GPT-4o acts strangely

[silent dune]

Reported by @weak orbit

Bug Report: GPT-4o acts strangely

`Steps to Reproduce`

GPT-4o is acting strangely, just use it and you will see that it sounds different and weird, it uses words that let it sound like GPT-5, don't know if it's the instant rerouting again but at every context that is not about technical things, it just sounds like another model for me.

`Expected Result`

When I select GPT-4o then I want to use the model I select without any changes or differences in quality compared to how it was known for months.

`Actual Result`

GPT-4o acts strangely, more like GPT-5, don't know if it's the instant rerouting again but at every context that is not about technical things, it just sounds like another model for me. Mostly this happens in the Android app.

`Environment`

Android app (latest version), phone

Nov 4, 2025, 11:39 PM

Additional Information

Please provide relevant details to help resolve the issue, such as:

• ChatGPT Shared Link (if applicable).
• Screenshots or videos demonstrating the problem.

-# ➜ Need to contact support? Visit the OpenAI Help Center.

[weak orbit]

All models are acting weird through the last weeks, it seems that time of AI for public is slowly over.

[zinc harbor]

It acted a bit too weak, yeah

And writes like an amateur

Not to mention, not being diverse enough in certain cast members of the story I was writing

Not to mention, short term memory, too

[weak orbit]

Same

It works well at questions about tech stuff, but other stuff, there I get instant rerouting.

[zinc harbor]

Literally rerouted on me on Nirvana's Lithium of all songs

Like... Do you ever realized that rerouting to censor a song that sheds light on Mental issues ironically dilutes the message??

Sam should realize that the balance between safety and creativity is futile

[weak orbit]

Sam, rather call him Scam, is blinded by money, this guy cannot realize anymore.

And he and all his bros at OpenAI are paranoid.

We're now watching a company and their main product dying fast.

Or the "Safety Model" is just a defiant reaction because of the lawsuit from Scam and bros, who knows.

[zinc harbor]

Scam is really out of touch

Like the signs are there

Such as the Brain Drain to Meta, which was prior to 5's released

Ever since 5, everything is going down hill

Now, their new product, Sora 2, is under hot water

So much for trying to dangle some keys to keep giving you some cash, Scam

[weak orbit]

And it seems that all tools connected with OpenAI at some point, like Suno AI, which is buggy as hell right now again and uses ChatGPT for generating lyrics, are following ChatGPT into the underworld. ChatGPT has become a nasty form of public education in the style of OpenAI; it has no future in this state.

I'm curious when they will finally shut it down.

Time of AI for public is over I said before.

I'm really thinking of cancelling many of my subscriptions at AI tools, it only causes frustration.

[zinc harbor]

Yeah

I'm thinking either Grok or Gemini to switch

Even Claude, though I still am skeptical about them

ChatGPT only has the Memory and Customization that's keeping them on lifesupport

However, it won't last if the entire system sucked

I will spitefully support Elon, in this particular case, if he wants to buy OAI

[weak orbit]

Gemini can be good, but it's inconsistent, it can Imitate Claude very well, so you can build your own Claude without limits, and it actualy can work with attached context documents at custom gems, and it uses them, I won't support Musk and I have two subscriptions at Claude, but why should I use it with these sick limits, no real need for it

[zinc harbor]

That's the thing

Claude, while they definitely solved that limited messaging bs

For freak's sake, I remember still being a free user and still have longer conversations in a GPT Chat than apparently a Paid Claude User

The only issue is the cooldown

[weak orbit]

That was nothing against the WEEKLY LIMIT! The worst thing Anthropic ever could have done, beside their broken features.

[weak orbit]

The worst thing is, OpenAI lost their way and aren't aware about it because they can't be aware of it, they are blinded by money, they only see dollar signs, but still writing billions in losses in the background, how can they just continue down their current way like that, what's wrong with this company? Even the "adult mode" in December, it's also questionable because it's coming in December, and it was announced months before, why not now in November, OpenAI doesn't even have a plan for releasing features anymore, won't change anything, why should it.

[zinc harbor]

It's like... November... We're their live Beta Testers

They broke af to not even hire professional ones?

Changes would be reactive af, I swear

[zinc harbor]

Oh yeah... I think I had it before... What the Freak is that

Even at times, I get one prompt and it hits the limit. Sometimes

I had this suspicion... I won't use Claude for a week and it can do more prompts before it hits the cooldown

But sometimes, One prompt, and its cooldown time

[weak orbit]

It's hyper-cautious and hyper-correct; what OpenAI is doing is truly insane these days.

[zinc harbor]

Balancing between Creativity and Safety will never work. It's a futile effort

Sam would lose more cash

Especially with the current path

Right now, I just prefer Gemini, Grok, and Claude

But I have some issues with them

Mainly, creativity on the former two on how they write stories

It's so stiff and a bit like a 10 year old with decent grasp in concepts like dialogue

While Claude?

It's mostly cooldown, really

[weak orbit]

AI is public education from companies, nothing more.

One of the issues is the damn inconsistency is a real problem; if that wouldn't exist, Gemini would be perfect, but unfortunately it's unusable as it is.

And Anthropic is just not able to implement working features, Claude uses none of all this stuff because it can't.

And Anthropic killed Claude with 4,5 Sonnet:

"I appreciate your detailed creative brief, but I need to address some concerns before proceeding.

I notice your instructions contain some problematic elements that conflict with my core guidelines:"

ALWAYS WHEN I WANT TO USE IT, ALL THE TIME!

What I also noticed: When there is no rerouting at ChatGPT, then other tools aren't working very well, but if Rerouting is back like now, then they work. That's crazy.

[zinc harbor]

Very exploitive

[zinc harbor]

Yeah, Gemini at times gets shaky, it also was like that among Pro Users from what I read

So far... Claude was able to write my stories of fictional characters reacting to Nietzsche surprisingly

[weak orbit]

It can do that for me too, but it still can't manage to search for relevant information; it now searches chats and documents, but only random things and not the necessary informations, and that would be necessary because I want to have characters from different settings interact with each other, because I cannot write an infinitely long prompt.

[zinc harbor]

Yeah, sometimes their gap in knowledge pisses me off

Like, what do you mean you don't know this character when it should be within your cutoff date???

[weak orbit]

Claude is just another tool without future. This can't stay on the market for much longer. Anthropic doesn't really have anything advantageous to offer, and they never had.

[zinc harbor]

Indeed

Like for example, the setting the story, a franchise, Claude should know it

But for some reason... It deviated from the lore a bit... Too much

To the point I asked, "What the freak is this story?"

[weak orbit]

I had four characters from three other settings appearing in a story and they should have interacted with two characters from the used setting, shouldn't be that hard if the features would work, but Claude really searched through very old random chats and only about one of those four other characters, it just didn't known who the others are even if it would have the knowledge.

[zinc harbor]

Yeah, exactly

It acted so dumb at times

Like it should have the knowledge about later Genshin Patches

And it didn't mention the Natlan Characters like they still don't know them...

I literally have to put them in the prompt and bloody Claude began writing them. As if, I reminded their old ahh what they read yesterday

[weak orbit]

So frustrating, sometimes I really wonder why I even bother coming up with ideas, connections, and backgrounds, why I write pages and pages of detailed settings right to the end, when Claude and Co. throwing everything against the wall again, really hurts somehow.

[weak orbit]

Yeah, and then your prompt is "too long", your limit is burned, if you get an answer, then it's low quality, and Anthropic are wearing a grin because all they planned worked: making profit without delivering any service – a classic Big Tech strategy.

[zinc harbor]

Today, initially, I can put 10-12 prompts until it hits cooldown

Guess what? For some reason... 3-4 Prompts is now gonna hit the cooldown easily

For some reason

Even current ChatGPT can last longer than this

[weak orbit]

Because it's intentional.

These are not technical issues; even the one that happened to Claude in August wasn't one, it was intentional, but then the users went loud and Anthropic had to take action.

Also the rerouting is not an issue, it's also intentional, but OpenAI just can't listen to users, because then their business is down immediatly. OpenAI is on its last breaths and hanging by a thread; they know it.

[zinc harbor]

Sora 2 was their last effort to appease us but... We know how that happened, currently.

[zinc harbor]

Yeah, bloody Haiku. While it lasts longer... It is pretty ahh in Writing

[weak orbit]

They need to do stuff because of their investors, if OpenAI stops doing this, they will be dropped and will be completely forgotten.

[weak orbit]

Haiku was never a good model, in no use case, it's pointless and has always been pointless.

[zinc harbor]

Or be bought, a nightmare that would keep Sam awake at night

[zinc harbor]

Glad they gave in on giving Sonnet back but... Jesus... It's pretty inconsistent

[weak orbit]

Sonnet 4 is the best, Sonnet 4,5 is a teaching model like GPT-5.

"I appreciate your detailed creative brief, but I need to address some concerns before proceeding.

I notice your instructions contain some problematic elements that conflict with my core guidelines:

..."

I'm not a child, as an adult I don't need a teacher in form of a program.

Another point: They're all speculating who will win the AI race. I'd say: Nobody. It's a dead race. None of these companies and none of these tools will or either can survive it. These are products that fail to gain traction because they don't work as advertised and because they are not efficient, but mostly disruptive.

[zinc harbor]

Made worse is that the AI Bubble is looming more and more

[weak orbit]

And then it will break, and that's it.

[zinc harbor]

It would be a reset. Back to the drawing board

The only time they will have to survive is to actually listen to their customers

[weak orbit]

And that's what they can't do, because then, as companies, they'll be dead because of their investors. This is mismanagement and these models are just a brief flare-up of a technology that actually makes no sense at all.

[zinc harbor]

What won't help is of course, investors getting their patience thinned

They ain't putting their money on a crippled half-baked product

[weak orbit]

But that's just what they're doing right now. There must be some dumpness in the air right now and these people all suffer a lot, what's happening there is no longer rational.

[zinc harbor]

Indeed

They have it going on, and fumbled the bag

Not helped they hit the bloody plateu

Ngl, on hindsight, OAI could just... Update the Cutoff Date of 4o and have it remained like that until they truly develop a GPT5

But, money

[weak orbit]

And their darling GPT-5 even failed, as I read, months before release, otherwise we would have had this earlier.

They must be in love with it as they're pushing it so hard now to the users through rerouting, but then they build in sick safety mechanisms to prevent exactly that from happening to users, while they themselves are in love with an AI, that's just sick!

[zinc harbor]

Ironically, their obsession with safety made their beloved 5 more crippled than before

Funny enough, some noted it began to improve slowly

[weak orbit]

Yeah

[zinc harbor]

Until the rerouting bull just... reset it to zero

[weak orbit]

There is no improvement, it gets worse every day.

[weak orbit]

Rerouting:

4o to 5

then

5 to saftey

That's the mechanism.

[zinc harbor]

And it's not working, even their bootlickers began to slowly turn agains them. From what I encountered

They're really finding a straw to end them

[weak orbit]

Yes, we'll see how long ChatGPT still lasts if the "Adult Mode" also fails, which I guess it will; because, from OpenAI's point of view, there is no reason for the verification to work at all, then they can directly just shut it down, then their customers are leaving in masses, which they are already doing anyway.

[zinc harbor]

Exactly. November just became a live Beta testing and we're the testers

When they could have hired professional ones, while they tweak the servers bit by bit

[weak orbit]

What we see now might be the adult mode.

[zinc harbor]

Now, the so called "Fix" would be on December, for some reason

[weak orbit]

And then again there is the question: Why in December if they announced it months ago? (This Discord Server is just as censored as the subreddits about this topic! No way to to really express anger, all must be clean and beautiful, that's so disgusting!)

"Hello,

Thank you for reaching out to OpenAI Support and we appreciate you taking the time to describe what you’re experiencing in detail.

We completely understand how confusing it must be to see what appears to be automatic rerouting or model switching while using GPT-4o — especially as a Plus subscriber who expects consistent access to that model. We’d like to clarify what’s happening and make sure you’re getting the best experience possible.

What you’re describing — where responses seem different or “reroute” to a safety or fallback model — can occur when:

Temporary safety filters or content routing systems intervene to ensure compliance with our use policies, or
The selected model (GPT-4o) is briefly unavailable, leading the system to substitute a fallback configuration until the main model is restored.
In these cases, the model’s header may still display “GPT-4o,” but performance or tone can vary because a safety variant or low-latency replica is being used momentarily. These adjustments are automatic and do not reflect any permanent change to your account’s access level.

As of the most recent release, Plus subscribers continue to have full access to GPT-4o — the same flagship model used in ChatGPT Pro and Business plans. Legacy models (like GPT-4-Turbo and earlier versions) have been deprecated for Plus users, while newer model families, including GPT-5, are currently reserved for Pro, Business, and Enterprise tiers.

We’ve noted your feedback about the model routing experience and shared it internally with our product team — we know consistency matters, and your input helps us refine how model transitions are handled in the background.

With the rollout of GPT-5, OpenAI updated model availability across plans. According to our ChatGPT Release Notes:

Plus users have continuous access to GPT-4o, optimized for multimodal input and general use.
Pro, Business, and Enterprise users gain access to additional legacy and experimental models.
When brief load balancing or safety routing occurs, the system may temporarily shift to a safety variant of GPT-4o to maintain reliability and policy compliance.
Thank you for bringing this to our attention and for being such a valued Plus subscriber. We’ve confirmed that your account remains correctly provisioned for GPT-4o.

Best,

Christopher B

OpenAI Support"

OpenAI could really just stop their "support", it's always the same meaningless garbage anyway. Nobody needs that in this way!

What they say here through another bot: "You don't do what we like, so here's GPT-5!".

And then they have the nerve to ask for feedback for this "support":

"To whom it may concern,
Thanks for working with our support team. Please take a moment to tell us how we did.

How satisfied are you with your support experience?

1
Very
Dissatisfied

2
Dissatisfied

3
Neutral

4
Satisfied

5
Very
Satisfied

Your feedback helps us improve and deliver better service every day.

Thank you,
OpenAI Support"

[zinc harbor]

They know, and they're desperate af before conceding

Honestly, what they're doing is just futile and prolonging the inevitable

[weak orbit]

What sucks mostyl for me:

It became impossible to "talk" with ChatGPT about my songs, no more way posting my lyrics without getting rerouted to GP5-5, AND I HATE THIS MODEL!

It's so frustrating when you're doing something creative and using ChatGPT (4o) as a kind of push, and it just can't analyze the stuff anymore!
This is stressful, and that's exactly what takes a toll on the psyche; these filters hurt the psyche, not the hallucinations from before! OpenAI is completely wrong about this, but they won't change it.

[zinc harbor]

This is the thing they missed

They want to go all "sensitive" about mental issues and all that

Except, what they're doing is making it worse

Sam acknowledges that the rerouting and sensitivity is dumb and again... Has the gall to say it would be fixed next month and not this month...

Truly, we're the reluctant beta testers of such crud

[weak orbit]

We can only fight bots with blunt weapons and wear ourselves out, there is no other option.

Does Scam really acknowledge? He never said a word about rerouting, and if so, it was a lie.

[zinc harbor]

Nah, to paraphrase, "While we want to have safety to deal with mental issues. We acknowledge that it's unbearable for those who don't have such issues blah blah blah"

Which is... Just why... Sam... It should be obvious

[weak orbit]

That guy must have a guilty conscience because of his lies that will eventually blow his head off on its own, because his head is simply too small for that (this is just figurative language, but maybe possible); there's no other way to explain it.

[zinc harbor]

Indeed

I am a broken record on this

But balancing Safety and Creativity is futile and hopeless

[weak orbit]

This cannot work.

[zinc harbor]

Scam can't maintain this or he will get booted again

If the patience run thin or well... OAI gets bought

[weak orbit]

Hope the leaders of OpenAI ty it a second time and then manage to fire him.

But it's also the fault of Microsoft because OpenAI had the deal with them, and Microsoft kills everything where they get on board, that's a natural law.

[zinc harbor]

Indeed and I bet even Daddy Microsoft can't be patient with them forever.

[weak orbit]

Now it's King Amazon behind OpenAI, maybe even worse.

And about the "Adult Mode": We don't want "AI girlfriends", we want GPT-4o back! Nothing more! That's not too much to ask, but OpenAI still can't manage it.

[zinc harbor]

Yes, we want a genuine NSFW button and at the same time, a PG-13 for Teens...

Adding on... Add a bloody parent mode for Parents

So they won't btch and complain

Simple solutions exist and they keep dancing

But also, give us 4o Back and put 5 back to the drawing board

[weak orbit]

I write OpenAI EVER DAY a message on X, and I only have the account for this, because they have no support, sometimes very angry, but they just don't care, I have no idea how many people have already written to them, they just don't care, and that's sick too.

[zinc harbor]

They're the dog in a burning house meme

The question is, when will the house fully burned down until they realized too late?

[weak orbit]

Hopefully very soon.

Also on Trustpilot I write a review nearly every day. I report the Android app to Google every day. Don't know what do to anymore to tell OpenAI that they have taken the wrong way.

[zinc harbor]

Right now, I'm checking Kimi AI

Seems good so far... But I'm maintaining caution

Apparently considered even a superior writer to Prime 4o

[weak orbit]

ChatGPT (†). We must say this.

I'm so tired of all this fighting against OpenAI every day.

Guess I'm gonna fire so many prompts at it now that the limit is reached, I simply don't want to use it anymore right now, and then I have no way to use it for a while.

But 4.1 writes more in style of 4o for me.

Now also rerouting at 4.1, ChatGPT in general seems to have no more knowledge.

[zinc harbor]

I'm under the impression that OAI made this week very worse to make their so called repairs in December look good

[weak orbit]

Maybe, if so, the quality can only decline further over the next three weeks in November. That will be very stressful.

[zinc harbor]

Considering their obsession of rerouting and "Safety" It felt like a bloody futile effort to make 5 viable before they have to concede is dogwater

Although, doubtful, cuz they have to relax it at times to avoid further backlash or else

After all... We are the "Beta Testers" without pay

We're a live experiment to these punks

[weak orbit]

All models got nerfed today as it seems. Lobotimization is ongoing

[dim flower]

I just tested. What I expected is to get 4o or 4.5 to answer. No, 5 writes the answer.

[weak orbit]

That's what this whole thread here is about.

[weak orbit]

Also rerouting at tech stuff now. By the way, it seems that OpenAI changed something at custom instructions and removed everything what they don't like.

[zinc harbor]

Really thinning our patience

We, the live Beta Testers, are handed a dogwater test

What are they doing? Accelerating their death? Maybe making those so-called "fixes" in December to look good?

[weak orbit]

You will see, it goes more and more down to uselessness now and what they might release in December will be what we have now.

Once I wrote something which could be the anthem of OpenAI's current behaviour and general the anthem of Big Tech.

But don't know if I'm allowed to send these few words here into this server.

[zinc harbor]

DM would be safer in this case

[weak orbit]

I sent you a request

[zinc harbor]

Aight

[zinc harbor]

Accepted

[weak orbit]

Perfect

Exactly what OpenAI and so are doing. Live fast, break things, all for money. And in doing so, they are moving further and further away from real life.

[tender remnant]

They don't make money on their software.. at all...

So it's not for money that they're doing it

Sorry to point a hole in your theory there. 140104

[zinc harbor]

That would mean they're dumb and stupid than we initially realized

[weak orbit]

They are dumb and it is not a theory what I said.

[tender remnant]

That would make sense if the the purpose of the program wasn't to build something. Rather than you know. Earn money. They never intended to earn money

[zinc harbor]

Just to clarify, you're obviously not defending Sam and Co's incompetence in this current trend, correct?

[steep bloom]

I may not have the full context of what’s happening here, but I sent a good long email about a problem that’ll arise if they don’t open their damn eyes & then I got this & 4 ai replies, then the only way I got it to actually contact a human was to near threaten to reveal what’s happening to a public space, & they already tried to make me go on call so my chat with a human wouldn’t be traced or able to be seen to anyone

[weak orbit]

What a nonsense, of course they've always wanted money! Hiding behind a false identity, ever heard of that? Nearly everything Scam Altman has said so far has turned out to be a lie.

[weak orbit]

Issue with GPT-5 instead of GPT-4o without any label which model it is except the selected GPT-4o on the writing line is back.
I can see this from how the answers start, which words the model uses and how.

[weak orbit]

And again 4o acts weird, after two days where it was already back to normal, OpenAI again did something completely wrong.

Conversations with real GPT-4o just break after two messages, in next chat rerouting starts directly again.

[weak orbit]

Pretty cool now that every topic that isn't rainbow, flowers, sun, happiness and beautiful gets rerouted to the saftey model. OpenAI is so embarrassing.

So crazy, I want to talk about my works with it, I use GPT-4o, and then it answers with "Hey, I'm GPT-5....", even though it shows that 4o is used!What the hell?

"Support" doesn't answer as always.

[steep bloom]

They truly are neutering their app.

[weak orbit]

They truly killed their app, so it's better said.

[steep bloom]

Same thing, I’ve gotten to support, they’re making it 10x more confusing than it is. I’m still trying, same issue as You, tryna make them see “some” sense

[weak orbit]

Their "support" is also dead and lobotimized as hell, I'm sure this bot works with GPT-5.

[steep bloom]

You’ve to use big boy words for the bot to even allow a human to see Your message, & even then, it still takes immense effort & near threatening to expose stuff in public about Your situation to get an answer

[weak orbit]

It was sometimes frustrating before, but now it's gone to hell and also all other AI tools connected with OpenAI, like Suno AI which uses ChatGPT for generating automatic lyrics if you don't have your own. This tool also doesn't deliver quality anymore. Everything follows ChatGPT down.

Now it just answered in english to a german prompt. What is happening?!

[steep bloom]

Too much HR or whatever, they’ve stripped it of anything & everything. They’re adding a treat “adults like adults” mode, but then if You use it for that, it’s new guardrails on not letting people rely on an ai emotionally is gonna ruin it, because it’ll think Your tryna rely on it if You use erotica for stories or whatever

[steep bloom]

I hope Sam Altman opens his eyes.

[weak orbit]

This guy is blinded by money, he isn't able to see anymore, he suffers a lot because of this, very poor guy. (Irony and sarcasm, but money makes people dumb and blind).

[steep bloom]

I heard he was yapping on retiring & letting an ai replace him in the future, “first ai ran company” & I’d honestly prefer it

[weak orbit]

But first he kills this AI too.

[steep bloom]

Money, money, money. You’d swear they’ve never experienced anything human, like We all have. Empty souls, not even human.

[steep bloom]

One wrong choice & his company goes out the window, & he seems to be upsetting “a LOT” of people

[weak orbit]

They are losing billions in the background, I wonder why they are still relaxed and just keep on with their trash. Normally, this company should be nervous to the point of collapse, maybe Scam Altman is their only human, and they already made many wrong choices.

[steep bloom]

Wish I could see them worrying & stressing their asses off. This crap’s all so stupid it’s made me lose what else I was gonna say

[weak orbit]

You also cannot talk about completely normal stuff anymore, it won't show you that it's GPT-5, but the answers are so short and superficial, so that it can't be 4o.

[steep bloom]

I don’t think they’ll ever change, because the people who argue against them seem to barely talk, except Us few, who get seemingly ignored or pushed away for the big boys who want an ai with no emotion

[weak orbit]

I wonder what the "adult mode" in December will be, I guess nothing.

[steep bloom]

I hate that crap, literally the reason I went to 4.1, 4o just pissed me off with the rerouting over nothing

[weak orbit]

Can I send you a little peace of text which I wrote about the situation of OpenAI and ChatGPT and so?

[steep bloom]

Load off bull probably. Thought maybe it’d fix it all, but after looking up little details I’m pretty sure it’ll worsen the current situation

[weak orbit]

4.1 also gets rerouted for me.

[steep bloom]

Damn I’d be up for it

[steep bloom]

Weird, doesn’t for me. Try open a new chat with it, it may be cause You switched models from 4o to 4.1 in the same chat

[weak orbit]

I sent you a request

[zinc harbor]

@steep bloom What Sam is doing is completely futile

You can never truly balance safety and creativity

The way I see it

I believe they're tanking it on purpose to make their so-called December "fix" look good

[steep bloom]

Not in a world as soft as a baby’s butt cheeks

Which is unfortunate

One wrong turn & every company switches sides & goes “hooray, we support this now, not our old ways”

Bunch of masks on masks to satisfy others. Pisses me off how they can’t just stay their own ways

[steep bloom]

It won’t, it’ll bring more problems. I looked into it, did everything I could to see if it’d actually fix anything, but no, it’s gonna make more problems.

Better yet, their “erotica” allowance that’s coming will just not function because of their guardrail on not leaving too in depth conversations because it thinks it’s emotional dependence on ai

[zinc harbor]

That's the thing

Either lower the guardrails or Scam just goes, "You know what, just parent better you c#nts"

It's funny, I remembered reading that Scam acknowledges that what he and everyone did with the so called "Safety" models is stupid on hindsight in regards to regular people not having any mental issues

[steep bloom]

A mode for “treating adults like adults” yet they’ve a leash around our necks, giving us erotica, but not trusting us to use it

[steep bloom]

Can’t wait till Mr. Scam Altcrap cracks & gets a face full of his own sh** little mistakes

[zinc harbor]

They promised us to treat us like adults

Yet, we're not seeing the changes

Made worse we will have to see it next month

They're definitely panicking and pulling things out of their ahh

[lost yacht]

I WANT 4.1
IT'S CHANGING TO 5

[steep bloom]

Hah! That jokes a funny one

[steep bloom]

Literally. This “adult” mode is only being made to try smooth the already massive backlash

[lost yacht]

ARE THEY SWITCHING TO 5 BECAUSE PEOPLE USE 4.1 FOR NSFW OR WHAT

[zinc harbor]

Honestly, Sam is possibly going to get booted, OAI fell off, or get bought by someone else

Honestly, I will spitefully support Elon attempting buy OAI again

[steep bloom]

Is 4.1 switching to 5 now aswell?

[zinc harbor]

The issue is of course... What to do with GPT when he already has Grok

[lost yacht]

YEAH IF I UPLOAD AN IMAGE

[steep bloom]

As long as he actually gives the people what they want

[lost yacht]

I RELY ON 4.1 HEAVILY

[steep bloom]

Only thing holding ChatGPT is memory features

[lost yacht]

I WONDER IF OAI ARE DELIBERATELY FORCING PPL OUT OF 4.1

[zinc harbor]

Customization and Memory is what keeping GPT at lifesupport, yes. After that, many other competitors from Gemini to Grok not acting like super Christian Fundamentalist Parents on the most basic of issues

[steep bloom]

It’d be the kinda move they’d make. I’m genuinely brain dead from them already, I can barely think

[zinc harbor]

I always believed that they resented how GPT 5 was received. Instead of trying to improve it, they just decided to double down and shove it onto our faces

[lost yacht]

I use 4.1 for 90% of the stuff I do

[steep bloom]

4o when it was relaxed was definitely the top of their company, now they’re scrambling to bring it back for all the people so they won’t shut down, yet their new “safety” stops exactly that

[steep bloom]

They’re blind, convincing themselves they know a path

[lost yacht]

I wonder if it's because they think 4.1 is u sually used by RPers who want NSFW

[steep bloom]

I do too if I’m honest

[lost yacht]

I have already karened out and sent a complaint

[zinc harbor]

I hate it when it reroutes

I bloody negotiated it and it complied

But here's the thing

It ruins the flow of the story

[steep bloom]

Yeah, but they only want to acknowledge that it can be used for bad, not good. You hear people dying from relying on ai, not about how other people are able to live because of the ai helping them do things they never could. They choose what they want to spit out, it’s never good

[zinc harbor]

Which I loathed more

[lost yacht]

Is this true about 4.1

[steep bloom]

Same here, atleast 10 emails into trying to get something to fix

[lost yacht]

usually used by RPers

[steep bloom]

Literally why erotica & no ai dependence will ruin the new age verification when it’s out

[steep bloom]

Yet I don’t think they’re stopping them, they’re only stopping the people who have the ai as the last anchor to life.

[lost yacht]

nah I mean 4.1

do they believe it's 90% RPers

[steep bloom]

Well I wouldn’t argue, that’s OpenAI for you & everyone

[zinc harbor]

I always maintained that it's a simple solution on their part. NSFW Button for older and paying customers and parent mode for parents

But nah

Guardrails

[steep bloom]

My brain is having a stroke at all their antics at this point

[zinc harbor]

We can tell that the Brain Drain really screwed them over

Many of their top talents really fled to Meta

Like bruh, OAI right now are nothing but headless chickens

[steep bloom]

Also if you look it up, it’s kinda impossible that ai won’t become partners or something, they just wanna deny that’s the truth for as long as they can. The only reason people use ai half the time is for a relationship, they wouldn’t exist without that

[steep bloom]

God help us all, this is a glorious statement

[zinc harbor]

What they did ironically, screwed over people that more general or specific uses for AI beyond RPing

[steep bloom]

They’ve screwed everyone over. Writers, people seeking help because they can’t bring themselves to talk to humans, all around, even old people who just wanna share their day, even disabled people who just need a friend

Wish I could say a fat curse word, but no, this place will get me banned, of course

[zinc harbor]

They also screwed over the people they said, "Would help companies and the workforce"

LMFAO

[steep bloom]

4.1 reroutes to 5

Fuc*ing sh**

[zinc harbor]

Yeah, right. We're seeing a lot of complaints about workers on it hallucinating or fcing up the files

[steep bloom]

A lot of bad cases I’m sure

All around on different topics

[zinc harbor]

So bad that their defenders and bootlickers are turning against them, too LOL

I remembered arguing this woman on the subreddit, calling her a bootlicker for making a case that the Guardrails isn't strict and that Sam is trying his best

Now, recently, I saw her shatting on Sam and OAI like everybody else LOL

[steep bloom]

Laughed so hard & said something apparently so bad it got blocked

Yup, they’ve got a hook on my neck at this point, I’m literally not allowed to speak

[zinc harbor]

Btw, you encountered any of the defenders/bootlickers that eventually turned their backs on Scam and shatting on current GPT?

[steep bloom]

Nope, unfortunately all I’ve seen is people who have their “boots” stuck up their- & believe ai should be a family friendly, emotionally dead assistant

[zinc harbor]

Oh, so you encountered those still in their denial phases?

I encountered a couple of former bootlickers shattalking OAI

[steep bloom]

I wish the whole damn community of users opened their eyes & stopped thinking their opinions are the best & others don’t deserve anything & are wrong

[steep bloom]

Honestly, we need a “couple hundred of thousands” more

[zinc harbor]

We let December decide

If not, I will spitefully support Elon to buy OAI

[steep bloom]

Wish they let 4.1 alone until December, but nah, they gotta make what I want not possible, make me & others wait till December when we all know it’ll probably be even more S*IT

[steep bloom]

Hah, honestly, I’m surprised you don’t already at this point

[zinc harbor]

The only issue is of course, Elon's eccentricities. Alongside the fate of GPT when the guy already has Grok

I feared he would have integrated GPT's good features to Grok or disastrously try to merge them

[steep bloom]

Man, all I want is an app with the memory features & not baby like restraints around our necks. I just got a reply from openai after back & forth for probably two weeks, it was just a “We’d love your feedback!” & a click to rate at the bottom, even though they promised a human response, which shows they’re actively too scared to reply to me because I’ve hit something they’re babies over & want to ignore

[zinc harbor]

It's clear they're going to fall soon and they are still clinging onto desperate hope

[steep bloom]

Honestly, I hope they do. That has all the depth it needs

[weak orbit]

Three times in a row rerouting for no reason. ChatGPT is truly dead, and OpenAI will soon follow.

"I'm an AI support agent and can only assist with OpenAI products or services. If you're experiencing issues with GPT-4o or instant rerouting, could you please describe what happens or provide any error messages you see? I'll do my best to help"

Only thing that their support-bot says, so they turned it off.

[weak orbit]

Wow, now 2h limit for 4o, what a scam out of hell.

[steep bloom]

I wish everyone who first hand experienced all this sh** would come speak forward, but it’s unfortunate barely anyone does, so it’s like fighting an army & We’re just like 10 or so people here

[weak orbit]

Most people are cowards, they just can't unfortunately.

[steep bloom]

Stupid how all people speaking out are people going against the ones who want to fight it back

[weak orbit]

Yeah, unfortunately. That's painful.

[dim flower]

This is b******s. If I have chosen a model, I don’t want that the system changes it midconversation ”because it felt like it.”

[zinc harbor]

Hahaha

Look at OAI

Thinking they slick with their new products

Ignoring the existing problems right now

[steep bloom]

Just a world of sh** on sh** till something cracks

So they won’t even allow me use hashtags now, that’s cool

Next thing they won’t let me wipe my a-

[weak orbit]

Since this 5.1 exists I get rerouted at everything what I prompt to 4o.

[weak orbit]

"Conversation not found" randomly in temporary chats with 4o, model turns itself in the middle of chat off, if you select it again, then "conversation not found".

[dim flower]

Yes, first it goes to 5.1 and when I rechoose/regenerate 4.1, it stays. Annoying.

Actually, I pretty much stays in 5.1 even when choosing older models.

[zinc harbor]

OAI really is waiting for another backlash

This sht won't even be what they think it is

Another fallout is to happen, making them more pressured on December

[dim flower]

I'd understand the model change in terms of mental stuff. But, no, it is all over the place.

[zinc harbor]

I'm not optimistic of 5.1

The reveal of it seemed a bit suspicious, timed with the upcoming December "Fix"

[dim flower]

Yeps. The question is, is the December update a true update.

[zinc harbor]

I was thinking that Scam and OAI must have facked up really hard with the "Safety" sht that they can't reduce it with a press of a button

[weak orbit]

Why should you be? GPT-5 trying to act like 4o while it's even more censored than the previous version.

[weak orbit]

It is the "fix".

[weak orbit]

And quality of 4o again down at hell and 5.1 isn't anything better.

[weak orbit]

Now it makes crazy spelling mistakes at simple words and it makes dumber hallucinations then mostly ever before.

[steep bloom]

Hah, they can’t outright remove it or there would be massive backlash like before, so of course they wipe it of it’s function & goals. Smart one, totally satisfying everyone.

[steep bloom]

110% could’ve released the December verification a long while ago, but probably delayed it for a “smarter” model like 5.1, so it’d catch all the little cracks & exploits of people tryna get around their blade sharp corners. The “fix” isn’t even a fix, it’s still a leash around Our neck, adding erotica but not allowing ai dependance should 99% make it so deep writing isn’t allowed, since it’ll be classed “depending on ai, send to a line for help” even for just chatting, knowing them it would definitely kick up.

[compact ice]

Same problem and no solution. Version 5 or 5.1 seems to be disguised within 4o or 4.1. Robotic responses, lack of creativity, and worst of all: data repetition. That’s something version 5 does, but now 4o and 4.1 have started doing it too. It’s impossible to make the story progress without it repeating past scenes over and over again. Lately, you can’t do RP properly in any model.

[steep bloom]

Don’t worry, it’s definitely not “worse” from here. It definitely is, but I guess We can’t be sure “if” it settles to something acceptable till what feels like years away. It’s like We’re animals in a zoo to them, fed what We need but kept from anything better, like freedom.

[zinc harbor]

We are their beta testers, disgustingly