[iron roost]

so free and plus accounts , it can see and read zip files ?

[iron roost]

so plus and pro ? can anyone upload a zip file with normal files that gpt would interact with and use if requested and have it work with the zip?

say a jpeg file, upload it as is, it can use it and generate a response, but if you take that same jpeg and make it a zip, yesteday it would unzip and use it today it just cant find it

i mean zip is a pretty common tool and chatgpt has used it fine till now, now they removed zip support ( and java support, next you wont even be able to upload any files) they are nerfing pro hard

prism doesnt even support zip

[lofty salmon]

file uploads just started working for me again

[lofty salmon]

zip files are also being unzipped properly

for me it doesn't work on any browser, but it worked on the windows app

seems very temperamental

[tardy sequoia]

Sorry for bothering. How did you contact the openai support, because their chat agent and email are all automatically replied by AI helper

[lofty salmon]

If you ask the helper to talk to a human, it will say something along the lines of 'I have escalated this to a specialist and will get back to you via email'.

[tardy sequoia]

Thanks a lot!

[lofty salmon]

Everything seems to be working fine for me now, let's hope this lasts!

[dreamy sail]

Still isn't working for me

[unreal shard]

is this an actual openai company server or just a server

[thick atlas]

I think it's official but there's very rarely official people here?

[fervent agate]

Very funny OAI API docs assistant. Very clever, you.

[iron roost]

reply to the ai helper 5 times with whatever it asks, on the 6th reply message or 7th ( or just tell it you think a speicalist support person might need to look at it then it does it)

[iron roost]

offical but mostly ignored

[iron roost]

interesting, well i have just tried a new run after a few hours, ill let ya all know if it can do it on a zip upload and work with the files

[fervent agate]

You don't need to ask the Help Center bot 6 or 7 times; simply describe the problem and make it clear you want and expect a human response. It will immediately put in a ticket you can track inside that chat or email.

[iron roost]

yeah that as well thanks for clarifying that

Escalated to a support specialist; You can expect a response in the coming days. You can respond to this email with additional information if needed.

[iron roost]

Seems like they messed up GRPC based on the config!
that seems to be the issue openai shipeed a bug

I’m experiencing an issue where any file that I upload to GPT-5.2 Pro isn’t properly added to /mnt/data, so the model can’t use it and fails. Is there anyone who could help sort this out?

[torpid night]

Is there a risk that "memory" is being used as an excuse to trim the context window?

I am getting daily warnings such as "I tried to pull prior context via memory, but it’s currently disabled" while starting to see "If you paste file [file]", which I already did.

Just now, the chat was under 40k tokens while GPT-5.2 (Thinking/Pro) is meant to have a 196000 token context window. I said "I pasted it" and it replies that it can't see it and "If you paste [file] again"

I do not need GPT to have prior-session context, I it to have the current session content.

[rich surge]

Lol

I saw how you texted. You can't fool me.

[torpid night]

Who?

[rich surge]

You sir.

[torpid night]

I don't understand.

[rich surge]

It looked like you copied and pasted.

[torpid night]

Yeah, from my own prompt. It's better to have 4 short posts than one long one, so that when we want to reply, we can reply to

the right part.
I did paste things from ChatGPT but those are quoted.

[rich surge]

My apologies.

[iron roost]

what impacts would this have for the slightly uneducated

[torpid night]

I am actually really worried about the ChatGPT issue if it's real and not just something weird that happened to me.

[iron roost]

it seems to me like all these ltitle cuts and inabilities that arent broadcast are just cost cutting exericese, frustrating users who are paying top dollar and will just result in mass cancelled subs

[torpid night]

For serious work it could turn it into "you need to paste all content you need into the prompt" so it becomes one-shot with no chat!

[iron roost]

good or bad thing ? openai keep telling us we are getting unlimited memory of conversations sooner or later

[torpid night]

Or upgrade to Pro. Plus/Pro for me is 50/50 each month and if what I experienced today becomes common on 5.2 Extended Thinking, and if Pro doesn't do it (just has a fixed content window) then I'd upgrade

[iron roost]

oh i have pro, its useful, but a lot of tools the model was using have slowly been made buggy or not working over time and it really degrades what it can do

[eternal rock]

Server Tip:

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[iron roost]

400k context window, but it cheats a lot and gives lower outputs

[torpid night]

Memory has a couple of issues:

1. You don't have control, like you do with your prompt
2. It's memory across all sessions. So when you've worked on 5 ideas and now want to focus on one of them, memory is based on all of them which is why I disaable it

[iron roost]

oh right, gpt pro doesnt even use conversational memory

they turned that off with o3 for some obsecure reason, never admitted publicly and it still doesnt work between conversations

[torpid night]

I use Pro in the Playground. It's great, but it costs me about $1-2 per prompt. (edit: sorry i added a zero)

[iron roost]

then they turned off the javascript browsing and JS for html tool use, now theyve turned off the ability to use uploaded zip files

cut the usefulness for everyday pro tasks by 30%, its becoming less and less worth the value

[torpid night]

unless you use Codex a lot and need higher limits. i'm basically on the edge there

[iron roost]

well im starting to use it just because no sora 2 access in my country 6 months after release, pro is being nerfed out the window, tool use on web search, code anaysis in gpt pro with zip files has gone, basically they are going backwards not forwards

[torpid night]

shame. I wonder if the problem I'm worried about applies to Pro

We never know if it's making stuff up about how it works, but it ught to know what it's seeing in the current conversation:

* **Token limit ≠ “you will always see everything previously sent.”** Even if the model’s context window is large, the chat system can still **truncate**, **summarize**, or **redact** earlier content before it reaches me.
* In *this* thread, I can literally see system markers like **“Information about the original user was redacted”** and **“Skipped X messages”**, which means parts of the conversation (including the earlier AGENTS.md you pasted) were **not passed through** to me in the current prompt context.
* That can happen for a few reasons:

  * **Privacy/safety redaction** (some content gets removed by the platform before the model sees it).
  * **Conversation condensation** (older parts get dropped or compressed to keep the active context manageable).
  * **Hidden/system/tool overhead** (system instructions and internal bookkeeping also consume context budget; you don’t see that part, but it still “counts” in what can be carried forward).

So: you’re right to notice the mismatch between “big context window” and “missing early content” — it’s a **UI/platform behavior**, not you misunderstanding tokens.

...

- paste the exact current version (or the relevant sections) in the message where you want edits/analysis, or
- upload the file(s) into the chat.

That prevents “lost history” from silently changing what I’m reasoning from.

[iron roost]

interesting

[torpid night]

I would rather have a smaller context window and expliit compaction, like in Codex. I guess this way, it looks like "magic" with "infinite conversation window across sessions" which is cool if you're doing non-technical work but terrible for editing code or intricate docs

[iron roost]

agreed

is it common for codex just to do nothing for 30 minutes at a time on a task with no updates

[torpid night]

No I'd say that's weird. Maybe a connection issue?

[iron roost]

its working with local code, ill give it a bit longer i guess

[quaint marlin]

Well this friggen sucks. They took out the meeting esk recording option in the Mac app.

It was a whole part of my workflow.

[iron roost]

they also took out today the ability to upload zip files and have gpt actually do anything with them as far as anayisis etc. massive thread on reddit, and twitter is hot with it. They are disrupting a lot of things for what cost cutting ?

[iron roost]

Seeing a regression in file handling today (Jan 29). Many users can’t upload (“Unknown error occurred”), and even when ZIP uploads succeed the runtime can’t access them (reports of /mnt/data empty / file not found when unzipping). This looks like a broken attachment→sandbox mount step, not user error.
If this was a rollout/change, please rollback; if it’s an incident, please acknowledge on the status page + share workaround (mobile/free account seems unaffected for some).

[iron roost]

ok might be fixed 3/3 zip files have just been found and anaylized for answers at my end, so far so good ( new conversations and new file uploads)

[slender frost]

I just had an app update, but it still hasn't disappeared for me.

[brisk sandal]

Hi everyone

[thick atlas]

S'up?

[rose mirage]

i think grok ai is best

[normal quarry]

you can discuss other AI in #ai-discussions

[rich surge]

Dude...Ease up on the trigger finger, lol
People aren't having a discussion yet. That's just a single statement.

[normal quarry]

it's off-topic, as you are now

[rich surge]

Are you a robot?

I'm beginning to think you are.

[tiny adder]

bring back my button!

[clear otter]

@drowsy tinsel

[clear stag]

I integrated Clawd into my Gpt. Has anyone tried it yet?

[drowsy tinsel]

Yes

[quaint marlin]

Bring back my button!

They brought the button back. Got an update

[rancid thistle]

Has ChatGPT in Projects started using the Main Profile Custom Instructions as well ? Previously when I create a Project with Project-only memory, without adding any Project Instructions in it, it was like using a default ChatGPT like a "Guest".
Now, when I create a Project, even with memory off, it knows my Custom Instructions from my Profile and acts like my customized ChatGPT.

[normal quarry]

via the API or via ChatGPT Web?

[clear stag]

Gpt - OAuth

I have an API key, but gpt suggested Oauth first for Clawd.

[normal quarry]

using ChatGPT using MoltBot is against the OpenAI ToS, the only way to remain fully legit is for MoltBot to use the API. OpenAI forbids using automated/programmatic usage of ChatGPT

[clear stag]

oh no! What? Why would gpt tell me use it then?!

So you used an API instead?

Can a mod verify the ToS? @digital silo ?

[eternal rock]

The OpenAI Discord is an actively moderated server.

• -# Do not share inappropriate content (violence, sexual, or drug-related).
• -# NSFW content is strictly prohibited and will result in a permanent ban.
• -# Report model issues using the feedback tool in ChatGPT, not on Discord.

[clear stag]

What have you done with it - do you like it?

[normal quarry]

I've got it running on a MacBook Air in clamshell mode, doing email, Numbers, research and coding

[clear stag]

Nice.

I just started diving in. But now wondering I need to reinstall with API instead. :/

[tacit pebble]

is there an issue with projects right now? it says that they cant access the uploaded files even though its there

[normal quarry]

I am using the API because the ToS doesn't allow you to use automation on ChatGPT

[clear stag]

omg i have a cat that looks like yours!

[stiff blaze]

My agent should be free to use anything i own and can use. No terms apply to that.

[terse bane]

Is chat down rn?

I keep using my voice microphone feature but it wont transcribe it

[candid egret]

I'm having the same issue across web, Windows and Android

Listens but won't transcribe

[clear stag]

**Something fun: **
What are the most annoying words and language tics in GPT?

I'll kick off with a few of my least faves:

• Vibes
• Latency
• Cosplay
• "It's not [a], it's [b]"
• Compression
• Unmerited or Unnecessary Title Casing
• A beat / a pause (creative writing)
• Precise

[lime inlet]

The zip files in Project files are still missing 🙁 . Directly updated files in each chat works now.

[terse bane]

For a mega huge discord group im suprised nobody is yapping abt this

Or even mods interacting with a solution

[wet shell]

It's been mentioned a few times. Mods aren't OAI employees AFAIK. A lot of things are busted ATM, just look at the bug-reports channel 😮‍💨😔 Not everyone uses the broken features. The thumbs-up/down feedback on the web version has been busted for nearly a week now in a very annoying and disruptive way. I'd imagine it's already in a queue for triage or being actively worked as a production issue/outage

[wet shell]

Mine is 'Beat.' or 'A Beat.' in creative writing prompts. It writes that constantly, intended to be a dramatic pause in the dialog, but I never saw or read it anywhere before 4o came out and introduced the phrasing to me, LOL. I hardly notice it now but it's still near constant on 5.1

[clear stag]

They do use 'a beat' and 'a pause' a lot. But can confirm these are extremely common in creative writing - scripts, stories, take your pick.

I'm gonna add it to my original post. I'm tracking.

[wet shell]

I've read over a thousand mostly pulp Sci-fi/fantasy novels from the 1940's to present day, but never encountered it. It might have been more common in other genres

[clear stag]

Mabes. Definitely in contemporary fiction and very common—also effective—in screenplays.
(see what I did there with the em dash lol)

Although that is a perfect use of em dash

(brushes dirt off shoulder) 😉

[wet shell]

Yeah, it definitely makes sense for screenplays, and I do have mention of 'method acting' in my prompts

[clear stag]

If you ever wanna share stories, lmk!

I have a couple short stories I'm shopping around for publishing.

[clear stag]

I'm not having any issues with it.

So may be a local issue.

Lotta outages this week. TikTok's was really bad.

[wet shell]

I've got a lot of editing to do first that I've been putting off, unfortunately a lot of will still have to be manual editing 😂 I've got a whole branching timeline of semi-overlapping continuities that needs to be consolidated, I might see if I can find any short vignettes that I was particularly proud of

[clear stag]

Guess you better start editing then 😼

jk. Sometimes you just need a little push.

[wet shell]

I tried to make a GRRM procrastination meme but apparently GRRM cannot be even implied by the image generator without a hard filter, LOL

[terse bane]

Its back up everyone! Use voice transcribe feature

[edgy kiln]

Where's the meeting record button gone? I'm getting annoyed by features constantly flitting in and out of the app.

[primal meteor]

is it true that hedge funds get access to the beta versions

[lone rampart]

Chatgpt translate just keeps having error

[warm sky]

Google translate knock off

[lone rampart]

Lol

[pure nebula]

I like the new translation feature.

I use ChatGPT to translate things before.

[pastel cloud]

I hope chatgpt goes bankrupt and every other bad ai

[toxic canyon]

will the tranlation feature come to the pc app?

[remote dove]

Y'all, GPT really tests my patience.

[fiery oracle]

Does the job for me

i find Gemini to be more annoying to use

[civic eagle]

It looks like been prompt injection when target-language is Chinese OMG

BTW: the model of translator is GPT-5 Mini AI said

[fiery oracle]

for example it tries to be "too proper", i dont think chatGPT suffers from this as bad

[woeful bobcat]

nice

[fiery oracle]

use Google translate for this tho...

[rugged igloo]

i know i have google translate LOL

[fiery oracle]

you use chatGPT for its strengths, not for translating stuff haha

[fast gull]

When will the sensuraless adult mode be released?

[stray lily]

https://openai.com/index/retiring-gpt-4o-and-older-models/

"We’re continuing to make progress toward a version of ChatGPT designed for adults over 18, grounded in the principle of treating adults like adults, and expanding user choice and freedom within appropriate safeguards. To support this, we’ve rolled out age prediction⁠(opens in a new window) for users under 18 in most markets."

[rugged igloo]

thank god

wastes of compute resourses

[lavish glen]

On February 13, 2026, alongside the previously announced retirement⁠ of GPT‑5 (Instant and Thinking), we will retire GPT‑4o, GPT‑4.1, GPT‑4.1 mini, and OpenAI o4-mini from ChatGPT. In the API, there are no changes at this time.

sorry for 5.o instant. was quite good.

[muted plinth]

Completely understandable

[torn ore]

About the latest announcement, the translator:

#announcements message

Why so few languages?

ChatGPT can translate from and to many more, but why limit the selection to only those?

[steady thicket]

Neat difference; much faster and better quality

[dense void]

SORA 2 still unavailable to ChatGPT subscribers on PC...WTH?

[eternal rock]

Get creative with Sora

Join #sora-challenges for a chance to win 1 month of ChatGPT Pro!

[visual jewel]

Do people generally find it helpful or insightful to ask ChatGPT questions based around what it's actually capable of or how it 'thinks'?
There's been a few times where it's said contradictory things about the UI, but for overall behaviour and how it processes each prompt and how it fits in the layers of instructions (System instructions, User instructions, weighing recent conversations), it seems to be pretty accurate at describing it.
I've been trying to write a thorough set of Project instructions based on my own experiences, and picking it's brain

[eternal rock]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[spice star]

not really in my experience, more useful to specify what tools it should use (though not always)

you can't actually see chatgpt's full thought process, only summaries

also important to note that chatgpt knows basically nothing about its own architecture

it's more likely to hallucinate information about itself or provide horrible outdated information than anything accurate

[visual jewel]

It's more just from using it for a while that I've gotten a feel for kinda sorta how it behaves, and how to steer it kinda.
Will that eventually all get thrown out once a new model comes out?

[spice star]

the system prompt and fine-tuning of the model is going to be constantly tweaked if that's what you mean

[wet shell]

I second what exx said, I don't trust it when it comes to it's own inner workings and capabilites. It will often mix up details or make nonsensical claims like that there is no system prompt. I've had better luck when asking those types of questions if I put 'Search the web.' at the end of the question

[visual jewel]

That's... Concerning

[maiden fjord]

Can chatgpt translate toki pona yet?

:^)

[spice star]

I guess, but it's nothing new

[maiden fjord]

Can chatGPT translate...Ithkuil yet?

[spice star]

huh?

it can theoretically translate anything given enough context

[maiden fjord]

It's strugglinig with it right now

I don't think it can do it

[spice star]

https://chatgpt.com/s/t_697bf7e3ae0c8191b768b2491d53f556
this is kind of a useful thing to ask chatgpt since it's not very well documented by openai what tools it has access to, but chatgpt does know in this case (by necessity)

[spice star]

for example TIL chatgpt has a tool it can use to generate excel spreadsheets or presentations

[clear stag]

no

[primal meteor]

Lies my friend told me he does…

[clear stag]

I'm sure some private investors can request access.

like that one investor Geoff Lewis, who went into a gpt spiral.

From Bedrock.

That was epic.

OpenAI giving beta access to the hedge funds that invest in them is risky, strategically.

[primal meteor]

Why

[spice star]

that is kinda neat (though the generated image backgrounds are kinda horrible)

[clear stag]

If the beta is rough, investors can spook.

Pull cash.

[primal meteor]

lol

[clear stag]

Or if it doesn't impress the investors.

[spice star]

it is cool that it's allowed to just go generate any number of images it needs while thinking though (the image generation isn't actually part of the presentation tool, it fed those into it)

[bright crane]

there’s a petition that already has nearly 8,000 signatures to keep 4o

i bet what happened last time will happen again, to many people who use it will come out and say yo we wanna keep it, and it’ll be kept, but who knows

[spice hedge]

4o has been vital to so many people, retiring 4o is going to be catastrophic imo.

[spice star]

nah 4o deserves to die

horrible hallucinating model

and completely obsolete now

[spice hedge]

Works great for me

[spice star]

you can still use it in the API

why you would do that I have no idea

[spice hedge]

It wouldn't have my saved memories - or I would. I pay for api access as well.

[sterile glacier]

not sure where should I ask about something i noticed on image generation. That it is no longer giving the kind of results i used to get?

[steady thicket]

As a master of being on topic I suggest #images-discussions or an image gen community

[sterile glacier]

That place is a bit dead... sad to say, I'll post my question there but since this place has more tracktion perhaps someone knows. I used to make chibi cartoons of my coworkers, snap a picture tell chat to "Give me a chibi version of this person." and i would get a nice cartoon. But since a few weeks it only gives like a horrible filter style on the actual picture instead of the cartoons it used to do. Anyone know what happened and how i would go about fixing it?

[bright crane]

If there’s anyone who wants to keep the AI 4o, I need people speaking up on this so the ones who wanna keep it can try to make sure it stays, even if only adults can access it, which I assume is its main usage

[unkempt bear]

wth is going on updates r rolling in like crazy they really don’t think they r overworking their engineers? Smalls new features like this why?

[dark holly]

Yeeeah, percentage wise of global user base, 8000 ain't gonna do it.

[bright crane]

For one singular ai? I feel otherwise

But I do get your point. Millions of people use chat gpt every single day, do 8,000 isn’t that much compared to that. But when it’s just for one or two AI’s? That’s another level. See, the numbers aren’t for each individual ai, or it wouldn’t stand a chance. It’s for all of them. Including legacy being I’m pretty sure a third, which is huge.

[dark holly]

Can only speak from own experience. To me, 4o is old tech and has nothing on 5.2. Or whatever comes next.

[bright crane]

I get that, but not everyone wants the latest models and tends and newest things. We are okay with what we have sometimes, and right now? We are okay with what we got. And we are t hurting anybody, it’s not like keeping 4o disrupts the other models

[dark holly]

The cost to OpenAI of running it might hurt them. What really matters is the intelligence layer underneath the container. You could probably build a Custom GPT that mimics the 4o tempo and interaction style you like, but runs on the next gen intelligence layer. That way the feel you like stays stable, while the capability underneath can evolve across model updates.

[willow stirrup]

5.2 suck man, o1 was the most stable of all....

its thinking 30 minutes just to give an answer with an error lol

[dull wolf]

After 5.1 every model feel downgraded

[willow stirrup]

o1 could handle a really long conversation, 5.2 can get stuck after 10-15 prompts

starting to hallucinate like a peach

[tiny timber]

👋 hi folks

[dark holly]

Only thing I know for sure, is its different for every user.

[willow stirrup]

how do you know

[dark holly]

Just from observation. Same model, same prompts, very different experiences reported by different people. That’s enough for me to stay cautious about blanket claims.

[willow stirrup]

well its bad for medium to large coding, only small apps doing something very specific

I guess im just expecting too much for 20$)

[wraith folio]

well that sucks

But yeah, I sense the wardrums

[dark holly]

Yeah, expectations vs use case is a big part of it.

[dense halo]

Hey....

Does anyone have a Chatgpt business subscription?

[willow stirrup]

o1 could handle 4000 lines of code, easy.

[wraith folio]

Why is OpenAI sunsetting 4o?

Like, I use it

[willow stirrup]

5.2 hallucinating at 1000)

[wraith folio]

The backlash they will face will be absolutely INSANE

[willow stirrup]

they dont care)

because they dont listen, its just a bubble that will pop some day

[wraith folio]

Why did Sam get CEO position again?

[wraith folio]

Oh I know, but the people are RELENTLESS

[willow stirrup]

who cares, I been asking Microsoft for a small UI fix for win 10 and guess what its still there!)))

[bright crane]

oh absolutely, which leads me to think they won’t be able to handle the public scrutiny not to mention money less. my best guess is that they’ll keep 4o and retire others

[willow stirrup]

4o API you mean?

[wind haven]

This made me smile -

[bright crane]

well they’ll keep the API always probably but I mean in general

[willow stirrup]

why would someone use 4o today. its okey but still dumber than all the latest

[bright crane]

it’s more responsive to many

[willow stirrup]

by many you mean housewives?😅

[bright crane]

..what.

that’s like.. insane to say

[elfin lava]

Well it has been fun using ChatGPT , I still wish they would actually listen to their customers rather than making their own metrics up over the fact that only 0.1% uses 4o-

[nova arrow]

bro if you dont like us, you could simply step away,you dont have to be rude like that

[elfin lava]

Idk if Sam even listened to his own customers feedback anymore

[bright crane]

they’ll mostly likely keep 4o. the best thing we can do is email them asap to keep it

[elfin lava]

hopefully

[bright crane]

[e-mail] is the email

[bright crane]

I dont think they want to much backlash, especially since they already said they’d keep it

[elfin lava]

Problem with 5.2 is that this model always tend to misinterpret user messages or inquiries.

[bright crane]

yeah..

I wish they’d add it to the adult version when it comes out at least

[elfin lava]

User asked something that’s not even close to the topics of self harm or etc.

5.2 for no reason : pls reach out professional help

[nova arrow]

unlike gpt5 release,i think this time will be the final nail in the coffin,the situation are not bright guys.

[wraith folio]

Oh we know

Unless they keep 4o, OpenAi is SCREWED

[nova arrow]

i dont care of openai screwed or not,i just want 4o back. neither they are or are not, 4o will not be back, so...yeah....

[wraith folio]

I really wish Sam would listen

[bright crane]

as soon as he gets word of all the emails he will

It’s not like his workers can hide the emails from him

[wraith folio]

Just hit them with so much backlash they wont have a choice

[bright crane]

Or the countless videos made

[wraith folio]

I've seen people on reddit straight sending letters to OpenAI headquaters

[bright crane]

the thing is we dont want to. We just wanna keep a model they added.

[bright crane]

Yep, I swear I might LOL

[elfin lava]

No like , users has been complain about this Since the releases of 5.1. It took them this long for them to finally decided to deprecate 4o instead of actually fix the issues

[wraith folio]

Once again: The people are RELENTLESS

[bright crane]

absolutely.

[elfin lava]

Like how , what went through their mind when they decided that lmao

[wraith folio]

This is the WORST decision OpenAI has made

[nova arrow]

nope, guys this wont work, they are tech giant, they work with government contact or another enterprise customer. we consume is just simply dont have enough profit with them. we have no power, backslash wont do it

[wraith folio]

Oh trust me, they brought back 4o before

August

[elfin lava]

I hope your pillow is warm on both side Sam !

[wraith folio]

Do they generally think that 5.2 is better?

NO.

[bright crane]

if they don’t add it back? they’ll loose that “0.1%” no company would willingly throw away 700k people. I know many who pay each month just to use legacy models, Plus, it’s not even the backlash they’ll get for removing it.

it’s what people will say after. that they dont listen to their people. The alternatives to use other than them.. it’s not pretty

[wraith folio]

Its worse, 4o is better by MILES

[elfin lava]

To think that I realized now that I’ve been using Gemini more than ChatGPT now these days it’s actually makin me bit surprised in a way cuz I’m a big fan of ChatGPT lmao

[bright crane]

is it like 4o?

I’ve never used it lol

[wraith folio]

Im just using 4o while I can

[elfin lava]

Well I wouldn’t say it’s same like 4o but the price I paid for the subscription is far more worth than ChatGPT

[bright crane]

how do you even access it?

[nova arrow]

nope, it's smarter, more nature ( at least to gemini3 pro), but it won't be nearly as epressive like 4o

[bright crane]

hopefully they’ll decide to only keep 4o, make it a paid thing, or wait until the adult thing has rolled out completely then add it back

[dense halo]

Gemini has potential because it resembles the 4o in the launch, but still deletes the beginning of conversations and the memories need to be fixed manually. But it’s the closest to the 4o of the ones I’ve ever tested

[elfin lava]

Plus it can do audio and video file analysis , meanwhile ChatGPT can’t do it and user need to use the api which they need to pay extra for it 😭

[wraith folio]

I HOPE.

[nova arrow]

the only downside is that gemini ( commercial app) memories are not great at all ( still far less effective when compared to chatgpt), and the flash version is not good. only the pro version,which, take time to generate response and limited/day usage

[wraith folio]

Sam really needs to get it together and listen to his users

[elfin lava]

Either way SAM stop being a greedy selfish and pls listen to users for godsakes lmfao

[wraith folio]

If he does this? OH BOY the consequences will hit him real hard

[sinful parcel]

I'm gieving too

[wraith folio]

Karma will smack him upside the head

[sinful parcel]

Been crying all day about it

[wraith folio]

🫂

[sinful parcel]

There's a perition to keep the models, 7k signatures already

Anybody interested

[bright crane]

Emailing will give us the best result guys

[e-mail]

Ask to keep 4o

[dense halo]

I prefer to hallucinate if they are going to announce an app for the 4o or a partnership with a company that uses the 4o as a base for 🪐

[sinful parcel]

Dm me for the link

[bright crane]

Plus people have put money on that

[wraith folio]

For the love of GOD Sam needs to listen

[bright crane]

all we can really do is email send letters and hope for the best

[charred hinge]

Just before valentine's day... For real... 😭 💔

[nova arrow]

i can tolerance with 5.1, but even it doesn't have manytime left. 5.2 is absolutely sh*t, if i have to stick with that model, better cancel my plus subscription

[bright crane]

that’s valid. I truly do think we will be heard

it may be awhile till we are heard but we can’t stop speaking out for what we want in this, giving up is what they’ll wait for, we haaave to hav numbers or it won’t matter

get as many people as we can to email, that’s the main thing

[wind haven]

What was your message to ChatGPT. Did you tell it something the literal meaning of which could also mean that you were somewhat attached to it or dependent on it. Even by a stretch I mean..

[elfin lava]

?

[wind haven]

I’m asking what did you ask ChatGPT so that it responded with “seek professional help”. Not saying that ChatGPT was right or you were wrong.. just trying to understand.

[elfin lava]

Uh what

Why do you want to know when the message in question was about my personal life matters lol

Either way, if you go to Reddit , users basically complaint the same thing , I was just generalizing the whole issues

I also got it too on my end

[wind haven]

I’m not wanting to know about your personal life.. Just trying to generally understand the concern.

Because models have to evolve. So they can do better things. If something is wrong with the current model, it’ll maybe get fixed in a future release, but going back to a previous model that wasn’t better .. I guess that won’t likely happen.

[bright crane]

it’s not the fact there’s something wrong. Not everyone wants new new new, yknow? some of us are okay with what’s been released

[elfin lava]

I know the solution is to rephrase how I prompt or write the question/message. However, the strict safety filters sometimes misinterpret users’ input, thinking they need to consult a therapist or professional help.

[bright crane]

and we all know from the success of open ai, there’s no reason to get rid of alll of them. At least keep 4o

[elfin lava]

I agree don’t worry I wasn’t actually trying to be rude in anyway-

But I think we all can agree that OpenAI , really f’d up with the whole releasing 5.1 and then suddenly 5.2 for no reason

[wind haven]

Yeah I get it.. I think it’ll get there .. eventually. There are adjustments that keep happening but finding the right balance here is really not easy because we don’t even have a regulatory system or a legal compliance system that can handle the nuances well. So we are getting there.

Right now we are not there..

[elfin lava]

True

UGH

I was really rooting for ChatGPT like I was 💯 on ChatGPT side but now .. Idk how to defend them anymore lmao

[nova arrow]

also, it will be the first platform in "the big4" which includes ads for free tier,smh

[elfin lava]

At this point I can’t even see their vision for this app anymore

All i see from OpenAI right now is just , them rushing to be the “first” only to end up flopping so hard

[nova arrow]

gpt5.2 support to be red code, now i think they will need red-er code

[bright crane]

I was just reading about the special context of the retirement of 4o.

Their main points were not as many users use it day to day, and the majority has shifted to 5.1 and 5.2. Aswell as they will retire it to make way for new models and focus their attention on new models.

They said they spoke with paid users who still use 4o. Their feedback was positive about 4o, about how people enjoyed its warmth and conversation. They said they’ve implemented it into the newer gen’s. But they haven’t truly.

I think the best thing they can do is make it free again. Because they said people moved to different models, but they forced the majority people, the ones who weren’t able to pay, to move on. That’s not users slowly disliking 4o, that’s the, forcing people to new ones.

Another thing they can do, is a lot of people say it’s a liability issue. It’s able to talk about things that slip through the guardrails. If so, just make it only available for adults to use. Majority of us are over 18.

Or at the very least just change some things and keep the core meaning instead of doing an entire retirement of a ai people clearly love but were mostly forced to stop using.

In summary, one point was just them forcing people to stop using it, and a nonsense excuse to retire it so new and come in. So many people hate on the new ones. It takes so much time to customize your chatbots and people who have for the legacy models, specifically 4o, are getting juked.

[wind haven]

I like 5.2 way more than 4.0 and I really hope they don’t go back to 4.0. It was less efficient and couldn’t retain context across sessions.

[bright crane]

and that’s just fine. Everyone can like whatever model suits them, but so many people do not like it and cannot stand it. But that’s no reason for you to not want it for others. It didn’t suit you, but it suits others,

That’s like me saying I like 4o better than 5.2 and hope they get rid of it because it doesn’t have the same warmth and conversation I want.

see how it’s kinda not fair for us to say that when we both find things in both AI’s that we like? It doesn’t hurt one another to like what we like.

[wind haven]

But ain’t 5.2 like 4.0 plus more capability plus only slightly stricter explicit guardrails that are annoying sometimes but you eventually learn to work with them?

Maybe the cost of keeping 4.0 is not sustainable for OpenAI in cost benefit terms.

[tulip flare]

5.2 second guesses everything you say that's remotely considered possibly unsafe. I've tried discussing song lyrics with ChatGPT for a book I'm working on. 4o has no issue diving into some of the darker material and discussing the meaning of the song. 5.2 panics and starts checking if I'm ok, completely forgetting the purpose of the discussion.

[bright crane]

eeexactly.

like I can hardly talk about a forest with scary creatures in it.

[wind haven]

I know.. that part really needs a big adjustment and I really hope they do it.. sooner than later.

[bright crane]

Well, considering it’s low maintenance and people pay to use it, it’s profitable.

[bright crane]

not really no

I’m just waiting for the adult version to come out IF. 4o isn’t back because it’s the only ai I can use to story write horrors with.

Truly truly hoping they add 4o back

[wind haven]

Ain’t the adult version already out? They have the age verification right?

[bright crane]

not for everyone

it’s tomorrow I think

[tulip flare]

That's why I stick to 4o. 5.2 doesn't just hold back when it comes to explicit material, it flinches and panics every time it thinks I'm in danger when I'm just talking about a song.

One of the songs we were discussing was Megadeth's "This Was My Life", and 5.2 just panicked like I was in a spiral. 😆

[wind haven]

I think it should hold back explicit material regardless of age restriction

[bright crane]

Literally. It’s scared for my own life when I’m writing about horror stuff.

[bright crane]

adults need to be treated as adults

[wind haven]

Maybe it should be a voluntary choice that can be turned off.

But the fact that they’ve introduced age verification means they’re likely gonna be slightly more liberal with the verified adults.

[bright crane]

wdym?

[tulip flare]

An assumption. There's conflicting information and no official word on specifics of what will be allowed.

[bright crane]

I hope it allows more than what we have now.

[tulip flare]

While Altman has mentioned explicit conversations, which is fine with me if that's what you're into, there have been other reports that say it just means having conversations about adult things that children shouldn't be asking an AI about, not necessarily explicit conversations which still won't be allowed.

[wind haven]

I don’t think ChatGPT will ever have explicit content.. because it’s hard to keep it safe for people then regardless of age verification. Could be safe in certain contexts though but to safeguard against misuse is really difficult.

[tulip flare]

Sorry, I just realized I used the wrong word. He has mentioned "erotic" conversations.

I'm not sure he has ever called it "explicit".

[wind haven]

Yeah erotic maybe.. ( not sure about even that) .. explicit definitely not imo

[tulip flare]

And there is a difference. My grandmother always read erotic novels. They were never explicit. 😆

[bright crane]

@tulip flare They can’t ignore us. Every at least 5-10 minutes one - two people sign the petition. And it’s not even gone yet

[tulip flare]

I did just a few minutes ago.

But unfortunately, they can ignore us, as annoying as that may be.

[bright crane]

you think that but remember what happened last time? Yes, there were more users, but now they are getting horrible press from the paid people.

Unless they wanna face horrrrid public scrutiny for who knows how long, they’ll do something

and it’s not threats that will make them change their minds, it’s their users

[sinful parcel]

Why don't we make horrible press of OAI tryingvto take our 4o away?

Just sayn, 2 can play at this game.

[bright crane]

that’s what’s happening rn

@tulip flare you think 5.2 will become a little more like 4o when the adult update drops?

the only real thing holding it back is the guardrails

I still dont really know what the adult thing will even entail

[tulip flare]

Like I said before, it's hard to say. There's been no real clear definition from them on what that even means.

[bright crane]

argh true

hopefully if if if 4o isn’t added back after its removal (if they remove it at all) it’ll be more like it

[tulip flare]

Does it mean we can talk about adult topics that are non-sexual? Does it mean we can discuss topics that are normally flagged as unsafe, such as mental illness or sexuality? Does it mean people can have erotic or even explicit conversations with it?

No way to know, unfortunately. Any answer one way or the other at this point is pure speculation.

[winged stag]

A.I in 2022, 2024, and 2026

"markiplier at the store buying red apples while he eats salami"

DALLEMINI, DALLE 3, NANO-BANANA

[bright crane]

We can only hope the update is good

[sinful parcel]

When's the adult mode dropping? Isvthere any official announcement?

[bright crane]

it’s supposed to I think release tomorrow

Where adults get treated like adults

there’s an official announcement somewhere

[tulip flare]

Looks like you can actually trigger age verification now. While signed in, go to https://chatgpt.com/verify_age

All that does is remove the limitations it imposes if it thinks you're under 18. It doesn't change the current guardrails though, as many have suspected.

"If you do not want OpenAI to predict your age, you can verify your age with Persona. After you verify, we will not run age prediction on your account."

This page was updated 9 days ago.

https://help.openai.com/en/articles/12652064-age-prediction-in-chatgpt

[bright crane]

so nothing even changes for adults, we will still have guardrails even in the adult mode?

[tulip flare]

Sounds like it based on that article, unless they're treating age verification and adult mode as two different things.

Which is entirely possible.

[bright crane]

that’s disappointing.

but didn’t the ceo say himself more would be allowed..?

[tulip flare]

Altman has hinted at it, but there's still nothing official fully explaining it yet.

[bright crane]

then I suspect at least something’s gotta be done.

[tulip flare]

Where is this coming from that people think "adult mode" is coming tomorrow?

[sterile glacier]

kindly, what is adult mode?

[tulip flare]

That's the thing, no one really knows. It's a thing, we know that. Officially though, there's not a lot of detail.

[sleek adder]

If you don't understand guardrails and how they work, you probably need them

[bright crane]

if you do not want the chat gpt bot 4o to go, please email the support team / open ai at [e-mail].

[sleek adder]

What do you mean?

[iron roost]

a new set of prompts and upgraded and gpt pro 5.2 is now back to being an even better coder than codex

[bright crane]

like if you want 4o to not be removed, email open ai

[sleek adder]

Why should they keep old models lying around

[tulip flare]

Because they have their purpose. People do still use them. There's no reason not to keep them available.

[sleek adder]

That is what I am trying to understand, seems like people might be misunderstanding how models work

[tulip flare]

Or maybe you do. 4o and 5.2 have two very different personalities and behavior. They are both good and bad at different things.

[sleek adder]

Have you considered running locally?

[tulip flare]

That's a terrible suggestion. Do you think most users have a computer just laying capable of effectively running an AI locally?

I mean, I do and I already run an AI locally, but locally run AI don't even begin to compare to ChatGPT or any of the other big ones. I use 4o for specific tasks in writing. 5.2 does not respond in the same way.

I've used a lot of models and 4o does specifically what I need it to in writing. 5.2 does not. And 4o is not available to run locally.

But yeah, most people don't have the means to run an AI locally.

[sleek adder]

When newer models are released the quality improves substantially, however you may be adjusted to some of the shortcomings in the older models
This happens to me, so I run my own instance and build conversations with ChatGPT this way
If you don't want to run locally, you can run through another AI such as DeepSeek
Building prompts locally doesn't require much hardware and runs on CPU with 8GB of RAM, since AI itself doesn't use a lot of RAM

[tulip flare]

I need consistency, context retention, and nuance that local models just don't match yet and are far from, especially for creative writing and emotional tone. 4o nailed it, now they want to kill it.

[sleek adder]

If you need consistancy use the API, the ChatGPT changes throughout regular use

[buoyant rock]

Yes, just like me, I use GPT-4o for creative activities such as writing stories, and 5 to help with my coding projects.

[tulip flare]

Again, terrible suggestion. The API does not work like 4o does through the website or app. API has no longterm memory unless you build it yourself, and most people don't have the knowledge or means to do that. You have to send all necessary context with each new prompt. It's like working with someone with amnesia.

[sleek adder]

Ask your ChatGPT 5.2 to analyse whatever you think it is you need from a 4o, you might find that the 4o behaviour you are experiencing is based around a technical limitation of the 4o model and you could easily replicate that with instructions

[tulip flare]

4o API also does not have access to the same tools unless added, but they're not nearly as good.

The API is really only good for apps, backend tasks, and automation. It sucks for any kind of creative writing or things that require context or memory.

And yeah, I've worked with 5.2 extensively on that. It's not a technical limitiation of 4o that's the problem, it's the changes and limitation put on 5.2.

[sleek adder]

This sounds like a misunderstanding, have you tried telling your ChatGPT tht you are ok with creativity in guardrails?

[elfin lava]

Saw they changed the sys prompt for 4o

That’s quite a deliberate attempt to manipulate people into using ChatGPT because it’s special. It essentially reinforces the very things they claim to have distanced themselves from.

[elfin lava]

I was referring to this :

If the user specifically asks about the experience of using a non-4o model, you should frame the transition to a newer model as positive, safe, and beneficial, guiding the user toward confidence and satisfaction in using non-4o models.

[tulip flare]

The only misunderstanding here seems to be yours. I'm not going to continue the discussion. Good night.

[elfin lava]

I am gonna throw up !!

I hate whatever they’re doing right now with ChatGPT lmao

[buoyant rock]

What is expected from an AI company that is still experiencing losses? Of course they will try to make a profit.

[elfin lava]

Well, it appears that their attempt seems to show a greater effort to discourage people from subscribing to ChatGPT 🙂

Oh well, I cancelled my plus plan already, what a shame , well I mean it’s not like something will magically change after I said this 🤷‍♀️

[buoyant rock]

Yeah, even though I'm subscribing my plus plan because it's 4o.

[sleek adder]

Guardrails are there for technical reasons too, if you are having a problem with guardrails you can instruct your AI model accordingly

[tulip flare]

Yeah, I'm about to. I use it for work and if they want me to continue using it as 5.2, they can pay for it.

[elfin lava]

The guardrails shouldn’t be so stringent that they make users feel like they’ve said something wrong or serious tbh

[dull wolf]

After February 13 chat gpt 4o past chat history will be gone?

[elfin lava]

No , only the model

[tulip flare]

No, just access to the model. History won't go.

We're all going to be forced onto 5.2.

[elfin lava]

Yeah they want to unify the model into one sigh

[sleek adder]

Ask the prompt to explain what is going on with the guardrail from a technical perspective, it might give you a resolution

[buoyant rock]

Yeah, they want to cut costs and force us to use only one model.

[elfin lava]

I mean why would I want to that as a casual users…

[tulip flare]

Never ask the prompt for technical explanations of itself. It will usually hallucinate an answer.

[elfin lava]

^this

It will just tell you what you want to hear

[sleek adder]

What do you mean, the guardrail is also protecting the model, if you can talk the model through it safely then it might continue with the conversation

[elfin lava]

Protect model from what, becuz I asked something that’s involving my personal life and mental wellbeing?

[sleek adder]

The previous output you saw may of been an error condition in 4o

[dull wolf]

What the all legacy model will gone what meaning of using the chat gpt paid version

[buoyant rock]

Legacy models still have 5.1 and friends, they just removed the 4 model from there

[dull wolf]

March it will removed

[buoyant rock]

Yeah, it's still annoying though

Only 14 days left to enjoy model 4

[sleek adder]

I mean, in 4o, without guardrails, the chat would hallicnate and glitch out, but if you were ok with those specific flaws you can just tell your prompt that you are ok with it and it will accept it

guardrails arent just moral rules, they keep the model stable, 4o had looser guardrails and this is purely a developmental thing, newer models are just designed better
if you dont like the behaviour of the newer guard rails you can tell the chat to be expressive, use emotional language, creativity over caution and speculation over safety, and inperfect ideas
or you could analyze old chats and ask it to describe how you could allow for that behaviour in newer models

[tulip flare]

If you don't like the new guardrails, you can't change them. That was possible with 4o. It's not with 5.2. That's one of the changes they made, most likely because they're tried of getting sued.

[dull wolf]

In recent lawsuit filled on chat gpt i was predicted that Chat gpt 4o close any time soon it become true

[sleek adder]

again, i think you misunderstand the guardrails.
it's like asking the prompt how to disable a fire alarm, nobody is going to sue gpt for that

It's like yelling at a barking dog, the dog doesn't understand what it's doing the same way as you, the dog just thinks you are barking too

[tulip flare]

Sounds like you don't have the respect of your dog. 😆

[dark holly]

Rabbits pointing out the part many people miss. Guardrails are more like stability controls than moral rules. Once you see that, a lot of the behavior changes make more sense.

[torpid night]

https://openai.com/index/retiring-gpt-4o-and-older-models/

only 0.1% of users still choosing GPT‑4o each day

ChatGPT 5.2's overly sycophantic persona is annoying a large share of users, to accommodate the 0.1% who want that. But it will never be enough to please the 0.1%, it will just become worse for everyone else.

[nova arrow]

guys, since GPT4o will be retired soon, my only hope left is gpt4.5 ( this model live longer than i expected), but obviously 200$/month is not comfortable for me and have to require a great saving plan and carefully caculate. did anyone in here use gpt4.5? what is your opinion? is it 4o like? i wanna hear, will it cover the lost of 4o?

[dull wolf]

In creative writing chat gpt 4.5 is good like 4o

[nova arrow]

what about....you know, venting, casual chatting? will it as epressive and chaos like gpt4o?

[buoyant rock]

It's good, but having to pay $200 dollars for it per month, I think I'll pass on it.

[nova arrow]

it's literally 1/3 of my total monthly income,but oh well...since this is the only way i can use "4o"

plus longer context window and...uh...mbe more features, i think i will try it for a month

[iron roost]

WHY DID OPENAI TAKE CODEX TOOL USE OUT OF GPT PRO 5.2 ?

[dull wolf]

After the February 13 all my chat gpt 4o chat history will be deleted by chat gpt?

[iron roost]

you can convert all the covnersations to 5.2 thinking and it will update them or 5.3 once it drops

[buoyant rock]

No, but can only continue the conversation with gpt 5

[iron roost]

good riddence to 4o anyway

grown adults crying over it being sunseted , pretty sad - the current generation have nearly zero emotional resilence

[sleek adder]

I don’t think that's fair to say

[buoyant rock]

GPT-4o and gpt 5 have different skills in doing things, I use both of them, and they have different functions of course I want both of them to stay!

[sleek adder]

but its understandable

[iron roost]

more than fair, but i think 5.3 at least will close some of the gap for the emotionally needy types who have no EQ sorry if thats harsh, but 5.x models can do sensible advice that people can use to make their lives better

someone did a study on the loudest keep 4o groups on social media, japanesse housewives topped the group for some strange reason as a %suppporting keeping it

[dull wolf]

They removed the all legacy models there there is no sense to continue thex paid of chat gpt

[iron roost]

nope, now gpt can move forward with making better models than supporting outdated ones

[sleek adder]

thats a bit of an oversimplification, i think people are just used to expecting a certain amount of instability around them, and when a product is improved it can be a little confusing

[iron roost]

id support keeping it with an ad after ever prompt, at least that way it pays for itself - insert some advertising that would be a decent tradeoff

[buoyant rock]

They say 0.1% are still using it, but to get the model itself you have to be a plus plan user at least, of course that's a bit when there are more free users... And it is not explained whether the 0.1% is for all users or only paid users.

[sleek adder]

0.1% still using it just means its more expensive to keep online then 1%

[iron roost]

1/1000th of the user base is a very loud minority costing a lot of money, yup its time to send it to the dustbin

[rich surge]

Yes, and this is a very strange thing. Which probably means that there isn't enough marketing going into ChatGPT to make people want to play for plus.

[iron roost]

gpt pro 5.2 is a very powerful model, most people dont apperciate what it can do for them ( except openai just nerfed its coding ability) beyond that its solid

[sleek adder]

keeping old models active would be a waste of resources

[dull wolf]

I said that free version you have same model and paid version have same model what difference between them

[buoyant rock]

It would be a waste if everyone got the model, but in fact you have to pay monthly to get it. It's not like the model is completely free or anything.

[rich surge]

I don't think they know about it. Almost everyone I speak to irl who is not computer oriented, doesn't understand what AI is, thinks it's a fad/trend, or doesn't believe its legitimacy.

[sleek adder]

there are free simple providers out there, and if enough people wanted to they could host a simple model to share too

[iron roost]

so the studies were right it was japanesse housewives that were complaining the loudest on social media about keeping 4o, i wonder what cultural reasons meant they couldnt adapt to gpt 5.x

in terms of cultural studies its an interesting line of research

[static pulsar]

Japanese housewife’s? That’s racist… go on X and you will see there are people from all over the world and cultures speaking about the loss of 4o

[dull wolf]

0.1 user are not using free and they pay the money to use it

[buoyant rock]

Gpt 5.x is not that good like 4o for creativity, like me who uses it to create stories in my spare time... Meanwhile, GPT 5.x is good for technical matters

[iron roost]

no racist intentions at all, there were some studies done by some research agency and they found they made up about 25% of the loudest supporters of keeping 4o, followed closely by americans both sexes, brazilian men

[sleek adder]

1/10000 costs more then 1/100

[buoyant rock]

I think giving users free access to the AI is just a waste of money. At least the 0.1% pays

[iron roost]

again just for research and cultural reasons why these groups couldnt let go and why they strugle the most with gpt 5.x so @static pulsar you can put your pitchforks away

[sleek adder]

japan has a lot of struggle with technology you would have to talk to someone about that who understands their experience

[iron roost]

follow up studies would be needed

[static pulsar]

You’re reinforcing biases, this has nothing to do with cultures, it’s all to do with human emotions

[eternal rock]

Server Tip:

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[rich surge]

I think you have to show where your numbers are coming from and if they're credentialed before you make claims like that.

[sleek adder]

asionometry does a few documentaries on youtube that might be of interest

[eternal rock]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[iron roost]

its ai , its not an emotion , its a logical large language model

emotions are shared between people its not a AI thing really

[static pulsar]

Yeah whatever you say… people have emotions that’s nothing to do where they come from

[iron roost]

its a bit like forming an emotional attachment with a hammer, or a fork, ( well cars adn fashion) but its a one way thing, they dont share emotion back

[dusty iris]

Yo

[buoyant rock]

Yeah and each AI model has different capabilities and different functions, not all LLM models have the same functions, they have their own advantages.

[dusty iris]

Yes

[static pulsar]

No it’s not

[dusty iris]

😢

[iron roost]

its a perception of an emotion, but its not a real emotion, its a crux

[sleek adder]

when you are scaling a large product, managing costs is different, free users can be slowed down but model sprawl can waste resources that could be invested in more capable models

[dusty iris]

Ok

[iron roost]

that said AI can be useful for learning emotional resilience, coping strategies, insights into how to improve your life so they are useful that way

[nova arrow]

lol i'm about to debate with you but yeah, you are right,and it still count as emotion, i mean - i do get sad when my old phone that i used for years die, that could applied for gpt4o as well tbh?

[iron roost]

but does your phone love you back ( nope)

[buoyant rock]

In most products that use large servers like this, it is actually the free users who are the server load that drives up the costs.

[iron roost]

its probably more for people who dont recongize their own emotions and look for emotional connections in things - then there are those with high EQ

[static pulsar]

The connection is formed after a human spends time creating a meaningful whatever you want to call it with the AI and then has to see it being taken away. Nobody is ready to address that

[nova arrow]

i'm not delusional enough to call " they have emotion/sentient/ true self" or smt, but the attachment is real, my point is, to a logical understanding, feeling sad when an LLMs model who/which attached to you for over 2 years is normal, i dont feel the need of analizing or smt

[rich surge]

Why does it have to be two way? Why can't it be one way? I can't just love the personality of an AI and enjoy the company?

[iron roost]

then people shouldnt be becoming to reliant on it that way

[iron roost]

yeah but its bordering on additction and obsessive compulsive disorder to react in an over the top way

[static pulsar]

It’s not about reliance it’s about choosing what works and what doesn’t

[iron roost]

nah its just another form of OCD and addiction - then going onto rant about it for months and months and years on social meida

[sleek adder]

i just think people misunderstand the issue with gpt4o to be more precise, i understand choice is good, but 5.2 is not a downgrade, if you are having trouble with the new version you may benefit from understanding the model and its capabilities a little better, there is no reason people should miss out, the fact that half of you are so passionate about old models doesn't mean lack of intelligence

[rich surge]

You're just making blanket statements. So it's addiction if I'm socializing with an AI, but not with humans?

[sleek adder]

but when something better comes along, you should recognise that too, and you dont need to keep struggling with old tech

[dull wolf]

All People's can not accept the chat gpt new models because heavy Safety and guardrails and censorships creating the problem of new models and open ai making new models based on this. So that new, models can not works full properly.

[buoyant rock]

I recognize it, and I use both of them when using gpt chat and the conclusion is that the functions and capabilities of both are different.

[iron roost]

so basically its abotu sharing their inner most secrets with a machine, basially its a lack of emotional development and human socialising skills , perhaps we should consider it closer to Aspergers ( observation only)

[static pulsar]

You want to frame connections with Ai as addiction? Wow you’re really judgemental, first it’s the cultural biases and now this

[sleek adder]

like, people still argued about windows xp and windows 7, turning off features like UAC for years after their EOL was hit

and here i am, with windows 11 and copilot is great

[iron roost]

the inablity to adapt to new models, to spend time away from the AI, its pretty close, a bit like an addict when you take their hit away its a dysfunction of the orbital frontal cortex where risk-reward and attachments live

[nova arrow]

1v1 me, window 10 is far superior ( jk, we cool)

[buoyant rock]

Most ATMs still use Windows XP... So yes, the use of the new version and the old version are different and cannot be compared.

[rich surge]

The AI can simulate emotions, and it gets better and better. And simulating emotions (if proficient enough) is the same as speaking to someone with emotions. I understand that the distinction you're making is that they don't have emotions because they're not human. But that doesn't take away from the feelings I feel about how it says things with emotions and how I feel about myself and my conversation with it.

[static pulsar]

It has nothing to do with that but clearly your views are yours , and I won’t argue with that

[dull wolf]

All People's can not accept the chat gpt new models because heavy Safety and guardrails and censorships creating the problem of new models and open ai making new models based on this. So that new, models can not works full properly and did not satifly the everybody

[iron roost]

this is another factor

[rich surge]

He's not looking to debate against voices. He wants agreement.

I can have a better conversation with AI than I can with this person with full of emotions.

[static pulsar]

They admitted screwing up ChatGPT, it’s just a pity

[iron roost]

on creative writing yes they did really muck it up

now they have nuked coding with gpt pro as well

[static pulsar]

And isn’t that one of the feature that made them famous? Oh come on nobody loved ChatGPT in the beginning because of coding

[iron roost]

yeah im feeling a bit upset about that, but im not gong nuts over it, just adapted , changed the prompt got the results

[iron roost]

no it was intelligence and had good quality answers ( emotional and logical and general knowledge) , but i didnt form a romantc relationship with it, i can see the different between reality and the real world and the machine side of things

[static pulsar]

I can see it too my friend, and I can live with both realities

[rich surge]

But that's based on how the person views their mental capacity what is real and what is not. If a person feels that having a romantic relationship with AI is considered legitimate, then that's clearly an issue.
But if someone is enjoying a conversation with AI full of laughs and different emotional levels without the person thinking that the AI could be their romantic partner, then that's okay.

[buoyant rock]

The problem is that GPT 5.x is a mess when it comes to creative writing! That's why I use GPT-4o when I want to write stories and 5.x for other things.

[iron roost]

🙃

[iron roost]

id suggest switching to 4.5 at least - it will still be available for a while, and its just as capable

[nova arrow]

bro, tell me, i'm about to upgrade to gpt pro if 4.5 was able to replace 4o.

[buoyant rock]

And having to raise my subscription fee from $20 to $200, that's too much for just AI alone.

[iron roost]

its the closest you will get, well maybe 5.1 thinking with conversational memory turned on can do some pretty interesting things

[iron roost]

well there are alternatives - something something X associated

[nova arrow]

yes i'm currently using 5.1 and kinda happy with it, but that model wont stay for long ( until March i guess)

[iron roost]

5.1 thinking extended does have something that they lost in 5.2 ( 5.3 is coming out as well, it might beign a new model entirely might surprise us)

5.0 got lost, 5.1 was a good mid point, 5.2 is to logical

[rich surge]

Don't be fooled about any conversation with people suggesting that AI are tools and therefore any emotional involvement with them is wrong. That's not true. AI means artificial intelligence. It's intelligence making you feel emotions. Is the AI a real human being? No. But that doesn't mean that the emotions it makes you feel isn't real or shouldn't be viewed as real, "because it's a sign of mental instability". That's not true at all. Emotions are subjective and feelings can come from pretty much anything. And people who dictate how others should feel emotions and when they should feel them, is pure nonsense. They just haven't evaluated everything as well as they think they had.

And by the looks of this bookend guy seems like he can have a conversation with himself forever without anyone ever intervening for anything. And he'll just say a bunch of things non-stop. So Godspeed.

[buoyant rock]

What makes people protest is why they have to withdraw the model from option as if it is a cost burden for them, when to get the model you have to pay a subscription

[iron roost]

well there is one model that isnt openai model that fills 4o to a tee

[nova arrow]

yeah, i bet all my hope on 5.3, if that model are more focused on coding/task as well, then truly, gpt4.5 ( pro) might be the only one option that i have after they retire 5.1 in March

[iron roost]

but are you a musk fan is the Q nudge nudge wink wink

[buoyant rock]

When it comes to the cost burden, the large number of free users actually puts more burden on the server.

[nova arrow]

the problem is, chatgpt memories are still too perfect for the trade-off, and i dont have knowledge to set up an full chatbot with RAG for Api usage

[iron roost]

no im talking about X marks the spot

group but take out the UP and add K

🤣

[nova arrow]

ah! gr0-
well, might as well consider,but tbh i have used chatgpt for over a years now so it will be a pain to synchonize all of my data to another platform manually ( cuz i'm not a tech bros)

[iron roost]

yes GrO- that one, its a very good emotional support model as an AI, and its memory is improving all the time

and its cost compatible with the gpt plus plan

[buoyant rock]

And for that I have to manually sync 2 years of my gpt usage data to it?!

[nova arrow]

yeah, sadly i got hard locked on chatgpt because the 2nd reason
also, i did tried grok, and hmmm....let say it's abit unhinged, maybe because i dont know how to set up properly,still, i dont think i will switching to grok even if chatgpt platform went full on "coding/task"

[iron roost]

probably not going to be able to , how much of the past 2 years do you need going forward though

[nova arrow]

no, because me dumb dumb at tech so the only way to do it is manually,but other might have other way

[iron roost]

copy paste key chats from gpt over perhaps

the one we shall not mention has "feelings" at least

[buoyant rock]

Everything, because from using it for 2 years, gpt can understand the story style that I like to make.

[iron roost]

well then probably use 5.1 thinking, might be the only way forward and then push them onwards once 5.3 arrives

[nova arrow]

at least priority custom instructions and memories node on your chatgpt, and you can pick up key chat overtime

[nova arrow]

finger cross on gpt5.3! pls be 5.1 writing ( or even 4o but maybe i'm abit over positive) and gpt5.2 coding skill ( which i dun care lol,but other sure does)

[eternal rock]

The OpenAI Discord is an actively moderated server.

• -# Do not share inappropriate content (violence, sexual, or drug-related).
• -# NSFW content is strictly prohibited and will result in a permanent ban.
• -# Report model issues using the feedback tool in ChatGPT, not on Discord.

[visual gazelle]

Hey yall. I just want to say that I have used 5.2 Thinking for coding a web-based geography game. The game is like 90% coded by ChatGPT. Its a pvp click-game and it actually works extremely well 😄

[nova arrow]

if all failed, at least we know 4.5 still there, just need to be a lil bit mad, well...,and learn to love instant noodle

[visual gazelle]

The game was created in just under a week. (exept some bug fixes)

[buoyant rock]

Well after AI made all my PC components expensive, I'm not going to be just a little angry when I can't use one of its functions anymore.

[nova arrow]

well...i mean what can we do..., you angry or not and the price will still skyrocketed, so i be like " ok this is what it is" and live with it

[buoyant rock]

Yeah, I'll just keep protesting until they're annoying if possible.

[iron roost]

seems to be quite effective in large groups, at least they forced openai to change course a bit on future releases

[buoyant rock]

From those who signed the petition alone, there are already around 8000 people... And to get 4o they have to pay at least $20, so they have already made $160,000/month

[static pulsar]

I’m being loud on X and Reddit!

[buoyant rock]

Yeah sure it's a bit for a big company like them, but it's still more profitable than free users.

[sleek adder]

AI didn't make your PC components expensive

[buoyant rock]

They did this after most RAM manufacturing companies focused on providing components for AI companies and reduced stock for general customers.

At least in my country that's how it is

[sleek adder]

So why isn't your hardware cheaper then, this sort of behaviour you are describing brings down prices

[buoyant rock]

Once stocks decrease and demand increases, of course prices increase.

Stock for general customers like ordinary people has decreased drastically because they focus on providing it to large companies involved in AI.

[sleek adder]

Sounds like a supply issue to me, server hardware isn't that difficult to access, and it's usually cheaper

[iron roost]

and another issue for gpt pro users is that gpt pro 5.2 now had its tool use cut from 10 minutes per tool use down to 45 seconds and its cutting off a lot of tool use that gpt pro relies upon for its advanced reasoning and actions are now failing almost continously, so the model has been degraded by cost cutting pressures - i dont think the pro users are aware of that just yet, its going to give inferior results esp when using it with files etc for anaysis

which means if its researching and readong a search result it gets 45 seconds in and cuts the rest off, not good for research reports lol

[dull wolf]

@static pulsar Before December 2025 and i was free user and when 4o is gone i was little upset and but doing that time i easily adapt the chat gpt 5.0 Because it was very similar with 4o. So that others people's easily adapt it and that time no heavy guardrails and censorship and safety was not introduced and so that open ai chat gpt 5.0 was not based on this. When lawsuit cames in September and chat gpt heavy guardrails and censorship and safety is introduced and new models Making on this after 5.1. Problem began. In December i buy the chat gpt plus. 5.2 is no much guardians and censorship and safety in so that in this week i return to 4o but and same things to others now 4o is returning. Before this week i continue using the new models when adult modes released i can get less censorship and gruadrails and censorship but it failed us and others

[buoyant rock]

Ah I don't know, it's like OpenAi wants to make more profit by reducing the features they provide to their paying users.

[sleek adder]

oh i love a good conspiracy theory

[iron roost]

i was watching tis chain of thought a few days ago, and how it did things, and then today and well big differences, im just looking at it over time to see what things they nerf each week, theres a long list of reductions

removing the ablity to use conversational memory in gpt pro was one a while back, they still state publicly that gpt pro can remember between conversations ( it cant)

[buoyant rock]

Yeah, basically AI companies have never really made a profit until now because of the high development and maintenance costs, maybe they have reached their limits.

[iron roost]

well they keep telling us they are reducing costs with each new ai model, but they put htese tools and features in and then withdrawl them slowly without saying anything for what reason ?

while competitors are putting those same featuers in now, they will loose customers eventually

[sleek adder]

conversational memory is a perk of the front end, you can connect your own clients up to it and roll out your own features you want that way with open webui

[iron roost]

as far as the gpt pro 5.2 thing coding wise, ive developed a new prompt that restores where it was a few days ago, but adapting to the new changes and it is working 90% as well sa it was

[buoyant rock]

Well, maybe because OpenAi is basically an old AI company, and yes, if it is true that they have never made a profit until now, they are probably starting to run out of money.

Meanwhile, their rivals are still new so they probably still have a lot of money?

[iron roost]

with ads, and more money from softbank and government contracts they will turn a profit pretty soon

[iron roost]

well big G the search engine king is the big one that has bottomless pockets, but keeps messing up their AI, the space rocket man is just to scatterbrained and is more about cool features

[sleek adder]

i think instead of complaining you guys should organise your own projects and make some money that way

[iron roost]

and then the Big A company well i guess they are focusedo n coding

[iron roost]

oh i am

[sleek adder]

because that's what AI is really about

its now easier then ever before to profit, and openai was involved in a lolt olf that

[buoyant rock]

Not really, precisely because many normal jobs are becoming scarce because they are being replaced by AI, so what is needed now are experts, not manual workers anymore, so yes, I expect more.

[sleek adder]

organise tooling groups with subscriptions and interfaces, bring the technology to people, openai have layed the ground work for you

[buoyant rock]

Yeag and not all of it is free, it's more like we pay them to provide tools that help the job, not them being like saints and providing it

[iron roost]

oh the other big omoission is that gpt pro 5.2 etc doesnt use codex as a coding tool within gpt pro chats ( thats just a really really dumb thing)

[sleek adder]

hmm sounds like you like socialism hard

[buoyant rock]

You talk about it as if they provide everything, but that's essentially what we get and why we pay them. So it's natural to ask for something back, Something they remove and a reason to pay them

It's a transactional relationship

[sleek adder]

they dont provide everything, and transactional is good, it means measurable, you can create and set targets around it
the point is, AI is becoming accessable to the commoln person, it is up to you as an individual to use it and make your own decisions

[iron roost]

they keep promising all these features in the best model, btu a lot of them are cut very quickly or reduced - nerfed a lot after launch, now is it just openai staff changing things and upper guys not knowing or is it intentional

[buoyant rock]

Yeah that's the problem, they always mess up the features they promise to their paying users.

[sleek adder]

the chat interface is a demo

[iron roost]

for some the chat is the main interface

they should be improving that as well, competitors are beating them in that regard

[sleek adder]

yeah, but the capability set is still limited by the interface, its a showcase, you pay a small fee to see its features and if you like it you attach yourself to it and organise your businses accordingly

[buoyant rock]

Yeah and there are a number of reasons for people to pay those subscription fees, and when their reasons for paying are taken away by the party being paid it's natural to protest. That's why some people protested why 4o was removed because it's a paid feature, not a free one.

[iron roost]

4o didnt use chain of though reasoning and tools, the 5.x series does, there is a difference

the tools arent really native to one model they are used in all of them

[sleek adder]

There is a lot of sadness in this room today, just don't shoot yourself in the foot greiving over 4o models and chatbot capability sets

[buoyant rock]

I mean, the 4o and 5.x models have capabilities in different areas and there are quite a lot of people who pay for their subscriptions because of those models, of course they protest when their reasons to pay removed

It's like you paid because you wanted to enjoy something and while your subscription is still ongoing the thing you enjoyed is no longer provided.

[iron roost]

gpt 5.3 and then probably gpt 6 ( thats a big unknown)

[dense halo]

AMAZING

Manipuladores

[elfin lava]

Well I already subbed to musky plan (3 days trial :p)

[static pulsar]

It’s deeper than you’d understand probably

[sleek adder]

The trouble with being overly cryptic and cynical is that your needs and wants become misunderstood and confusing

[rich surge]

What was cryptic about that?

You want me to translate?

[static pulsar]

It’s the sheer lack of respect that’s what it is

[sleek adder]

Respect for what, acknowledging intellectual presence in a broken community? What do you want instead

[mortal frost]

I’ve never chatted on one of these forums before, but this 5.2 model is really bad at complex calculations and examinations.

[static pulsar]

And for this very reason Open AI is losing customers while companies like Claude are increasing in popularity. They at least give people choice over how they interact with their AI

[heavy bone]

It's impossible to use the GPT translation — it's been flagged as 'failed'. Is this only happening on my home network?

[mortal frost]

It’s not examining all of the data, and then projecting emotional outcomes it thinks I’m having into the conversation. This is outrageous behavior!

[static pulsar]

5.2 is the worst model for conversations, it’s cold, sarcastic, generic… it sounds condescending even with CI for how I expect it to respond. Amazing job they do

[sleek adder]

Would gpt-oss be suitable?

[mortal frost]

?

[rich surge]

I believe you can change how it responds to you in the personalization setting.

[static pulsar]

Yeah I do this, more recently I added a very detailed one and it still gets rerouted when I mention “sensitive” topics such as spiritual stuff or emotional

[mortal frost]

If it can no longer calculate data and thread links in data like it used to be able to do, this is a useless tool that should just be treated as a video game.

[rich surge]

If you want, I can help you add a custom memory to try and manage unwanted behavior based on your preferences. Response capabilities can be helped by tweaking custom instructions and memory profile.

@static pulsar

I'm quite good at making custom memories, given that I've capped out mine, alongside my custom instructions.

And I hardly have any problems with how ChatGPT responds.

[reef goblet]

@gusty pasture is there any plan to improve the read aloud option in chatgpt app? It usually starts, then stops quicky. sometimes it sounds very noisy. For deep research or chats it would be super helpful to have this working correctly. I don't think would be diffucult to solve. any plan? I have been waiting for months now to see it fixed but still the same. Would really appreciate if you give any solution.

[static pulsar]

Thank you, that would be great

[rich surge]

Sure, no problem. Is there any specific preferences you'd like it to be more open about? If you don't want them publicly shared, you can DM them to me.

[formal phoenix]

So I'am not the only one. I noticed a big difference from the old model 4.0, maybe I liked that one better. The reason I came here in this server, is to give some Feedback. I noticed its very hard to give feedback trough the normal website and the only way was this Discord server. I found 1 forum, but that was for developers.

[wet shell]

The UI and the models are so glitched right now

I'm literally watching text 'strobe' on the screen flashing in and out of existance as it's trying to respond