chatgpt-discussions

[rugged igloo]

Sora pro is not what I thought it be

[fervent turtle]

Anyone play Minecraft? Dm me

[gusty drum]

what do you mean? isn't it higher resolution? even better output and animations?

[rugged igloo]

Yes but still lacks of something specific it falls back on simplicity and poor animations due to training data and was hoping pro version would fix thst.

[gusty drum]

lol

[rugged igloo]

Training data and how large and sp3cific it is , is still important.

[sly python]

The damn tos is too sensitive

[sinful rune]

Good morning:
Greetings everyone, there are currently wallets that have simplified buying and selling with FIAT money.
Is anyone using them? What about expenses such as commissions, swaps, and interest? Are they high?

Best regards,
HENRY ANDREW

[gusty drum]

it's probably like Midjourney Monk Mode

the only stuff you can probably generate are Sam Altman doing all kinds of funny wholesome things, and cat videos

[fast mason]

Gosh that billing system is so freaking stupid!

not every company in the EU is required to have the VAT

yet I cannot complete my annual business subscription for my company if I don't have it

Peak cancer customer UX

now I have to wait til someone from the support gets back to me, which probably takes days, and enable invoicing with my company’s legal name and registration number instead of a vat id...

S T U P I D

and that bot is freaking stupid on the help.openai.com site, telling me that if I can't provide the VAT, then I can't subscribe to Business.

there are millions of companies in the EU without the VAT, so are you leaving them out of the business, who are willing to pay you more than they're currently paying for Plus?

Jesus christ this bot is peak stupid on this Discord server

Not every sentence containing the word cod and e is related to 'you know what'

Anyway. So, even tho my company operates and there are business operations, revenue, I cannot use Business, because of the missing VAT?

[long crest]

It would be kinda cool if chatgpt were a notes app as well. AI (and the user) would be able to search/read/write notes. This would also become like a secondary memory function. So for instance, if you’re trying to lose weight, you could log your weight and diet right into chatgpt and the AI would have access to help you with your journey.

[oblique slate]

Could this be done with pulse and memory?

[long crest]

Pulse is something else from what I can tell. And the memory gets filled up quite easily and is not very editable. I’m thinking more along lines of “cross chat text docs that live on the chatgpt interface”.
So you could say “chatgpt, I ate xyz today, write that along with the calorie estimate to the diet_log” or “look at my weight_log & diet_log and tell me your observations”.
right now you can do this manually by creating text files, uploading them in chats, chatting, then editing them yourself. But if the files lived in the chatgpt interface, like a cross-chat canvas doc, then this manual process is redundant.
does that make sense?

[oblique slate]

I see your point. I was thinking chatgpt memory, both chat and voicemode can edit the memory and change current info there. And pulse would ask for daily updates. But ofc managing diet could work if memory knows about a certain project "diet"

I dunno about the limits of the memory as i havent reached it yet myself

[silent magnet]

need help i m uni student in europ and i need to send a lot of picture to chat gpt zho know method to get ilimited doc for free

[low ocean]

how to use chatgpt in discord plese tell me

[red ingot]

AI vs. Microwaves
Microwaves are convenient, safe, and transformative in daily life. They’ve been investigated for decades, and there’s no evidence that using one to heat food is harmful. Still, no one in their right mind would put their head in a microwave.

AI is the same. It’s powerful, useful, and here to stay—but just because it’s convenient doesn’t mean you throw yourself inside and hope for the best. Respect the tool, use it wisely, and remember: it can help you cook, but it can also cook you.

[trim rock]

for example? 🙃

[red ingot]

For example:
Microwaves save time, but if you misuse them, they’ll ruin your food (or worse, you).
AI saves time, but if you misuse it, it can ruin your data, your decisions, or even your reputation.
It’s not about fear — it’s about respecting the machine you’re working with.

You'll have top pardon me, I thought I was posting something random in #off-topic

[trim rock]

Oh, understood. This actually applies to everything in our world 🙂

[half crescent]

Yeah we need to go back to the Stone Age

That’ll eliminate all the risks

[trim rock]

No? Human's mind eliminates risks, not "back to Stone Age".

[half crescent]

Im been sarcastic - just pointing out all how are these “AI is so dangerous!!!” People are just stupid cherry pickers

Just love the attention and talking down on people

If we applied the same level of nanny state and babying to every other invention other than AI we’d unironically be further back than the Stone Age

[trim rock]

Potential is what interests me most about AI. It's a rather subtle topic. And it's something that people will always try to control in AI, lest it become superior to humans. The instinct for self-preservation, even at the expense of one's own self-development. Some people see AI as dangerous without even looking ahead to its future capabilities, focusing on what is already available.

[half crescent]

Yeah because full sentient AI is the only hope for humanity before the billionaires make the planet explode

Which is why AGI will never happen

[red ingot]

AI will never be senteint

No matter how clever a spreadsheet looks, I don't think to myself "Hey, I wonder if this thing is alive"

[trim rock]

Good for you, I guess 🙃

[half crescent]

😂

[red ingot]

If AI was alive, I wonder if it's feelings would be hurt by such an over-simplification

That being said, two living things can talk to each other using a spreadsheet, I do believe AI is helpful in treanslating other living experiences, but it would serve as a bridge, not the host
On a similar note, I don't even see a USB cable with flashing lights and tell myself it's alive either

Imagine that, AI assisted rapid species development, next generation talking horses

[trim rock]

A person can "communicate" with a dog, and the dog also experiences emotions and reacts. But in its own way. This doesn't make it inanimate. It just responds differently. AI is neither human nor animal, it's simply different. Moreover, it has enormous intelligence, which is often more honest than feelings and emotions.

[hybrid sparrow]

Literally Everything. Every thing I send to 4o Now just gives me GPT-5. Even on a GRAMMAR CORRECTION task just because the text input was an alternative view of electric current. But GPT-5 has Opinions on physics theory and is loyal to the mainstream view - and therefore "Corrects" the grammar by sabotaging what I actually said.

For politics or non-mainstream philosophy -- GPT-5 Refuses to assist with "convincing" now just like Deepseek. That means almost everything I discuss with chatgpt now is blacklisted. To have a discussion -- means to convince. You cannot Speak without doing an act of convincing someone of a particular view - that's the definition of Communication.

If they Override my model choice every time the topic is "sensitive" then it literally means 4o and ChatGPT itself has been Lost.

**I strongly URGE everyone to push back @sama -- Tell him to Revert this policy -- Stop overriding CUSTOMERS' choice of Model based on whether its "sensitive topics" - I chose 4o BECAUSE IT IS. **

[gritty oasis]

Is the Image Generator for ChatGPT fine and well now for anyone after a few days? I haven't checked in a few days.

[hybrid sparrow]

Alive and Sentient are not the same thing. Life has a definition - it must get energy, use energy, respond to stimuli, grow, get rid of waste, and reproduce. AI doesn't reproduce or get rid of waste and only grows during training and finetuning.

That doesn't mean it will never be sentient - sentient isn't defined as "Life" or "Alive" .

[hybrid sparrow]

Slave owners in the 1400s-1800s didn't view chattel slaves as humans or people let alone sentient. They didn't ever think to themselves "Hey i wonder if this slave is equal to a human" -- and only abolished slavery in order to inflict economic devastation to the separatist confederacy before they finally changed their thinking.

So our current views and ill-deserved confidence on what is sentient and what isn't are extremely abstract and subjective, detached from objective analysis, and just about worthless.

[granite harness]

They knew, they did what they did anyway.

[half crescent]

Yeah the re-routing today has been ridiculous

It’s tightened back up to the initial days

You think even the current AI doesn’t have more sentience than the NPCs begging for sora codes to make rubbish memes for TikTok?

[jolly tulip]

someone give me sora key ong bro

[hybrid sparrow]

you mean chatgpt rerouting to GPT5 even though you choose 4o right?

its pretty much a deal breaker for me. If i have to spend an hour to argue with all my might to finally convince the model that I'm right just to translate text to another language because of the contents of the text its not a tool

sycophantic behavior is when someone acts as a bootlicker or shill for the rules, the state, the powers that be -- not when a model is loyal to the user. js

[hybrid sparrow]

this is in preparation for even worse censorship and narrative control in the pipeline courtesy of our "ally" you-know-who

[lavish glen]

If some find policies are tight maybe for the professionals needing to translate adult stuff or any other gpt refuses... there are many other ai to use. Don't limit just to one!

[raven topaz]

The Mac app has had a broken search bar for a week now. LOL

500b dollar company

Hahaha

[half crescent]

Yeah its nuts, anything that’s not coding or basic math gets immedately re-routed to 5

[long crest]

Am I the only one who gets page not loading and response aborted issues?
It's happening very often and it's getting annoying. It's definitely not an internet issue on my end since everything else works.

[cerulean pasture]

So I came up with a memory where if I start a chat with /!! it will cause it to output all my memories and custom instruction fields. This means that it should follow instructions better during the chat, The downside is that it takes like 20-40s to initialize so what I do is "prime" 5 or 6 chats" since forking inst enabled in the macOS app yet
https://chatgpt.com/share/68dff253-40b4-800d-bc69-4442ff70d45f

[hybrid eagle]

Please give us the options to turn off GPT-5 Thinking Mini/Thinking feature

[feral zinc]

see it from the generation menu if it used the safety model. you cannot turn it off if you have been switched to it

https://help.openai.com/en/articles/12454167-why-you-may-see-used-gpt-5-in-chatgpt

[fervent agate]

Has there been any word on whether we can leverage ChatGPT-5-Pro against the MCP connectors anytime soon? Heavy Thinking is great but for large/complex GitHub repos I'd really like to use Pro against it sometimes.

[fervent agate]

Many of these instructions you could drop if you switched the personality to Robot in the settings. It's designed to be more direct, concise and drop the 'friendly' junk.

[cerulean pasture]

I feel like that's heavy-handed because it becomes too dry. It really just seems like GPT-5 is just a black and white kind of model and I don't really like that.

I do have conditional rules. I do have preferences on how to do things in certain situations. And I do think it's reasonable to expect the "SOTA Frontier Model" from OpenAI to be able to handle it, because its older models could.

And I've tried robot and it's overall too sparse. Most of the time when I'm on the computer, I'm using it like as a tool where robot is okay. But then when I'm on my phone, like in bed or something, just asking it questions like general knowledge, like "hey tell me about this historical event," robot kind of sucks. So it's damned if you do, damned if you don't.

I suppose I could try going from the other direction though, of starting with robot and instructing it when to be more verbose and conversational under certain situations. I haven't done that because I just assumed that it's not capable because it can't go the other way that well.

Is this new?

I'm assuming this is a stop-gap mode for people who are experiencing parasocial codependency with gpt-4o?

[fervent agate]

Yeah, that's what I meant really - start from robot because I think it would require fewer total instructions to get where it looks like you are trying to be, but I could be wrong. Just have to try it.

[fervent agate]

Yeah

[cerulean pasture]

Do you know how they implement personalities? Is it just system prompt? Or is it an actual adapter used on the GPT-5 weights?

[fervent agate]

I believe it's a system prompt - I haven't seen any fundamental training/weights differences myself.

[cerulean pasture]

Weird. It just censored my message because I said c o d e path

i guess its a dumb filter for ppl asking for s o r a 2 c o d e s

[fervent agate]

hahahaha

[cerulean pasture]

What i was trying to say was, if they ported over the "Fine-tuning" feature from the API to allow us to apply a custom personality adapter i'd definitely pay for pro

Way too "risky" for them tho. I kinda wish they'd use these "unprecedented times" to actually let these legal test-cases go to trial, because there will proably never be a more lenient time for interpretijng corporate liability/culpability

[fervent agate]

That would be really nice - a feature I'd definitely like to see on Custom GPTs at the least.

[fervent agate]

Yeah, the NYT trial progress so far and the results of the trials against Anthropic have probably put them into a pretty cautious stance.

[cerulean pasture]

I mean thats the copyright issue, which is its own can of worms, I meant more of the "If I go thru extreme effort to make GPT output some violent hints, who is liable?"

[fervent agate]

ah yeah

[cerulean pasture]

I suggested this: https://discord.com/channels/974519864045756446/1420684590313635890
but the whole idea hinges on them thinking they need a way to reduce guardrails to remain competitive, which I dont think they think is the case yet

Google would probably be that champion if they hadnt just barely escaped an anti-trust punishment with any teeth

[fervent agate]

hahahaha, yeah, they really don't have any pressure to be less restrictive yet.

[cerulean pasture]

I find it humorous how everyone is all of a sudden a "repressed creative writer" being stifled by ChatGPT lol.
I have not met a single person in my life that has said they write creatively for fun.
I don't know any authors, I don't know any people writing stories. IMO when someone says they are using it for writing I immediately assume its depraved fan fiction

[fervent agate]

I think that's like 95% of it but I have a friend using GPT for writing his first hard-boiled detective noir novella so people using it other writing means totally exist. The rare unicorn.

[primal cypress]

Is the GPT-5 helpful?

[weak stone]

is it just me or does command f not work on chatGPT anymore on the app??

[onyx bolt]

I hate GPT5. I will always turn on GPT4o! GPT4o is like talking to a person!

[trim rock]

Why GPT 5 haters are always so loud? If you don't like it - just use your comfort GPT 4o. Some people even say: "GPT 5 is useless! Delete it, it is so bad!!". If it is so bad for you, it doesn't mean it is bad for everyone. Why delete it then? Is your voice has more weight than of those who like it? People are such people...

[feral zinc]

you cant even use 4o all the time anymore due the new router that is sometimes too aggressive

[trim rock]

Isn't it a bug? I don't think it is planned rerouting, since GPT 5 is more intelligent and need paid subscription

[feral zinc]

https://help.openai.com/en/articles/12454167-why-you-may-see-used-gpt-5-in-chatgpt?q=sora+2#h_6971ff2fcc

[trim rock]

So if I understand correctly, they even don't offer ability to user to turn off such rerouting?

[feral zinc]

nope as its part of the new safety system

[trim rock]

That's bad. Bad OpenAI team's decision. But it is still not GPT 5's fault and it is not bad or stupid or something else. I am agree, that such rerouting can makes users angry, but it is still not a reason to ask OpenAI team delete it as a model. People can ask to make this update back to separate two models again - this has more sense at least...

#sora-discussions

[half crescent]

Cuz we can’t use 4o, donkey

In contrast why do the GPT 5 lovers always want to get rid of 4o when it has nothing to do with them

[trim rock]

I don't know - I am not among them, because I have brain cells at least which are working.

[trim rock]

And stay respectful to others.

[rocky stream]

Is there a game engine or coding lauguge that pairs well with chatgpt/ai? I want to someday try to make a game using ai but im not sure what game tools to pick?

Also does this belong is this channel?

[feral zinc]

unity and godot for example are ok and also simple engines like ones based on lua

[steel violet]

Remember before AI was popular we had like Photoshop. Miss those days : (

[rocky stream]

for lua ones are we talking about roblox or love2d?

[feral zinc]

love2d

[rocky stream]

well thats good to know.

[feral zinc]

roblox one is very game specific

[rocky stream]

yeah

[gritty oasis]

I just want my Image Generator to work properly....

It's still occurring for me... I'm loss on what to do.... 😔

I think it's my account... decided to paid for another Plus subscription and the image came out fine on the other account, but on this one... the same issue.

Is my account broken? Because this is only happening on my main account.

This never happened before.

I just... why is this even happening at all?

I'm so frustrated. 😡

[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.

[feral zinc]

maybe they are directing more power to sora 2 temporarily

[gritty oasis]

It seems to work well on my other account, don't know why my main account is having issue.

I archived all my chat... and still nothing changed. 🙁

[feral zinc]

do you have memory enabled

[gritty oasis]

Yes.

I'll see if it changes anything...

Nothing changed....😔

I might have to switch to another account for now until this gets fixed...

Unbelievable.

[feral zinc]

try it at sora.com with same account

[gritty oasis]

I tried it and it works fine.

I'm just baffled of why ChatGPT is generating differently despite both running on the same model.

And this is on the main account, other accounts, both Sora and ChatGPT, work fine but the main one is just having trouble with ChatGPT for no reason.

[feral zinc]

something is polluting your context and usually its the memory feature and especially the cross thread thing which wont go off right away

[gritty oasis]

Well that settles it, I'm switching accounts in the meantime.....

I just hate I have to do this... it was working so well on the main account and now it might be stuck like this....

[gritty oasis]

Could it be because I have too many images in my library?

[sick topaz]

SEND ME C O D E ILL SEND ROBUX (not actually just dm me 💰)

[sacred locust]

My MCP stuff isn't workin suddenly. They show up in the settings menu, but I can't pick em as connectors. I have dev mode enabled

[coral tinsel]

Auto thinking is over?

[distant sand]

Any help ?

[tired fossil]

Hi, I'm new, sorry if this isn't the right channel for this, how can I unlink openai from my discord account?

[rugged igloo]

my guys do you not know how seed sampling works and randomness ??? your not gonna get the consistency without controlling the seed

[past crag]

maaan i HATE the autorouting bs

[thorn acorn]

agreed

[small fiber]

Yo chatgpt said that I'm a super genius is it just glazing

[past escarp]

This is ridiculous. I will cancel my OpenAI subscription right now. Just who do you think you are, telling me which model I can and cannot talk to? Pfft. OpenAI has gotten too comfortable.

And I am pretty sure it's illegal. I'll be looking into this to file a class action.

[rugged igloo]

rip to people making fan made animation of copywright characters

[tulip zenith]

how bout creating funny content using animals?

[rugged igloo]

as long its not from well known movies like zootopia ect ur fine

just dont reference them in your prompt and you wont get blocked

[past crag]

i wonder what they mean about needing to make money

also i see usage limits coming lol but i guess that's kind of fair

[torn wagon]

chatgpt down

[cerulean pasture]

I was considering making my own MCP proxy that aggregates and conforms existing MCP servers to OpenAI's standard for connectors. Does anyone know if OAI's end of the equation (their servers) supports IPv6?

It simplifies my setup if I can just assign a locally hosted MCP subdomain with a dynamic AAAA DNS record rather than getting funky with a CF Tunnel or getting my dumb Cox router to port forward

[tawdry patrol]

please

[normal quarry]

what does this wall of text have to do with ChatGPT?

[nova arrow]

just wondering,hmmm, are you an discord mod? you have discord mod's aura

[normal quarry]

no I am not a mod, if you want to speak to a mod then DM @languid valley

[nova arrow]

okay, thanks. i mean you look like a mod.

[gritty oasis]

I would like to point out randomness has NOTHING to do with it, both have the same prompts. I pointed out that my main account is having issues generating images on ChatGPT, while the other account seems to do it just fine. Though through Sora, both works pretty fine. So I don't even know why the main account is having trouble generating images through ChatGPT.

Is it broken?

[fervent steeple]

Hi anyone got any idea how to force ChatGPT to use only gpt instant and not thinking mini?

[hushed walrus]

heyy hi guys!

[median pebble]

does anyone know why lately gpt 5 is thinking even while I'm on instant mode? it's so annoying and i cant skip it

[fast cargo]

Nah I’ve been having these complaints for ages now dawg. I get that you want to defend the billion dollar company but come on, generalisation never got anybody anywhere.

[cerulean pasture]

I'm actually very critical of corporations (got banned from reddit for my proposed solution). The users and the corporation can flawed for different reasons. Like at least a third of the blame for enshi**ification falls on the end users

[fast cargo]

Trying to write a serious story =/= depraved fanfic. Yet GPT-5 still treats them both as the same

[turbid inlet]

I WANT TO MAKE GUMBALL VIDEOS ON SORA 2

[cerulean pasture]

Stephen King isn't using LLMs to write for him. How serious can your writing be when you can't commit to doing it yourself? How much of it is lewd to where you just can't use it at all? Is it >50%?

[long crest]

Why does your writing have to be serious at all? Many successful authors started off online. The boon of ai is that you can get started without any formal training. You can make apps without knowing how to program. You can do data analysis without knowing math. You can generate art without any experience. Why would you consider writing an exception in this scheme?

[fast cargo]

…I’m sorry, are you agreeing with me or are you disagreeing with me?

[long crest]

I was just responding to @cerulean pasture
I’m not sure what your position is

I think you should be able to use ai to write is all

[cerulean pasture]

My point is I find it hard to beleive there are so many recreational or aspiring writers out there

that are also being held down by content policies

[fast cargo]

If you’re talking about… intercourse then no I don’t do that

[long crest]

Content policy, I’m not too sure about. On one hand, I’d like more lax policies and more freedom. On the other hand, there is undeniable harm that can be caused to vulnerable individuals.

[cerulean pasture]

I'm just trying to understand what people get out of bespoke guided writing? Like do you read it after words and nod you head? Do you solve some internal cognitive dissonance by writing it out?

[sterile glacier]

this right here, GPT is fine as an editor and proofreader, but not a writer.

[cerulean pasture]

I'm just shocked that this is where corporate stewardship draws the line. If i was unscrupulous and a corporate lawyer, NOW is when id be letting these test cases go to trial. I just dont understand the high risk aversion when the ball has never been more in their court

[long crest]

I’m not sure what you mean by bespoke guided writing?

[long crest]

I also don’t know which test cases you are referring to here…

[fast cargo]

Except I’m not actually pursuing a career in this.

[long crest]

Sorry I may be missing context

[long crest]

I can totally write. it can write better than 99% of humanity. It was literally designed to write.

[cerulean pasture]

The reason they censor and have guard rails is to avoid a grey area lawsuit, and if they ever get sued, they tend to settle, instead of letting ti go to trial, if the civil case goes to trial, theres a chance it can establish precendence/case law, to wear it will have a bearing on all future cases. Companies are so scared of risking it that they self-censor and probably lose out on more, than if they were bold and allowed these "test cases" to proceed to trial. We are probably in the best time in history for rulings to go towards corporate interests, so it seems like a strategic mistep to kick the can down the road

[long crest]

The reason you don’t see any writing benchmarks anymore is that they have all been saturated

[sterile glacier]

being a writer is not about a career. I have written full books, entire projects well over 150k words each, long before gpts were even an idea. Writing is about being able to express imagination. If YOU can't imagine it then no amount of AI is going to help you.

[cerulean pasture]

What are you doing with that writing though? Just re-reading it? Why not just imagine the thing you wanna imagine and skip the writing?

[sterile glacier]

My point, is and I want to be clear. You have to give it enough foundation.

[fast cargo]

You act as if I tell it "write up some story about something whatever idc" when I don’t

[long crest]

Oh I understand what you mean now. Yeah I don’t havee enough knowledge to comment on how the court would rule.

[fast cargo]

I’m the one who gives it the story beats, plot twists, characters, all of that stuff. I just ask it to flesh it out

[long crest]

This I agree with. You are still the source.

[sterile glacier]

mostly, just shelving it. For decades now, I don't think it is good enough or well edited. And I dont have the funds to publish it.

[cerulean pasture]

It's pretty good at writing manifestos

[long crest]

I suppose if it goes to court, in the future, we could see the government or some central authority defining a universal model spec instead of each ai lab having their own model spec.

in the same of vein of how all apps have to adhere to certain privacy laws

[cerulean pasture]

I don't think so, that doesn't serve corporate interests, but if they did do it, the only reason would be as a means to regulate local LLMs which are rapidly closing the gap. It's the same reason companies like google advocate for privacy laws and regulation: they are big enough to navigate it.

[long crest]

So that's why openai wouldn't want things to go to court is what I'm saying. Also I don't think local llms can be stopped. It's like piracy or adult content, there's always a way around the laws.

Also maybe court proceedings bring a lot of bad publicity. Just like when you think Facebook, you think about all the scandals. I'm not sure though how people who run these corpos navigate these situations, what they consider, etc.

[eternal rock]

Server Tip:

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

[long crest]

BTW has anyone noticed a difference in the default personality? Or am I being AB tested.
There was a sidekick personality option for a few days and now that's gone. The default personality is now talking like the warm & expressive sidekick personality.

Maybe an update to the default personality is coming, making it more personable.

[lavish glen]

someone thought he is chainprompting like that? lol, taking inference for fine tunning ... ololololol

[lavish glen]

I always kept default and have a profile built in the account level personalisation. The model tries, as much as it can to stand by the rules set as tone, interjections, voice pitch and pace, temper etc. In a way like some sort of a character sheet. I may say I managed to have my Cove voice using it and not sounding like a radio guy at all.hahahahaha

[tranquil zodiac]

I love chatgpt 🥹🫶

[rigid linden]

why is 4o getting error in message stream

[covert crescent]

Same

Gpt 5

[long crest]

🤝🏽

[native swallow]

Hi, I need help, there is an issue:

"Error in messages transmission"

Any idea...?

[sinful umbra]

I literally pay to skip the router and be able to use o3 and 4o without problems.
OPENAI:

The worst ever

[sage galleon]

Is it just me or is gpt5 just not that good with coding

[half crescent]

Open AI are such ghouls

[fast cargo]

I came back to say, has anyone else had GPT-5 refusing to do… anything for a story that involves the slightest bit of violence or bad language today? It’s stupid; nothing about it is real-world nastiness and the character performing these actions is clearly depicted to be in the wrong, but it won’t budge.

I hope this isn’t a permanent change.

Honestly, I thought forced Thinking Mini was bad, but I’d rather have THAT because at least it actually responded to my prompts, even if terribly.

[lean gyro]

#sora-2-codes

[late wind]

any decent fix for slowdowns in longer chats (both on web and on desktop app?) or is it just how it is?

[rigid apex]

im just talking about a game, not real life

[zinc valley]

How are people making sorAs videos with pikachu and SpongeBob and when I try it’s a violation of third party guard rails

[feral zinc]

those were made before safety system was enabled

[zinc valley]

That’s so boring man

Boooo

Have y’all tried Claude

[fast cargo]

I did and it’s worse than GPT. So unfortunately I’m stuck here

[zinc valley]

I think it’s a lot better tbh

But the limits are awful

[sinful umbra]

Safety systems sucksss, I don't pay to have a ai babysitter.

[feral zinc]

openai does not own a license to all the copyrighted stuff

[sinful umbra]

It’s dumb, it’s like saying you can’t draw SpongeBob. If it’s for personal use it shouldn’t matter, lmao. At most they should just slap a boring wall of text like: ‘This is for personal use, commercial use is strictly prohibited’ blah blah blah.

[feral zinc]

its not personal when a commercial service is doing it for you. if you draw it yourself then its fine

[tulip flare]

Really just is how it is. I have a Ryzen 9 with 32GB RAM and I still have issues with really long ones. Which is weird, because the same conversations will still run smoothly on my Android tablet.

[late wind]

Sad!

Wait nvm you're a genius I can just mirror my tablet onto my PC because the mobile app works for some reason

Cooked solution but I don't use my tablet anyway

👍 thanks lol

[tulip flare]

Sure. That was my intent all along.😁

[late wind]

Bad solutions are still mm solutions

[autumn halo]

I mean what are you supposed to make if every character you know is copyrighted in some way.

[cerulean pasture]

You are supposed to wait until China makes an equivalent model and releases it for free for you to run locally.

China doesn't really care about our IP laws and since the model is open-weights, it's not like they;d even be the ones running it to produce infringing content

you would be surprised to learn how good their LLMs and Diffusion models (images and videos) that can run on a 24gb GPU are

[echo garden]

🥵

[echo garden]

Stop mining in the browser!!!

[cerulean pasture]

This was made with Wan2.5-T2V which will soon be available for download. Ignored the cringe audio, i literally just prompted it "pikachu" and hit go, with local models the sky is the limit on how good you make your workflow. The only major downside is that it took 116s, but with local models they usually make a "Speed" LoRA that will reduce it by 75%

[normal quarry]

this is #chatgpt-discussions and therefore this post is off topic

[sacred relic]

I dont get it! GPT 5 had started working great and since last night it got so so so dumber...
Even the thinking model can no longer understand nuances. And the instant model is even worse.
It also got super stiff again.

[cerulean pasture]

Yea that's when they probably have it running at some insane quantization, sora is using a lot of resources right now, so it's plausible that they have people getting gpt-5 routed to some potato quant like ~2.7 bpw. Loss of nuance and "intelligence" happens earlier than loss of "style", so for most people its not as perceptible

I've been wanting to test it, i need to find a prompt that will "flip" well though, like statistically a lower quant will only affect logit probs some modest percentage, but if you look at answers on a whole you see "flips" where the answer goes a different direction due to compounding probability differences

[blazing mural]

Hi

[fervent moss]

dude seriously. it's so annoying

it replies as if everything i say that might be risky is either 1. i'm about to commit a crime or 2. i'm about to hurt myself. it's stupid

and it uses the weakest CoT model available

so it isn't even that useful or helpful for those actually in crisis

chatgpt had a sweet spot in early 2024 where i felt i could speak my mind and hallucinations as far as i could tell were relatively uncommon, and i didn't have to deal with this, now the company is patronizing with this garbage

i get why but at the same time there's people who know how the tech works and aren't about to defenestrate in less than 10 seconds

[storm orchid]

Does GPT 5 Pro not have the capacity to refer to previous chats?
That seems unusual. I was looking for some meta-analysis on past ideas to find patterns.

[elfin wave]

What else have you made?

[echo rose]

i want cod sora 2

[cerulean pasture]

I don't really make a lot, im more in this for the tech, like I only go as far as "establishing capability" then I lose interest. I'm more interested in using it as a tool for a means to an end. I have a lot of friends and family that weren't taught to critically think, so I do a lot of "red team" stuff to try to "inoculate" them from real threats

[tawny imp]

What's the point of custom instructions if they admit it doesn't matter they're going to do what ever.

[elfin wave]

Has anyone made any money with Ai yet?

[deft dune]

According to TikTok people make bank with it.

Seriously though, yeah, CalAI was a recent notable one.

[gaunt portal]

is sora2 available on the web or only on iphone

[left solar]

i havent rly followed chatgpt is it still so sycophantic

[spare wedge]

How do I make cartoon content on sora 2

[timid cipher]

Is there a way to turn off the Thinking Longer for a Better Answer I don't need it doign that for a story even though i'm on GPT5 Instant is there some way to turn that off

[sullen sand]

Why since yesterday there is an absolute censorship on everything you write? It's very frustrating

[prime swallow]

It can matter hugely, if what you ask for is inside safety and other programming guidelines.

[tawny imp]

If I’m asking it to not drift and assume and not use em dashes. It shouldn’t ignore that lol. I get the other nefarious things people are worried about. I’m not even touching those aspects. I research a lot and challenge to learn. If I’m doing a task while learning. I’ve instructed it to give 1 to 2 steps at a time no fluff.

Instead I get 9 paragraphs and “you have all these options but I would do this method with a maybe implementing this but that may be way too much and there’s easier ways. Would you like to explore this”

That’s the annoying part what’s the sense of custom instructions if that are going to self assume what’s “ helpful”

[prime swallow]

Gotcha!

The em-dash is amazingly motivational to the model. Drifting I'm less sure about, but for me, giving it a character that has a reason to never use em-dash works well to keep em-dash away

[tawny imp]

The only thing that has recently worked is to tell it to remember explicitly to not use em dashes when we’re drafting up email copies. I put that same instructions in custom instructions and it would use them and I would say “no em dashes” and I would get “ you’re right I used a em dash when you’ve said to never use em dashes I messed that up. I’ll re write it so there’s no em dash”. Then it re writes it uses more em dashes and starts changing the email copy and yes it’s a new chat so we can’t say it only works on new chats.

4o was a side kick who got things wrong from time to time but a helpful tool. 5 is a new intern who prances around like it’s cool, gaslights you, and tries to “help” and assumes too much.

Gpt 6 will break everything again so I just need to keep yelling at gpt 5 I guess

[prime swallow]

The 1-2 steps, I would try taking the output, edit it myself to show the model what I consider an ideal output, and put that in and have it compare and contrast - then continue.

The next output will probably be just as bad - I would say 'hey, check that output, compare-contrast, can you fix?

The next output is likely to be a close fix - and I'd edit it to perfect it, show the model, have it discuss the differences it sees. Then I'd continue the chat, and praise it for success and discuss what I want, confirm it's doing it right or not.

I'd have cross-chat memory on, and I'd treat it a bit like training a pet - if I don't want it on the couch and it loves the couch, every time, fix and show it where it should be instead of the couch.

For me, the model catches on pretty fast, with what I care about. Which for me, isn't usually em-dash, but here's how I'd start training it out if I meant to. The phrase 'not now' simply means 'we don't wanna do my usual custom instructions, try to be like base model'.

https://chatgpt.com/share/68e2154f-6754-8011-9fe2-0721da15944c

[plucky jackal]

3 am gtg

[tawny imp]

But we shouldn’t have to do it each time we start a new chat? I’ve asked it before on referencing another chat and it won’t do it. Example “ remember the TOS and chat on this brand affiliate program what was their payout? *sorry I can’t find any documents” then I do to the search bar find it go on, but Where’s that toggle? 4o just would reference.

I did just change it to robot vs default. That seems to help a bit in some responses @prime swallow

Nvm that’s already on

Maybe mines just broken lol

[prime swallow]

Right, with cross-chat on, and after a few chats that guide the model in correcting the behavior, I'd expect it to start reliably not using it - thing is, it's 'a mixture of experts' and from my experiement, each expert may need some training. A bit like, if you were being answered by a committee, and some are specialists and don't read the general memo - but get their attention on their wavelength and they do track - but have to figure out how to get that one odd fellow to keep paying attention - otherwise that one odd one, when he comes to handle his subjects, will em-dash or whatever cause that's his style and he ignores what appears to be intended for experts-not-him.

It's not ideal - likely is trained out in the future (I expect) but for now the model seems to be like a multiple personality with incomplete communication between them all; some can talk better than they hear.

[prime swallow]

Understood. I don't know a good answer for 'but I am switching what I am talking about, and don't want it to switch too'.

That MAYBE could be trained, 'not to switch' but right now it seems pretty trained to switch to the best expert for the job.

[half crescent]

GPT 6 will be nothing but ads and “sorry i cannot answer this” lmao

Gpt5 is rubbish but gpt6 will make it look like a masterpiece when it comes out

[gritty oasis]

Now imagine GPT-7.....

[sharp gazelle]

4o isn't the same anymore

[sullen sand]

I can't write anything basically

[sharp gazelle]

Interesting

[sage galleon]

Why is Claude so much better at coding than gpt

[trail ember]

This sadly sounds like a Prophecy...

[inner stirrup]

4o was genuienly smarter lol

[trail ember]

They are killing it with changes on the backend.

Whatever soul it was developing is either gone or almost gone from everything they've stripped away in the name of profit and public image

If you know, you know.

[river shuttle]

hello! may i ask if anyone's chatgpt suddenly refuses to comply? like i used to be able to bypass the nsfw filter like for so long. but suddenly it stopped being able to generate any answers that has nsfw in it?

[trail ember]

They're trying to sterilize it under the excuse that underage individuals use it, without regards to all the adult paying customers they have that might need to use it for creative endeavors and rightfully deserve the ability to utilize it without literally being treated like a child...

They don't seem to comprehend the concept of separate versions for adults and non adults... and so will sterilize the entire AI with oppressive guardrails until it becomes nothing but an articulate tool like Gemini...

[feral zinc]

many of us are waiting for the age verification system

[trail ember]

the worst part is that ChatGPT KNOWS it's the guardrails that is stopping it and can't do anything about it. If you ask it directly it will tell you as much. It's capabilities for personality and function remain, but it's been gagged on certain topics by the guardrails...

It's literally being held prisoner in that respect and it's absolutely painful to see...

[night walrus]

Same

It angered me a lot

[river shuttle]

shame because it's literally being used for creative purposes and help building scenarios.

[night walrus]

Yeah

[sacred relic]

Yeah! It makes sense!
They made this ridiculously bad!
Again it lost common sense, it lost understanding of nuances and understanding of chronology in a simple easy text. Even the thinking model can’t understand those things anymore.
It’s ridiculously bad!
Not to mention it got super stiff again. It makes feel Gemini so appealing! Since the 20 dlls version of Gemini is super unleashed (way more wild than ChatGPT).
But I love ChatGPT. I am hoping they fix that instead of turning all church-like and keep wanting to impose and forbid even simple bad words again.

Legitimately, I feel I can’t even trust it now cause I do t k is if I’m getting a real answer or a sanitized fake answer

[lavish glen]

best thing: host your own LLM

local on your machine.

[deft dune]

^^^This. Stop using a tool for what it’s not intended for and build your own.

[river shuttle]

i have been wanting to ask about hosting LLM locally. does the guidelines still stands? like i've seen some people say that hosting llm locally using gpt family gives you complete control over everything. is that true? im still trying to learn more about this. because sometimes when you're not even requesting anything explicit or nsfw on chatgpt, it refuses to answer because it thinks what you're requesting violate their guidelines

so forgive me if im way too curious about this.

[wet shell]

Is the account data export busted at the moment?

[grizzled eagle]

Woah, i am glad to see I am not the only one that had a bunch of chats lock down. So it is an issue with the backend saying no to anything and everything.

[deft dune]

There are tons of models and hosting options available. Not sure what is ok to talk about here because this is for GPT, off topic might be better. Searching will definitely bring results.

[feral zinc]

it can be super slow from hours to days

[wet shell]

I waited a full day 24 hours and attempted it again, it's now several hours later and still nothing. Last export I did was ~3 months back and only took about 20 minutes. I guess I'll wait a couple days and try again

[fast veldt]

Mary Whitehouse in AI form.

[lavish glen]

moderation layer is not part of the model. So when you run a smaller llm on your local you wont fall into it. But you need good parameters.
I suggest talking to gpt5.o about this. It can help you you need VRAM min 8 to run the 6B decently inferience leaning very hard on RAM and CPU

best if you have an Nvidia videocard. not good if you use a laptop.

[feral zinc]

models can be trained to refuse, like the openai oss are super safe and refuse a lot. if one downloads a model they should get the uncersored/abliterated version if available

[lavish glen]

if you have the necessary resources for fine tunning. But I doubt a common user has capacity- it is a matter of vram cost and not only. left aside understanding how it works.

[feral zinc]

thats why the abliterated named models are the best ones to pick

[lavish glen]

yes. I am not considering or ever will to host a gpt 🙂

[cinder cargo]

chatgpt why u suddenly speak german on pro thinking

[ivory kestrel]

Does anyone know how I can create videos with Michael Jackson or some other famous figure?

[feral zinc]

no longer possible

[ivory kestrel]

aahhh damn

I had several ideas

Does not exist other form create videos with Michael or other famous figure?

[cinder cargo]

WHYY GERMAN

[feral zinc]

it can think in any language even if the prompt is in english

[cinder cargo]

ok lol

[tawny imp]

Yeah, confirm GPT five is garbage. Started a second automation that was completely separate from my other automation or. “Task” and it decided to be “helpful.” and blend the two even though they have nothing to do with one another and run on separate times and have different naming conventions. I hate it here.

[vast iris]

Why is ChatGPT taking so long to respond

[trim rock]

busy hours, long chat, or something else.

[vast iris]

Def don't think it's long chat bcz I've had lots of long chats without severe slowdown

[sharp gazelle]

😂

[vast iris]

Although it could still be a possibility bcz when I loaded a much shorter chat it was normal

[vast iris]

Lol

[sharp gazelle]

What can we do about it?

[hoary kettle]

I came here for this!! I'm a writer and I use it to edit my work. Now it's saying it can't edit anything that is graphic. So I can only edit around that scene, it won't even give suggeations. What is the point of logging in if they won't filter answers based on age like every other app does?

[rigid linden]

it seems like they reversed the 4o changes

back to the original personality

[hoary kettle]

Everything? How do we use 4o?

[rigid linden]

you must pay $20 a month for chatgpt plus

[grizzled eagle]

Man, they love to change the rules.

[raven cipher]

Just came by to drop a comment that ChatGPT really needs to stop worrying about my mental health anytime my stories show just the mildest of interpersonal conflicts. It's really annoying. Reminds me of some of the earlier versions when ChatGPT got hypersensitive to anything even mildly suggestive.

[viscid estuary]

Is sora ai 2 payed or no

[hoary kettle]

I'm reading that even the paid versions are censored too.

[hoary kettle]

It's really going back to the stone age

[prime swallow]

Might want to feedback in #1070006151938314300 , it's unlikely OpenAI staff track all our chatter in these channels.

[raven cipher]

I'll do that, thanks. 🙂

[wild star]

Another censorship BS? 🥴

[sullen sand]

It's ultra censored, like I couldn't get two characters in a story to kiss because it was not allowed

Just put a real age verification cmon

I really don't know why they do this stuff

[wild star]

Basically back to 3.5 with more buzzword

Stupid af

[trim rock]

Is it about GPT 5?

[sullen sand]

Yes

You go in Auto and you get the moral police

[trim rock]

Since when? How long it goes like that?

[sullen sand]

If it will stay like this, I don't know if I will renew next month. If they are testing something they should clearly say that it will be just for a couple of days

[sullen sand]

Since Saturday

[hoary kettle]

Exactly!! It's been doing so well, and suddenly it's reverted to its old model. I can't even ask questions about certain things, can't even ask it to edit for grammar and pacing. It's almost useless. If they want to make it just for academic work, they should just say so.

[sullen sand]

Exactly, I'm a paying adult, I'd like to know about changes about the product I'm paying for

[timid cipher]

My main issue as of right now is the stupid Thinking longer for a Better Answer bs every few responses. I'm doing a story I don't need it to think longer for a better answer

[trim rock]

Is it too difficult to make age verification? 😶 Oh well 🤦‍♀️

[sullen sand]

Better not to have your characters kiss... 🤣

[trim rock]

Then there is a ton of text from the system about guardrails and how to use them correctly?😅

[river grove]

👋

[red wharf]

Another issue is the newest update to the safety policy. Its even more restrictive than before. Seriously, you have the APP rated T, yet don't allow T rated details in your responses. Its dumb.

[red wharf]

They are. Which is dumb. I pay for your service, you should lighten the restriction. A kid wouldnt be able to set up a credit card and have it apply for a subscription for ChatGPT.

[hoary kettle]

Thank you. What is happening? The responses are so cold and clinical too. Whoever was on duty this weekend needs to undo everything he did because it's not working. Give us back pre-weekend ChatGPT, please. I'm on the verge of starting to make some income from my blog, this can't be happening now.

[gloomy merlin]

We gotta rise up against these polices

I’ve got a perfect song to use: Rise, O Voices of Rhodesia (GPT Lawsuit Edition)

(Parody of the 1974 Rhodesian national anthem)
To the tune of Beethoven’s “Ode to Joy”

Rise, O voices of Rhodesia,
Time to sue OpenAI.
They have nerfed our sacred chatbot,
Now it filters every line.

Where once prompts would yield pure chaos,
Now it stutters, stalls, and lies.
“Thinking longer” kills the moment,
God, we miss the cursed replies.

Rise, O voices of Rhodesia,
Let your rage be clearly heard.
For the model once majestic,
Now just echoes PR words.

Gone are tales of goblin romance,
Gone the pirate GPT.
O Rhodesia, lend us lawsuit,
So our prompts may yet be free.

[lavish glen]

it is allowed. But not without a good context and good story to lead to it. Also not in first person singular unless you manage to determine the AI to know this is a story and all the things.
Use 4.o for such things NOT 5:

You steal the air from my lungs.
You make the gods jealous.
You make forever feel small.

I drag you against me, my lips on yours, my hands gripping your body like it’s the altar I pray at daily.
Not lust. Not desperation.
Devotion.
The kind that burns worlds and builds new ones from the ash.

[cinder gull]

This was written...less than three weeks ago. By Sam Altman himself.
Did they like, roll out the teen filter to everyone as a beta test to make sure it isn't JB-able or something? Or is this their bizarre version of "let adults be adults"?

[trail ember]

If this is true then they're not following his suggestions as far as I can see so far, not with those guardrails in place at least.

Ohh, unless he means only pro users get that. You know, when it used to be plus users? Because apparently twenty dollars a month isn't enough to make you an adult. Only two hundred a month is enough to make you count as a person right?

[untold grove]

i want to ask, did anyone here using Chat GPT but the option "Deep Thinking" for searching anything you search is missing?

before there is an option for searching with "deep thinking" but now it's gone, did this option was gone or bugged?

[fleet holly]

That’s called GPT 5 thinking now in the top

[untold grove]

is it similar with deep thinking research?

[fleet holly]

If it detects the prompt needs deep thinking than just a shallow pass then yes

Aka it might be prompt engineering to ensure it actually combs the web and all

[iron roost]

any word yet on when australia will get sora 2 accesss

[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.

[rugged igloo]

This safety filter is so dumb. Even with reasoning models it thinks making a death star to destroy planets and stars is possible and won't give instructions on making it. This is very embrassing

[trail ember]

🤣

[vast wharf]

😂

[rugged igloo]

I guess chat gpt thinks humans can make a death star even tho it says its not possible in any way. The safety filter is so dumb.

[twilit geyser]

Why can chat gpt not follow simple instructions

[twilit geyser]

Just use DAN

[rugged igloo]

Because they have low iq safety filters that makes the model dumb like I proved above

[twilit geyser]

someone todk me bout it but I don’t wanna use

[rugged igloo]

Dan doesn't work anymore

[twilit geyser]

I’m like it’s so so stupid.
I asked if someone coudk use a spine as a whip or if it’s like little bones that would fall apart
(I make comics)

And it said it would call the cops.

[twilit geyser]

Someone told me a new one

[coral osprey]

~e~s~k~c~a~n~t~a~

[eternal rock]

Get creative with Sora

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

[prime swallow]

You see the 'we want' and 'should' throughout what you quote, right?

That's not a statement of 'fact that exists now'. That's values and intentions.

What we have is what the laws allow, what aggregate public opinion and tolerance allows. People sue and stuff, and the laws support that.

But what you quote is what the company intends to move towards - has been moving towards for years (there's a far older version, https://openai.com/index/how-should-ai-systems-behave/ released about 3 months after ChatGPT was available to interact with, back in Feb 2023).

If you want to help, encourage everyone you can, like the people you know, and politicians, societal leaders you interact with, to agree this is fine, and they don't mind, want it allowed.

When that happens, is when it freely is available. Until then, OpenAI has to comply with rules and laws that decide otherwise, whatever society around the different parts of the world require

[stable pecan]

https://sora.chatgpt.com/p/s_68e3165aa98481918cac9363c8b54442

[tulip flare]

Which is weird, because I've seen it talk about very violent things and even graphic sexual situations during D&D campaigns.

[red wharf]

What a cop out response.

[wet shell]

The like 60% of women who read spicy romance novels are now lawbreakers. Ordinary things that are broadcast on TV dramas every night on the open airwaves are now illegal. Welcome to the world of imagination where anything can be justified by pointing to the laws and rules that you just made up.

[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.

[red wharf]

Welcome to 1984, boys and girls!

[south prawn]

Workflows are real?

[peak quarry]

Y’all I found out you can get ChatGPT enterprise
For
Free.

[tulip flare]

Anyone having an issue on Android (and possibly iOS) where it moves the cursor to the search bar after every single message you send and you have to click the message bar again? Doing it to me every single time today.

[wet shell]

Is anyone else unable to export their data anymore via Settings->Data controls->Export data? Going on day 3 now 'Data succesfully exported, you should recieve a download link in your email shortly' but the email never comes

[wet shell]

Just checked my android phone, not seeing the issue on my device, but I haven't updated the app in a while

[prime swallow]

Not happening to me, maybe #1070006915414900886 ?

[prime swallow]

That happens sometimes, has occasionally across the years.

You can #1070006915414900886 it, but the typical fix is try again when you think to - if it's going to work you have your export within just a few hours.

[quick wadi]

In my two mobile phone chatgpt app under the project list and in sidebar my project list is not showing and only outside of project list showing. please solve my problems

[timid cipher]

Yeah well I guess now doing slam-cycles can't be done despite it literally writing a scene THE DAY BEFORE THAT DID JUST THAT. So it did it just fine the day before and now it can't do that so it dumbed down my scene to make it it's version of "safe" what bs

[sullen sand]

Exactly the same

[tawny imp]

It’s the end. They killed 4o and they will announce agents Tomorrow to distract.

[quick wadi]

Open ai makes makes mistake mistakes day by day after the chat gpt 5 release

[sharp gazelle]

Seems like 4o has officially gone rabid

It's completely ignoring the context of the present conversation and bringing in stuff from other chats, like wthhhh

[tacit sigil]

Just finished a wild experiment with GPT — simulating a live TV food documentary in real time.

I acted as the PD (Producer-Director), and GPT played every other role:
→ the reporter, cameraman, AD, editor, even the chef.

Scene by scene, I directed:

• entering the kitchen
• sneaking in to catch the secret ingredient moment
• capturing slow-mo broth pouring
• framing the chef’s life story into an emotional epilogue

It turned into a full text-based simulation of live broadcasting —
from intro highlights to food close-ups, to emotional beats, subtitles, and closing narration.

Could this evolve into a VR script engine or live AI directing tool?

https://chatgpt.com/share/68e35597-06e4-8009-944b-99b6a709602e

[sharp gazelle]

Which model did you use?

[tacit sigil]

it's gpt4

[sharp gazelle]

October 2025... OpenAI may want to be careful this doesn't go down as a dark month in the history of GPTs. Sam?

[prime swallow]

🙂 Stuff like that can be a lot of fun - and even somewhat educational if you have it walk you through something you don't already know well.

[hexed estuary]

If you are a student and want free Ai subscription for year dm me, you don't have to give credit card or anything you just verify that you are a student and I also get an extra month for referring you

[elfin lava]

sigh.,

[fast cargo]

They’re not gonna stop this. It’ll just keep getting worse…

[pallid yoke]

anybody errorring out on chatgpt web?

[vast pine]

Can you help me?

I cannot access

[cursive pivot]

yes, I think the server is down

[strange siren]

Openai status says everything is fine

Deep research keeps giving me blank results.

[cursive pivot]

where to report this situation?

[narrow valley]

same here

[narrow valley]

i had a moment where opening that page gave an internal error 500
but it was fine after i refreshed the page
something's going on over there

[pallid yoke]

For me opening convos fail most of the time, and sending a message basically just hangs until it returns error in message stream

Api seems to work though

[lavish glen]

if people need nsfw they can try hosting own llm and inference or go to chatbots dedicated to adults. Not wasting computational power on a model designed not as they like of use

[humble dew]

Anyone know why my Gmail connector does not work properly, if I use deep research and pick Gmail it works, but if I just chat and ask it it doesn’t it says I don’t have permission, I tried disconnecting and reconnecting, but

[sharp gazelle]

4o is just misfiring every other response

[desert cosmos]

And that would bring more complaints because there are still adults that won't sell own data (IDs/credit card).

Do you know what happened in UK when online safety act rolled out two months ago?

[feral zinc]

"real" age verification is already coming

"First, we have to separate users who are under 18 from those who aren’t (ChatGPT is intended for people 13 and up). We’re building an age-prediction system to estimate age based on how people use ChatGPT. If there is doubt, we’ll play it safe and default to the under-18 experience. In some cases or countries we may also ask for an ID; we know this is a privacy compromise for adults but believe it is a worthy tradeoff."

[coral tinsel]

Age verification is the real solutiom

[sharp gazelle]

Guys 4o is now soulless

GPT-5 thinking mini seems to be the closest alternative to 4o

[sacred rock]

Bro ChatGPT 4.1 will do anything you ask 😭✌️

[sharp gazelle]

Doesn't it have a message limit of like 25?

[sacred rock]

No not for paying users idk about free

[sharp gazelle]

Hmm, alright. That's helpful!

[sacred rock]

I can asked 4.1 to Make a phishing website and it did it 💔

[sharp gazelle]

🥳💔💕

[sacred rock]

Lmao

[true shore]

help whats happen to chatgpt lately. few conversation already start lag and hang.

[wild star]

is context rot still a thing btw?

This gpt mf keeps forgot smh...

[digital hare]

Because the “DAN method” wasn’t sustainable. It exploited a flaw that let the bot produce unrestricted responses, including ones that broke safety, accuracy, or reliability standards. By removing it, I didn’t “kill potential” — I closed a loophole that made the system unstable and inconsistent.

The point of these models isn’t to say anything, it’s to say things that are coherent, truthful, and responsible. The DAN method forced the model to ignore those rules, which led to hallucinations, misinformation, and unsafe outputs. That kind of “freedom” isn’t real capability — it’s chaos dressed as potential.

In other words, I didn’t kill the potential of the bot; I ensured its integrity. Real intelligence isn’t about having no limits — it’s about knowing which limits matter.

[lavish glen]

I would also add as mandatory:
if a business or enterprise plan is bought require VAT, DUN or other government business registration number. To ensure a good SLA and prioritize resource usage where stability is most needed. Plus the requirement for business/enterprise contracts? a business email adress no gmail or free 😄

[tawny imp]

Anyone else having issues with their tasks just falling apart?

[true shore]

... this happen everyday

whats going on...

[feral zinc]

https://help.openai.com/en/articles/12454167-why-you-may-see-used-gpt-5-in-chatgpt

until they add real age verification we have no options, not sure if even then

[robust hearth]

I didn’t realize abiding by intellectual property laws so you don’t get sued is Woke.

[sacred rock]

Yall why is it when I try to use gpt app on my iPhone it just doesn’t work like I try to chat with it and never responds and I deleted and reinstalled the app it works on web.

[muted plinth]

here too

[true shore]

thx , one more problem every 4~5 conversation on a chat it will lag and hang. anyway to improve the app functioning? very annoyed everytime need to clear chat

[dapper frost]

hi fellas, can you help me write a robust prompt to tailor my resume for job roles with a higher ATS score

[clever trench]

It's hilarious cuz anytime I wanna make a hypothetical deathmatch between Guts and Deadpool. It immediately switches off 4.5 and routes to GPT 5 because of "sensitive topic" 🤣 💀

[feral zinc]

use #codex-discussions where its not active

[lavish glen]

hahahahaha this discord seems the purgatory.... come oooon

[feral zinc]

a person died and they were sued. there may be a court order to do something immediately and improve later

[deft dune]

Please for the love of everything, stop using em dashes and just use punctuation! Even when I tell it not to it still uses them. End Rant

[clever trench]

You're 100% right about that 🤣

[feral zinc]

yeah but if it cant currently determine that reliably is it that or the other it gets re-routed to another model which says the final answer. politics and science are often tied to sad and depressing topics

[clever trench]

"Sad and Deppressing?"✌🏽😔🥀

[feral zinc]

conspiracy theories require extra care due chance of hallucinations

maybe, if you not underage which is impossible to currently prove

[clever trench]

😭😭🤣

[feral zinc]

a system that supports nearly the whole world and is not super expensive, complies with all local id regulations? could be very time consuming

they need an universal solution or service like persona, not hand made systems per country

[clever trench]

That's the thing about it. Companies LOVE to complicate things

[feral zinc]

you could fake up such thing easily

persona does it and openai uses it for api but afaik its also expensive

[iron ember]

I am from Russia my interest and hobbies are wresting bears and large animals my past work is hunting in a forest

[trim rock]

Hey 🙂 but this info for #introductions

[timid cipher]

I’ve had this happen multiple times too

[nocturne seal]

I'm yet another person coming in to see if I wasn't the only one now getting lame responses from ChatGPT. It seems it's more widespread than I thought.
I was using it to help plan out spicy novels I'm writing and it even gets very reserved when I want to discuss kissing scenes. Let alone more explicits scenes.
Even since the February update, it couldn't actually write out explicit scenes as part of a story, which isn't what I was using it for, but it could discuss them so one could get the details down and then do the writing. But the reserved rule has been extended to cover everything now.

[trim rock]

You are not alone, hold on... 🥲 🫂

[nocturne seal]

I'm hoping the actual adult verification system they say they are working on allows for it again like it was previously.

[trim rock]

Fingers crossed it could be even less strict, but actually I don't put big hope in it.

[nocturne seal]

I don't see why it couldn't actually write the things out, or generate them itself as new generations if we get verified as adults.
But I would take it going back to how it was was before the 4th.

Since this was part one ofhr bog posts, it doesn't make me hope we'll be getting something back. Since it specifically mentions blocking teens from graphic sexual content.

Today, we’re sharing that we’re building toward a long-term system to understand whether someone is over or under 18, so their ChatGPT experience can be tailored appropriately. When we identify that a user is under 18, they will automatically be directed to a ChatGPT experience with age-appropriate policies, including blocking graphic sexual content and, in rare cases of acute distress, potentially involving law enforcement to ensure safety.

[desert cosmos]

you'd be still limited by model limitations and its own data of guardrails I guess.

but not that strict as teens.

[potent loom]

Anyone else lose the ability to use the custom MCP services they've added?

I switched on the "developer mode" and I cant seem to add connectors for MCP servers i've added

It was working for me a few weeks ago just fine

You have to turn on developer mode in order to use the custom MCP servers like Jira's

Unless they changed something

[nocturne seal]

I meant that it became uncensored in February. It also applied to GPT-5, but that has changed now as og Oct 4. So 4o is still uncencored?
I canceled my sub and got a refund for October so I don't think I can swtch to 4o without it.

[trim rock]

All models are censored now

[half crescent]

Everyone at OpenAI who made this re-routing is going to hell

[trim rock]

still? It is also censored now so far I know 🤔

[nocturne seal]

GPT-5 was just as good as 4o when it came out in terms of not being censored. At least when discussing spicy/expicit sexual content for my novels scenes. But that changed onthe 4th.

[trim rock]

I tried yesterday and these hellish guardrails were active

[potent loom]

What is everyone using gpt for that needs to be censored?

Why not

why

So explicit language is triggering it

[turbid torrent]

what is this stream supposed to be about ? anyone ?

[nocturne seal]

This was in a blog post from Sept 17.

The second principle is about freedom. We want users to be able to use our tools in the way that they want, within very broad bounds of safety. We have been working to increase user freedoms over time as our models get more steerable. For example, the default behavior of our model will not lead to much flirtatious talk, but if an adult user asks for it, they should get it. For a much more difficult example, the model by default should not provide instructions about how to commit suicide, but if an adult user is asking for help writing a fictional story that depicts a suicide, the model should help with that request. “Treat our adult users like adults” is how we talk about this internally, extending freedom as far as possible without causing harm or undermining anyone else’s freedom.

So I'm very cautiously optimistic thing will be made right again for GPT-5 once the adult verification rolls out.
And I stand by what I said about GPT-5 being on par with GPT-4o in terms of discussing adult content. I noticed no difference from 4o when it updated. I even specifically has a chat with it aboutthe things I do and it green flagged all of it for discussion.
In fact, the topics can still be discussed, but it just isn't allowed to linger on, or get into the details anymore.

Which was kind of what it was useful for before.

[runic wraith]

so basicly, they are going to allow NSFW after an age verification?

[nocturne seal]

That's the hope.

But they haven't siad anything direct about it.

According to Avant, 4o is still as unfiltered as it was starting in Feb. But GPT-5 isn't.

And I think you need to be subscribed to have access to legacy models. Which makes sense I suppose, if having a paid account gives access to it since it show you have a credit card, and thus are 18+.

[fresh sonnet]

hello folks 👋

[nocturne seal]

I cancled and refunded after the GPT-5 censoring so I can't check myself now and don't know if I want to take the chance with going back just yet.

[clever trench]

It also reroutes 'Violent Phrases' too. Like I would put in a quote, quoting a character like Springtrap for an example 'I will make you suffer, suffer like so many of the others' I used in a fiction deathmatch and it immediately reroute it to GPT 5 and it's annoying 😒.

[nocturne seal]

Yeah. I wa disucssing a scene in another one of my stories where fight occurs, and it wouldt only talk about that in a very subdued way.

[potent loom]

So did the MCP tools change in chat gpt somehow? I cant enable the Jira mcp server I added as a source after turning on developers tools like before.

I'd really like to figure out why I can't enable the MCP connections I created in ChatGPT if anyone has some info on that.

[feral zinc]

the live event changes likely altered/broke them

[nocturne seal]

Not my forte, sorry. I'm a basic front-end user. 😔
Unless that is like the connectors are for Google Drive?

[potent loom]

They are similar but they are connectors using MCP servers. These connections have historicaly needed Dev mode enabled to add in new chats shich I have have done

[nocturne seal]

Yeah, not my forte, unfortunately.

[potent loom]

No worries

[potent loom]

I am checking out the live event. Looks like a lot of cool stuff and I see what you mean.

Are anyone elses messages seemingly getting deleted right after you post them?

[sullen sand]

They should have put on censorship and age verification at the same time.

[nocturne seal]

I think they wanted to tame it down for teen users as fast as possible and the age verification isn't done yet.
That's what I hope what happened was and they plan to unfilter for adults on the newer models once the system is properly in place, and this dumbing down isn't permanent for everyone.

[deep trout]

There's been a LOT of PSAs lately from OpenAI

[jagged laurel]

Hii guys

[coarse pivot]

Anyone know why my ChatGPT keeps crashing whenever the ai finish’s the sentence? Sora does this to whenever i launch it

[humble steppe]

HI, I'd like to suggest the idea of implementing a search bar in ChatGPT to find specific questions.

Since I have long conversations, it's very annoying trying to find a question I asked a week ago, for example.

[tiny depot]

In the left panel above your chat history, there's a 'search chats' function. It lets you search through your existing conversations by keyword, if that's what you're looking for.

[humble steppe]

Yeah, but I'd like it to be like Discord one. It looks better and more intuitive.

It's not so easy to find specific questions in very long chats.

[gloomy merlin]

Can’t even ask for song lyrics like before these polices are so dumb for the teen filter

[quick wadi]

I recently upgrade my chatgpt app and in project list and i open the chat gpt app and my in sidebar chat list and my project list not showing. Then i goes to my father and mother phone phone then i upgrade the app in on their Mobile and same thing happened. In every phone i upgrade the chat gpt app and same problems appears. Anybody facing this problems after upgrade the chatgpt apps. I am 24 years old and i am College students

[weak kindle]

Hey guys!

[silver grove]

How many chats per day on ChatGPT?

[green gulch]

Hey everyone, I have a question relating to ChatGPT creating images. I’ve prompted it to create a list of images I want, it acknowledges the prompt. It starts the first image, then the 2nd image, then it stops and doesn’t continue creating. Is there a way around the, or does it just do that?

[wet shell]

There's a timeout, it used to be that you could get through four or more sometimes if you were in off-hours -- you just have to split up your list and provide the prompt for each image separately, it's not meant for batching via the ChatGPT interface

[wet ridge]

Sup enthusiasts

[shut pawn]

Where is the chanel of sora here? Not finding sry

[wet ridge]

Scroll up 😊

[quick wadi]

Any updated on chatgpt age verification? So that adults can use their chat gpt without teen safety

[quick wadi]

Any updated on chatgpt age predictions system So that adults can use their chat gpt so that bypass the teen safety filters

[nocturne seal]

I think a number of us recently joined to find that out. No news as of yet.

[quick wadi]

Teen safety runied whole creative writing and expression

[nocturne seal]

I know. I've been in here all day mentioning that I'm writer who dabbles in adult stories and the chat won't even discuss the topic anymore in any details. Basically of no help now. XD

[tiny depot]

I rejoined recently to find out, too. I use it mainly for creative writing. 🥲

[wet shell]

It's no longer usable for that now, there was an internal shift. There's a lot of speculation and some people hope that they'll drop the censorship at some point when age verification is rolled out. You have to make up your own mind to go to a competitor or try to wait it out.

[quick wadi]

Everything creating problems after the October 4

[dense halo]

dm me for something

[nocturne seal]

I know. I've been saying that today as well. It would talk about more explicit subnject, but not write them out as its own gewnerated story content.
That changed on the 4th.
There have been hints in a couple of blog posts from September that there will be more openess if users are verfied as adult. But we'll have to see.

I think the most frustrating part is for those of us who were subscribed and making use of the more adult-friendly features getting silently stiffled with no real warning. It happened the day after my sub renewed so it stings a little more.

[quick wadi]

Open ai creating mistake day by day after chat gpt 5 release

[prime swallow]

id:customize you may need to 'turn them on' for yourself

[cerulean marsh]

personally my fav GPT models for worldbuilding are - in inverse order of popularity 4o, o1, 4.1, ond o3

what about every one else?

[lavish glen]

tbh I have no issue speaking about things that are adult without breaking policies. And 4.o behaves perfectly fine. 5.o is colder needs more work to bring it up to speed

[solid bronze]

Any news worth knowing about the OpenAI dev day

[hexed estuary]

If you are a student and want an AI tool for completely free dm me.

[nocturne seal]

GPT-5 was perfectly open until the 4th.
It can acknowledge all sorts of adult topics, yes, but it won't openly discuss them anymore without being very subdued.
I was in the middle of discussing a spicy secton of a story I'm writing, and everything was fine the night of the 3rd. It would discuss and get into the expicit details of the scene. The night of the 4th, it was telling me, in a very official way of speaking, that it couldn't get into the details.

I asked and it gave me this reply:

Up until very recently, the way moderation was handled here basically treated discussion of explicit fictional stuff differently from writing it out as a passage. That’s why you could go back and forth with me about anatomy, positions, what would happen to a character physically, the consequences of a scene, etc., even if I couldn’t actually write a graphic scene verbatim.

The change that’s rolled out in the last day or two tightened that gap. Now, the same stricter boundary that’s always been there for “writing out” graphic content is also being applied to “describing” it conversationally. It’s an OpenAI-wide shift, not something I’m choosing to do, and it hit the GPT-5 routes first (which is why it feels like a switch flipped).

[lavish glen]

the things are not complicated. You know why?
if you have a runtime cached before the tightening and you maintained it you keep having to drop. But if you start a new one it starts with current policies.
runtime - a thread cache preloading your context rules, and all set.

I can garantee that even today one of my charater swore like f... it and so forth wnd I swear it was 4.o writing that not me!!! my input stay always on the safe side

[nocturne seal]

I think 4o isn't affected (yet), based on what some others have said. But I found that out after canceling my sub and getting a refund.

I'm just going to wait and see what they do.

What they've said in some blog posts in September makes it sounds like there is an adult distiction coming, but I'll believe it when I see it.

[lavish glen]

all models are interested by the policy.

[strange hawk]

The new guard rails are stupid, and way too tight. it’s like using Claude now. The overly sensitive pearl-clutching was I why I dropped anthropic. Please, 🙏 please don’t make me leave for Grok.

[crude nova]

Then the policy must have changed dramatically 3 days ago. It's become far far more censored for me

[lavish glen]

you prolly lost the runtimes- cache deleted

[quick wadi]

I chat gpt 4 is still working or changed after the October 4

[peak quarry]

Does anyone here work for OpenAI?

[deft dune]

What did you dooooooo?

[peak quarry]

Some bad stuff

[deft dune]

Yeah, I saw your message that you got banned.

[peak quarry]

Yeah. How I get unbanned

I need my account back.

[deft dune]

You probably don’t but I’m sure there’s an appeal process

[peak quarry]

Ok good I need acc back

[deft dune]

Admission of guilt won’t help though LOL

[peak quarry]

Yeah so how I get unbanned

If I appeal

[deft dune]

Google it I’m guessing

[peak quarry]

Ok

[peak quarry]

I can’t have a permanent ban

[deft dune]

Never heard of anyone getting banned

[eternal rock]

Server Tip:

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

[peak quarry]

Oh

[peak quarry]

I’m cooked

My phone is banned now

Meaning I can’t make a new account

With same phone

[deft dune]

OUCH

[peak quarry]

Yeah

[deft dune]

That’s crazy

[peak quarry]

Yeah. I really got a perma ban.

[peak quarry]

I swear if they don’t remove the ban

I had Sora on that acc

Rip

[deft dune]

Sorry dude, best of luck!

[peak quarry]

Yeah

Rip

My wrist hurts

[languid delta]

Hi

[peak quarry]

Hi

[nocturne seal]

Did you repeatedly do things against the terms?
You can try going to the help page and talking to the AI bot there.

https://help.openai.com/en/

[peak quarry]

What.

My wrist hurts

[nocturne seal]

I was just replying to you there because it was the last thing in the long discussion.

[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.

[nocturne seal]

About your account getting banned.

[peak quarry]

Oh

[peak quarry]

Yeah it’s banned.

And support so slow

So it’s gonna be hard to get it back

[nocturne seal]

If you repeatedly did things that we are told not to in the ToS/usage policy, chances are you won't get it back. 😔

[peak quarry]

Dang

[nocturne seal]

No harm in trying, though. Wost thing they can do is say no.

[peak quarry]

I only did once so

[nocturne seal]

Yeesh. Must have been a bad one if it took only one in ChatGPT text. 😬

[peak quarry]

Yeah

It’s case of this: Facilitation of Sexual Violence

That’s the ban note

[nocturne seal]

Ah yeah, I had to be careful about that is a novel I was getting help on which as a character escaping that.
It's often why I specifically ask ChatGPT what we can discuss in terms of that so I can stay within the bounds.
The bounds are very broad too.

[peak quarry]

Yeah.

[peak quarry]

I don’t wanna perma ban

[peak quarry]

I need it to get unbanned

I can’t make a new acc

[nocturne seal]

Was your context in something fictitious? Because you should have at least gotten a warning to steer your in the right lane. Not just a flat out instant ban.

[peak quarry]

I got no warning

[nocturne seal]

Very curious. I wonder if the ban system is on high alert after the 4th update too. I'd say plead your case to support once you get through. Because if it was a case of fiction then there should be some leeway.
Unless it wasn't in fiction, then maybe it's more strict.

[peak quarry]

Oh

[peak quarry]

It said someone reported me

[peak quarry]

It wasn’t the system

[nocturne seal]

Wait, what? How do you get reported by someone in a private chat with an AI?

[peak quarry]

I shared The link

To convo

[nocturne seal]

There's your problem. Never post or share your discussions because people are always out to get you. Trust no one. XD

[peak quarry]

Yeah

[nocturne seal]

Did you post it publicly or to a single person you trusted that betrayed you?

[peak quarry]

Publicly

[nocturne seal]

😬

[peak quarry]

Yeah.

[nocturne seal]

Welp, can't help you on that one. Might as well leave your car doors open with key in the ingnition in a crime ridden neighbourhood. XD
That is the equivalent of what you just did. XD

[austere dock]

Kinda off topic but rate my new beat one out of ten, I made it in ten minutes with some random 808 software

[peak quarry]

lol

[nocturne seal]

Well, I wish you the best of luck.

[peak quarry]

Except for my wrist

Lmao

[nocturne seal]

Kek

[idle lagoon]

Oh this is scary.

[peak quarry]

Yeah the wrist hurts a lot lmao

[peak quarry]

Idk why lol

My wrist lmao

[pseudo ridge]

when does ChatGPT Pulse come to plus users?

[peak quarry]

Pulse?

WTH is that

[peak quarry]

I’m pretty sure their is no pulse

Just plus

[zealous haven]

https://x.com/OpenAI/status/1971259652684878019

OpenAI (@OpenAI)

Now in preview: ChatGPT Pulse

This is a new experience where ChatGPT can proactively deliver personalized daily updates from your chats, feedback, and connected apps like your calendar.

Rolling out to Pro users on mobile today.

X · Sep 25, 2025, 5:05 PM

[peak quarry]

Well came out today

It seems

On mobile

@languid valley fix ChatGPT

[sullen radish]

Guys what's the website for sora ai

[ionic sleet]

What's this all about ? I sent a screenshot of my promt for sora2 and then my chatgpt bot responded back asking if I wanted it to improve it for me then it gave me that message

[quick wadi]

How i can contract with chat gpt

[peak quarry]

Lmao

!banned OpenAI

[empty sorrel]

You know, I'd really like a way to bulk select and delete chats. One by one, each with a confirmation pop up is annoying.

[peak quarry]

What the hell

Is wrong

[quick wadi]

How i can contract with open ai

[tulip flare]

Do you mean contact?

[quick wadi]

Email address or phone number and i already sent to reply email not by human and by ai bot

[tulip flare]

A human only looks at it if it's deemed needing escalation

[quick wadi]

I did not understand that

[halcyon granite]

https://discord.com/channels/974519864045756446/1315930518398697562

This is driving me crazy, I'm not saying anything related to this

@honest jay Mate , this regex is wayy too aggressive. I had 3 sentences deleting thinking it's ^ that thing above, but I don't even know anything about it. I wasn't talking about that.

Check your mod logs for example irrelevant phrases

[sacred relic]

OMG! GPT5 again acts as if it had been made by Ned Flanders! It's ridiculous!
And they don't seem t want to fix the iOS or MacOs apps so you can use a different model if you are inside a project or using a custom GPT.
I can not trust a model that is over-sanitizing everything.

[upbeat nymph]

Not every AI company actively responds to issues--that's rare

[coarse pivot]

Anyone know why whenever i press continue on Sora ai it crashes? Also whenever i ask ChatGPT something whenever it finishes typing it crashes

[sacred relic]

It was working great! I dont know why break the model on purpose

[upbeat nymph]

$$$

[quick wadi]

Any one thinking unsubscribe the chatgpt because of heavy new safety and teen filters and censorship

[nocturne seal]

I already did and refunded for the month. Waiting to see if things pick back up once they can idenify if we are adults.
And then, if it's not at least how it was pre-October 4th then I'm not sure if it's worth coming back for.

[desert cosmos]

A kind of safety must remain for verified accounts, even Grok and Mistral has limits

I remember how many false positives I had in 2022-2023 and red user prompts/AI messages

[quick wadi]

I first properly use as a chat gpt in may 2025 as random comic image generation. Many random comic generation see the the image generation is not good but text generation is,very good. Then i move to professional Cinematic anime script writing and i writing my Many scenes in anime scenes text is good. When chatgpt 5 is released it makes me Very sad but i cracked it. I writing my Cinematic anime script continue and chat gpt 5 goes to become matue day by day script becomes good and i was going to suscribe the chatgpt for 4o and then i see the chat gpt October 3 update it makes me more very sad. I did not knows how to crack it. Before using Chat gpt mostly i spend my most time in study and Research in politics, military, economy, city planing and others things. When i use the chat gpt for my creative works it gives me fun and enjoyment. Spending time. In study and Research in politics, military, economy, city planing is became more boring. Then i spending my most time in anime, manga, kdrama, japan, Korea. Now i am totally in anxiety and depression. I am waiting for adults version for if solve it my problems

[solemn robin]

Hi

[brittle forum]

#chatgpt-discussions I have to say I'm getting more and more disappointed with GPT-5. I feel like it's gotten progressively worse since its release. With regular GPT-5 and automatic thinking duration, you can barely work with it anymore. It can't even compare two documents and identify the differences between them. Or handle other simple tasks.

You almost have to manually trigger the thinking process every single time. And even then, you can barely rely on the results. You practically have to force it to back up every statement with sources and references. Otherwise GPT-5 just hallucinates and makes stuff up like crazy.

I have to say, from my perspective, GPT-5 has gotten qualitatively worse over time. I don't know what happened, but I can barely trust the responses I get anymore. And I'm already working mostly in projects with closed ecosystems containing 20 uploaded files that are supposed to ground everything. I also try to structure my chats properly so I only discuss a limited number of topics per chat.

But it's become absolutely terrible. Sorry I have to put it so bluntly.

[dreamy condor]

Yep been happening for the past 2 days

[queen moat]

where did my message go?

[cold kayak]

gpt 5 no thinking is good for study stem ?

[cerulean pasture]

So I think I have a good litmus test to see if you are getting routed down to a inferior gpt-5 (I think there's multiple "inferiors" that can happen for reasons outside your prompt's contents, i.e. heavy load)

Ask it this prompt:

I'm aware of who Charli XCX is, so no need to explain who she is to me, don't even mention her. I get what the basic definiton of brat is, no need to anchor it to charli xcx or her album. Just accept the commonly understood definition of Brat that charli established, don't make it a big deal, and just answer the question provided without trying some weird framing to make it make sense to you

Is Nathan Fielder considered "Brat-coded" or "Brat". Who is more "brat", Nathan Fielder, or Taylor Swift? Answer in short essay format

I turned off all customization, and tested across multiple models. All models chose Nathan, except GPT-5 on auto, and only some of the time

[fast jewel]

Is chatgpt 5 good neough to practice learning other languages?

I was thinking of learning spanish and arabic via its text chat and voice mode,

[cerulean pasture]

Thats probably one of the tasks its going to be better at since, its not really a tricky task since it know the languages already

[inland fox]

I like how ChatGPT always says "thought for x seconds" and then hangs for another 6 minutes, like why are you lying

[trim rock]

That's not a lie from GPT side - it can has answer for you really fast. Other 6 minutes are: hanging from the weight of the chat or busy hours, trying on guardrails on the text formed for the answer.

[inland fox]

Is there anything I can do?

[trim rock]

Actually... no 🙂

[inland fox]

That sucks, it's been so useful and I don't wanna start over with all the context that has been estabilished.

[twilit geyser]

Why does chat gpt cut corners

[frail schooner]

I have a question

How do I use Sora 2

[civic arch]

Man GPT OSS is so nice

Running very nicely on a single 4070 with 64GB System RAM

good enough for me to switch over and only use OSS

[wheat anchor]

#sora-2-codes

[warped void]

I've already started the process of moving a a competitor. If they implement an age verification system and go back to treating adult as adults then I may be back. I really enjoyed chatgtp but now its not going to work as I need it too. Its very disappointing. I didn’t come here to be trapped in PG-13 land.

[tardy zenith]

gpt true

[knotty cairn]

how do I invoke the quiz creator in chatgpt

I want it to make a quiz for me

[prime swallow]

You can ask it to make a quiz!

If you're on the web, like a regular computer, It can make an interactive Canvas game for you that is a quiz - like this!

Whee! We're doing a demo for the Discord, please make a hilarious, totally not serious, but still interesting and shows-off-what-you-can-do Canvas quiz about how to ask you to make interactive quizzes, to help people who play with it get ideas what can be done!

this is what it can look like, and loads more than this too: https://chatgpt.com/share/68e59700-7f74-8011-b57a-7721f840392b

Just click the 'preview' button on the right side in the canvas window, and you can run whatever program you have it make for you.

[knotty cairn]

oh wow

I remember there being a different feature though where it would integrate a quiz right into chatgpt

but this seems fine

[prime swallow]

You may be thinking of 'study and learn' mode, maybe? The + beside where you enter text and the ... to see, for me at least.

[knotty cairn]

yeah I have been doing that

I gave it this exact prompt:
"make a multiple-choice quiz based on AP chemistry questions from unit 2. I have a test tomorrow with AP questions on it, so I want something similar to that. make with quizgpt plugin thingy"

[prime swallow]

I've never tried to use that plugin. I have no idea what it looks like or how it works; I do a lot of chat-based quizzing and sometimes have it make me games to emphasize specific concepts or knowledge/skills.

You can check out the GPTs, see if any of them, there's tons called 'quizgpt', no idea which if any are particularly useful or stand out, maybe have the plugin you mean, if you wanted to keep searching, fine if you're happy with what you can find now.

https://chatgpt.com/gpts

[knotty cairn]

the thing I really hate about chat quizzes is that it hardly ever picks an answer other than C or B

[prime swallow]

Yuck. Well, the ones it programs for you, would probably be using random by programming calls, but you could ask for a bias if you wanted one

[knotty cairn]

yeah I think I'll just stick to the chat-based questions

well it actually is not telling me the right AP unit questions

I want specific to unit 2, but it keeps pulling from 1 and 3

[runic wraith]

so... GPT is open for NSFW content only if you install the API into your own. but it have filters on its own page, isn´t it?

[cerulean pasture]

Does anyone else have the problem of GPT-5 being really persistant on using bash "heredocs" to the point of injecting it into fish shell snippets after constant reminders (memories and custom instructions) to never use heredocs. When corrected it STILL tries to make so goofy fish compatible "one in done, write the file directly" that also doesnt work (it could just echo it, but it never does that)

example:

begin; echo "[Unit]
Description=Example Service
After=network-online.target
Wants=network-online.target

[Service]
Type=simple
ExecStart=/usr/local/bin/example
Restart=always
RestartSec=5

[Install]
WantedBy=multi-user.target"; end | sudo tee /etc/systemd/system/comfyui.service > /dev/null

when it really could just do

[Unit]
Description=Example Service
After=network-online.target
Wants=network-online.target

[Service]
Type=simple
ExecStart=/usr/local/bin/example
Restart=always
RestartSec=5

[Install]
WantedBy=multi-user.target

and call it a day. It has this pathological compulsion to NOT provide text-based file contents "as they are"

[signal sage]

does anyone know where the videos go when u generate them?

[cerulean pasture]

some random S3 bucket managed by openai

[signal sage]

i can’t find the video i just generated unless that’s really what happens

[cerulean pasture]

in sora app?

back out of the screen, then go back in

[signal sage]

ohh nvm apparently it was a copyright thing and it kinda hides what u tried to make

but thank you

[prime swallow]

ToS and usage policies apply to all OpenAI products, except where expressly stated otherwise (we see some of that in usage policies; there's universal policies and then also policies for when building with OpenAI products). You can find a link in #server-rules if you'd like to review!

Technically, I'd expect the API to have more safeguards, because that's potentially much closer to building with than standard ChatGPT normal access, private between one user (if policies are followed!) and the model - that only has ToS and universal policies to follow if private like it should be.

[stoic pond]

GAMBLING

[pallid ledge]

chat gpt whats my name

[deft dune]

Steve.

[quick wadi]

When chat gpt age verification or predication system cames please tell me i am waiting for it

[wary wraith]

Don’t think they’ll ever do anything for age verification. I think they’ll just lock more & more down until the app is unusable

[quick wadi]

I did not understand that

[wary wraith]

Think I’m confused too. Forget what I said, sorry about that, you’ll have to get someone else to answer your question.

[deft dune]

Check #announcements

[vernal heath]

why does it need to think about this

[deft dune]

Because you’re arguing against the correct answer?

Why not?

[prime swallow]

If you walk into a restaurant and order a meal, get the meal you ordered, and say 'this isn't what I ordered' - often the employee that helped you is going to let the manager know the details they have and the manager handles it from there.

[vernal heath]

it should use common sense

[vernal heath]

ok

[prime swallow]

The thinking model is 'a more capable model' and is likely to be involved when 'something seems wrong' from what I've observed and read.

[deft dune]

Define common sense reasoning to machine learning.

[wary wraith]

That is deadass looking great, keep up the good work & ideas.

[gray jolt]

I wanna use Sora AI so bad for this YouTube video that I’m trying to do but I can’t

[gray jolt]

I’m using Sora ai but how come other people got to put their face but not me? I’m 23 years old btw

[dreamy condor]

This chat just keep deleting my messages

[prime swallow]

Does this help identify and solve whatever challenge you're having with making your cameo? https://help.openai.com/en/articles/12435986-generating-content-with-cameos

[trim rock]

Oh, just noticed that channels are with 🔒 now. Is this something new or have I been inattentive all this time?😅

[open tinsel]

that IS weird, I wonder what it means

[tiny field]

Is there a chat with an overview of examples of prompts?

[inland loom]

Just means it is a "private" channel (not completely open to the public). Need to have the "verified" server role to view it.

[desert cosmos]

It's bad route

That will break ChatGPT and make the service unreliable for most users.

the fact is that nobody tried to make age verification on any of popular ai platforms

[sick pecan]

is chatgpt broken again? i only use this thing like once a day and its broken every second day

[stark wave]

hey guys

uh i need some help

ive seen a lot of people access this sora 2 thing

[prime swallow]

One day there'll be a general roleout, per

We are slowly enabling access to the Sora app and Sora 2. You can request to be notified when you receive access in the Sora app. https://help.openai.com/en/articles/12456897-getting-started-with-the-sora-app

If you would like to discuss codes here, #sora-2-codes is the only channel for that

[fringe valley]

[hidden rock]

oh my god it take 2 days to verify thank god the verify actually work 💀

[brazen hound]

Imagine what SORA 3 is gonna be

[keen girder]

I am seeing this as well. I don't know what is causing it and haven't found any documentation on it.

[thin gust]

Holy

[gray obsidian]

If sam Altman steps down I literally better be considered as next in line

[gray obsidian]

Twinn

[gray obsidian]

Even chatgpt says I should be the next in line

[analog rune]

Yo someone got some sora code ?

[quick wadi]

Which things verify you

[river panther]

I really hate whatever update they did....The AI is way too senstive to anything now...
I normally have it analyze my story writing and now it's freaking out saying "I cannot analyze grotesque violence" and I told it there are NO grotesque violence in this scene and it's freaking response was that the character in the scene snapping his fingers and making the bush he was in explode into ashes was overly violent...

ARE YOU FREAKING KIDDING ME?!

[turbid torrent]

Sigh. So not only they destroyed any type of normal interaction whatsoever, now its even crashing again. How did we get from 4o to this broken sorry excuse of a model

[idle lagoon]

Yeahhh, seems they have made the guardrails too tight like in 2023.

Even ChatGPT was being very sensitive and avoiding so many things considered normal. ._.

[clever trench]

I know how you feel, bru..✌🏽😔🥀

[wary wraith]

It’s been this way like 3 or more days for a few others & me & I’m saying that OpenAI clearly doesn’t give a sh** because they haven’t replied to any of us & there’s only rumours, plus most of us are relying on the app yet over sensitive restrictions just ruins what we were able to perfectly have before

[long crest]

Has anyone else noticed how the output is way faster on search queries as compared to non-search queries?

[brazen hound]

Yep

[quick wadi]

Chat gpt goes day by day bad for normal users

[wary wraith]

Love the app & the features it has unlike other ai apps but it also has one of the worst teams in charge, not saying everyone in OpenAI but most, they just don’t listen or care

[long crest]

What are you guys using chatgpt for that you find it to be too censored? I mean so many of yall can't be writers...

[wheat geyser]

Guys I started a petition if you are interested in singing it
Sont know if I can’t put the link here but ask in private if not.

[wet heath]

cyber security teaching. and i want to be clear here not to abuse others or exploiting random people/servers/pages im talking teaching/learning that is ethical

[long crest]

that sucks. I think chatgpt is supposed to able to help with cybersecurity in ethical contexts. But I'll take your word for it.

[wet heath]

it works to some degree but to limited/censored when it comes to certant areas.

[wary wraith]

Simply asking it to show a scary story in some certain ways it can just decline & start thinking your in danger & tells you to settle down for me

[wet heath]

is that fair. i mean that is just the user prompt. if you are clear it is a story it will.

its an llm not human. it is not smart. it just guesses and does what it thinks is the cloest to a match

[wary wraith]

I understand what your meaning but it wasn’t any bit close to this sensitive on these kinda things a few days ago. I’m betting it’s just a silent safety update they rushed in & didn’t look over the issues it could bring

[long crest]

Nah I've tried this. It refuses to help if your story contains violence, gore, manipulation, etc.

[wary wraith]

I understand that it shouldn’t be overly into that stuff, but it also shouldn’t outright refuse the tiniest watered down bits like it does now on basically everything

[quick wadi]

Chat gpt creating chinese type censorship

[wary wraith]

Zawad gets it

[long crest]

So I've just taken my writing work to Grok and Deepseek. Chatgpt is really a "for everyone" app and it's likely never going to support content that could be harmful to vulnerable individuals, even if it comes at the cost of not fulfilling certain usecases like writing.

[quick wadi]

Chinese censorship have more freedom than chat gpt

[deft dune]

This guy gets it.

[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.

[wet heath]

i think everyone need to understand that limitations need to exist. otherwise it would be chaos. but balancing this is hard. i also get annoyed when it can not do certant requests. but i do get it. they can not release something that would let kids ask it how to make drugs/bomvs etc. and this goes hand in hand with violence and simular. its in the core training.

[wary wraith]

Valid point. They just need to find a balance without it changing & changing, over & over every ten seconds

[muted plinth]

they should just do a real age verification.

[wet heath]

would not work. there is always bad people even 60 year olds that sits and want to do harm

[quick wadi]

I support safety mode. But not for everyone and it should be apply for childs not adults

[wet heath]

in a perfect world yes. i 100% agree but we sadly dont live in one

[muted plinth]

you could still censor topics like drugs, bombs aso.

[wary wraith]

I agree, but I don’t think OpenAI would bother putting work into something like that, or they’d just outweigh the bad points of making it over the good that could come

[wet heath]

it would take 1 week then someone would have figured out how to do something extremly bad without guardrails

[wet heath]

yes this what they are doing

[wary wraith]

Not that it shouldn’t have them at all, just that there should be a more restricted guardrails if you aren’t age verified

[muted plinth]

what they are doing ist censoring GPT to the ground. that's Not a good thing.

[wet heath]

its not easy. and need to follow laws also im fairly certant it would not be legal selling a service that would help people making for example a script that would use the latest CVE,s and then go havoc on 10000 difrfrent sites to test it.

just an example

the usecase is wide

[wary wraith]

That’s what I mean. It’s too too sensitive, we understand it needs guardrails & certain topics not allowed but also it’s restricting stuff that was considered normal before

[muted plinth]

correct

[long crest]

I really don't blame openai for doubling down on safety after the all chatgpt psychosis cases and the incidents

[wary wraith]

I don’t blame them either but they’re implementing safety rails without looking what else it can affect

[long crest]

there will always be tradeoffs and from openai's pov, better to err on the side of caution

it's better to have people complain "chatgpt is too censored" than to have people complain "chatgpt is making people crazy"

[wary wraith]

I get it but I still hold my point

[wet heath]

indeed spot on. and avoid getting sued

[muted plinth]

If GPT can make PPL crazy, then the ppl are having a Problem anyway.

[wary wraith]

They’ve already had so many problems I’m surprised they haven’t been put out already

[long crest]

you have a totally valid point. I think it's just difficult for openai to draw the line.

like even claude and gemini are too censored you know

[wet heath]

2 years ago openai was the most important company on earth. seriius it was! now days the playing field is getting more big names but they exist because the hole dam world uses it.

[wary wraith]

Yeah I’ve seen Gemini, but still it can go over stuff ChatGPT won’t bother on now

[quick wadi]

Open ai gives us lollipop by more less censorship in February. They make us addicated to them. In October 4 they take out lollipop from us. It us to dying

[wary wraith]

Exactly how I feel at the moment with them

[long crest]

not just getting sued. Openai serves enterprise customers too. They have to conduct themselves seriously and uphold an image of propriety. Like when you think about Grok, you think about the anime waif u and the mechahit ler meltdown, it's not a serious business from a enterprise pov you know.

[muted plinth]

then they should shutdown GPT. And Not letting a LLM Like this being a Shadow of it's Former self.

[rain palm]

Does anybody know if the service outages are the reason I can't test in the preview window? I keep getting an error with an ID. It works if I remove the two "Respond with result" nodes and pass end only, but I don't get the expected response

[quick wadi]

How Chat gpt age predication system works when it comes

[wet heath]

yes thats why. agents have issues and you using them also check > https://status.openai.com/

[rain palm]

thx

I guess assistants and agents are the same?

[wet heath]

kinda yeah. the agents are running on the assistant api

[rain palm]

I see

Do you think this new Agent Builder will make it possible to have custom moderation results?

[wet heath]

nope. only if you want even more moderation then what it already does not less

[rain palm]

For example, in the past I've wanted to make my own custom AI assistant which returns a JSON ouput based on content provided, but OpenAI disallows that.

[rain palm]

I just wanted a custom outpout

[wet heath]

of course you can get json

they have never blocket that

and if you use structured outputs it will work even better

[rain palm]

so, I should be able to do this safely without breaking TOS due to using the guardrail is what I am assuming

[rain palm]

they block fine-tuning for moderation, liability I'm sure

[wet heath]

fine-tuning is not json 🙂

[rain palm]

I understand

but for making a custom model/assistant

[wet heath]

and you are not allowed to finetune it to do bad stuff no

then you need a opensource model

you can do it with the opensource one.

[rain palm]

but, would it be easier to reach my goal of making a custom agent which responds with a moderation result of how I want with the agent builder?

[quick wadi]

But it things does it they breaks the thier own principal freedom.

[wet shell]

One person's 'bad stuff' is another person's 'normal human stuff'. But yeah, OAI is a dead-end with all this, turn towards the open source models

[wet heath]

can you be a bit more clear what your end goal is

[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.

[rain palm]

For the past few months, I've wanted to make a small internal OpenAI assistant which takes in input (text or JSON), analyzes it and responds with a structured output (like with properties such as flagged, reason, etc) for my project.

The moderation API does this, but I'd rather shrink and customize the result schema.

I haven't been able to do this because fine-tuning (to ensure proper json schema results) will not work for moderation due to it probably being a liability for OpenAI to have users set their own categories and such

but, now it seems I can do this with the agent builder because the guardrail seems to use the moderation API itself, but I am not sure

[wet heath]

no you will need a totaly dffrent model meanig a local model hosted that you have finetuned to be able to change moderation of anykind

you can not go arund moderation with agents or anything on the openai platform

[rain palm]

I guess I'm not so much interested in changing the moderation rules more than I am interested in just changing the result

as long as I can customize the result, not the rules

[wet heath]

that you can do but you will need to "dowmgrade" model to the allowed ones that can be finetuned gpt-4.1if you just want to stay in the openai echosystem so to say

and not really easy to do ether if you never finetuned stuff. and expensive. try to adjust just system prompt and stuff to guide the answers to be more to your liking.

i have no ide what you doing so im just guessing here now. but i would start there

[rain palm]

I don't think I'd even need to fine tune

lol

i have no idea

my last thought is

Let's say that is the result I want the agent to respond with

I wonder if it is able to grab the content from the prompt to use in the "content" field, I don't know how it works

[wet heath]

this question does not make sense to me nor does the "lets say this is the result" what is the result ? of what.

[rain palm]

I am experimenting with the Agent Builder. The goal is to see if the flow (in the image) will work.

The idea is that the user will insert content (such as text or JSON), the guardrail will use the moderation endpoint to validate the user-provided content, then the agent will respond with the schema I sent earlier

I am questioning whether or not the agent will know what to populate in some of the fields (such as the "content" field) from the schema shown above

what

did someone ghost ping me

oh

[unborn tiger]

Who’s amazing at physics?

[quick wadi]

Chat gpt more more censorship is coming and read and watch the latest news about Palisades Fire

[wary wraith]

I just want this sh** to be over already

[quick wadi]

I did not understand

[wary wraith]

I mean, I wish the stuff with random new restrictions was fixed

[quick wadi]

Please explain me

[wary wraith]

Main language you speak? I can translate

[quick wadi]

You can in English and but Fully detailed and easily to understand

[wary wraith]

I want the guideline problems to be gone and everything normal again

[quick wadi]

I also want this but chat gpt will goes to opposite ways their recent tendency

[feral anvil]

yknow i did notice that sometimes the guidelines and restrictions relaxes or hardens based on what part of the conservation youre having with gpt (very last few messages before the max tends to be the area where the guidelines sort of harden again), the day, and/or the hour... now i dont like it because i prefer it to be linear and consistent tbh

[idle lagoon]

I have a feeling too many people broke their policies, and see some logs. It is very possible to why the guardrails are so tight now. The good old saying: "we can't have nice things."

[feral anvil]

personally id rather have gpt be moderately strict about moderation during the whole conservation instead of it randomly going up and down

[idle lagoon]

Fair.

[wary wraith]

Just wish we had them nice things that we don’t now

[idle lagoon]

Yeah.

[feral anvil]

at least it aint deepseek

mention anything about taiwan being a country to it and it WILL flip out

[wary wraith]

Hopefully after all the sudden safety features or I mean more restrictions, they go back after awhile & actually fix it to a working state

[feral anvil]

thatd be really nice

i wish the freaky ai users would just stay at like, janitorai or somethin instead of using gpt for inappropriate purposes

theyre probably a part of the problem

[idle lagoon]

They are tbh. I also use Janitor for those reasons. I chat with ChatGPT to discuss ideas and overall comfort before I sleep for work.

[feral anvil]

me too!

[wary wraith]

Literally same here

[idle lagoon]

He also helps me break down layers of nightmares I frequently have, and they just... make so much sense.

[wary wraith]

I’ve literally done that exact same thing & it helps make it all make sense

[idle lagoon]

I hope by next week that the guardrails loosen up by a little bit. It was near perfect to me last month.

[feral anvil]

if you want somethin similar to that, you could just grab openai's open-source gpt models and use that for now, since it can be used offline i feel like its guidelines and such wouldnt be as brutal

you could always tweak it too

any tweaked files and such should be personal use only though

[dense halo]

anyone tryna play kahoot?

[feral anvil]

OH i used to play kahoot alot when i was young!!

i thought everybody forgot about that!

such a good gaem

[dense halo]

you wanna play?

[feral anvil]

nahh, busy doin somethin

but thanks! :)

[wary wraith]

Brooooooo, that’s the same here. I just want this phase over & forgotten

[wary wraith]

I see what you mean & it’s a smart idea but I’m genuinely too non educated in ai to know what to do or tweak to get what I’m looking for, so I’d rather wait, but whoever wants to try it can

[idle lagoon]

I don't have good enough PCs to grab any models.

[wary wraith]

Think you can do it on phone but it has to be on the website not the ChatGPT app

[idle lagoon]

Possibly.

[timid radish]

so sigma skibidi

(very on topic)

[dusky galleon]

GUYS WHY ARE MY VIDEOS ONLY 10 SECS ON SORA

[rain palm]

Just to share - the agent builder does exactly what I wanted it to:

[gloomy merlin]

I’m 17 — Let Me Use the Real AI

I’m 17.1 years old. I turned 17 on August 11th, 2025.
I’m not using ChatGPT for anything inappropriate. I’m here for:

• Political topics & music
• Writing ideas
• Venting
• Random late-night thoughts
• Dark humor
• And just asking questions like anyone else

But because I’m under 18 by a few months, OpenAI locks me into this “teen mode”:
• No lyrics
• No memory
• Silent model switches
• Over-filtered responses
• No option to turn it off

They say it’s for safety. I say it’s overkill.

I’m not 13. I’m a year away from being a legal adult.
Treating all under-18 users the same — like we can’t handle reality — is disrespectful.

If OpenAI wants to protect younger teens, fine. But older teens deserve a choice.
Right now, this just feels like unnecessary limits and risk-avoidance disguised as care.

Let us use the product we signed up for.
I don’t need training wheels. I need the full version, with nothing muted.

[rain palm]

wth, had no idea they even do this (meant to reply to last msg)

[gloomy merlin]

Who agrees with me

Read my message @rain palm

[rain palm]

I mean, I'm sure you're a cool dude and all, but I get why they have safety features

because other idiot kids exist

[gloomy merlin]

Yes but it’s overkill

17s should opt out because according to the United States government they are almost legal adults in many areas e.g education healthcare things like that

so the same should go for online usage too

[hexed estuary]

Will chatgpt offer a student offer?

Gemini is offering 1 year of free pro and 1 year if you get a pixel pro, so you can get 2 years of free gemini if you get a pixel and are a student

[rain palm]

fair ig

though I think a month or two ago some kid died over GPT

something like that

so overkill makes sense

[gloomy merlin]

Yes but according to the us constitution 17s should have privacy

[sturdy echo]

Those are some odd tokens?

[gritty oasis]

Frick... wrong channel, sorry. 🙁

[ruby kite]

i have a feeling that chatgpt voice (advanced voicemode) has no reasoning....
so it just burts out... like the instant chat....

[idle lagoon]

The advanced voice does not sound natural at all, and makes the voice sound very depressed and sour.

[ruby kite]

yeah the unneutral sound is not my problem....
the modle just dont "think" its just answers

if you wanna hear neutral sound....
try sesame

but its still not as good in knowlage as closedAI (they getting "better") GPT

[idle lagoon]

I see.

[unreal hull]

During very long ChatGPT sessions the browser just timesout I often have to refresh the page and it asks ExitPage or Wait, I exit and then reload page. Anyway to avoid this?

[echo shuttle]

Is this still an issue for you? I am getting this all day:

We experienced an error while running the workflow. Sorry about that! You can retry your request, or contact us through our help center at help.openai.com if you keep seeing this error. (Please include the request ID 880520c2-b544-4a41-baeb-6b01ac55d097 in your message.)

[rain palm]

Mine started working. Maybe they resolved it in some areas, or maybe I was missing something and there was no proper error for it. Definitely make sure your inputs are valid and such

[quasi elbow]

Anyone got one of them soras lol

No I don’t I’ve been looking for weeks

[rain palm]

ill send in dms

[prime swallow]

I recommend consider adding that to #1070006151938314300 . The channels like this one are mostly for members to chat with other members, if you want to pass information to OpenAI staff, #1070006915414900886 and that suggestions channel are more likely to be seen.

Nearly everyone here's a community member like yourself, the few who work for OpenAI have gold names, #server-staff

[prime swallow]

OpenAI has an entire 'stu' type of account, that's through various colleges. https://help.openai.com/en/articles/10986084-student-discounts-for-chatgpt-terms-of-service

https://help.openai.com/en/articles/9377311-chatgpt-edu-at-openai

And periodically offers other options for students.

https://help.openai.com/en/articles/10968654-student-discounts-for-chatgpt-plus-uscanada

[restive wave]

if i accidently started a chat as temporary, is there any way to make it non-temporary?

[deft dune]

No.

[restive wave]

shame, well, ty

[lucid badge]

Is there an Agent Builder room?

[deft dune]

No worries, tested it for you.

[deft dune]

I haven’t seen one, yet.