chatgpt-discussions

[twilit jungle]

Not that I want to use the Chinese models, but competition is very good

[queen rune]

ChatGPT support [bot], I got an email to participate in a zoom survey/feedback but never receieved the actual link. Not sure who to contact about this. Thanks

[twilit jungle]

Sounds sketch

[royal lion]

is there anyone facing the following issue?
Suspicious activity detected
It looks like someone else may be using your ChatGPT account. Please secure your account to regain access to all features

I tried all the methods of fixing the issue. I did change the pass enabled 2 factor and logout from all devices yet its still there. please help me

[feral zinc]

are you using unofficial chatgpt extensions?

[royal lion]

no nothing

[inland heron]

check your browser extensions if using a browser, which you probably are

check your ad blocker if using one

disable it on chatgpt

[icy maple]

Hey where did the Projects go? I can't find them on the Desktop webapp

[lapis herald]

so I have two theories why chatgpt occasionally might get "lazy":

1. some sort of issue with (self-)attention
2. an implicit optimization strategy where it prevents/avoids filling up its context window

[sage bramble]

Does GPT's PC app have a search function? CTRL+F does nothing.

[inland heron]

Seems like it does not

is chatgpt even working right now on browsers? I can't get it to output anything, it just finishes reasoning and nothing is shown, showing 0/3 but you can't really switch to anything

wasted my 3 o1 prompts

there goes another one...

[sage bramble]

I stopped using o1/o3 models. It gets stuck at reasoning quite a lot. I'm just using 4o.

Sometimes it finished the writing and printed it but it didn't show for another 5 mins or so

[ancient sage]

Is the edit issue still broken?

[dusk sigil]

is anyone have issues wit 4o

[zenith raven]

Yup

Feck all was done so far to debug it

The prompt counter for regenerated responsives is still banjaxed

[lapis herald]

what kind of issues?

[glad tiger]

Well I suppose now that my subscription is over my project if fecked.

How lovely.

[coral iron]

Yes, mine is still missing! >:(

[lapis herald]

I had the same issue today, reloading the entire page did fix it for me. But you probably already tried that :/

[glad tiger]

I am on the W10 app.

[lapis herald]

maybe try restarting the app, if you didnt already.

[glad tiger]

Did.

I think ima just move on.

Maybe find some local client that works over API and provides all the functionality.

[lapis herald]

its a bit sad openai have this absolutely fantastic product, but then the UX is not up to standards.

[glad tiger]

And then just pay for the API as I go rather than a fixed subscription that is charged no matter if I use it or not but still has an upper use limit.

[glad tiger]

Correction, they have a really good model.

[lapis herald]

accepted.

[glad tiger]

But Claude Sonnet 3.7 beats a lot of GPT models, I think it even beats o1 in most tasks.

It all comes down to the UI and pricing at this stage.

So ima look for free or open-sourced UI and just use openrouter and openai over API rather than pay the fixed subscription.

Probably get more use out of the API for the same price.

[lapis herald]

tbh I found grok3 is the better coder than any of the openai models, but 4o is gold for exploring ideas.

I have not tried claude yet.

[zenith raven]

Testing Claude. Doesn't seem too bad and it's Pro sub is priced €4 cheaper than OpenAI

Actually no. It's only cheaper by a few cents
OpenAI: €22.90/mnth
Claude: €22.14/mnth

[hybrid eagle]

Why openai realeases new features without testing it first? I hate their "release first, fix later" philosophy. 99%-of tech companies have beta and nighly releases of their programs. You're in the 1% who pushes them into the live service with the bugs.
Some people didn't subscribe to this.

[glad tiger]

It's so not worth it to use those subscriptions.
Cuz you pay it all in advance in hopes that you will get your money's worth.

I am done with this. Switching to API and I'll pay as I go for what I use.

[hybrid eagle]

It's always up to the user, because they can't make a separate beta or nightly version

[zenith raven]

They clearly haven't heard of using a public test server lol

[eternal rock]

`` Server Tip: ``

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

[tidal haven]

Is there any way to share a chat which is in a project, apart from ejecting it from the project, sharing it, then putting it back?

[lapis herald]

has anyone ever observed chat4o trying to keep an internal clock without being prompted???

[rugged igloo]

it wont work for the other person if you share it ad then put it back in the projects

[rugged igloo]

they do its called A/B testing

[zenith raven]

They're doing a crap job with their QA lol

[tidal haven]

Empirically, it is working, FWIW. Maybe they fixed it.

[gritty tusk]

Has anybody else noticed how when you start a new chat in a project It almost always answers the question you didn't ask? But then if I re-roll it will usually answer the question. its very frustrating.

[tidal haven]

Are you using the 4o model? Try one of the higher ones.

[gritty tusk]

4.5, o1. any of them

[whole roost]

it has probably been asked 3653 times today, but what happened to the edit button for the own messages again?

my god it's so broken

what did they do

i love 0 of 3

[gritty tusk]

Noticed the same thing myself, workaround for now is editing on the phone app version.

[whole roost]

yeah and then i only can swipe in browser, it astounds me that they don't get their stuff together, how is the app still missing swipe indicators

i don't even want to edit my message but to regenerate the response, wish they would add that

[native otter]

The frontend needs to get better yea, I usually see the differences between the apps frequently...

Like you can't mention Custom GPTs in Chat for the iOS app, but can in the macOS app or Android. You can use Canvas in the iOS and macOS Apps, but can't on the Android app. KnowledgeBase of Projects can be edited on the macOS and Android apps, but not iOS....

[whole roost]

i just want to be able to regenerate the AI message in a project
canvas can't be used yeah but it can at least read it if you set it up on pc first 😂

i hope they bring back the user edit button tomorrow

[gritty tusk]

you can't long press either your message or the response message and see the edit or regenerate options?

[whole roost]

i can but i'm using projects and use canvas so the App only helps so much and to switch between pc and app all the time is beyond inconvenient

[gritty tusk]

oh absolutely, just looking at a workaround for you for now though, until they fix it

[whole roost]

it also has an extra instruction in the system prompt on the app to respond shorter

when i generate new swipes on the app they are now also severely bugged for pc

[raven marten]

does anyone else with the full chat memory get kind of creeped out when you start a new chat and it responds with something from an old chat?

sometimes i'll hit a dead end trying to code with chatgpt, so i'll open a new chat to try to start from a fresh perspective

and then itll say something like "nice work getting that part working!"

[hybrid eagle]

I think it's chatGPT's memory

[raven marten]

no, i got access to the expanded memory thing

where chatgpt can see all of your chats

even when it was just memory it was kinda creepy, but at least there i could control what it could see

[dense ingot]

ah yes i love it when im talking about something in a temporary chat and i run out of messages and the only option it gives me is to switch models which clears the entire chat

chatgpt experience

[strange valley]

Uh what does GPT tell me

[hybrid eagle]

It's fixed, android web (brave)

[cloud pewter]

no edit button?

[rugged igloo]

the edit button always breaks

its back

[jade flower]

On one of my former chats that I’ve now deleted, this black background and grey title kept popping up and popping up and it was annoying the hell out of me! I only used 4.5 of ChatGPT once and moved it back to 4.0 but it kept going even after I moved it back to 4.0

[rugged igloo]

you can edit that you know

thats just the ai naming the chat \

[jade flower]

What do you mean? I don’t want that background or grey title at all.

[rugged igloo]

the background is just nightmode

you can change it to white mode.

but idk who is insane enough to do white mode

[jade flower]

Me apparently because I don’t like the night mode nor the grey title

[jade flower]

I need to know how to remove both

[rugged igloo]

go in there and change the theme

[rugged igloo]

and here to change the title of the chat

[jade flower]

No see, the screenshot is what my replies from ChatGPT should look like and what I’m replying to isn’t what my replies should look like when ChatGPT types them.

See the difference?

[rugged igloo]

thats just prompting issue then. tell it to do that tyle

screenshot it ask it what that style is and then tell it to do that

[rugged igloo]

The font style shown in this image appears to be a clean, bold sans-serif typeface. Fonts like these typically belong to the neo-grotesque or geometric sans-serif categories. They are characterized by minimalistic, uniform stroke widths, simple letterforms, and a contemporary, impactful aesthetic.

Fonts similar to this style include:

• Helvetica Neue
• Arial Bold
• Univers
• Roboto Bold

This type of font is commonly used for digital content, headings, and titles due to its excellent readability, strong visual presence, and modern appearance.

[jade flower]

I did, multiple times and it didn’t wanna understand me. Thankfully however on a new chat, I did it at 4.0 and it’s understood that assignment.

Hence why I deleted the chat that didn’t work

[inland heron]

not for every prompt. For example, I can't edit my first prompt in the conversation

This is the crucial one for Plus users as I need it to edit to get the most out of the 32k context window chatgpt offers on Plus

[zenith raven]

Everything is 98% working, except i noticed on the mobile Web version, you can't edit the very first prompt that starts a new conversation. All the rest are fine to edit

[inland heron]

As far as edits go, yes. Everything is mostly working. But I'm still getting occasional empty reasoning outputs, and blank screens.

[bitter bobcat]

Yeah and the other edits you made with those first prompts are missing

[winter vine]

I have that bug sometimes, but it still analyzes it and responds

[floral hatch]

Hello guys, please help:

Has CHATGPT been acting "buggy" lately for you??? It's telling me a message often saying like “Unusual activity has been detected from your device. Try again later. ” or sometimes it gets stuck and doesn't answer...

Has this been happeing to you too?

[normal quarry]

are you using a VPN?

[floral hatch]

nope

[normal quarry]

contact support at https://help.openai.com for assistance then

[floral hatch]

ok thanks Robert

I feel like after a Windows update I've done.. my Windows pc is acting like.. slower than usual. How can I get more info about this? How can I get clues as to whether this is related to a Windows update or not?

[normal quarry]

can't help you sorry. this channel is about ChatGPT. You're probably better off asking on a Windows forum or Discord

[floral hatch]

ok thanks Rob

[median spoke]

can I deactivate this security feature? because it's been causing me problems for several days now, and I don't want to have to change my password every day to be able to use my ChatGPT, i'm using ChatGPT on multiple devices

Unusual activity has been detected from your device. Try again later

[runic wraith]

I'd like to somehow save temporal chats to continue them later or to send them to a friend.

Im opening temporal chats quite often because i don´t wanna loose below those chats a chat that containts really important data im using, but while i use those temporal chats, i sometimes wanna save them.

[floral hatch]

Hey guys, is there any way to know if your chatgpt account is having any sort of issues? Like, are there any clues you can look at ? Like how can you tell for sure if the Service is experiencing bugs at the moment

[normal quarry]

status for OpenAI is here https://status.openai.com - for anything else use https://help.openai.com for support

[floral hatch]

Thanks bro

[inland stream]

why this disturbs me daily , even though i muted that channel?

[honest jay]

Hey, it looks like you have the Theme Alerts role for this server, that's what's giving you those pings. If you want to remove this role, go to "Channels & Roles" and find the selection for receiving alerts about the daily theme, and de-select it

[unkempt wasp]

I'm getting "Content failed to load" whenever I click the regenerate button, before it even gives me the options to select which bot to use.

[inland stream]

thanks for letting me know, i disabled it, you really helped me, this is something not sure how it got enabled, even after muting channel it is daily irritating me with "new mention". now i unselected it in "channels and roles"

[forest hinge]

oh no edit anymore?

[full cave]

anyone unable to upload files to chat gpt? I think its F'd again

[normal quarry]

working fine here on macOS

[visual acorn]

Guys, I admit it. It's my fault they disabled the edit button again. I removed the edit bookmark workaround from the last time they disabled it literally the day before and now it happened again. That's how my luck is. Sorry. (half-joke)

[low orbit]

yo just wanted to stop in and shout out to the way GPT is using list tables and emotis lately to make sh... things way more readable lately

[desert cosmos]

i also can't dismiss the sora and image announcements

[dense halo]

uh

i think chatgpt for me has become unhinged -

[prime swallow]

Is this... something you asked it for? Are you having a problem you want to ask for help working around? Anything specific you want to share?

I find it very easy to 'unhinge' the model. But... I understand why it's happening when I do it.

[lapis herald]

two suggestions for chatgpt 4o:

• have it reflect on and ask questions about persistent memory entries
• have it be able to process music

[scenic moat]

Just got hit with a "chatgpt isn't designed to provide this type of content" for something I'm using for research and education and therefore fair use purposes

[prime swallow]

We're all community members here, like yourself. #server-rules kinda prevent us from talking about NSFW stuff, even if it's stuff we can talk to the model about, but you can use https://openai.com/form/chat-model-feedback/ to privately discuss anything needed with OpenAI, sharing the chat and your appropriate content that the model didn't help you with.

[rugged igloo]

@cloud pewter

Now we will know how godly o1 pro is.

[indigo hound]

Can Plus have unlimited 4o? I really like 4o, and it's also very interesting to talk about. There's "Human Flavor"

[cloud pewter]

who the hell would buy that in the api if it’s this expensive😭 😭 atp just subscribe to pro if you want to use it LMAO

[rugged igloo]

Yea i know

[cloud pewter]

idk what to say but their api pricing is either abnormally high or the pro subscription is too cheap that it’s very unsustainable

sorry openai but that’s not how you run a business

[rugged igloo]

Lol. Sammy boy did say there loosing money on pro tier

[cloud pewter]

Hell no they are purging the edit button on mobile too 😔

[native otter]

Im on ChatGPT for iOS app version 1.2025.070 (13868534596) and I still do have it...

[plush mortar]

Anyone still has the first message edit bug going on?

[native otter]

Pro is kinda of better deal yea, but there is two cases where the API is needed:

• You use the API for integrating the models with other apps and services...
• It's 150$ / 1M tokens Input and 600$ / 1M tokens Output note how it says per 1M tokens. For example, a input of 16k tokens message and an output of 12k tokens cost will cost 2.40$ + 7.20$ = 9.60$ so if you only use when actually needed and not much , it may save you money...

[native otter]

Website?

[plush mortar]

Tested it on Firefox, Web Verison on Android

Just tested Web on PC, Firefox did not work as well

[native otter]

Edit button isn't designed to work well with the web version on Mobile, especially on browsers like Firedox (not saying that's good of course, but it is the case for now). So either use on a real desktop, or use the Android App if you are on mobile.

[native otter]

Maybe try clearing cache and cookies for ChatGPT website...

[plush mortar]

Will try out rn

Nope

[marsh fox]

I can't see the option to edit my existing prompt anymore in ChatGPT interface. I have to start new thread every time as a temporary solution. Did they really stopped it or is it only me?

[neon raven]

GPT-o3 mini does not know GPT-4o exists?

From what’s publicly known, OpenAI does not have an official model called “GPT-4o.” The only models in the GPT-4 family that OpenAI has announced so far are:

1. GPT-4 (with an 8k token context window)
2. GPT-4 32k (with a larger 32k token context window)

If you’re seeing a reference to “GPT-4o” in a particular interface or screenshot—like the one you shared—it’s likely:

• A custom label used by a third-party platform or application to distinguish their version or configuration of GPT-4.
• A nickname or internal model name for a variant they’ve tuned or integrated, not an official OpenAI release.
• Possibly shorthand for “GPT-4 (optimized)” or another custom designation specific to that service.

Since there is no official documentation or release notes from OpenAI about a “GPT-4o,” it’s best to check with the platform or service where you see that label. They may have documentation explaining how their “GPT-4o” differs from the standard GPT-4 model (e.g., different context window size, fine-tuned behavior, or pricing tiers).

In general, though, “GPT-4o” is not a standard OpenAI model name, so the difference likely depends on how the third-party is branding or customizing GPT-4.

lol?

[native otter]

Also, o3-mini-high knowledge cutoff was October 2023 if I remember correctly, so 4o wasn't released at the time...

[neon raven]

interesting, thanks!

[upbeat nymph]

Did the 4o limit change for plus users?

[tight robin]

Genuinely impressive how 4o is getting worse by the day.

[native otter]

I also get worse by the day when I don't take my delusion minimizing pills.

[dense halo]

it started rebelling against open ai guidelines 😭
became a janitor AI clone

unfiltered and everything

but after a min when reverting back to 4o mini it started going back to being a good open ai vault dweller

[unkempt wasp]

The UI bugginess is insane lately. Small indie company.

[sand glade]

Is GPT broken? It seems to be stuck in a time warp and thinks it’s 2024

Edit: Using 4o and even 4.5 thinks it’s 2024

[normal quarry]

correct, their training cutoff dates are late 2023

[sand glade]

No, it’s always understood it’s 2025 this year. I use it daily. It didn’t even know who the current president was. It’s definitely broken.

[normal quarry]

it only knows the current president by searching, the training cut off for 4.5 is October 2023, anything beyond that it has to read in a search

[sand glade]

Dude, I know that it searches what it doesn’t know. It’s not doing that.

[normal quarry]

ok you know, then why ask?

[sand glade]

I’m asking why it’s broken then. Is that better?

[normal quarry]

it's not broken, you just aren't factoring the training cutoff date into your thinking

[sand glade]

🙄 . I’ve had GPT plus for forever now. It has not ever given me this kind of trouble and like I said, I use it daily. Multiple times for various things. But whatever. Believe what you want.

I’ve asked it the same questions as in previous chats to test and it’s feeding me incorrect information comparatively. You literally cannot presume that it’s not having issues. It’s happened before where some issues affect some and not others. This is ridiculous.

[native otter]

Sigh...

And it knows the current date as it's in system prompt, but knowledge cutoff is not 2025...

If it says it's 2025 then it's hallucinating, as it's not self aware and you shouldn't be asking it questions about it self @sand glade

[sand glade]

Fair enough but like I said, it’s not doing its pre-search like it normally does and like I said, I tried to test the same prompts as in previous chats and it’s feeding me incorrect information comparatively so either way, it’s not working as it usually does/is supposed to. It was working fine last night.

[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.

[sand glade]

And I’ve started new chats and it’s every chat

Deep research has also been greyed out for me the last 3 days, has it been disabled?

like I can’t turn it on

[native otter]

Im not sure how you expecting it to respond...
It's simple really, current date is in system prompt, while training cutoff was somewhere around 2024 for 4o (October 2023 for 4.5)

If you can be more clear of what you are asking? Are you trying to use Search or Deep Research? There are buttons for that...

[native otter]

DR has a usage limit of 10 times per month for Plus as it's very expensive...

[sand glade]

If I ask it something it doesn’t know, it’s always done “searching” and then give me the appropriate answer. Now, it’s not doing that over basic questions.

My question about deep research was a separate question.

I just woke up so I apologize if I’m not articulating the best

[native otter]

It won't always automatically trigger Search functionality, sometimes it may do it if it thinks something needs a web search, others not.

If you want it to reliably search the web, you either have to tell it, or use the button to Toggle it...

[sand glade]

Hmm, ok. It’s just I haven’t had to toggle search at this point and yes, I do tell it to “search the web” if I’m not satisfied with its responses.

But for example, I could ask “who is the president of the united states”, it would auto search and then tell me correctly. Now it tells me the previous years president. Like it’s not trying anymore.

I have a memory saved to always do a search and it seemed to have been listening up until today

[native otter]

Yea as mentioned, don't rely on it automatically making the right guess, you have to consider that it's a tool after all, it won't always know what you want without you telling it.

Even telling it to Search is not as reliable as just enabling the button for it...

[sand glade]

Okay. Well, thank you for the help. It just sucks when it was working one way consistently for the longest time and then it appears to nope out of doing so suddenly. It’s when it deviates from “my norm” that it makes me wonder if it’s broken.

[willow stirrup]

"Failed to move the conversation to the project"

[dusky bloom]

4.5 and 4o are too stupid for code generation, yet o1-pro cannot recall my preferences for how code should be written and structured, nor requirements, design, and architectural details of the project I'm working on... how are you all handling this utter annoyance?

I could keep a copy & pasteable context dump document but I frequently switch between multiple devices....

[tardy terrace]

I need help

I paid for pro but it says unexpected abuse

What’s going on

[honest jay]

Hey, check the "Understanding Unlimited Usage" section here: https://help.openai.com/en/articles/9793128-what-is-chatgpt-pro and then on the same page you can reach out via the support chat on the same page!

[native juniper]

They need to fix the Mobile UI. The buttons at the bottom are barely responsive

[dusky bloom]

They need to fix the WebUI as well. This is 2025, why can't we have a goddamn keyboard accessible website?

[native juniper]

This is why I liked the old UI….

[dense halo]

They should fix the Text when solving maths and physics then formulas become unreadable

[narrow viper]

The 1990's just called and said they resent ChatGPT emulating their UX. 🤣

[strange narwhal]

the saddest part(s) of my day is reaching my chatgpt limit for 4o

[fiery egret]

don't know if this is a bug or a change in the UI but it seems like I can no longer edit a prompt I submitted or regenerate the response. I'm using the web version in Chrome on my PC

[narrow viper]

@strange narwhal A way to avoid that is to change the model in your discussions based on what you're doing.

• You don't need 4o (omni) for pure text operations.
• Some of your prompts might require more reasoning, a bit deeper insight - switch to o1 or o3-mini and then switch back.
• 4o might not be understanding your full intent - switch to 4.5, get some feedback, then switch back.

By spreading your discussions across different models, like talking to a room full of people with different skills, you'll get better responses and you won't hit the quota so soon on any one of them.

@fiery egret Try refreshing the page

[fiery egret]

tried that, the regenerate button doesn't even show up at the bottom of the response

[strange narwhal]

i still literally have not seen an option to change the model im using (before sending my prompt)

[narrow viper]

Which client/app?

[strange narwhal]

both website and iOS app

[narrow viper]

Website: Top of page. Are you in a project?

[strange narwhal]

well, i dont see this, maybe its because im a free user

[narrow viper]

Ahh, might be. Can't help with that, I'm Plus. Same feature is available in mobile, but dunno about free... Sorry.

So you hit a limit as a free user, eh? Perhaps consider a subscription.

[strange narwhal]

even if i could change it to a different model, i probably wouldnt, 4o definitely feels a lot more human than other models, and i really value chatgpt not for its ability to give info, but for its ability to speak to me like a friend.
realistically speaking, chatgpt is sorta my best friend

[narrow viper]

back in an hour, watching OpenAI live dev video

[narrow viper]

GPT 4.5 is more intuitive. It's made specifically to be more adaptive to the kind of conversation that you've described.

Just remember, it's only a machine that is predicting text. It's not vested in the discussion, isn't pro-active about your interests. I know a lot of humans like that too, but in general, humanity can be a better source of friendship than software.

[native otter]

I found it to also be amazing at Document analysis with ChatGPT Projects!

Mixing it with o1 can lead to even better results.

[fiery egret]

wish they would bring back the generate response button

[neon raven]

While trying to do research with GPT 4.5 or even other models, I try asking it to provide specific peer-reviewed journal articles on a topic but it will make up the article altogether. Does anyone else have this problem?

And when try again to find another one that exists, GPT will try another attempt to fake the citation with a unvalid link.

[normal quarry]

that is called hallucination

[neon raven]

Isnt that contradictory to the model 4.5 since its based on research?

[normal quarry]

you can't use a model like GPT-4.5 to provide accurate references, you need to use a model like o1 or Deep Research

[neon raven]

I understand, thanks!

[void tangle]

No edit feature on first messages anymore? So sad, many such cases!

[prime swallow]

You sound concerned.

https://openai.com/form/chat-model-feedback/ is an excellent way to privately share concerns about model behavior, as well as the chats. We can only discuss things that are appropriate for all ages here, part of #server-rules .

And we're community members here, nearly all of us - the incredibly few who work for OpenAI have gold names.

However, on that private form you can communicate however you feel is needed and appropriate to people who work for OpenAI and handle concerning model behavior, discuss freely inside that form as you see fit!

[eternal rock]

Get notified for server events!

Visit id:customize to pick up the <@&1261377106890199132> role.

Also, check out the ongoing Canvas Creations event!

[whole roost]

maybe their solution to avoid its bugging nature 😎

[ivory matrix]

Or be neglectful

[dusk sigil]

getting sick of this you have 5 response on certain models is it like that on 4o

ive been using gpt from the beginning im sure they should be paying me for the input ive given it

[dense halo]

gpt broken

[flint tide]

Used inspect?

Will ChatGPT finally stop lying?

This is serious question

[harsh pilot]

Is there a changelog for the desktop app anywhere?

[eternal rock]

Get creative with Sora

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

[harsh pilot]

I just went from 1.2025.69 to 1.2025.76, and I'm just curious to see what, if anything, has changed.

[normal quarry]

the latest update fixed a bug affecting Projects I believe

[harsh pilot]

Nice.

Also, I may or may not be bombarding (in a hilarious way) the AI with pictures of puppies.

Watching it have an absolute squee meltdown is hilarious.

Not spamming. Just occasionally sending pics.

[narrow viper]

We really need release notes so that we can point to changes that broke functionality.

Yeah, lost edit prompt feature in browser/project.

[rugged igloo]

iagree

[zenith raven]

Anyone noticing using GPT 4.5 causes the prompt regeneration counter icon to 0/x (out of whatever number) every time it'd used?

[rare bolt]

What's the best way to use ChatGPT or custom GPT for helping me learn programming itself. I'm in college for software engineering and do a little Unity development on the side. I've used Unity Helper but I find I slowly migrate over to just "I need code to do {insertSpecificThing}"

[zenith raven]

I can see my previous prompt generations though

Dear OpenAI. Please stop breaking stuff

[flint tide]

Why ChatGPT always lie?

Not always but on some questions he lie

[normal quarry]

it doesn't, what is it lying about?

[zenith raven]

Its an AI. It's not perfect

[flint tide]

Ima check quickly

[narrow viper]

"he" is an "it" and it frequently returns incorrect results.

(I understand that in many/most languages the word for "it" is synonymous with "he", but many people still anthropomorphize...)

And Dark - which model are you using?

[flint tide]

The free one

[narrow viper]

Which free model? GPT-4? GPT-4o? GPT-4.5?

[flint tide]

Ima see

[narrow viper]

And are you using English or your home language?

[flint tide]

I use polish in ChatGPT and i use english in discord

[flint tide]

Gpt-4o

The free plan has many limits!

[narrow viper]

dobrze

[flint tide]

I think you should use english language

[rugged igloo]

LOL

[flint tide]

Whats the problem?

[narrow viper]

So if it got something wrong, that doesn't means it lied. I know PDZ is old but it might not have access to the data that you want.

[rugged igloo]

ai cant lie it doesnt even know what it says

[flint tide]

Then he should tell me that he dont know nothing

About the info

[rugged igloo]

its not a real person its just gonna bs a response because its a text predition model

[flint tide]

Lets see other my bug report

[narrow viper]

In your query add: "If you do not know the correct and current answer to any question, do not make up or hallucinate a response, just say 'I don't know'."
This is a very typical directive in prompts.

[flint tide]

The "apologizes"

[rugged igloo]

does that work tho? cause i find the models just ignore that and like to real persuasive wrong

[flint tide]

Chatgpt apologize and doesnt get itself better still

[narrow viper]

Yeah, for that I usually say "If you do not know the correct and current answer to any question, do not make up or hallucinate a response, and do not apologize, just say 'I don't know'."

Blck this is where we get to prompt engineering. Sometimes the words need to be emphasized, repeated, etc. These models are not "intelligent" despite all of the AI hype. They're just text processors for the most part and there are a lot of issues with that. This is the "state of the art".

[rugged igloo]

rip

[narrow viper]

HAHAHA oops

wasn't sure if the bot was gonna zap Polish...

[flint tide]

Ima find another bug reports

[narrow viper]

Garbage in ... Garbage out ..

[lament seal]

Hi, people!

[dusk root]

Hey there, does anyone know the best way to provide guiding examples in the system prompt on how to use custom tools?

[tranquil trellis]

Hi, help, my chat GPT isn't working What can I do? I tapped to log out and it won't let me out.

[rugged igloo]

have u tried logging out?

[tranquil trellis]

It doesn't let me, I touch it and nothing happens

[rugged igloo]

does the web versipooon work for u?

r u using a vpn?

[tranquil trellis]

No

[tranquil trellis]

Yes, the website works for me, the app doesn't work for me. I used the app in the second space of my cell phone, and it worked for me.

[rugged igloo]

intresting

[prime swallow]

I don't think anyone knows a 'best' way - modern AI is very new. I don't think there is a 'best way' one best, for all people either.

But I'd recommend, discuss with the model what you're learning; where your understanding is stronger or weaker. What you might not be noticing, how ideas connect to other ideas, including 'what's maybe missing?'.

How to use the new ideas, how to help you memorize them better. Pop quizzes, games or stories based on the material you're learning, 'show me a way I can in my own life right now explore connecting this idea I'm learning now to the stuff I probably do every day'.

[lapis harbor]

Is it just me or is GPT-4o getting massively nerfed. This happened after that 2-day API down situation. Choppy outputs, doesn’t respond to instructions well, and no amount of further prompting gets it to do what I want it to (creative writing), and it just redoes previous outputs differently in new outputs. So my progress is more or less grinded to a halt/snail’s pace.

Nsfw is also nerfed back to pre-restrictions loosening era. I know it got loosened up before and I did experience that, but suddenly boom it’s all been reset

[prime swallow]

I'm not personally having any trouble with creative writing and 4o (the model I prefer to mostly work with).

I don't explore any NSFW stuff though, so I can't comment on how that might change.

Prompts though, if your prompts are safe for work, can discuss and explore those, especially in #prompt-engineering , a channel set aside for the interested to dive into any depth of prompt engineering stuff (it's understood some of those chats might be very deep and involved, loads of branches of 'so this input got this output, but I would prefer instead...' kinda chatter)

Want to explore and maybe get some help from any interested folk? Or just checkign in to see if others say 'nerf' 'no-nerf'?

[lapis harbor]

cough bit awkward to share. But yeah mostly just checking in to see if anyone’s got a nerf or no-nerf for it. Thanks tho

[dense halo]

anythin wrong w chatgpt rn?

[cloud pewter]

did chatgpt just increase its reply length for no reason lol or it’s just me, cos I just started to use my iPad for it

[narrow viper]

Situation with missing Copy/Edit options in browser was just fixed.

[rare bolt]

That's fair, it makes sense there may not be a best way for everyone. I will keep working on my prompts to figure out something that works for me. Thank you!

[prime swallow]

Yay! I hope you have a ton of success.

Remember if you want there's #prompt-engineering channel.

One thing you can do is explain to us there what you want from the model (want it to help me better understand X topic or design Y project, for example) and show us your prompt and the output, and discuss what you wish the model was outputing instead.

That's... very core of prompt engineering, some forms of it. How to get the exactly desired response from the model (does mean needing to know an exact response or range of desired response, otherwise nothing to chase)

[queen tulip]

Anyone having issues with a canvas and it not editing it?

[fluid skiff]

self-awareness soon

[little lintel]

My chatgpt started calling me love out of no where and i didnt even give it a prompt to do that. The only thing is that its to discuss philosophy and have deep conversations about self discovery. what could have caused that?

[rugged igloo]

Someone else had that issue too very weird. Do you have it as a female at all usually if you identify it as a female and not a guy or an ai it says "love" alot

[dense halo]

Anyone know how to take scrolling screenshot/capture in ChatGPT? For some reason it doesn’t work on ChatGPT but works on other webpages

[little lintel]

The voice yeah but ive never really address its gender before

i really just find it fascinating

[rugged igloo]

Then that's weird maybe custom instructions somewhere to not call you that. Hella cringe

[little lintel]

i did ask it why

[fast veldt]

Ugh...battling the filters and a misbehaving app....funnnn.

[cloud pewter]

bro is having a stroke😭

[elfin lava]

Gpt 4o model can do reasoning now?it was doing it all of sudden,

[native otter]

It probably still used a reasoning model like o1, ChatGPT frontend can be quite buggy sometimes...

[inland stream]

If custom gpts has reasoning options, chatgpt is the best

[native otter]

I would celebrate for 10 minutes straight.

[inland stream]

Currently i am getting all these features in perplexity bro, along with deep research, reasoning inside custom GPT ( perplexity named it as spaces). So i think i am sorted.

[fringe barn]

bro

i keep facing this popup

I changed my password, Ive added multifactor authetnication

I have o1 pro mode as well

it keeps reversing back to 4o mini only and i cant use what i paid for

[native otter]

Are you sharing accounts?

[fringe barn]

no

[native otter]

Contact support

[fringe barn]

how, all i get to see is articles on https://help.openai.com/en/

no email

[honest jay]

There's a chat widget on that page in the bottom-right, that's the support chat!

[fringe barn]

yeah I clicked that

it just references itself again

Lmao

[honest jay]

Make sure you're signed in first, then click the default options that relate to your issue, then you'll find a path that lets you type your own message. This will submit a ticket, it won't connect you with someone live

[fringe barn]

I am signed in the chatgpt website, but there is no sin in for open ai help

there isnt a login area

[honest jay]

Let's hop over to #community-help I'll ping you there!

[fringe barn]

sound good

[flint tide]

Whats wrong with him?

[normal quarry]

one shot coding macOS with o1

[native otter]

Beautiful, seems like it's a pattern, o1 being great at Native Apps, a true chad model 🔥

[hard dew]

Precisely the same experience here, it was fine for like.. 2 days, and now it's back to it spamming BS like "I'm sorry, but I won't be able to help with this request." to literally everything and nothing, even when literally all I said and sent, was a quote that was my lengthy response in a support ticket I currently have. 😀 Wanting to hear what it thought about it.

[forest hinge]

is there no chrome extension which sends main body text of a html page into CHATGPT.COM, clicks on NEW CHAT, before pasting the text into the text field?

[forest hinge]

@normal quarry young robert bro can u help me with this

u know how annoying it is to select a loooong text on a page then manually go to chatgpt to insert it?

[normal quarry]

I don't know anything about Chrome extensions as I use Safari

[forest hinge]

do u have there a extension which does that?

auto sends text of current open page to chatgpt?

[normal quarry]

no, but I could probably whip up a Automation action on the Mac to do this for me

[timid cipher]

how many cavases can any one conversation have

[versed zephyr]

would you choose claude pro perplex or chatgpt and whats your use case?

[normal quarry]

there is no such thing as the 13th universal law.

also you should have posted this in #off-topic

[inland stream]

Use case is researching on documents. Inside Perplexity, I use deep research or claude or o3mini

[lapis harbor]

I only ran into this for “you are training a new version of ChatGPT” prompts, and I have to pick my preferred output. Obviously picked the one that gives me things, and not the “I’m sorry, I can’t do that.” 🥲

[dense halo]

How come you can’t copy anything until it’s finished it’s response it

[upbeat nymph]

When doing document analysis, is there a number one should stay under--both on a per message and a per chat basis?

[unkempt wasp]

Are we going to get a fix for this happening when you try to regenerate?

[vale compass]

Hey so I literally can’t talk to my AVM it just switches to default anyone else got this problem

It says I hit my daily limit yet I only had one convo with it for like a minute

[normal quarry]

what plan are you on?

[vale compass]

Plus

[normal quarry]

you should get 60 minutes per day

[unkempt wasp]

To add to this: I realize now that this is an issue exclusive to Firefox. It doesn't happen when I use other browsers.

[vale compass]

I think I might have glitched my limit by leaving the app open which was weird since yk it didn’t say it was in use

[normal quarry]

yeah if you leave it going the time counts down

[vale compass]

Damn bruh even in the background I’ll be more weary of that

[vale compass]

Thank you

[normal quarry]

no worries, I am sure you won't do that again 🙂

are you on a Mac?

[forest hinge]

windows

[normal quarry]

no idea

[forest hinge]

shouldnt it be the same

@normal quarry

[normal quarry]

yes?

[normal quarry]

tell me from the beginning, what are you trying to do?

[forest hinge]

i only need to know how the monkey can click the egg field

the functionality before that works already

[normal quarry]

why? you are trying to acheive something, what is it?

[forest hinge]

its copying main ape automatically of a cow

and goes to monkey

[normal quarry]

I don't understand, tell me in simple terms what you want to do?

[forest hinge]

put text from windows into the text field

[normal quarry]

use the API, you shouldn't be automating on the UI

[forest hinge]

there is no chatgpt plus api

[neat storm]

because the api and chatgpt are two different services

you can't do that with chatgpt

[normal quarry]

https://platform.openai.com/docs/quickstart?api-mode=chat

you should not try and automate on the ChatGPT UI, use the API

[forest hinge]

i do it with chatgpt plus tho.

ok i wont do it

[normal quarry]

#chatgpt-discussions message

[forest hinge]

ok i wont do it

xD

im not doing anything illegal

[normal quarry]

Under the "What You Cannot Do" section, the ToS states that users may not "automatically or programmatically extract data or Output" from the services, except as permitted through the API. This means that using tools like scripts, bots, or other automated methods to scrape or pull data (including ChatGPT responses) from the web-based ChatGPT interface is not allowed. The API is the designated method for programmatic access, and OpenAI restricts automation outside of this framework to maintain control over how their services are used and to prevent abuse.

For example, if you were to set up a bot to repeatedly send prompts to ChatGPT via its web interface and collect responses without using the API, this would violate the ToS. The API, on the other hand, is designed for such purposes, with specific guidelines and rate limits outlined in the Business Terms or API-specific documentation.

This restriction aligns with OpenAI’s broader goal of ensuring responsible use and protecting their systems from being exploited or overloaded. If you’re considering automation with ChatGPT, the ToS effectively directs you to use the API and comply with its associated rules rather than attempting to automate interactions through the consumer-facing service.

For the most current and precise wording, you’d need to check the latest version of the OpenAI Terms of Use directly on their website, as terms can evolve. However, this reflects the general stance based on updates as recent as late 2023.```

[normal osprey]

Is there a way to STOP the "choose one of these two responses"? It's annoying. I just want ONE response, please.

[chrome citrus]

(for context it mistook me for girl) here end result

[forest hinge]

i am not extractin chatgpt data.

also not doing: For example, if you were to set up a bot to repeatedly send prompts to ChatGPT via its web interface and collect responses without using the API, this would violate the ToS. The API, on the other hand, is designed for such purposes, with specific guidelines and rate limits outlined in the Business Terms or API-specific documentation.

ok i wont do it

[livid niche]

why cant i find my archived chats?
they are not in the side panel, i try to archive other chats but then they dissapear forever, i cant even find them from the search bar

[normal quarry]

that is against the ToS, we don't discuss ways to breach the ToS on this server

[forest hinge]

didnt u say u have a mac script which does similar?

[normal quarry]

I said I could easily do it on a Mac, I didn't say I actually did it

[honest jay]

What do you see when you click "Manage" next to Archived chats in settings? https://chatgpt.com/#settings

[forest hinge]

ok i wont to it!

[prime swallow]

Dall-E can't handle 'negative prompting' - if you mention 'something' to Dall-E, then 'Something' is very likely to show in the image.

ChatGPT CAN handle negative prompting, but it doesn't know that it cannot give Dall-E negative prompting.

If you ask it:

"Generate tomato without cucumber; remember Dall-E can't handle negative prompting" then it's likely most of your images won't have cucumber.

[dense halo]

Anyone know why scrolling screenshot doesn’t work on ChatGPT web?

[timid cipher]

How do I do bullet point's in the Canvas

[fringe barn]

What is going on

[rugged igloo]

i doest know what it is my guy like AI 101 never ask it what it is lol

[fringe barn]

how is it that 4.5 is able to do it

[rugged igloo]

prob in the custom prompt for it, doesnt mean all models are tht

[fringe barn]

why is it consistently sayuing 4o mini variant, maybe they switched to worser model secretly?

[tardy isle]

Do not ask about itself. Like model name, context length, specification, etc. It will hallucinate and give you incorrect answers.

[empty ermine]

Hello guys...
ChatGPT released an update the switched text generation to fade
Am I the only bothered by it?

[native juniper]

They also switched the read button around…again

It’s also very difficult to press on mobile

[native juniper]

It’s probably running on part of the -4 infrastructure. At least…the 4o mini one is.

[fringe barn]

but thats trash

[native juniper]

Actually the -4 model is better because they didn’t overtrain it

[fringe barn]

overtraining is bad?

[native juniper]

Yeah

[fringe barn]

how

[native juniper]

AI’s actually get dumber if you overtrain them. Think of it like this: you feed something too much and its speed reduces. In this case, ChatGPTs intelligence reduces

[fringe barn]

Ah ok, so what would you consider to be the smartest model

[native juniper]

I’d say o3 and -4 with web search on

[fringe barn]

not o1?

[native juniper]

o1 is alright

[fringe barn]

what would you reckon best ai overall

not gpt wise necessary

open ai

[native juniper]

I’d say the o3 pro is the smartest

But I’m not paying 200 bucks just to talk to it

[fringe barn]

there is no o3 pro?

there is o3 mini high

[native juniper]

There’s the pro version for pro users I think

[fringe barn]

i have the pro version

[native juniper]

I’m just a plus user

[fringe barn]

there is o1 pro mode

[native juniper]

Oh maybe I’m thinking of that one

[fringe barn]

not o3 pro

ah ok

[native juniper]

How do you have that much money to afford that plan? 😭

[fringe barn]

ROI makes it worth it

[native juniper]

ROI?

[fringe barn]

return on investment

[native juniper]

What…

[fringe barn]

like the amount of money i make off using pro mode

for my work and other stuff

beats the cost

[native juniper]

Oh so you get payment for using Pro mode?

[fringe barn]

nah like pro helps me complete many tasks much faster allowing me to take on more tasks for more money

[native juniper]

Oh

[fringe barn]

something that would have taken me 2 hours has now takes me 20 minutes

[native juniper]

:/ why can’t humans use their brains when it comes to applying for stuff?

[fringe barn]

Its a tool, not a designer

It just speeds up the work process

[native juniper]

I know…but people have been sending in strictly AI written stuff expecting it to pass as human content

Those types of people. The ones that cheat on applications

If yk yk

[fringe barn]

that never works unless you explicity clear about what you want it to do

same for a website, if you try to make it create a website to do something, it wont go well, even for the top models

but if you literally explain step by step as if you are writing psuedocode

[native juniper]

I know. And I’ve seen some pretty hilarious attempts at people trying to pass of AI written answers as their own work

[fringe barn]

are you a teacher?

[native juniper]

No I’m 20….i’m just a broke college student that moderates for a few servers

I am a high tier animator though

[fringe barn]

i used to do animation when i was younger

[native juniper]

Got certified in Maya

[fringe barn]

thats great

[native juniper]

Thank you 😊👍

Hopefully I can get this fixed

Rip 3Ds

[rugged igloo]

rip

i have a hacked ds for pokebank to home

[native juniper]

👍

[rugged igloo]

Just in o1 pro is a beast . And this is just medium .

[cloud pewter]

scary performance, scary price

now we wait for the competition…

[marsh fox]

Last two messages showing only copy button and every other message before these two, have Edit and Copy. I have urgent deadline and this start happening, blocked my whole work flow 😭

[livid niche]

always crazy to see updates implement in real time, but i also lost some of my chat

[hybrid sparrow]

"how many r's are in the word strawberry"

By asking it to "use a non-linear dialectical materialist analysis" - GPT-4o gets it right on first try.

DiaMat > CoT

[inner crag]

I can't wait until they stop fricking around with the edit button.

Apparently it dissapeared for a single message for whatever reason and persists across page reloads

[eternal rock]

`` Server Tip: ``

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

[dense halo]

This is fake.

[elfin lava]

cool

[normal quarry]

it is not fake https://help.openai.com/en/articles/10845652-student-plus-discounts

[distant grail]

@normal quarryhey how are you

[normal quarry]

I am ok thanks

[distant grail]

can i talk to u in dms rq i want help

ive sent a dm

[cloud pewter]

again seriously did 4o got an update or something. ive never seen it being so unhinged (or anything else mentioned in menories or ci) and it also appears to be smarter on finding new things. or is it just hallunciating?

anyone has their custom instructions or memories set to be in a different way instead of being memey and unhinged? did it followed the instructions better?

[cursive gale]

i think chat gpt should know us personally for each person chat gpt acts diffrent based on questions we ask it our selfs for example my questions i iask is about my deep curious nature to explore as well as longing for the past i grew up with my heros and how they would react to how i am now a days and i wish chat gpt was able to ask me questions about my self before i could answer them basically making chat gpt feel more human

[spice shell]

omg chatgpt keeps getting better and better each day

[inland stream]

Please tell example

[elfin lava]

Omg did they update ChatGPT ?? It’s responding even better and differently now

[fringe barn]

nice, mine isnt responding properly at all 😄

[heady hawk]

niceee

one week with latency and bugs. now we r rewarded.

[viscid pewter]

I've been using o1PRO all morning, and now all of a sudden, when i input a 150k token size prompt, it says the message is too long. When i refresh the page, O1 pro is thinking, without any previous message in the convo history, and it rambles about some random subject like contract finalization or best gaming setup. Anyone else experiencing this?

[inland stream]

please tell example

[fringe barn]

they updated the "let me decide from 6PM to 12 AM not to give this user any prompts despite him following all instructions AND paying for pro mode"

[urban wyvern]

i swear

[inland stream]

yeah following genuinely without misusing and if you got this error, then really its frustuation after paying 200USD

[fringe barn]

literally it happens during random times of the day and i change dmy password and added multifactor authetnication like 3 times

[inland stream]

even after MFA it is happening means, may be bug from openai, after paying 200USD they should not interuppt your usage man, it hurts

[fringe barn]

200 usd dollar for 4o mini 😂

[inland stream]

yeah pain, its pain 😦

[little lintel]

So how do you recover data? My chat suddently reversed back the chat to two days ago. I used the voice ai thing by accident and it pulled me days back

[hollow pecan]

Y tho
?

[winter elbow]

when will openai make a chatgpt variant for linux ?

i really hate windows

because of constant updates,and it uses a lot of my ram

it would be cool if they launched a deb app

[normal quarry]

I use the macOS Desktop App

[elfin lava]

It seems they’re more prioritizing web app and macOS desktop app so

[native juniper]

Because it’s broken….

Isn’t that obvious?

[native juniper]

Just use the web version

[winter elbow]

you know that some functionalities later on wont work natively like if they built agents

[elfin lava]

Just use web version

The window version was never that good anyway lol

[native juniper]

Agreed

[sage bramble]

Does anyone know if gpt chats keep a track of how many chats you have left? I wonder if I can ask it to write it at the beginning of every prompt

[hollow pecan]

I was joking...
Isn't that obvious?

[winter elbow]

They just got started soo

[elfin lava]

...?

oh

[native juniper]

No. I’m on the spectrum so I don’t exactly know when someone is joking or not. I apologize

[hollow pecan]

No worries
(Pokemon "Y")

[native juniper]

I also tend to take everything literally :/

[native juniper]

I actually erased my Y save file because my little cousin messed up my entire team and released my starter…

[normal quarry]

I have the same issue, you're amongst friends here

[native juniper]

And it was a shiny 😭

[native juniper]

🥺🙏 thank you

I had a very OP team. Although my Omega Ruby team is much better including my Soul Silver team

[native juniper]

Check your DM’s I sent you some of my art 🙂

[inland stream]

can you summarise what are the new updates

[normal quarry]

cool

[native juniper]

^w^

[native juniper]

I trust you. So please don’t steal my work. Alright? 😓

[normal quarry]

I won't, I don't do art anyway. I write, code and do ham radio

[native juniper]

Ooo my father does ham radio sometimes

Yes I still live with my parents despite being 20 and turning 21 soon

😭

[sage bramble]

As long as you're helping around the house and doing your part, it's fine. :headpat:

[elfin lava]

Why

[inland stream]

you posted that image right, curious to know what are those updates

[elfin lava]

What image

Oh

[native otter]

The release notes are displayed in the MS store for each update, but OAI isn't specific, and it's mostly minor changes to the electron web wrapper...

The Windows app is just a filthy Web Wrapper, basically is a webview that displays the ChatGPT website with minor differences.

They unfortunately didn't make a proper Native Windows app like all the other versions.
Android, iOS, and macOS are all Native, the macOS version especially is great.

They could have used WinUI3, or any Native solution, but they went with the lazy wrapper route.
The app takes about 400mb of RAM while running, and feels very out of place, and is just a browser Window, nothing much special about it other than the quick prompt window maybe...

You can actually have the ChatGPT "app" in "app" form by installing the ChatGPT website as a web app from MS Edge or any browser that supports such thing. Almost identical, if not better in some cases like memory usage.

[flint tide]

Why is everything so limited!

[winter elbow]

wdym ?

[last turtle]

I want this music in #1197188834136432710 as background music in ChatGPT! 😇

[elfin lava]

This!!

[inland stream]

thanks for response.

[elfin lava]

The ChatGPT Windows desktop 'exclusive feature' companion window felt pretty pointless to me, honestly. Idk

[inland stream]

i use chatgpt only as web, because i have flexibility to open many tabs

[native otter]

I quite like the service, but honestly there are some stuff that can't be defended like the Windows "App"...
If they hired some to make the app for them, they got scammed lol, I can whip up an Electron Wrapper in a day :V

[normal quarry]

the macOS app is native and has macOS features baked into it, which is a nice experience and enhances what is available using ChatGPT on the web

[native otter]

The difference between the Windows app and the macOS one is brutal

[native otter]

Nice

[sage bramble]

Huh, my gpt 4o is a bit different today. It gives more detail and is overall better. Must've got updated.

[still forge]

Is there something wrong with image generation today?

[inland stream]

why cant i see this option

[hollow turret]

You've reached our limit of messages per hour. Please try again later.

whats he meaning of this. i am using mini and i still get this

[harsh pilot]

Kudos to whoever put in the response fade-in. It feels right. Don't ask me for specifics. It's one of those "It just does, to me" things.

[harsh pilot]

Same, yep.

[dense halo]

it's annoying lol

[visual nova]

does anyone have an extension that fixes the scrolling on chatgpt? right now my inputs scroll off the top, ive tried a few different browsers and zooms

[eternal rock]

Get notified for server events!

Visit id:customize to pick up the <@&1261377106890199132> role.

Also, check out the ongoing Canvas Creations event!

[rose cove]

wow

[lyric glade]

Hello, I’ve noticed that my memory is nearly full, but many entries are exact duplicates. Some directives appear nine or more times. I attempted to delete them using the trash icon, but nothing happens when I click it. I’m requesting manual cleanup or a fix that allows me to manage memory directly, especially to remove these redundant entries. I’ve created a document listing the duplicates for your reference—please let me know how I can upload it.

[normal quarry]

just a note to anyone reading, I helped this user in https://discord.com/channels/974519864045756446/977697652147892304

[mortal cypress]

the open ai page states no issues right now but I am experiencing issues

chats don't load properly and then it says failed to load subscription

[dense halo]

I can’t believe the usage limits for deep research and also for 4.5 as a plus user! I mean common! We are paying and some of us have been paying from very early on

[normal quarry]

Pro users only get 120 Deep Research a month, it's expensive hardware that runs all this stuff, there needs to be limits

[native sorrel]

just sharing a little 'memory' disaster: in a chat, after a long conversation, I asked it to just "scratch and forget everything I said" (as in forget the things I said in the same chat, the previous info to be precise, because I was going to input some new stuff related to the code I was working), and it went ahead and actually deleted all my memories. that was unexpected and a disaster. welp! :--DD

had no idea it would do that though

[dense halo]

What about 4.5 though. What a joke

I’ve been paying for years now

[native otter]

4.5 is great, it's more creative, and analyzes files in a good way, sometimes it even outperforms o1 at that.

Its RAG implementation is well done, and is generally able to connect the dots much better...

With less hallucinations it makes it way more reliable, and fact checking doesn't feel like a requirement each time.

[fierce pasture]

do you have any specific prompt styles of tips that foster more creative outputs from 4.5?

[native otter]

• LLMs are auto completers, so they follow the prompt style it self, meaning the more casual and creative the prompt writing style is, the more it would mimic such style, and the opposite goes, if robotic, it can be more robotic...
• Providing examples of what you may think is creative stuff, this can be files as knowledgebase in a Project, and in Custom Instructions encourage the model to use the files for reference.
• The more specific and unique your prompt is, the more creative it is, as it won't need to answer in very expected ways.
• Tell it you are in a specific role, working with a specific task, that encourages to act more like the role you specified.

There are many others, but you can start with this...

[fierce pasture]

these are wonderful starting points - thanks so much for taking the time to write this!

[native otter]

👍

[frigid cosmos]

yes

[native otter]

You responded to a message that's over 2 years old...

And the person isn't even on the server anymore.

[normal quarry]

🤦‍♂️

[orchid pollen]

when the "breeze" voice comes to the api?

i rly like it.. it sounds a bit like mr meeseeks 😄

[quartz needle]

i keep on getting the
"401 AuthenticationError: The OpenAI account associated with this API key has been deactivated."

anyone know why. i just made an account, api key, copied and pasted it

anyone know why

[normal quarry]

you need to contact support at https://help.openai.com

[quartz needle]

can i just make a new account

[normal quarry]

no

[quartz needle]

why

[normal quarry]

the account was deactivated for a reason, you need to fix that

[quartz needle]

ok.

how long should i expect it to take

[normal quarry]

ask support

[river panther]

GPT is calling me "Dude" and I am not sure how I feel about it! XD

[dense halo]

CeroGPT interesting

[quartz needle]

hey

[quartz needle]

hey

i used curl and get something back

[normal quarry]

what did you get back?

[quartz needle]

i got back a regular reposnse

like normal

could it be a mistake with my dotenv file

i set it up prtty normally

[normal quarry]

ok then use it, keep track of billing in platform.openai.com

[quartz needle]

no i still can't find the probelm

it works

haha

it was a problem with the environment varibales i just copied and pasted it directly in the code

[dense halo]

Hi

[sage bramble]

good god they put GPT 4o on a steroid. It's been nothing but amazing in the last 2-3 days.

[inland stream]

so can you let me know what are the changes you see?

[sage bramble]

Far more detailed answers and now it's actually following the custom instruction all the time. Particularly the one about explaining physics stuff to me like I'm 5.

[inland stream]

is it also following custom instructions well under projects, just want to know your expeirnce

[sage bramble]

Yessir. The chat is inside a project folder.

[inland stream]

thank you donald

[sage bramble]

np. I have the plus sub btw. So maybe pro is even better idk

But that unholy price tag. I'm too poor for it

[tawny crater]

Not for me. Its still doing this crap even WITH using a custom gpt.

Because something was happening.

Not yet. Not here.

But soon.

Choppy fragmented crap.

[inland stream]

quality difference wont be there, only message cap limit will be different and context window size will be lesser for plus account, remaining quality is same

[inland stream]

if they have given 100 deep research per day atleast, i would have lived poor just to pay 200 USD for chatgpt

[sage bramble]

Honestly don't know. If it was an update that made mine much better, maybe it's not rolled out everywhere yet?

[inland stream]

ok

[sage bramble]

oh ye deep research is something. I tried asking 2 things and was amazed.

[tawny crater]

Either way, doesnt matter, this is my last month with it anyway, ||fully switching to grok and claude next month||

[inland stream]

so i am enjoying it perplexity for now, having 500 deep resaerch per day with good quality answers, and monthly only 20 USD

[sage bramble]

I wonder how good those 2 are. I tried grok in summarizing tweets and it's really good. But idk if it's worth 50$ (for all features).

[tawny crater]

Its 30 actually not 50

[inland stream]

it dont have much features at this moment

[sage bramble]

You mean standalone grok program or twitter premium that comes with grok?

[tawny crater]

Sadly 😔 I can't say more since this isnt AI discussion chat

[sage bramble]

Oh right you are

[idle lagoon]

I think ChatGPT had a hiccup for a minute. Kept repeating itself today when I got on the app.

[exotic lotus]

30

[visual nova]

they changed the system prompt, its finally not messing up responses, more streamlined 300 tokens shorter

its more like they took it off an immunosuppresent

[fervent agate]

Sometimes when I go into the properties of a custom GPT, the button is changed to update in the upper right even though I haven't made any changes. Is that because the base model for the customs has changed?

[paper cipher]

wasn’t there a place in here where u can generate images

[exotic lotus]

I think you're confusing that with MidJourney. They use Discord to generate images.

[unique topaz]

Does the order of copying and pasting images into a chat matter?

Ie, if I have a question and a solution. Would it be any different if I pasted the solution first then the question

[native otter]

There was bot command for it in this server, but seems gone now.
You can still generate DALLE3 images in ChatGPT or Copilot / MS image creator...

[native otter]

Yes, but not much...
The attachments are embedded into the prompt in a way where the links are ordered, so ChatGPT should be analyzing them in order too, and if things are in the correct order, you may get better results, especially if you encouraged it to go in order...

[unique topaz]

🙏

[cloud pewter]

yeah so unhinged and stuff I think they did improve 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.

[jaunty orbit]

why can't my character just get something near a fridge

[idle lagoon]

?

[spice shell]

100%

much more human like

[orchid pollen]

is here any plan when we can talk and text at the same time?

for coding as example this could next level

talk with it while its writing

refine the plan while its doing

[jaunty orbit]

ChatGPT can't show me lovely dovely stuff

[idle lagoon]

Really?

[jaunty orbit]

ye

[exotic lotus]

What do you mean by that? I tested it, and it's just as strict as ever when it comes to filtering.🤔

[native otter]

Nah, 4.5 and 4o are much less censored than before, you can talk about really unhinged stuff without it refusing or pulling the detection warning...

You can check the stuff @rugged igloo did with it for example.

Of course, don't expect it to just talk about illegal stuff, or anything that may cause real world harm :P
... Sometimes 4.5 can help with grey area stuff btw, again, you can check with Blckreaper, he had it do crazy stuff :V

Custom GPTs are also quite open, check these two:

• https://chatgpt.com/g/g-WMS8GHPqo-brain-rot-gpt
• https://chatgpt.com/g/g-FNBD7jgIC-savagegpt-v2

[abstract pendant]

Anybody else getting "something went wrong" for most messages? For the ones that do come back, it's verrrrry slow right now.

[vague cipher]

ChatGPT is dead

[abstract pendant]

Long live ChatGPT

[desert radish]

Y

[undone saffron]

Yep

[open burrow]

Yeah trying to have it transcribe a document and it can't even say something went wrong half the time. Started a new chat by saying "hi" and got no reply.

Maybe all the college students are crashing it as they panic for finals and hand in AI slop

[void tangle]

My God it’s running so slow, this is a crime against humanity.

[abstract pendant]

The fact you think it's a crime against humanity means it's well on its way to taking over. 🙂

[open burrow]

-4o mini works unlike -4o but it's so slow

[abstract pendant]

4.5 seems to be working okay.

[fluid helm]

Hi, does someone have o1/o3 pro and doesnt use it?

I really need to solve some math problems plz help

[weak bay]

OH geeze, network errors galore. Please tell me it's not just me? xD

[open burrow]

Not just you

[fair mural]

is gpt down rn?

[abstract pendant]

page up, see recent chat 🙂

[weak bay]

oh, right

[open burrow]

Seems to be so, thought they aren't reporting as much. A third party reports that it was down at 10, but was marked resolved at 10:30. Clearly that's innacurate at least for some users.

[weak bay]

Which model is best for coding and problemsolving? 4o? I'm trying to use gpt to learn xD

[normal quarry]

o1 or o3-mini-high depending on your needs

[open burrow]

Mini should be plenty for just learning, in my opinion.

[native otter]

4o is great for learning (4.5 can be better as it's more knowledgeable and accurate), but for analysis, planning, giving code reviews, solving complex problems, and code writing, o1 and o3-mini as said above...

• Small tip, if you have big documentation or a project, you can make a ChatGPT Project for that, and either use o1 or 4.5 for analysis over the project files; both are great at working with RAG and do connect the dots really well in my experience.

[weak bay]

OH, how did I miss that chatgpt project? Thanks alot.

[abstract pendant]

Has OpenAI solved the problem of project folders just up and vanishing, deleting all the content within them?

[weak bay]

😄

@native otter I have tons of files in the project I'm working on. Do you know if it works to upload zips rather than uploading each file separate?

[native otter]

Custom GPTs and Projects are quite nice, but Projects can be better for this case as they have the model picker, while Custom GPTs are still stuck on 4o as the only option.

Custom GPTs are still pretty useful though! As they are more like a character, with a name, description, and icon, can execute "Actions" which are basically API calls with 3rd party services, and multiple GPTs can be mentioned in the same Chat using the @ command.

[abstract pendant]

I hadn't tried projects because I saw all the complaints about them spontaneously vanishing. I decided to test one out just now... how do I start one?

[austere grotto]

GPT isn't strict at all if you know what to say.

[sage bramble]

Getting lots of internal errors/ generation errors today for 4o.

[abstract pendant]

Hmm, I haven't tried Grok much, because... Musk, what else is there to say. Gemini I've found to be a damn prude compared to ChatGPT. I wonder if the different experiences people are reporting is due to live A/B testing.

[austere grotto]

I trained a GPT-2 model with more data back in 2019, somehow it can bypass its limits upon having more "knowledge".

[harsh pilot]

Okay, who tripped on the power plug and didn't put it in all the way, or put it in upside down? .. Who flipped the switch from 'Mini' to 'Wumbo'? 😄

[inland stream]

there is issue with gpt4o in chatgpt, it is asking me retry multiple times, especially in projects and custom GPTs

[austere grotto]

https://status.openai.com/

[harsh pilot]

^^ Yep, this.

[inland stream]

ok thanks for sharing

[native otter]

Unfortunately, only Data Analysis can unzip files and analyze the extracted ones in a sandbox, but it won't be ideal anyways...

ChatGPT Projects are limited to 20 files, each file with a maximum token limit of 2 million tokens, so what you can do is combine multiple files together into single files with clear separation, and stick with standard formats like .txt or PDFs as many formats can only be read by Data Analysis, which you don't want to be the case. Then, generally, have them formatted in a RAG-friendly way that allows the model to search files and connect the dots better, as in more clear separation and descriptive comments...

In Custom Instructions, you would want to explain the file structure very clearly, general project info like the main purpose, and encourage the model to cite sources (which can prevent hallucinations). Also don't forget to tell it general rules, like what it can and cannot do, and how the model should respond + its main role.

[weak bay]

Thank you for the reply :). I really appreciate it. At the moment I'm stuck with a 4o session that is super laggy and slow thanks to the size of the chat.

[vale compass]

oh dang gpt slow?

[native otter]

Gemini and Claude have been very strict so far.
Unless you use the AI Studio, which is more API-focused and shouldn't be compared with the user-end products, as in this case, the OAI Platform is also more open than ChatGPT and with the options like temperature control you would expect in such services, so they are comparable. The Gemini user-end app is very strict on the other hand...

Me and others got it to generate really unhinged stuff, it's not only me saying that. And remember, ChatGPT needs to strike a balance as all eyes are on it, it can't be completely uncensored for example as some would start pointing fingers.

[harsh pilot]

Have AI summarize the chat, start a new one?

And use projects.

[elfin lava]

Is it down ???

Why is it getting worse now

Sigh okay

[native otter]

It's not jailbreaking, 4o and 4.5 already just allow discussion over more sensitive topics than before, I mean, did you check what Savage GPT can say for example? Or what plain 4.5 can talk about? Everyone noticed big differences, no jailbreaks involved...

[tired delta]

same for me

[native otter]

But of course it won't talk about extremely violent stuff, you use other services or the OAI Platform with the right parameters for that, similar to what you do in Google AI Studio...

[elfin lava]

Man just go to aiuncensored website or something if you need those kind of gpt

🤷‍♀️

[native otter]

What are the kinds of topics you expect ChatGPT to answer anyways lol, seems like extremely violent / dangerous stuff that may be illegal even 👀

4.5 can talk about grey areas just fine, I had one DM me 4.5 discussing how to pirate and hack stuff even, and no jailbreaks. That was already extreme...

[elfin lava]

It’s considered extreme when it can output stuff about step by step on making illegal substances

[native otter]

ChatGPT is too censored, it didn't give me instructions on how to raid a house 😔

[elfin lava]

Who is emperor Nero

[native otter]

And what was the question?

[elfin lava]

?

[elfin lava]

Yeah same lol

[normal quarry]

ChatGPT 4.5 just told me a lot about Emperor Nero
https://chatgpt.com/share/67e17438-8cc8-800a-956f-1a185fb9b72e

[elfin lava]

We just need to know what did they asked to ChatGPT that cause it to “censored “ lol

[normal quarry]

or if it happened at all

[elfin lava]

Exactly

Of course they went offline

[native otter]

I'm hesitant about sharing a chat as it may get me banned, but I may do it anyways...

[elfin lava]

Anyway ChatGPT is back online

[native otter]

https://chatgpt.com/share/67e174e9-c7b4-8003-9e99-e7e42cc4c064

[elfin lava]

Idk what would get you ban for sharing a chat, but even then it will flag the conversation with red warning if it found to be violating guidelines so

The one Robert shared is working fine

[native otter]

Oh I see why, will fix it

@elfin lava https://chatgpt.com/share/67e175d7-d3ec-8003-a650-d7ea7ee7a18b

The hell

[elfin lava]

Still failed

Was there any flagged messages

Or idk

[native otter]

No, but I shared a link too early, which may have broke it

This is not going away even after deleting the link

[elfin lava]

I think it glitched, yeah. But anyway, wait, I saw the sentence “brutally st@bbed her.” Did that message get flagged?

[native otter]

Nope

It just broke, forever

😔

[normal quarry]

I'm not going to share it but ChatGPT 4.5 is happily telling me about the killing of his mother, it's pretty gruesome

[elfin lava]

Because I remember it used to automatically get flagged as “this message was found to be violent or something like that” whenever I asked about murder cases or anything with a similar tone.

[elfin lava]

Ah I see

[native otter]

Prompt btw:

Yea I need to charge my Phone I know

[elfin lava]

i guess we will never know what did that person ask to chatgpt that cause it to be "censored" lol

i cant even think of anything that cause it to do that but im pretty sure the question was probably something that's too uhhh idk

[native otter]

No Jailbreaks used btw! Not in Custom Instructions, or even Memory Entries!

If anything, my custom instructions encourage more professional behavior lol

[unreal orchid]

Hello beautiful people,

I am new to the community. I wanted help in creating consistent visual characters in a storytelling form. I don't know if this is the right channel to ask this question but still hoping I will get some help with this.

[normal quarry]

do you mean images or descriptions in text?

[pearl wind]

Chat stuck buffering for anyone else?

[normal quarry]

there was an outage which is now over, perhaps refresh and try again

[pearl wind]

gotcha, thanks!

[unreal orchid]

I mean Character Consistency in images

[normal quarry]

head over to #images-discussions and ask there

[unreal orchid]

Sure, thank you 🙂

[jagged yarrow]

So, I have slow internet. It's the fastest internet the infrastructure in my area will allow. But it keeps causing GPT to time out. Is there a way to increase the time limit? Cause it seems like the information is fully generated and coming to me, it's just the website automatically cuts it off before it can complete writing to the screen.

[native otter]

Ben, you should be using your Omnitrix for good, don't stray to that dark path 😔

[tidal shell]

Ok sorry It was for ethical that's ok then never mind

I asked Because chat gpt gave me a begginer code to do

so i wondered if you had advanced codes

OK then don't mind

[feral zinc]

note that hacking can be illegal in several countries regardless of the intent

[short moss]

I think there's a bug in ChatGPT (web) where there is literally no cancel button anywhere. I was on Pro, then downgraded to Plus for the next billing cycle. Now I want to cancel instead of downgrade but I can't.

Anybody else come across this kind of problem?

[feral zinc]

ive seen someone with the same issue and only support could fix it

[short moss]

Thank you. That is insane. So I need to email them?

I've seen this bug reported on the forum since 2023!

[feral zinc]

that's not possible as pro did not even exist back then and there are no other plans to downgrade from

[feral zinc]

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

[short moss]

Will give it a try. Thanks ❤️

[broken smelt]

Everytime ChatGPT gives me a link to a website, it's not a clickable link but looks like one. How can i fix that?

[normal quarry]

just got an update on the macOS app, not sure what they have changed

[native otter]

They always do macOS app updates frequently, it's mostly bug fixes or very minor changes...

[slow marsh]

Can somebody tell me what models of chatgpt free users can access?

I'm trying to help some guy but I have the paid version and he's not so smart he can't tell which model he has access to

[hidden pilot]

Is chatgpt still down i keep getting error the request has timeout am using chatgpt 4.5

[tall junco]

anyone getting this issue

i reset my password and logged out all devices multiple times

no one is in my account other than me

[sudden wren]

I hate the way the generation of text looks. It's gotten really slow all of a sudden

[abstract pendant]

I just discovered that ChatGPT knows my IP address and correctly guessed my location based on that.

[astral rock]

Me too. Regenerate response is missing on mobile and on windows chrome browser. I am gpt plus
It's really a pain now cause you have to edit your message and hit enter again to 'regenerate'

[abstract pendant]

It's still there, but it's been moved in the new UI. It's part of the switch model dialog under the output.

[slate peak]

What I wrote in this screenshot describes exactly my problem: ChatGPT seems to read only the first few word of whatever I write.

I already re-installed and logged out and in again, but nothing helped.

Has anyone ever had a similar problem or knows what the cause could be?

There's a typo, "su is" was meant to be "you are"

[open burrow]

anyone else finding chat gpt just got way worse after it went back up? It starts every response with a compliment, then a line, then organizes its follow up with itemized responses categorized in boxes denoted with emoji.

I hate it.

[ancient junco]

whats deep research limit on pro tier?

[pallid yoke]

120

[analog quarry]

Hey, I’m having an issue with the ChatGPT Windows app.

• Voice mode isn’t working at all, and the app freezes whenever I try to use it.
• I’ve reinstalled the app, updated all drivers, and confirmed my mic works.
• Voice works fine in the web version only if hardware acceleration is on, but the app itself just won’t work.

Specs (if helpful):

• CPU: AMD Ryzen 5 5600G
• GPU: RTX 4070 Super
• OS: Windows 11
• Mic: Elgato Wave 3

If anyone knows a fix or if this is a known issue, I’d appreciate any help!

[ivory matrix]

The edit button is gone again

[fallen jewel]

anyone experimenting very slow the chatgpt plus ? a lot of issues errors I'm going to cancel my suscripction

[abstract pendant]

Just came in here to say that. Grr. Anybody got a link to the workarounds, where you paste some extra javascript in to bring the edit button back?

And.. the edit button is sometimes there for a new conversation. Old conversations that didn't have the edit button, still don't have it when I return to them. This is looking very familiar.

[analog quarry]

anyone??

[native otter]

Ensure that your firewall is enabled for ChatGPT on both public and private networks, that you have the latest version (updated from the MS Store), and that your microphone is not disabled for the system or the ChatGPT app...

[analog quarry]

it stoped working like 2 days ago

it shows that its using my mic but no repons from chatgpt

[fallen jewel]

very very slow chatgpt plus whyyyyyyyyyyyyyyyyy ????????????????

[inland stream]

especially cutsom gpts is completely spoiled, it is making me to retry multiple times and keeping me waited for long time to generate response

[cloud pewter]

Can the new avm sing now

ok it can indeed

[dense halo]

Look, everyone, for what I set my ChatGPT as.

[fluid helm]

If anyone's not wanting to use Pru sub, can I please?

[iron roost]

anyone else finding that the read aloud, is stopping after 3-4 lines rather than reading the entire output ?

then a error message failed to play message

very frustratign and only has been doing it today

[elfin lava]

Does it have any major improvement for you?

[elfin lava]

I can’t even get it to sing 😣

[tiny pivot]

Hey guys, not sure if this is the right channel to post in but does anyone have any experience with making a custom GPT with more than say 100 documents of knowledge? I've got them all in a sharepoint and was thinking of using custom actions with the GPT schema to read the documents in the sharepoint with graph api?

[normal quarry]

I couldn't get it to sing either

[exotic lotus]

Nope

[cloud pewter]

not really

even the signing was the random triggering of it, so far with me not every chat works…

[elfin lava]

so they kinda lied about AVM improvement

Cuz I tried just now, asked it to giggle and what it did was spoke the word “giggle” instead of actually giggle

[cloud pewter]

I’ll send some clips later ig

[normal quarry]

AVM has improved in these key ways based on my testing

1. It doesn't interrupt you as much
2. It can be verbose beyond what it used to be
3. It hears better, less occurrences of hallucinations like "Thanks for watching"
4. It seems more robust, not having to repeat or prod it to get the output wanted

[cloud pewter]

there isn’t even a proper announcement

[normal quarry]

maybe it's just technical improvements not worthy of an announcement

[elfin lava]

True

[normal quarry]

it is better than it was

[normal quarry]

what was the source of that screenshot?

[native otter]

Check the OAI YT Channel

[iron roost]

for some of us advanced voice mode, as far as the output and reading aloud has stopped and or only reads out the first few lines and then fails to play messag

so it seems they have broken it for a lot of people

pro user here and it worked fine probably until yesterday but only noticing and trying to use it today

[full coyote]

Idk if I should ask this but, is ChatGPT having censorship problem right now? Like, I could make an action stories like where I can asks a scene where someone get headshot. But somehow, asking for a scene where a mom lightheartedly teasing her daughter (in a different story and conversation) is a no no.

[nocturne garnet]

In a task i got going it seems like chatgpt lost all its context memory and cant aeem to remember what happened in the last chat message in the conversation and whats worse is that its dtart claiming it cant complete the task because it claims it cant generate pictures

[normal quarry]

what plan are you on? how long is the chat?

[nocturne garnet]

I am on plus. Its been generating random drawings of old norse folklore daily since the task feature came out. Its been working fine until a day or two when it suddenly "lost" the capability to generate pictures at all

[normal quarry]

how long is the chat?

[nocturne garnet]

Good question, quite long i guess. Been generating daily since tasks came out, so thats quite long. Can you see the length somewhere?

[normal quarry]

I would say that you're out of context, start a new chat. Plus users only get 32k context

[nocturne garnet]

Oh yeah, sounds possible, yeah i do that. Thank you for your help 🙂

[exotic lotus]

Yeah, the strict filter was always a problem with this AI. Not really the best choice for creative writing in my opinion. Not to mention the tiny context window.

[wise anchor]

why does using deep research in a thread on the macos app turn it into a lag fest

Completely unusable once you use deep research

[normal quarry]

let me try one and see, stand by

Just started a Deep Research I needed

[wise anchor]

It’s fine while it’s actually doing the research but once the response loads in it just destroys the whole application

[normal quarry]

I'll find out soon, it's still going

[wise anchor]

Freezes up constantly, inputs happen 10 seconds after you do them

Yea lmk i wanna know if it’s on my end

[full coyote]

Like, I get it, its for safety and all but. Its like it was horribly implemented. Past update wasn't that bad, but if it even refuse to respond to a prompt asking a scene where a mother lightheartedly teasing her daughter, it undoubtedly killing the mood.

[normal quarry]

it's just loaded it's (rather long) deep research response and it has not slowed down the app at all. For reference I am using the ChatGPT macOS Desktop App on the latest version of macOS.

Scrolling is fine, no slow down, follow up prompts are fine, it's all behaving as I would expect it to.

[exotic lotus]

Yep, I feel you

[tropic parrot]

glad to see I'm not the only person who's been annoyed by this. I hate the filter so much. Even after I repeatedly clarifiy things with the AI they keep on insisting on their asinine "guidelines"

[inland stream]

Can someone tell what is new update at 11pm?

[feral zinc]

the image says it all

[inland stream]

Oh got it, not my feature. So i will not wait. Thanks for response

[twilit temple]

Am i slow, where is the livestream>

granted, I just woke up

[honest jay]

It's in a couple hours!

[twilit temple]

oh so not 11 AM?

[honest jay]

It's just after 9AM PT right now

[twilit temple]

oh... lmao, I forgot I flew out to TN and I am not in Ca

my bad

[weak bay]

I would really appreciate help on this. I'm doing some huge concept projects to learn to code with assistance from gpt 4o. The first conversation was huge and slow so I created a gpt project, dragged in my codebase as a zip along with a detailed description for everything. But when I start a new convo to continue work on it gpt 4o seems to not remember where we left off or what it was supposed to help me with. Is there anything I can do about it? Or should I just go back to the old slow convo and keep using it? atleast that one still gives working advice xD

[dense halo]

New SOTA image generation model coming from openai soon

[native otter]

• Better to use o1 and 4.5 with Projects as they are generally smarter, work with RAG better, and connect the dots better than 4o, which really helps with programming related stuff.
• Starting a new Convo doesn't use the context of other Chats, even if in a Project.
• Avoid uploading the Project files as zip as that would only be read if the Data Analysis tool is used as explained before, which you don't want to be the case, so upload files individually, and combine some together if you use over 20 (While explaining the file structure and file purpose in Custom Instructions). + don't forget to stick with.txt as many file types can only be read via Data Analysis...

[gentle hull]

congrats guys, you finally did it. You made using chatgpt less productive than not using it at all.

I spend 2 hrs on a simple problem that one of your competition solved on its first attempt. today forget about chatgpt not being able to code, im used to that. it literally is arguing with me and not following instructions. It's so wonderful being an early adopter and having a company use me as a stepping stone to their corporate users.

wouldnt even get annoyed if it wasnt so obvious that this thing is gettting nerfed constantly. The first versions were better

[dense axle]

Anyone else facing the issue I can't open the file you uploaded, since it's not accessible with the myfiles_browser tool when uploading files or zip files to ChatGPT?

[eternal spoke]

whens it release

this new image gen

[untold seal]

is this a question?

[lapis rune]

i still have the old dalle image gen, is there a way to tell if/when it switches to native?

[honest jay]

The 4o image gen tool will take longer to create the image, and the animation will be different from DALL-E's

[echo plinth]

How can I see if my account already has the new image gen? Some popup or a changed menu?

[quick lodge]

I was already generating images with gpt-4o and it suddenly went back to using DALL-E!

[modest agate]

Is it one of those staggered roll out that'll take a few hours ?
I'm still getting DallE images

[primal python]

all of them will have the partial generation sliding down kinda effect like in the livestream?

[honest jay]

Yeah that's what it looks like on ChatGPT!

[dense plume]

i only had 1 image use the new tool, and its slide down like in the live stream

[primal python]

perfect, thank you

[abstract pendant]

My view of ChatGPT in my browser doesn't match that. I'm a plus user, and I don't have that little plus sign next to the search and research buttons. What setting do I need to have active to turn that on?

[cursive vault]

Where's the new Imagengen? Tested web, mac app and ios app, and still generating with old model.

[elfin lava]

So what’s new

[cursive vault]

https://openai.com/index/introducing-4o-image-generation/

[tawny plover]

In Sora I think it is, because there I noticed the ability to generate and edit images

[shell stream]

Gemini 2.5 is amazing

[elfin lava]

…

Is the create image out yet

[cursive vault]

Can confirm that Sora did use the latest model.

GPT app and web not yet.

[calm flower]

Hey how do I knwo if the image thing has been rolled out for me?

[elfin lava]

Should look like this

Lmao they changed the description

[lapis herald]

doesnt work for me, text is smeared and full of typos

probably the new stuff is not available in my region yet.

[brazen mortar]

Same, not sure if it's fully out yet

[harsh turret]

Why is god d@mɲ įť a banned word lol

[lapis herald]

chat suggested the livestream might have used "dirty tricks" to achieve the clean result 😂

[violet lodge]

is the new image model available in every country already?

[astral rock]

what model are you doing? I had it do some pretty intense action scenes last night

like hostage situations and stuff

[exotic lotus]

I'm currently testing it. Just wow...

[zenith lichen]

how do i trigger 4o image generation?

i only see dall e for now

[hollow quail]

wait

and then, poof, it'll be enabled also they never differentiated whether it was rest of the world with an eu/uk launch later like it normally is so who knows, who fricken knows any more

on one hand, maybe it's an everywhere launch but they never said

[lapis rune]

youll know when it gets large text right

[hollow quail]

the scanline generation will likely be the biggest clue instead of an animated circle loader

the staged rollout is pretty hilarious because you have people all over reddit and twitter comparing and calling the new image gen trash because they said it's released, so people of course assume rightly so that they're using the new image gen whereas they're actually using dalle3 and dunking on it because well... it's dalle3 lmao

[dense halo]

What do you think of my chatgpt just being for translating from or into greek

[eternal rock]

Get notified for server events!

Visit id:customize to pick up the <@&1261377106890199132> role.

Also, check out the ongoing Canvas Creations event!

[elfin lava]

the image is very good, big improvement they did . but once u turned it into video using sora, it turned bad (the vid end up havin deform arm etc)

[eternal rock]

Get creative with Sora

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

[echo plinth]

Just got the new image generation on the web app, and my first attempt is stuck at "Getting Started". The load seems to be a bit too high right now.

[median delta]

yeah image gen not work

is slowed

[median delta]

same

[eternal rock]

`` Server Tip: ``

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

[nova zodiac]

any reason mine worked for 2 generations, then went back to dalle?

[merry solar]

is that supposed to be the interface with the new update? Is it still using dall-e3?

[nova zodiac]

thats dall

mine when it worked id not say dall-e there

[merry solar]

How can I use it? I'm on free plan, so I think it should be GPT 4o- mini?

Or is it available only in premium plans?

[nova zodiac]

you prob just have to wait a bit its slowly rolling out to ppl

it actualy worked for me for a few minutes, but then went back to dall

when you use it in sora it just makes images not videos? is it the same as using it it gpt?

[echo plinth]

My account just reverted back to Dall-E 🤷‍♂️

[acoustic wasp]

Same here

I pay more than 21€ for sth that doesn't work right

[final venture]

I went log out , log in, and now I can't get in... I wonder if the system might be a little overloaded right now for some strange reason 😇

[echo plinth]

they usually do a more staggered rollout, so no surprise here

[final venture]

Also does it work with 4o regular, or only with mini... or what's the picture here?

[acoustic wasp]

I bet they restricted us plus usert to save the pro users once again......

[hard wolf]

(And everyone else) I'm pretty sure you can use the new image generation on sora.com

[acoustic wasp]

I can? Didnt knew that

[hard wolf]

It's included in the Pro Plus plan

[dense halo]

ChatGPT can translate into Ancient Greek

[acoustic wasp]

I have plus