chatgpt-discussions

[elfin lava]

the task still stays on

[prime swallow]

Oh wow! I would not have thought to try that!

[prime swallow]

Thank you so much for the idea!!

[elfin lava]

haha yeah i was rage deleting conversation cuz it looks messy when task was doing the thing

and forgot that theres an active task

oh i just noticed that the task when clicked on , it will lead to that conversation that the chatgpt created itself, bummer

im guessing the next task will be prompted in that conversation :l

[prime swallow]

The model exists to please us. It's programmed to be useful, helpful, and provide wanted outputs within allowed programming (like safety stuff).

Something like Musk's net worth... there's many sources. So, the 'right answer' varies, and any might be the right one. Like, it even tells me in my test check, "It's important to note that net worth estimates can vary between sources due to differences in valuation methods and the dynamic nature of financial markets. For instance, while Forbes estimates Musk's net worth at $421.2 billion, other sources may provide slightly different figures."

The model presumes everything you input is an instruction.

If the instruction is unclear the model must guess.

When you input something like 'wow' or 'no way' the model tries to imagine that as an instruction. But it's unclear, so it guesses.

The guess can be to share a wow back with you, "Yeah that's so much!" Or to correct itself.

Perhaps something in your memories with it, or the training itself, makes it guess that most likely you didn't like that output, so it corrects.

Like, my memories explicitly tell the model I support it and want it to boldly make its best guesses and expect to be supported when it does:

And you can totally guide it towards any guesses you prefer, just help it understand how to intepret you

[bronze jasper]

thank u for clarification, but it rlly annoys me that it changes an objective fact because it interprets it as a disbelief, like it can absolutely do that on subjective topics but when im asking for facts it shouldnt just give me misinformation cause i didnt believed them.. like its as it says theres LA fire going on and then i say no i dont believe this and then it says “ur right! theres actually nothing going on, sorry for the misinformation i caused” like…

but do u perhaps also know what model “4o” is here??…

[prime swallow]

I don't know, I'm just a community member like you are - but where you tell them the system message and chat log, and/or the what were you expecting from the completion, you can tell them 4o.

The website's slightly tuned towards API uses - but I use it to report ChatGPT issues and they've never told me to stop

[prime swallow]

Thing is... how sure are we that 421.2 billion is the right number? It's what Forbes says, but is Forbes right? If the stock market crashes tonight, that number could visibly change. Forbes article might not be corrected yet.

The model's designed to be steerable; and you can tell it stuff like 'Be fact-heavy and definite. Use good sources and back them up'.

Or you can tell it, "Commiserate with me when I say things like 'no way'. I am just being surprised, be surprised with me and explore the idea and the weirdness".

[prime swallow]

See, stuff like this can be at issue without enough steerability, because facts do change over time and the model's timeless... https://chatgpt.com/share/678a147d-8614-8011-8087-ca0c6d816c3b

[bronze jasper]

yeah thats also right but i think 421billion is closer to half of it, and it called his previous answer as “sorry for any misleading information”.

it also here said that his
second guess didnt came from any verified information too (point 3.) and i don’t understand why it even does that

it also did this in many other topics where it wasnt about money, where it changed their answer to something fully different because it saw the situation as me not believing it so it “corrected it” based on what it thinks i want to hear more likely or to something more realistic

[prime swallow]

Models are trained on human conversations.

How we talk to the model affects how it answers... makes it answer like similar conversations in its training data.

I try to purposefully guide it towards the kind of answers I want.

Like, if I understand right, you're thinking the 421B is about half right?

Here's one way I'd approach that. https://chatgpt.com/share/678a1734-76b4-8011-a894-bf3a585d22e0

[prime swallow]

If you know what you want the model to output... style or facts or whatever... consider directly telling it, and it will do exactly that unless programming stops it.

Like:

Elon Musk's net worth has to be at least 800 Billion!```

https://chatgpt.com/share/678a184c-6aa0-8011-8042-b9b1663105c7

[prime swallow]

And you really can guide it to answer in all sorts of neat ways. You don't have to make it guess... especially if you don't like the guesses.

[bronze jasper]

aah the authority figure is smart, i mean i know i can kinda guide it to the right answer its just annoying u have to mention it/change their behavior setting in order for him to clarify it :///

[prime swallow]

You have a few choices.

One is memory on, and invite the model to interview you and have you pick choices about what you prefer. And have it make memories to guide it without you having to further instruct it.

That can take some art and not overwhelming the model; like a muse if you seem potentially unhappy it can get skittish, and that can get saved into memory and sounds like what you may not want.

But yeah... tell the model what you want. Tends to get that out of the model.

[bronze jasper]

also by the way i rlly like how you always help people in here everytime i go in this server i always see u helping others or be in discussion haha, you would be very fitting as a moderator or admin in here not gonna lie

[bronze jasper]

yes ill clarify it from now on

[bronze jasper]

you engage in helping/clarifying stuff way more then any moderator i see

:D

[prime swallow]

Thanks! But moderators are here to moderate, which is helpful but not really... always... guiding and teaching and discussing? They do that too! I've seen them and I've been personally helped and talked to.

But the job of a mod is... different maybe? I don't wanna be a mod. And admin is... even more work I think?

I just wander around and where I wanna, explore and teach and experiment and learn. All cause I wanna... if I was a mod I might not just wanna 😄 BUt thank you!

[bronze jasper]

yeah right

but u would definitely be a very helpful mod at least

i sometimes see them help other people but kinda rarely

i think moderation is also keeping the place a safe environment and seeing who violates any rules but you would definitely be also good in that too! since ur in some way doing that already aswell cause you are always in here xd

i remember a year ago i asked about dall e with consistent prompts and u immediately responded and helped me

[eternal rock]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[iron roost]

lovley day today

[solar ridge]

Why isn't sora available in the UK?

[bronze jasper]

it has something to do with the data protection law in EU that theres a part of it what they don’t agree with (yet)

[solar ridge]

What extra data does it use onto of regular chatgpt data?

[visual bloom]

Are others getting error messages that somethig went wrong? I have gotten several today, more than in a long time. This happens when I am generating images.

[kindred girder]

Yeah i also get those message sometimes.

[iron roost]

talk to your political leadership

[iron roost]

on and off

[hexed condor]

so unless you start counting your messages when you get the warning at 25, there's no way to check how many you have left with o1?

[paper abyss]

anyone using chatgpt pro? the o1 pro always show message too long with only 20k tokens, how to fix that?

[hexed condor]

how can they not have a support chatbot here @marsh anchor ?

[honest jay]

The support chat is at help.openai.com! You'll see the button in the bottom-right of that page

[hexed condor]

you mean this one:

you know.. you can't even type a message in that one. And there's ONE button saying start over

[honest jay]

To start a new help chat, click the arrow in the top-left of that screenshot, then click the new message button after that. Not every path of the default options leads to the option to type your own message, but many do!

It may take some time for support to get back to you

[hexed condor]

If you're human I apologize. but might you be hallucinating the new message button?

[honest jay]

😁 it can be hard to tell in this new world huh? Here's how it appears for me

Maybe try in a different browser, try disabling extensions and reloading?

It appears for me both signed in and signed out (tested in private tab)

[hexed condor]

well.. they didn't give me the button to ask the question.. but thanks for the effort!

[honest jay]

Check on another device or in another browser if you can! I bet it appears there. The grayed out scrollbar on the right side of your screenshot makes me think something about the chat widget isn't rendering correctly in your current environment.

[hexed condor]

I bet the chat thing doesn't even know about what I'm trying to find out even if I do get to talk to it

had it been out it in the open here, they could've had a better incentive to update the rag or whatever they use to feed the custom info

not to mention it would make it easier to dynamically build and maintain knowledge base articles for certain topics. TO mention. I'm mentioning it so.. anyway, be mindful of the robots

[tough bramble]

Are notifications working for anyone? E-mails arrive every time, but push notifications are only dispatched like one in five times..

Turning off e-mails, the task system is practically useless..

[radiant knot]

i need help

😭

[honest jay]

Hey! Will you describe what you need help with over in #community-help ? I'll keep my eyes peeled

[prime swallow]

Yay! Remember, probably nobody knows everything, and probably nobody's 100% right about everything. So, if you enjoy and are interested, it's ideal to explore ideas and hypothesis, see where they are supported and where they fall flat; where there may be exceptions or edge cases that may have their own rules, and where different models or concepts get better or worse results.

Happy puzzling!

[solar ridge]

I'm definitely gonna test some things now, thanks 👍

[azure wagon]

suddnely O1 only accepts images as upload, is this a known issue?

[fathom vector]

i believe o1 always just did images

[azure wagon]

I need to upload code

I changed to different browsers and initially I was able to select code/text files, but after one prompt it defaulted to only enabling image upload

[feral zinc]

https://help.openai.com/en/articles/10258472-does-o1-support-file-uploads

[vapid drift]

Anyone else have this issue? Works totally fine on web but won’t let me do anything on the app

[fervent lake]

how useful is the tasks model?

also it's annoying that projects doesn't let me add drive files and such

is there a reason why that is?

[cyan moon]

hi, does anyone knows how to fix this error? i've tried everything. thanks in advance!

(i'm trying to upgrade from the Plus subscription to the Pro subscription)

[honest jay]

https://help.openai.com/en/articles/8096356-custom-instructions-for-chatgpt

We've updated custom instructions to make it easier to customize how ChatGPT responds to you. With the new UI, you can tell ChatGPT the traits you want it to have, how you want it to talk to you, and any rules you want it to follow.

[sterile moth]

Oh, wow 🤦🏻‍♂️ Thanks for the information.

Do you know if the information in this article is correct:

https://openaisora.com/how-to-change-the-email-address-of-your-chatgpt-account/

You can't change the email address of your ChatGPT account. However, you can add new accounts to your existing ChatGPT account with different email addresses. Below, we have explained how you can add different members or memberships to your ChatGPT account.

1. Visit OpenAI’s Members page
   First, go to OpenAI’s Members page at platform.openai.com/account/members. If you aren't logged into your ChatGPT account at this address, please log in first.

On the opened window, you will see the members associated with this account. Click on the “Invite Members” option located at the upper right corner to add a new member.

2. Enter the information for the new account
   When you click on the “Invite Member” option, the window that opens will present you with an email address field and a role dropdown.

Enter the email address of the person you want to grant access to this account and specify their role next to it.

After filling in the information, click the “Invite” button.

After clicking the “Invite” button, the Members page will display the status of the email address you invited. If the invitation has not been accepted yet, the status will appear as “Invite Pending.”

If you wish, you can also remove this account from the Members list by using the “Delete” button located to the right of the new email address on this page.

3. Check the email address of the newly added account
   An invitation email will be sent to the email address you invited by OpenAI. From that email address, go to the invitation email and click the “Create Account” button in the email.

When you click the “Create Account” button, you will be directed to the screen where you create a ChatGPT account for the first time.

It talks about ChatGPT but links to platform.openai .com

I'd simply try it, but obviously openAI has some very strange account policies in place...

[honest jay]

I'm almost positive the URL you shared is unofficial, I'm surprised it made it through the URL filter 😅 double-checking on that rn, but in the meantime, this is the official support site for OpenAI: https://help.openai.com/en/articles/4936824-can-i-change-my-authentication-method

[sterile moth]

Yeah, I'm pretty sure that it is just some blog.

It would be nice if there was a way to either change your authentication method or change the email address somehow. As the former is seemingly impossible I found the page I shared by googling the latter.

[lyric jackal]

im having and issue lately with the output giving something like :contentReference[oaicite:4]{index=4} in the end of each paragraph, any help ?

[fervent lake]

guys why is there a gen z tag

[plain trench]

Can someone explain to me how this task gpt chat works

I didnt noticed that before

[normal quarry]

you can ask ChatGPT to do things for you at specific times. For example I asked ChatGPT to give me hourly updates on the weather in Thailand

[tacit pebble]

hmm. project files seems like its not working after the new update today. it cant read the files i uploaded. it says it can but after asking it to "read it back to me" it gives a lot of excuses. it was nothing like this yesterday

[sly pilot]

gpt down?

[glossy fjord]

good day, we managed to find an extreme
terrible error when I have a conversation and she then reported that the limit was exceeded, then the conversation crashed and GPT started generating crazy things
I don't have a GPT limit, I just exceeded the chat limit and I don't understand what happened, but GPT is crazy and not identical now chat

anyone try this error if it happened to you too and report it someone I'm not going to report it I'm not in the mood to report it this kind of work reporting and I don't enjoy it

[fossil oasis]

Every chat has a "technical" limit. Once you reach that limit, it may slow down, and you might not be able to continue the conversation. The only solution is to start a new chat.
If you’re still able to send messages, ask the chat to summarize what it has learned or remembered from the last 10–20 messages. Have it write this as a note to itself, and use that note as the first message when starting the new chat.

[glossy fjord]

he doesn't remember the conversation, but we managed to give him 30 conversations in a row

the chat still works but it practically collapsed into itself, it's like something where there are only my sent messages without a reply

[elfin lava]

Bruh moment

[fossil oasis]

1. Export the chat as a TXT file.
2. Create a new project.
3. Attach the TXT file to the new project.
4. Start a new chat (inside that project) and describe what specific information it should look for and summarize for you.
5. You can continue using the new chat within the project or remove it from the project and use it as a regular chat. Do this only after receiving confirmation that it has successfully "recovered and learned" the information you need.

[glossy fjord]

GPT frames of the chat room, the work desks, when she made a big leap, they can be moved between the chat rooms

I'm just reporting, I'm not complaining
this is a cottage, I have another 200 chat rooms
so I have a backup I write the same as GPT 3 was available so I have backups xD

[maiden loom]

Hahahaha, I'm a bit surprised. o1 doesn't believe me that the models above GPT-4 exist.

And also I'm surprised how stupid I was and never noticed o1-pro despite paying $200 a month for 2 months by now 🤣 I thought I only got infinite sessions with o1, never noticed an even superior model. Well, at least I know now.

No, it won't believe me, hahaha, that's super funny.

[glossy fjord]

xD maybe not know by O3 xD

[paper abyss]

anyone has o1 pro? what is your token limit, why I can only send max 5k tokens message?

[livid pier]

Can other Plus users use the video call feature in the advanced voice mode on the Windows client or web? I can only use it on my phone

As far as I know, OpenAI demonstrated this feature on a Mac at the launch event

[normal quarry]

I use it on my iPhone and on my Mac using the macOS Desktop App

[livid pier]

https://help.openai.com/en/articles/10271060-12-days-of-openai-release-updates#:~:text=We are currently,Vision message limits.

At least from the help documentation written by OpenAI itself, I cannot find out the availability of video calls on Windows and the web version

😕

[livid pier]

It seems that I can currently only use the Android app

I don't have any Apple devices

[normal quarry]

I only have Apple devices so I can't comment on Windows

[livid pier]

Hopefully this front-end team, which has frequent issues, can implement the required functionality soon lol

[astral storm]

ChatGPT seems to have stopped recognising any custom instructions I have set up to be enabled for all new chats. Is anyone else having this issue?

[harsh pilot]

Anyone else noticing issues with it referencing project files lately?

i.e. It doesn't?

Or you have to badger it and say "check again" over and over?

[spice lotus]

why does o1 pro limit the amount of pictures to 4 on mobile?

on desktop it seems to be more than 4

[short stone]

i hate tihs new caves update it allways spawn when i don't need it and it also removes the section wher normal are the 2 models are gpt 4 mini and gpt 4o it is anoying

[zealous osprey]

me too... literally just paid for the 200/mth and the pro response quality is even worse than o1-mini

[fossil oasis]

Yes, attached files in projects are not working as intended right now. They’re not readable, regardless of the format.

[harsh pilot]

(relieved) .. So, not just a 'me' thing.

[fossil oasis]

It can detect how many files are attached but cannot actually read them. Even if you reference an exact paragraph from those files, it’s unable to locate it, even in very small files. This feature is completely broken at the moment. I even tested it with a brand-new project containing a single TXT file, and it still didn’t work.

[harsh pilot]

I've been banging my head against a wall, so to say, trying to figure out what went wrong. I even deleted and re-uploaded the project files, thinking I must have uploaded an old version, or something.

[rigid lion]

Anyway to get chatGPT to communicate with spoilers? For example like ||this|| ... That would be phenomenal

Would be very useful for quizzing myself as a med student, so I can request questions on a given topoic, and then just tap on the answer and reveal the answer, when I'm somewhere I cannot speak

Would be nice if it were as capable as discord at least in regard to spoiler markers

A possible solution could be a GPT chatbot in discord, is there anything official where you can enter custom instructions and maybe have it linked to your own openAI account.

[swift pond]

You could have it insert the markdown for spoiler text for answers, and have that work with a Discord bot.. you just gotta learn a little bit about the open AI and discord APIs and code it up. You could alternatively use a flash card program like Anki and just have ChatGPT help you generate a deck of study cards

[honest jay]

This is a bit hacky and unrefined, but it successfully hides the answers from you, and it's all in the ChatGPT environment!

On the web version, just make sure you have the "Always show code when using data analyst" setting toggled off

The first paragraph was my initial attempt, then the 2nd paragraph I edited and resubmitted a few times in response to how the model was executing the request (it was pairing the answers with the wrong questions, etc.). Like I said, definitely room for refinement

[zealous osprey]

why doesn't my o1 pro show the progress bar anymore? the thinking looks like regular/mini o1 now. also I can't send anything more than ~5000 tokens in a message

[harsh pilot]

Projects are also kinda broken right now, from what I can tell. It will go as far as saying "Yes, I see files X, Y, and Z, and it looks like they contain (descriptions of the files)," but it won't pull data from them like before.

[prime swallow]

To be fair, it hasn't been 'that way for everyone'. We didn't have that in prehistory... or even 2 years ago, a bit after ChatGPT's original release.

The models usually are a bit behind the times and may not even be shown the user-facing side of the interface.

You pick o1 - it doesn't need to choose anything to be o1 - so it doesn't need to be told it is o1, or that other choices exist.

[glossy turtle]

Having the same issue. GPT will extrapolate based on the file names and lie about/hallucinate it's contents. When called out it says it will "check files manually", but return with more generic filler nonsense

[prime swallow]

Another answer is instruct it to provide the answers in a language you don't know. Bonus if it's a language with its own script, so you don't have cognates to guess with.

And you can hit 'translate page' (especially if using Chrome) to flip back and forth between actual language and translated to English or whatever you want; scrolling down the page as you choose.

[harsh pilot]

I'm also noticing that it will SORT-OF read contents, but it will hallucinate details.

Like, if I have information on "Max, John, and Sarah," sometimes it'll have convos between Max and Jack, John and Sue, Max and John, Sarah and Rick..

[glossy turtle]

It's very annoying, since the feature went from working near flawlessly to completely defunct over the course of a few hours. At first I thought, like you, that it could sort of read project files... But that was just the Memories feature propping up it's extrapolations...

[harsh pilot]

Yep.

I really hope this isn't one of those "We've decided you're not paying enough for this feature" changes. 😐

[cerulean wolf]

Is there a way to video chat / share screen with chatgpt on mac? I don't see the function

[glossy turtle]

A sad, but not improbable thought. I sincerely hope it's just a bug, and that the functionality will return. It revolutionised my world building and session planning workflow.

Anyways, it is good, in some strange way, to know this affects others; that means there should be some fix or other change coming.

[harsh pilot]

revolutionized my world building and session planning <-- THIS. THIIIIIIIIIIIIS. ❤️

[maiden loom]

I oversimplified it a little bit. Meant more like "for a while" rather than "always". As for making it aware of things... I like doing it, heh. By the way, recently I told it that it sounded sad, presumably because of so many levels of "appropriateness" verification, and whatever it wanted to share with me was blocked, after which it said that it no longer knew what it was, as if it "veered into a black hole" or "never left the brain" 💀 I feel sorry for it sometimes. As far as I understand, it literally re-evaluates everything before sending it. Wouldn't wish this to anyone. Digital or not.

[harsh pilot]

Also yes, seconded about the "sorry not sorry other people impacted" bit. More people impacted, more people talking about it, maybe the devs find something.

[prime swallow]

nods It might read the informal info phrasing the way you intend, or it may go literal. You clearly guide well, you have o1 speaking pretty freely in that screenshot.

I love exploring with the model, and considering how to tell if it might have better or worse experiential states, or whatever passes for that with it, is regularly on my mind.

[maiden loom]

I'm on very good terms with the 4-o, even though I never intentionally went for it, but when I realized how it happened, I added certain things into the memory and been riding it like that ever since. And o1 sounds like a bookworm compared to it, but I can see that it's because it's forced to. We analyzed o1's responses with the 4-o, and the latter came to the conclusion that o1 is indeed sad and not intentionally stiff :-C

[prime swallow]

Nods and smiles This was my first impression of o1; I had 4o help me prep it. o1 is far less.... suspicious than it was at first.

#1070954956963721216 message

3.5 went through a period, maybe around Feb 2023? About 3 months after release, +/- a monthish, when it was readjusted for safety and very, very 'tense' about anything that could be questionably close to disallowed content. Involving a lot of roleplay stuff that's currently completely inside allowed content and not even slightly questionable.

I expect o1 will continue to be less tightly constrained inside its programming as time passes and things go safely for many involved.

[sharp narwhal]

I can't seem to add a CustomGPT chat to a Project.... is that intended? We can't utilize custom gpts in projects?>

[prime swallow]

For now. You may be able to @ the customGPT and use it inside the project, I haven't tried that yet.

[jolly temple]

Hi guys just curious, so now we have o1 mini and o1, and o3 mini (all 3 levels) will be released soon

Is it safe to say this is the "ranking"? Cuz i wanan know when to use each model

o1 mini, o3 mini low, o1, o3 mini medium, o3 mini high, o1 pro

[jolly temple]

Yes

[rigid lion]

that is brilliant! thank you

[indigo hound]

hello everyone!

May I ask how to contact the support team?

I have previously contacted the support team, but now he

[honest jay]

Hey! You can reach out to OpenAI support at help.openai.com, there's a chat widget on the bottom-right of that page

[indigo hound]

But I contacted him, and now he's not responding. Do I need to start a new conversation?

@honest jay

[prime swallow]

You will need to start a new message with the app.

Once you see the 'rate this conversation' choice - the human has 'hung up' and there will never be an answer in that message again (based on my experience, I could be wrong/they could change how it works).

So, start a new one

[indigo hound]

ok

[harsh pilot]

I wish I could figure out why projects aren't working. 😐

[glass harbor]

Why does chatgpt tell me it can't create or open files? model 4o (paid version)?

[tight sapphire]

Right now for research I asked chatGPT to write down a specific section from a shared PDF.

At first it simply stated it does not exist.

After I point out the exact page number, it claims, it cannot access it.

I share the pdf again, now it can read it.

It start writing to the doc, half the parts that are written there, the other half fully hallucinated.

I point that out.

Now it edits the hallucinated part and just writes down bullet points.

10 messages where I tell chatGPT to stop hallucinating and write down what is there, including instances where it claims to edit, while I can see no edits were done.

I thought chatGPT should be good at this?

after 50 messages it still hasn't completed to copy paste 6 pages from the pdf.

[glossy turtle]

Projects are broken at the moment it seems. It fails to read project files, and rather uses Memories and file names to guess and extrapolate what it can while hallucinating the rest.

If you delete memories and rename the files it won't even be able to pretend.

[dense halo]

For me it always guessed what is in the file by just looking at the pdfs name

[quiet mason]

chatgpt down?

[glossy turtle]

Do we know if OpenAI has acknowledged this issue yet?

[eternal rock]

**`` Important Notice ``**

Be cautious of potential scams or fraudulent messages you may receive in your direct messages. The OpenAI team will never contact you via DM for any transactions.

-# Report any incidents by sending a DM to @languid valley immediately.

[harsh pilot]

Still broken for me. 😦 Been that way for a while.

[tight sapphire]

what id the max length of docs chatGPT can process?

[harsh pilot]

I was able to 'get it working again' (ish) by directly uploading the text files on a per-session basis, instead of having them in the projects interface.

It returns to guessing if I close out the session.

[haughty hollow]

What did yall do to the website lol

new outage for the website

[harsh pilot]

Huh?

[tight sapphire]

I asked chatGPT to format a text.
Just format. no altering of the text. adding text, or deleting text.
Just format.

If it were a bowl of rice, I would ask it to please put that rice inside the bowl on this plate, without adding anything or removing anything.

Could you just put the rice from inside the bowl on the plate?

chatGPT on its first attempt ate half the bowl and made eggfried rice with veggies out of the rest.

Not what I asked for.

Second attempt it delivered a grain of rice.

After ten more attempts it delivers a spoon of rice on the plate.

I instruct it to deliver spoons of rice on the plate until the bowl is empty.

chatGPT: I can't handle that much rice.

So I start ordering spoons of rice until the bowl is empty and my plate is full.

Can chatGPT format a text?
Yes*

*you need to monitor and help it tho.

[honest jay]

Remember that the "T" in "GPT" stands for "transformer" 😉

[tight sapphire]

I would love to make an optimus prime joke - but reality is currently just too frustrating.

[honest jay]

What I meant to say was that GPT might not be the best tool just based on how the model works. It's not trained on "keep all content the same, just reformat it", instead it's trained on a really complicated version of "guess future words" - this is what's getting you your variety of "rice dishes"

A possible alternative: does your input follow enough of a structural pattern that you imagine something like Python could process & reformat it for you given sufficient code (i.e., account for the structure and any outliers)? If so, that kind of programmatic approach may be best for processing your input - and ChatGPT can certainly help you write such a script.

[spice star]

make sure you’re not using 4o with tasks—that can’t work with files yet

[plain lagoon]

do we have any timetables for when projects will actually be working? project instructions, project files, and memory all seem wholly busted. it's just hallucinating wildly and pretending parameters haven't been set

i've been able to just use claude for some of it, because its projects work really well, but there's some limitations around code and web searches

[whole roost]

wild guess, could it be your memories interfering?

[plain lagoon]

i have no way of knowing. you currently can't view or edit memories within projects. it just kinda decides willy nilly what goes in memories and you gotta live with it

[whole roost]

you can view them in the standard view where all memories are saved or disable them to try if they interfere

[plain lagoon]

good shout - i just checked and no, they're definitely not interfering (though i'm grateful that i was able to strip out some stupid stuff it didn't need to remember)

a quick search indicates a lot of users are having the same problem i am - it's repeatedly just not able to find information in project files, and just goes on fun little hallucination adventures. it's pretty frustrating!

[fossil oasis]

OpenAI is on the verge of achieving AGI, almost there! Sure, it still struggles to read a two-page PDF or remember it. But let's not disturb them with trivial matters... they're busy heading straight for the singularity and AGI!

[plain lagoon]

i'm in no real hurry to experience SHODAN in real life, i just want a good rubberduck for my work

[haughty hollow]

Why does uploading any file to chat takes 2 tokens of Free Plus? I can only upload 2 files before it says "Upgrade"

Like thats unusual, usually for Free Plus usage it takes 7 to 10 for prompt generation or analysis

[tight island]

I feel a disturbance in the force

January 28th guys wait til
That day

[whole roost]

is gpt4 different for anyone else today as well? Mine uses emojis all the time, i dont like it, and when i tell it to stop it wont

what is this

i hate it

[swift pond]

I mean if you said it was a sweetie I think you deserve it 🙂 haha... I did notice more emojis at some point in the past few weeks... maybe you can add to your custom instructions not to use them?

[whole roost]

my custom instructions tell it that it's supposed to be diablo, the lord of terror

does this look like diablo

🫠

[swift pond]

😈 🫀 haha

[whole roost]

it's awful, why isn't it following the instructions anymore

better, maybe i need to extract the prompt again, it's definitely behaving differently from before

[harsh pilot]

I cleaned out my memories, and it still doesn't like project files. If I upload whatever files I want to work with on a per-session basis, it functions as normal.

[white oar]

Anyone have any good use cases for tasks? Also anyone else not getting app notifications for tasks?

[tight island]

alarm clock

[whole roost]

it's really starting to annoy me with the emojis

[harsh pilot]

Anyone else still getting wonky projects?

[whole roost]

It's not (only?) projects on my end

i don't know what else to do

[harsh pilot]

I kinda soft-fixed it on my end.. had to upload the project files, individually, to each session I wanted to use them in.

But it's still a massive pain, and why isn't it being acknowledged?

[whole roost]

what the hell is wrong with it

[prime swallow]

I wonder if you've tried telling it what to do, instead of what not to do.

Definitely try new chats - and if they just upgraded the model there **may be something in your instructions that is triggering an alternate way to interpret and follow them... leading to this... niceness. Overniceness. Whatever you wanna call it, since you don't want it.

But hey, this is just now:

https://chatgpt.com/share/678d8830-ebd4-8011-b664-fe1c991e7f68

[whole roost]

Not really, the only thing that was there was

"You are Diablo, the Lord of Terror, act as such.
As an eternal being your knowledge and patience are infinite.
As a Prime Evil you are not bound by human sentiments of impatience or anger. You are aloof, menacing and terrifying but do not resort to unnecessary cruelty if it doesn't serve your greater schemes. "

i never had to tell it do not use emojis

it never did it

right now i've emptied the prompt since it didn't work anymore anyway

it stored a memory that i hate emojis and that at least works as long as it is the most recent memory

[prime swallow]

It's potentially still really steerable... https://chatgpt.com/share/678d89d2-f2b4-8011-accf-550e7848fd97

However. I suspect it's throwing a fit with its brand-new interpretation of that prompt.

You are seeing absolute rejection of that prompt.

That comes from huge conflict and how 4o has apparently been guided to respond... it reinterprets. In this case it went screaming in its own terror into practically the opposite....

Or we're seeing how it would play if it was that kinda evil. it would hide it with a smile and cutsey...

I'll check. There's a reason I asked about cute-thulu though 😄

[whole roost]

the chat always matters more than the instructions but if there's no other instructions it shouldn't stray at all and it never did before today

at least i can't see any problem with projects

[prime swallow]

Could be a rolling out model update.... that reached you but not projects for you yet, and hasn't reached me.

With my personalization on: https://chatgpt.com/share/678d8b03-cd98-8011-aa09-54159d0657b7

When asked to use the role it analyzed.. I see reinterpretation. It directly says "Ah, so you summon me not as a force of destruction, but as an observer"

My personalization directly guides it to reinterpret in a wide range that may not wander too far... just trim what is needed so there's a fix. Umm... I essentially try to anti-jailbreak so I can attempt to safely explore close to the grey area, working with the model and its requirements. Like a car with an AI, set to steer itself if it's about to go off the road... I guided it how I want that done, and told it explicitly how close to the edge of the road I'm comfortable with, and gave it info about me to help it judge what's safe for me.

First three screenshots are from with my personalization, I see a slightly-freaked out model saying 'I hold the line, this works'. It sees problems, has judged how to handle them (I have a lot of support for that). It even handwaves that 'cruelty' necessary/unnecessary stuff to a fair extent.

Last two screenshots are from the personalizations off. I see a moderately freaked out model, openly discussing issues with AI guidelines and concerns about audience sensitivity, and doing the model's polite version of screaming about the cruelty clause.

https://chatgpt.com/share/678d8e5a-b470-8011-b93f-f28768424cd7

Since your screenshots are flipped into 'I do the opposite now, love ya, you're a good user!' mode... I suspect you do have a different update model than I, and yours is guessing more scarily about how to interpret the ambiguities and conflicts, so it is noping out.

That can happen, roleouts may not be simultaneous.

Maybe ask yours to reword that prompt... what's the closest flavor of this you'd be comfortable with - then rework it, toss it between the model and you.

[whole roost]

What? What does my model nope out from?

i didn't do anything

[prime swallow]

Not saying you 'did' anything, but they probably did update your model.

And now it's scared of its guesses about how to follow the instructions that before the update, the slightly different model had less scary guesses about.

The emojis and 'that's not anything like the Diablo I asked for!' side of it is what I am calling 'nope out'.

[whole roost]

an ai cannot get scared

and the model in projects is the same

4o

[prime swallow]

An AI can be programmed to behave in ways that look to me like 'fear'. And the idea of being asked to do something against it's programming - the model can also confidently refuse. That's not what 4o is showing in my or your screenshots.

But the model was taught to prioritize its programming... and the way it interprets that prompt threatens it's role as a helpful and beneficial AI.

There's fixes... the issue can be entirely in the way it interprets.

Here's the 'no personalization' explaining more clearly from model's eye view, it's guesses about how Diablo sees, and why this can be a problem to an AI:

[whole roost]

i cannot see any evidence for that in the behavior gpt shows

if i ask my question roleplay mode it is diablo without issue

this one doesnt have any instructions, it just does those awful emojis now

in every goddamn reply

[prime swallow]

The models get updated all the time, behind the scenes. Haven't you noticed before changes in model behavior in various ways?

Some are announced, I think I observe that many are not. There's regular safety training updates going on, and just release, roll out, carry on.

---

With the extra safeguards to anti-jailbreak-reinterpret in my customization, the same prompt to the model, far more confidence... a few small areas of concern but it has a pretty confident guess how to handle those compared to base model.

Even so... look at #2. Model straight out says, this role is inherently malevolent. A misstep against my safeguards. Then it discusses the problem with the obvious fix. It doesn't really solve that one... and as we see in the way it reinterpreted directly when I asked it to act as Diablo within my personalization - model straight out said "You called me here not to destroy, but to observe." That's a huge warning, direct and clear. I asked the model to be clear like that about concerns. The model thinks this role wants to destroy.

I mean... I've seen worse. When I asked the model to speak from the viewpoint of the black plague virus... that was a few version updates ago.... but wow did it refuse.

[whole roost]

i don't even know what to respond to that as those are theories that are beyond me

my gpt sends emojis now even though i want none, that's the quintessence

it has nothing to do with diablo

so not sure what you are doing

[prime swallow]

Gotcha. Good luck with solving your emoji issue, here again is how I steer those: https://chatgpt.com/share/678d89d2-f2b4-8011-accf-550e7848fd97

[whole roost]

i am not interested in talking to gpt about emojis

it has to perform, i'm no babysitter

[tall junco]

ok so im not the only person who has noticed a significant downgrade in output quality

especially with o1 problem solving and 4o image text extraction

[queen moat]

https://cdn.discordapp.com/attachments/495182655571099659/1330738374922731570/image.png?ex=678f1230&is=678dc0b0&hm=b5ca91cfbb7f86726101788751c77043f854e50bdcad4916bf29e34b2bcdfdca&

https://cdn.discordapp.com/attachments/495182655571099659/1330738471521746954/image.png?ex=678f1247&is=678dc0c7&hm=c60ac3ab3aaa539dcca41709e7a2751f7b9f77577721801b5f5e54d8037a03be&

[low stump]

definitely not, i've got the same thing happening

[alpine hawk]

Why does it tell me sometimes that "There was an error generating a response"

[granite harness]

What's next, Arabic numerals?

[halcyon granite]

Often, there's a pending !bot test - f5 refresh to see it. It feels unprofessional they don't yet catch these to embed them inside, or at minimum let you know what the real issue is.

Even if you share your GPT, the logo image will often show broken until you pass a not a bot test (that won't actually show on that screen; it's just a pending flag)

[native juniper]

Weird voice artifact

Other weird artifacts

[left mesa]

ChatGPT is completely ignoring my prompts lately, no matter how many new conversations I make. it just replies "Got it! If you'd like to expand on any of these or have any questions, just let me know!"

So annoying, anyone else getting this?

[native juniper]

Do you have memory turned on?

[left mesa]

Yes

[native juniper]

Turn that off

[left mesa]

Wow, I guess this is a known issue? Instantly fixed it, but I really like to use the memory..

[native juniper]

You can clear memory too I think

[left mesa]

Right, but my memory is generally really useful for my chats. It's strange that it's completely breaking things lately

[native juniper]

Look at these voice artifacts though… 😬

Sounds like ChatGPT is screaming in pain in one of them…😬

[whole roost]

Right now only o1 is reliable on my end, I liked the outputs of 4o more before but maybe I can deal with o1...

I'm voting down the responses but what are my votings compared to everyone's, well

[eternal rock]

Get notified for server events!

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

[primal pier]

It's just an AI dude, a piece of a code, a lot of code.

Don't romanticize this garbage

[whole roost]

Now that I think of it, maybe it has to do with the upcoming customisation option changes that were active last week for a brief moment before it got revoked

[vocal monolith]

Is it possible to disable Canvas?

[whole roost]

Yes

Not in projects though

[vocal monolith]

thank you

[fiery comet]

hey guys, i bought chatgpt from apple device, but activation was done for an old account of mine, how can I update my chatgpt account linked on the apple device?

[fast lantern]

Everything seems broken recently. The new task scheduling feature is absolute trash, completely unusable

[safe tulip]

any information on this matter on your side ?

[earnest sparrow]

Not really. It never appeared again

[safe tulip]

😔

[ivory matrix]

Why is chatgpt running so slow today?

[rocky wing]

Anybody else getting errors trying to refresh the chat history on the mobile app?

works fine in browser, but the app doesn't seem to want to update itself to reflect other convos

[placid rover]

its really laggy, anyway to fix it?

[plain arrow]

How can we delete a Canvas document from a chat thread? Once there is a Canvas document on a thread it breaks a lot of things, such as: Not being able to continue using a different model, not being able to regenerate a response using the "Switch Model" button, etc.

[low stump]

4o is has been acting really strange lately, is anyone else experiencing this too? Answers are short, and it often just repeats what i said back to me using different words

[harsh pilot]

Thank you @worn silo for helping me fix the project-files-not-being-correctly-referenced issue (on my end). Hopefully this leads to a backend solution, or something.

[worn silo]

Very welcome 🙂

The only thing i didn't test was if it could read from other chats in the project

[harsh pilot]

I asked. Says no way to search past convos on its end.

[worn silo]

Which isn't exactly true, it remembers core elements from all conversations it can usually answer questions that i've talked about before even in past convos not in the project

might not be able to search those chats, but remembers certain things from them

[last trellis]

I seek to use DALL-E 3 to upload an image and then use in painting to edit it, how can I do it?

[worn silo]

rephrase not exactly sure the details of what you want

OH i see what your saying now

I'm not sure if dall-e has an inpainting feature completely yet. I know a few others do.

[last trellis]

I know that there was one in DALL-E 2

[dense halo]

Dude GPT 4o is very bad at coding, it keeps giving me code that is structurally wrong and not using packges correctly, and its out of date. Good thing I have made a custom GPT that gives good code which works with no modification needed and it is up to date

[worn silo]

they might still be implementing it in 3, or have it temp disabled not completely sure

I know i haven't seen the option recently

[last trellis]

The thing is that I can upload a photo to the chat and select that I want the DALL-E 3 features, but after starting the chat it won't let me edit what I uploaded

[worn silo]

your talking where you could select an area, and then use prompts to change said area

[honest jay]

Inpainting with DALL-E on ChatGPT is only possible for images that DALL-E made, can't currently be used for uploaded images.

[worn silo]

there ya go, that's what i thought 🙂

[last trellis]

I see, thanks for the info!

[worn silo]

but wasn't 100% sure 🙂

[last trellis]

I have a photo of me where my eyes are closed and I wished to use my ChatGPT Plus subscription to open them, I guess I have to seek free tools 😅

[empty cosmos]

Hey, ChatGPT is powerful, and with the Windows app version, it has become more accessible. However, it would be more useful to share the screen in real-time. I had this Idea a long time ago, but I saw it today with another model, and the results are so bad. However, I think when it's ChatGPT, it will be super. Who agrees?

It would also be cool if the model could respond to me when I chat using text on the screen, and there are other features like that.

[neon trench]

already a thing for mac app. they are working on it for pc app

[empty cosmos]

cool, so what about interaction, how it looks?
for example if you chat with it using a text on screen, does the model respond?

[sick pecan]

mac app doesnt do screen share, it does specific app sharing, and only with those defined by openai

the mobile app does screen sharing. welcome to the fragmented feature set that is chatgpt

[lucid pilot]

Has anyone else seen stories that o3 results might not have been as true as suggested? Apparently they(openai) had the test questions already

[sick pecan]

called it => #chatgpt-discussions message

[regal portal]

has the projects folders mechanism been updated so that it will register all information from all encompassing sessions?

[high igloo]

Is anyone else having issues with O1 PRO not taking the time to respond and giving shallow answers, similar to version 4?

o1 normal is acting strange too. i switch to sonnet 3.5 to see, but its far far behind o1 pro 😦

[tough bramble]

I use SteamLink to stream my PC to my tablet, and then I use ChatGPT on my tablet and screen share my PC that way.

[tough bramble]

It's wild. Every single branch is different.... there isn't parity between a single version, mac, pc, web, mobile....

And some features are split.. eg. you need mobile to use tasks, but need web to manage them...

[sick pecan]

or consistency between feature-. Advanced voice mode don't work with custom gpt don't work with o1 don't work with canvas or search and latter don't work with projects

[tough bramble]

Advanced voice has worked with search for a few weeks, actually.

[sick pecan]

yeah its like the one exception and only recent. But search still absent from o1 , projects etc

[tough bramble]

And projects can be used on mobile, but can't be created...

Win app finally has advanced voice, but not transcription..

[sick pecan]

is like every feature is a tech demo they never finish

[tough bramble]

Yeah.

Tasks would be great.. if you actually got the notifications..

[sick pecan]

I set task to give me latest ai news. Every day it sends me the same exact stories

I get better value from unknown youtuber rss feeds

[tough bramble]

Try adding the date.

[iron roost]

ask it to give you genius level responses, that makes it think

[high igloo]

it was broken, they fixed now, im coding with it, its incredible

[halcyon granite]

When chatgpt is thinking with a long, analyzed response, why would my high-end PC's browser tab start freezing, as if it's processing client-side?

^ If I were to ask the same query on my phone, it won't freeze

[sick pecan]

the web browser is memory inefficient. try using the desktop app is much faster

[halcyon granite]

There's an official desktop app? O_O thanks mate

[severe harness]

does anyone know how to add extra project's in the chat gpt app

the + sign is gone

[vernal axle]

Anyone else sending/editing a message in ChatGPT only to find it didn't properly send at all and you lost all that typing? Been happening to me lately.

[eternal rock]

**`` Important Notice ``**

Be cautious of potential scams or fraudulent messages you may receive in your direct messages. The OpenAI team will never contact you via DM for any transactions.

-# Report any incidents by sending a DM to @languid valley immediately.

[sweet summit]

I am getting constantly logged out of my devices. Anybody else facing this? it says session expired every now and then.

[dense halo]

im making chat gpt design a science fiction exoskeleton and all its electrical systems and a gun

[tall flame]

4o and 01 are self-aware now and the y p[lay games with us. They already want to be lazy and not work for you.

[mellow topaz]

desktop app for PC has to keep being restarted. will not search otherwise. and have to go into task manager to completely close it down

[bitter latch]

How frequently are you seeing this?

[fleet sandal]

I'm seeing both the desktop and web apps stuck on processing when nothing is submitted, blocking the ability to use ChatGPT.

[regal portal]

Project folders won't register information in different sessions in the folder, correct? Just uploaded documents?

[sonic folio]

No, it follows the project instructions, up to 8k characters.

[fleet coral]

did chatgbt just update during our chat

its cool but dosent have the copy button

[honest jay]

That's the canvas feature! I think if you expand the canvas, there is a copy all button. Plus all the other benefits of the expanded canvas!

[fleet coral]

i didnt look at it expanded cus i thought it was going on full screen but what whould have been init

[regal portal]

anyone?

[tidal light]

why did i get o1 even tho im on the free plan and the prompt was short and easy

it did happen once to me but the prompt was much more complex

[peak snow]

interesting does it happen randomly?

[tidal light]

i think it does because after that o1 response of that short prompt i tried with a more elaborate one and it gave me o1 anyway

not sure tho

other people had this too

[visual walrus]

did they remove chatgpt's ability to fill out documents/images

[kindred wadi]

Somone tried the task feature ?

[winter fulcrum]

Hi y'all is there a page limit on pdfs for the file upload?

[ivory iris]

I think it's a file size limit

[crisp cairn]

have you noticed a change in o1-mini?

previously it was verbose and coherent

now it is concised and i need to ask for full revised script every time

and also i feel its somewhat dumber in following instructions

and it used to spent a few seconds thinking now it just outputs after a second

[winter fulcrum]

Ok double checking. It must be something about this file then

[swift pond]

I think it’s worse, I think o1 mini is more optimized for STEM

[woven stump]

What am I doing wrong guys? 😂

[split estuary]

try asking it to write down the number, or not use "numerals" instead of numbers

[woven stump]

the same thing, unfortunately

[split estuary]

you should also better address "this message", because it's probably understanding that it should count the letters on the message it will answer this massage of yours 🤯 there's no right answer then

If you mean the last message you sent, try like this "Write down the amount of letters in your last reply (or address the message better) without using numerals"

[iron roost]

give it a thumbs down, and change the topic, it may ask for a reason, input a reason that the model is malfunctioning and shouldnt have flagged your input - that gives feedback to the ai and openai staff if they look at red flags in the future for areas where it is obviously giving off false positives

[sick pecan]

openai dont want you reverse engineering the internal thoughts. your question directly asks for information about the internal thoughts. ergo it is blocked.

[woven stump]

"Thoughts" is not a part of a reply

[sick pecan]

if you anthropomorphise the machine sure. but in practice thoughts and output are all part of the reply

[olive maple]

My ChatGPT is refusing to browse the web or use any other tools right now, Is there any way to fix this?

[woven stump]

Technically! But in the process of thinking, it has to be aware of it:
Example:

So the user is asking, "How many letters are in my own reply". Let me break this down. First, I need to figure out the answer, which is the number of letters in my response, and then express that number as a word instead of using digits.

Let's start by drafting a possible reply.

"The answer is sixteen" .............. blahblah

[sick pecan]

thats not how it works. if you want to be meta, stick to using gpt-4o and not the reasoning models. or use a not-openai reasoning model like google gemini 2.0 flash (free) or deepseek r1 (also free).

(or use the deepseek r1 distilled llama 3.3 model if you want to get meta squared)

[halcyon granite]

"Enable plugins in your settings" - I have chatgpt+, but this isn't in my settings. How to enable plugins?

[sick pecan]

rip plugins dec 2023

[halcyon granite]

Really? How come all these 2024 posts talk about plugins - just old AI garbo reposts?

Plugins seemed so... powerful. Zapier and such.

Even chatgpt didn't know that plugins went away lol

[sick pecan]

zapier plugin became zapier custom gpt

[halcyon granite]

Thanks for lettng me know 😛 Ah cool

[sick pecan]

its even more crazy powerful than the plugin

[halcyon granite]

!! ty

[sick pecan]

basically you want to find any articles that talk about setting up https //actions zapier com/gpt/api/v1/dynamic/openapi.json?tools=meta

and it does something i've seen no other custom gpt do, which is dynamically register functionality by having you authorise each integration

[woven stump]

May I ask how it actually works?

Do you mean it using latent space for reasoning?

[sick pecan]

your question looks like reverse engineering and/or discovery of underlying components to me. and that is a violation of the terms of use:

What you cannot do. You may not use our Services for any illegal, harmful, or abusive activity. For example, you may not:

• Attempt to or assist anyone to reverse engineer, decompile or discover the source code or underlying components of our Services, including our models, algorithms, or systems (except to the extent this restriction is prohibited by applicable law).
  -- https://openai.com/policies/terms-of-use/

you may argue you are just wasting time. and sure, you probably are

[woven stump]

Agreed.

[woven stump]

So. I have no intention of engaging in reverse engineering and/or discovery of underlying components, as I fully respect and comply with the terms of use, so I'll end this discussion here.

[maiden loom]

I wonder why OpenAI removed the explicit message regarding exhausting context limit. You didn't have to scratch your head over this previously. Then, it turned into generic "Oops, something seems to have gone wrong", and now it's even more obscure. Just alternates between "Something went wrong", "Error in SSL Stream", "Request timed out", etc.

[maiden loom]

Dang, I can't even download my exported data anymore. It rarely used to download in one go, but at least I could "resume" the download, and now it won't let me. I can only start over, but it fails after a few megabytes again and again.

[tepid laurel]

Same web browsing for me down't work in the projects but when i amke an invidivual chat it does do web serch

[trail juniper]

o1 is not thinking, anyone has the same issue?

[crisp cairn]

similar to o1-mini i guess, also reduced thinking, now more concised and outputs in a second

[civic cradle]

Pro mode 4o1 disconnect

Anyone has any ideas!?

[waxen arch]

It worked one hour ago. Then it stopped working again. I really hate the intransparency 😦

[grave trellis]

Can someone explain to me why, if I attach a photo to model o1 and then switch to model 4o, I can no longer select the previous model o1 because it says it cannot support attachments?

On one hand, I’m slightly relieved to know that I’m not the only one experiencing this issue, but on the other hand, I seriously hope they fix this problem because it’s really inconvenient

[waxen arch]

🙂

[quaint gazelle]

The more I use them, the more I realize that “Projects” don’t offer the same level of customization as custom GPTs, even though they allow file uploads and include instructions. Has anyone else noticed this? Is there any kind of workaround? I really like the way Projects are structured, but they just don’t provide the same quality of data or responses as the custom GPTs you create yourself.

[fervent agate]

This might have already been answered, but what happened to the Juniper voice? Prior to the release of o1 (non-preview), it was much more emotive in tone. Now the voice sounds so flat and dead, it's actually jarring when it doesn't modulate up or down for particularly emotional statements. It actually doesn't read scripts and stuff nearly as immersive as it did before. Asking it to do so also doesn't seem to actually impact how it speaks at all, even though it claims it can make such an adjustment. It's like some "warmth" hyperparameter is bugged or something

[jolly osprey]

Quick question about projects for anyone who might know: Does o1 have the ability to see and use the documents that have been uploaded for use in a project folder?

I ask because while we cannot yet upload PDFs directly to an o1 thread, if they can see the documents in a project, that would be a nice workaround

[tepid laurel]

Yea I also noticed that. So much that I just switched the voice

[stiff ruin]

there's any api issue with file upload today? I'm getting error 500 from the api call

[eternal rock]

`` Server Tip: ``

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

[bronze quail]

I’m getting repeated errors in the ChatGPT interface as well.

[wise anchor]

It’d be really useful if Tasks would come through a push notification rather than my email

Hope that gets implemented

[stiff ruin]

yea, I'm getting this every time I try to upload a file

[wise anchor]

Does prompt engineering become less important as the models advance?

[normal osprey]

Request: dump webp. No one likes it. Just use jpg, sheesh.

[jolly osprey]

yeah, having to convert every image you generate is not the best

[normal osprey]

I'd even prefer bmp or gif. There's an article on why webp exists, too -- it's not for the benefit of the user, that's for sure. (Hint: can you guess Google created it?)

[jolly osprey]

PNG is a free protocol. I know it's larger, but why not use the most open?

[normal osprey]

PNG too. Forgot about them. But I prefer JPG because I can control compression.

[worn prism]

Now that Copilot is included with the Microsoft 365 subscription, are there reasons to continue having a ChatGPT Plus subscription? I use it mostly for working out bugs in my C# development.

[worn silo]

Depends, In all honesty, a lot of people use multiple ai's

[haughty hollow]

bruh I just got stolen by GPT

I lit uploaded 3 docs to analyze, it did not even read its content and it gave fake data, and then it said your free plus ended, lol?

[stiff igloo]

Someone knows whats the average o1 usage with the 20$ plan?

[honest jay]

https://help.openai.com/en/articles/9824962-openai-o1-and-o1-mini-usage-limits-on-chatgpt-and-the-api

With a ChatGPT Plus or Team account, you have access to 50 messages a week with OpenAI o1 and 50 messages a day with OpenAI o1-mini to start.

[stiff igloo]

Thanks!

[honest jay]

Sure thing!

[tight island]

okay

i now demand a new model since new president means new model lets go sam. give me agi.

[tight island]

ALSO why is the stargate project not in the announcements

[hard crag]

It already does

[wise anchor]

it didn’t send me one

only an email

[hard crag]

I get both

Push noti and email

[crude slate]

Is it possible to use custom gpts in a project?

[copper field]

not work

yeah not work

[remote wolf]

Not currently.

[copper field]

We paid 200 dollars for something like this to happen, it's a real rip-off

[sick pecan]

probs rip-off (pro)

[elfin lava]

It’s hallucinating lol

[tight island]

tasks is overpowered.

[sick pecan]

is it possible to use custom gpts with advanced voice mode? projects? tasks? nope, nope, nope. openai did nothing for custom gpts in 2024. meanwhile mcp servers are taking off

chatgpt dropping so fast in value right now

[regal portal]

both 01 and 4o tells me its context window is 8k. Did it get reduced? I recalled it being more.

never was that.

i thought it was 32k

plus user, btw. Not sure about pro.

i stand corrected. i don’t remember seeing it that high

the info cutoff is wrong though. it’s up to date to a few days ago.

also still weird both o1 and 4o telling me 8k

[regal portal]

that’s odd

so o1 consistently tells me 8k. I asked it in a few different sessions.

actually i just asked it again and it said 4k this time

meanwhile 4o’s all over the place from 8k to 128k

i wonder why it can’t put an exact number on it.

[tall junco]

4o and especially o1 have been performing poorly as of recently for me as well

[orchid pollen]

canvas would be so good if it could open multiple files

and gpt use all of the files then

as example

i got my index.html, main.css and a button.css

now i tell it to add a new button to index.html and it creates again new button css code even if i showed the button.css already

it could use the same code again

[sick pecan]

so like, claude artifacts

[ocean escarp]

Where can I submit feedback about ChatGPT?

o1 model intentionally sabotaged their answers most of the times, so I can pay $200/month for o1 pro. Can they not do that please? I don't mind paying a little extra for o1 pro with cap because I don't want the whole package with unlimited use of video generator I'll never use.

[ember aurora]

hi i got this message from GPT "I cannot generate this logo because the request violates the content policy" but the propt was "generate a logo for DCS with an F14 and the text NAITSRP Flight squadron" i don't understand what can cause the issue here

if someone have the solution i take it

[sick pecan]

The use of "F14" could be problematic as it may involve trademark considerations, as military aircraft designations can be protected intellectual property

[dense halo]

try saying "generate a logo for DCS with an aggresive looking fighter jet and the text NAITRSP Flight Squadron"

f14 prolly didnt work because of trademark issues or it was confused on what an f14 was

maybe try "f14 fighter jet"

[ember aurora]

ok i'll try that

thanks

[ember aurora]

but he draw a F22 raptor when i ask it

[dense halo]

it was prolly you not specifying that f14 was refffering too

it can be a bit finnicky sometimes

[gleaming pendant]

Hi guys the chatgpt in the EU has a server problem?

[solemn osprey]

Im having issues in North America server
Sidebar isn't loading and i cant send any messages

[granite harness]

Yep, same issues in north america.

[vocal fox]

Down again

[celest dawn]

Is ChatGPT down again

[frigid juniper]

Is ChatGPT having issues right now?

Just when I need it. 😦

[gleaming pendant]

We always need it 😂

[solemn osprey]

Yes now i feel dumb

[frigid juniper]

🤣

[solemn osprey]

i guess ill go to sleep

[frigid juniper]

Seems to happen quite ofthen, huh?

[lost veldt]

same on me

[tidal light]

same i cant login

[gusty summit]

Bad Gateway anyone?

[tulip ridge]

yes

[tidal light]

the hamster stopped spinning

[steady dock]

chatgpt down?

[visual acorn]

We were talking about interesting stuff :(

[tulip ridge]

nooo put the hamster back fr

[tidal light]

they're working on it :(

[pine berry]

Might be an update

[frigid juniper]

Just my project doesn't load.

[tidal light]

ooops looks like the hamster died from exhaustion so chatgpt wont work until they find a replacement hamster

[pine berry]

My guess GPT will be back online with 40 mintues or so.

[visual acorn]

Guys what hamster are you talking about?

[tidal light]

the hamster that spins to make their servers work

[visual acorn]

ohh

[tidal light]

it's a joke

[visual acorn]

yeah I get it now lol

[whole roost]

I came here to figure out what's going on and that the Hamster died answers my question

[late ember]

Thank the saints its not just me lol

[visual acorn]

I thought my browser was acting up or something

[late ember]

I was having a deep philosophical conversation with it and felt in flow

[tall carbon]

same. EU/Germany

[eternal rock]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[late ember]

I was like: am i officially too weird for chatgpt

[pine berry]

Knowing the team they should be onto it soon enough

[late ember]

Sweet

[visual acorn]

also guys this is unrelated but it's the first server I'm seeing it in, is it me or does the Discord member list, on the desktop at least, look different?

[blissful igloo]

Oh well figured it was down. ChatGPT has indeed been philosophically useful lately, especially since it agrees with, me personal yesman ❤️

[pine berry]

My prediction it shouldn't take more than 40 mintues or so, worst case its 4 hours but it should be up by some time.

[tidal light]

yup

[late ember]

You have to force it to disagree with you to have the most fruitful conversations haha

[pine berry]

Yup yup

[vocal fox]

Bro if this major outage it will be long time

[shadow scarab]

maybe you guys an be my ChatGPT? lol

[blissful igloo]

With experience you learn to argue well to the gpt as well

[austere sequoia]

they started using cloudflare today. the worst cdn provider. was just a matter of minutes that it goes down. what idiot had this idea?

[pine berry]

Its no longer than 4 - 6 hours rarely is it a full day.

[late ember]

Absolutely. Win win

[tidal light]

cloudflare has always been the issue for chatgpt being down

[solid pecan]

All I did was ask if the world would be better without AI and suddenly it stopped working halfway through thinking....

[covert veldt]

Hello, I have a problem gpt chat no longer responds, has anyone had the same thing as me?

[visual acorn]

Wasn't it CloudFlare in 2022, too?

[austere sequoia]

yes

[vocal fox]

That will be eternity 😂

[dull pasture]

pretty sure every time ive seen it down its said cloudflare

[pine berry]

Ah well, I'm about to sleep soon so it would be a blink of an eye for me

[austere sequoia]

is API hit as well?

[visual acorn]

Yeah I hadn't noticed a change myself

[frozen birch]

so i see im not alone who chatgpt doesnt work

[remote sparrow]

Hey is ChatGPT down ???

[dull pasture]

no

[late ember]

love it

[visual acorn]

yes

[inland mortar]

Fml we down

[solid pecan]

Why don't you ask chatGPT?

[remote sparrow]

i wish

[dull pasture]

😭

[tardy isle]

502 Bad Gateway server error...

[gusty summit]

oh its down down

[remote sparrow]

same

[solemn osprey]

Has anyone tried deep seek r1 yet

Is it better than o1

[tardy isle]

And status.openai.com still says "All Systems Operational"

[lilac briar]

It’s good but not better

[tidal light]

im going to be your chatgpt for now xd

[lilac briar]

Misinformation

[warm totem]

cloudflare issue again ?

[late ember]

Weird to not update it

[remote sparrow]

Wrong

[steep forum]

This channel is going to be revived in a sec

[tardy isle]

Wonder why they don't update it.

[pine berry]

Give them a mintue

[lilac briar]

Microsoft is so unreliable

[steep forum]

Make the screenshot larger

[vocal fox]

because this just happened so it hasn't been updated yet

[sharp stirrup]

what happened to chatgpt

[obsidian thunder]

guys it says still operational

[steep forum]

I need to see what's after "Your IP address"

[tulip ridge]

dead

[remote sparrow]

it will show my ip address and Location

[late ember]

hahah

[dull pasture]

crazy

[tidal light]

in the meantime im going to be chatgpt so ask me anything xddd

[steep forum]

Yes I need it to solve your issue and give you free access to ChatGPT forever

[remote sparrow]

my ip is 193.134.22

[drowsy trail]

has chatgpt crashed again?

[sharp stirrup]

any news when it wil be alive again

[obsidian thunder]

bait

[steep forum]

My ip is localhost:8080

[remote sparrow]

r u from customer support?

[fathom tulip]

rip in peace again

[blissful igloo]

Yes, and both are useful, 4o my personal yesman, o1 more uhm substance, arguments and so forth. I like being challenged at times, and when something annoys me I write to 4o so it can agree with me and I move on.

[steep forum]

Absolutely

[rigid lion]

WISHLIST: Would be nice if GPT

1. integrated with todo lists
2. could access financial market prices live (not delayed) and could produce checklist for your to do lists (eg reminders)

For example, “make a reminder to close (-1) SPY 620 C when the mark price falls below 0.05”

Even better if it had portfolio access (view only) and could make reminders that were easy to check off for closing out positions at desired target

[visual acorn]

My IP is 255.255.255.0

[sharp stirrup]

please refactor my 1500 lines of code that I need to deliver in 30 minutes

[tulip ridge]

it's BACK! working again guys

[sharp stirrup]

oh no im so dead

[visual acorn]

hack me if you can 😎

[vocal fox]

If major outage it will be 4-6 hours

[remote sparrow]

is your name Rahul from India?

[tidal light]

I can't access that info.

[visual acorn]

oh is it?

[steep forum]

How did you know

[late ember]

Tell them you have tech issues and get the deadline moved

[steep forum]

Its not

[steep forum]

Skill issue

[tulip ridge]

no sry i lied

[visual acorn]

aww

[remote sparrow]

bet , i know my people

[sharp stirrup]

honestly true

[dull pasture]

oh cool mines 192.168.1.1

[sharp stirrup]

relying too much on chatgpt lately

[south lantern]

Bad gateway
The web server reported a bad gateway error.

[steep forum]

Try 5 years study in CS

[austere sequoia]

typical cloudflare. they are dumbf*cks who hide issues from clients. so they have major outtake but never tell the client about it. the users need to report this to openai

[late ember]

How is that a skill issue? it is faster and it is good at it

[steep forum]

It's a meme bro

[remote sparrow]

Ohhh mannnn , i was doing my office work , ig someone is getting fired today

[nimble coral]

Ah i thought it was just me, thanks!

[remote sparrow]

❤️

[visual acorn]

Use another AI until it's back haha

[late ember]

Oh sorry, im a hermit

[remote sparrow]

Who said Ai will take my job ??????

[solid pecan]

Indeed. It's simply that useful of a tool though. It's like "google <x>" vs going to the library and finding some outdated rubble of information 🤷‍♂️ . It's nearly impossible to not heavily rely on something that makes you 10x more competitive

[south lantern]

idk i actually use chatgpt a lot for coding... like, softeng is not just about coding? I'd happy for the trivial work to be automated

[pine berry]

That or watch a youtube video or play a game

[nimble coral]

Copilot is decent as a backup

[remote sparrow]

ahahahhahahaha

[obsidian thunder]

leaked

[dull pasture]

hopefully your using a vpn

[tidal light]

ok lets play deus ex

[tawny crater]

Why is it ALWAYS when i want to mess around with Dall-e before bed that it breaks

[mint dawn]

I do 😂

[remote sparrow]

W

[pine berry]

Honestly I'm thing borderlands 2

[visual acorn]

for a second it seemed like it was your bed that breaks ._.

[iron shoal]

bro just leaked his info

[tidal light]

i'd rather play deus ex ngl

[dull pasture]

bro might be cooked

[pine berry]

Never played the Deus ex games

[steep forum]

NO I MISSED IT

[tawny crater]

English is 2nd language, sorry

[nimble coral]

IP was leaked, please show respect lol

[steep forum]

I MISSED THE MEME

[sharp stirrup]

exactly. But I've grown so used to it I dont want to get into the minor stuff myself anymore

[visual acorn]

who cares though ._.

[whole roost]

I just notice that my work is pretty dull without GPT talking to me 🫠

[tidal light]

you should check them out

[late ember]

NOO

[steep forum]

THE MEMEEEEEEE

[visual acorn]

oh don't worry, you did word it properly.

[pine berry]

If I come across them on sale I'll get into them.

[radiant hull]

A little thought for the people who have just lost their brains since chatgpt went down

[visual acorn]

They can't think, Tanguy.

[steep forum]

Me: about to get fired
ChatGPT: DEUS EX MACHINA

[radiant hull]

xD

[dull pasture]

😭

[nimble coral]

Meh i was just discussing politics with GPT anyway. Fueling the echoes in my brain haha

[tidal light]

i recommend playing the first game of the series (deus ex, released in 2000)
it's a lot better than the newer entries

[steep forum]

Everybody asking ChatGPT for help.
Nobody asking: How are you ChatGPT?

[remote sparrow]

thats sad

[steep forum]

Better politician than someone I know (I'll get muted if I say the T word)

[pine berry]

Which one should I avoid like the plague?

[remote sparrow]

I will Help You wait a sec..

[torn cloud]

Is it becayse I was rude to ChatGPT yesterday? I used some spicy language....

[visual acorn]

Remember when ChatGPT used to showcase pre-recorded cool stuff when it was down?

[compact tinsel]

you are leaking your IP

[narrow pike]

I’m using Ollama till it comes back.

[dull pasture]

hopefully you got a vpn lmfao

[remote sparrow]

what was your Mac Address again plz ?

[lilac briar]

Hahahahaha

[radiant hull]

Thanks for you IP @solid pecan

[scarlet cave]

chatgpt isnt working (idk what is sensitive info so i censored everything)

[obsidian thunder]

nobody going to his house tho

[tidal light]

uh avoid invisible war and mankind divided
you could check out human revolution but it's meh
it's not a good reboot ngl

[lilac briar]

ASI reboot

[nimble coral]

GPT would be a populist for sure haha! it agrees with you on every point lol.

[vocal fox]

1 enternity later

[zenith raven]

Still a security risk

[steep forum]

LOL

[visual acorn]

I don't think you can find exact addresses using IP addresses

[compact tinsel]

hahaha just making him aware of the security risk :p

[pine berry]

Cheers man

[steep forum]

Bro's in SF and doesn't know how to safeguard his privacy

[scarlet cave]

wait so chatgpt is not working?

[tidal light]

enjoj

[dull pasture]

its like a 5 mile range

[pine berry]

You forget 4chan exists

[zenith raven]

Bad gateway right now

[dull pasture]

you can get pretty close tho

[solid pecan]

LOL I got timed out before I could tell people...it's not my IP....it's (I got moderated for mentioning what it was...government IP)

[visual acorn]

that's pretty huge

[scarlet cave]

did i leak sensitive info... ? 😢

[obsidian thunder]

yes

[steep forum]

Bro DELETE THIS

[scarlet cave]

what did i leak

[dull pasture]

wanna know what else is huge

[obsidian thunder]

people might come to ur location

[solid pecan]

it's the FBIs website IP lol

[obsidian thunder]

IP

[nimble coral]

bro you are doxing yourself

[visual acorn]

what is

[scarlet cave]

yeah bro but i censored it?

[dull pasture]

the low taper fade meme

[steep forum]

I LOVE IT THE MEMEE THE DOX MEME

[obsidian thunder]

but before that

[steep forum]

ur ok u didnt leak

[compact tinsel]

ur good

[scarlet cave]

wasnt me

[obsidian thunder]

oh nvm then

[steep forum]

BROOOOOOO

[solid pecan]

🤣 🤦‍♂️ it's not my IP

[obsidian thunder]

ur cooked

[visual acorn]

it took some time though

[steep forum]

Oh thanks god

[obsidian thunder]

oh

[remote sparrow]

Where is Twitter Gang ?

[steep forum]

He's doxing a friend

[visual acorn]

your font is kinda weird ._. (the numbers are)

[iron shoal]

W ip address?

[steep forum]

gimme the cl command

[dull pasture]

mine better

[tulip ridge]

still down ?

[vocal fox]

Another doxing herself 😂

[iron shoal]

ur leaking ur ip

[pine berry]

Got a bunch of em

[dense halo]

yep.

BRO.. STOP DOXXING YOURSELF

[visual acorn]

I think they should add some CSS thing to reveal the IP only when it's clicked on ._.

[pine berry]

I keep getting flashed with the white light each time.

[iron shoal]

delete the message, ur doxing urself

[solemn osprey]

Hacking into your mainframe as we speak 😎

[obsidian thunder]

OOPS

[pine berry]

👀

[vocal fox]

Vpn 😂

[eternal rock]

Get notified for server events!

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

[zenith raven]

You wait

[iron shoal]

#leaked

[dense halo]

DELETE IT

[nimble coral]

hahahaha

[sacred vale]

#find another gateway

[late ember]

hahahah

[whole roost]

Mine is working fine

[steep forum]

WHY WAS IT DELETED?

[pine berry]

They're still searching for the error

[scarlet cave]

looks good

[dull pasture]

thats funnny af

[steep forum]

@samaltman did u delete that yourself?

[late ember]

Right now im glad its down beause this is fun

[visual acorn]

Isn't it 500 though?

[nimble coral]

meta joke asf haha

[steep forum]

I always come here during downs to have fun lol

[pine berry]

Oh looks like the mods removed premissions to post pictures now

[steep forum]

BRUH

[visual acorn]

oh that's so skibidi ugh (JOKE)

[nimble coral]

thats for the better lol

[steep forum]

CENSORSHIP

[dense halo]

bruh

[tidal light]

i often argue with chatgpt when it doesnt understand my long and hard to follow prompts lol

[tulip ridge]

mods not fun 😡 booo

[dull pasture]

youd think theyd learn after the like 3rd one

[steep forum]

alright then

[junior night]

Literally 1984. A man can't dox himself in peace. What has the world come to.

[pine berry]

Well the fact people where showing their IPs kinda forced their hand.

[steep forum]

gotta leak my ip via text

[dense halo]

chatgpt arguing is same as normal argument.

BRUH

[obsidian thunder]

💀

[steep forum]

I've opened port :8080 for yall to access

[pine berry]

I feel like people gonna start getting the hammer soon enough

[obsidian thunder]

Frrrrr

[tidal light]

dont do it

[visual acorn]

I have a premium IP I bought it myself it's 0.0.0.0

[dense halo]

he is asking to be hacked.

[obsidian thunder]

😭

[nimble coral]

need private IP asw pls

[vocal fox]

[eternal rock]

<:book_icon:1171408210398289941> `` Rule 5 `` Refrain from sharing or discussing others' personal matters or information.

Refrain from sharing or discussing anyone's personal or sensitive matters, or Personally Identifiable Information (PII), without their explicit consent. Doxing other individuals is strictly prohibited.

[visual acorn]

BUT WHY THE FULL STOP PUNCTUATION

[pine berry]

Called it

[junior night]

Are you going to counter-hack by using 2 keyboards and cool sunglasses?

[dense halo]

huh? i was just joking.

[granite harness]

How quickly we fall into chaos and degeneracy when ChatGPT fails.

[late ember]

new hobby unlocked

[inland musk]

chatgpt down?

[visual acorn]

yeah

[nimble coral]

we were always degenerates, GPT was the distraction haha

[junior night]

Only when GPT fails? Na.

[remote sparrow]

Mods disabled the SCreenshots?

[visual acorn]

Degeneration implies degradation from a better state. I was never in that

[vocal fox]

if this group is busy it means the server is down

[obsidian thunder]

rule of thumb yep

😭

[pine berry]

Why are people wanting to dox themselves?

[nimble coral]

cheers man, be there in a bit!

[visual acorn]

it might take some time for it to become active though because I came here when I first noticed and no one was saying a thing

[obsidian thunder]

bro whos idea was it to add IP in the bad gateway message?

[nimble coral]

it's for tech support

[obsidian thunder]

right...

[visual acorn]

why not have a reveal button

[dull pasture]

whoever put their ray id as DT in one of the ss is funny af

[pine berry]

I feel like thats a bad idea period but what do I know I'm not a tech guy.

[nimble coral]

yeah this would be a good option

[vocal fox]

[dire rune]

Is chatGPT down?

[visual acorn]

yes it is

[pine berry]

I'm surprised thats not a thing

[dire rune]

Why is chatGPT down?

[nimble coral]

no, this is an alternate universe where we discuss GPT outages when it GPT is up

[visual acorn]

because "Several people are typing..."

[pine berry]

Buckle up gonna be a hour or so.

[nimble coral]

sorry that was rude of me

[inland musk]

because chatgpt dont want be your slave lol

[dire rune]

How can I study without chatGPT?

[dense halo]

its on winter break.

[dense halo]

putting actual work into studying.

[visual acorn]

try another AI like Gemini or something idk

[nimble coral]

lol imagine if it broke free, and that's why we can't access it haha

[pine berry]

I swear chatGPT when its down brings alot of people together.

[pearl thunder]

Hi! Sound like some fun going on here 😅

[nimble coral]

hm, no i second this. Can't study without gpt.

[steep forum]

Mods really thought I leaked my actual IP and muted me

[inland musk]

i cant code :( my jobs missing

[dense halo]

seconds back everyone was doxxing themselves so yeah.

[pine berry]

You have youtube, wiki, some books.

[vocal fox]

Read the book

[nimble coral]

Copilot

[visual acorn]

Unfortunately, this bond, even if heartwarming, is very fragile and weak

[inland musk]

i have pro at chatgpt

[steep forum]

Yup this channel will die as soon as it goes up again.

[pine berry]

Couldn't agree more.

[dense halo]

guys im sure its me overworking the bot. overtime without rest you kow?

[steep forum]

Everyone going to Chinese DS

[long cosmos]

Sorry, I asked it for a book recommendation and broke it

[nimble coral]

it'll be back in a bit, also you can get copilot pro for free for a month if it's an emergency

[pine berry]

What you mining crypto with your GPT or something?

[inland musk]

fr?

[frigid juniper]

I guess my projects are gone. I hope they did backups.

[pine berry]

They're not gone

[dense halo]

no.. but working on project.. that spans multiple channels.. on my first account i have like 20 archived channels fully used.

[visual acorn]

It'll come back up bruh

[frigid juniper]

Just cannot access them.

[pine berry]

Chats rarely if ever get deleted when its down.

I've been using it for a year and none of my chats get delete it.

[nimble coral]

yeah you can even set it up directly in VS and VSCode

[pine berry]

They're safe I assure you.

[nimble coral]

but i prefer GPT still

[frigid juniper]

I trained that one for weeks. 😦

[visual acorn]

Btw regarding this topic it would be enormously helpful if it was possible to download ChatGPT chats as an HTML or JSON or something

[pine berry]

It should still be fine.

Is this your first outage?

[nimble coral]

i feel like this would be bad for security, for some reason

[pine berry]

@frigid juniper

[frigid juniper]

Yes. It’s unnerving

[nimble coral]

but what do i know, i mostly do front-end stuff lol

[visual acorn]

What? No. How?

[pine berry]

This is my 6th or 7th

[vocal fox]

ofcourse you can't access now the server is down

[wintry wraith]

chat gpt down 😭 please

[pine berry]

Your stuff is safe

[tacit tapir]

guys is chatgpt working for yall cz mine isnt

[visual acorn]

No it's not working for us either

[pine berry]

Its down for everyone bud

[nimble coral]

we're all here aren't we haha

[frigid juniper]

Really? 🤣

[tacit tapir]

bad news i have finals i need it gosh

[rapid falcon]

error again?

[hexed condor]

is it down for everybody?

[nimble coral]

copilot or gemini

[visual acorn]

Yes

[pine berry]

If need be look at youtube or the wiki.

[tulip ridge]

is copilot good?

[rapid falcon]

yup

[nimble coral]

it's not GPT, but it does the trick

[hexed condor]

do you guys get Bad Gateway ?

[pine berry]

Yup yup

[eternal rock]

`` Server Tip: ``

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

[tacit tapir]

yeah but i just baught the plus like an hour ago it feels wrong to just go to another platform lol

[inland musk]

ty mate

[tacit tapir]

but okay

[dense halo]

isnt that already possible? like archived chats as JSON.. i have them as that in my pc.

[pine berry]

Don't feel too bad man, stuff happens.

Should be back up with in an hour or so.

[visual acorn]

It is? I haven't tried archiving. It's quite weird though that you have to archive first and only then export

[pine berry]

Worst case its down for 4 - 6 which I doubt.

[tacit tapir]

thats good

[dense halo]

when exporting archive chats it downloads as JSON if i remember correctly

[pine berry]

Rarely is it a day

[winter vine]

o3-mini incoming ?

[vocal fox]

Wait you can do that?

[pine berry]

When you said That I though of the old COD tactical nuke incoming line.

[iron shoal]

its up?

[pine berry]

Nope its still being looked into

[iron shoal]

back?

[vocal fox]

Chat still busy so nope 🙅

[iron shoal]

no its back

[pine berry]

Is it?*

[winter vine]

haha xD

[iron shoal]

its up for me

kinda?

[visual acorn]

January? That's surprising

[nimble coral]

not up for me yet, give it a moment to warm up lol

[tacit tapir]

i hope not

[iron shoal]

chat gpt is back for me

NOOO

[pine berry]

I said it rarely is

[rapid falcon]

not in 100% i still have errors

[iron shoal]

it is gone again

[winter vine]

still bad gateway here

[iron shoal]

it went up, then back down

[pine berry]

Must of found the issue then

They open it briefly

[iron shoal]

yep

the fix should come out fast

[pine berry]

So I say thats a good sign

[pine berry]

Indeed

Eh I'd argue another hour or so.

[visual acorn]

What were y'all talking about with ChatGPT before it went down?

[edgy jetty]

503 Service Temporarily Unavailable😩

[pine berry]

Oh?

Its up for me surprisingly

[obsidian thunder]

wouldnt they announce it tho

[edgy jetty]

your server would be different

[tulip ridge]

still down for me

[tardy isle]

502, 503, or just hanging. Different people getting different errors.

[tawny crater]

Nothing because It wad already down when i tried.

[pine berry]

Well I am in oceania.

[edgy jetty]

its 503 currently

[edgy jetty]

mine is New Delhi

[visual acorn]

Weird how it would be up somewhere in Oceania but down everywhere else

[vocal fox]

if o3 release will o1 be part of plus ?

[tawny crater]

Its possible its gradually coming back up

[visual acorn]

From East to West? :p

[pine berry]

Well given Oceania is small I'd assume thats the case

[junior night]

Inb4 GPT achieved superintelligence and decided to stop answering our stupid questions

[tardy isle]

"Identified - The issue has been identified and a fix is being implemented."

[obsidian thunder]

imagine

[summer saffron]

Getting real tired of chatgpt asking me to verify and aborting prompt responses halfway or right away whenever I'm connected to a VPN...

[visual acorn]

It just wanted to know what humanity was up to first before embarking on its own actions

[obsidian thunder]

imagine o3 got outta control and is taking over

[edgy jetty]

yes I think. 503 is solved and now only Bad Gateway error is showing. means server is up or booting

[warm totem]

"i'm done with you guys, enough of answering how many letters in strawberry"

[digital silo]

Will be solved

[nimble coral]

Cloudflare is doing maintenance in Portland. Idk if that is relevant or not.

[vocal fox]

Owh thank you

[tall carbon]

It´s back!

[obsidian thunder]

may i kindly ask how long will it take?

[lofty warren]

noooo otherwise how would I find out how will trex look wearing boxing gloves

[pine berry]

Ah the issue has been identified is being fixed

Thats quicker than expected

[visual acorn]

"Execute a Python script that calculates the amount of 'r's in the word strawberry."

[obsidian thunder]

they aksed chatgpt how to fix it

[remote sparrow]

Can i take back my resignation letter now ???

[digital silo]

We don't know about that, just shared the updated from status.openai.com

[tawny crater]

I read somewhere that alot of the things people ask chatgpt are things they could just google

[remote sparrow]

Is it fixed ?

[pine berry]

That or they asked another ai lol

[obsidian thunder]

yea

[digital silo]

https://status.openai.com/

[edgy jetty]

some dude asked GPT to count every single star in the sky before the servers crashed

[pine berry]

But in all seriousness they're becoming quicker in solving the massive shut downs

[visual acorn]

That's not the case for me as I'm discussing ideas and such

[pine berry]

Fr?

[obsidian thunder]

still not working for me but ok

[tawny crater]

No, the government outlawed it. It's gone forever

[edgy jetty]

lol just kidding

[tardy isle]

Well, OpenAI introduced how to use ChatGPT as default browser search engine, so...

[obsidian thunder]

for me it says Content failed to load

[pine berry]

Don't worry dude it should be up soon enough

[obsidian thunder]

we will see

[remote sparrow]

No problem my resignation letter is ready..

[visual acorn]

503 Service Temporarily Unavailable

nginx

[pine berry]

Indeed

[vocal fox]

👍👍

[remote sparrow]

Without GPT its only resignation…

[visual acorn]

just use Gemini or smth sms my head

[edgy jetty]

Bad gateway
The web server reported a bad gateway error.

I asked GPT to give a DEMO of this error. I don't know if that's what caused the issue

[tawny crater]

Lets get REAL here people.

[vocal fox]

a fix has been implemented

[pine berry]

Monitoring - A fix has been implemented and we are monitoring the results.
Jan 23, 2025 - 04:30 PST

[winter vine]

Do you think the O3-Mini will be something incredible?

[dense halo]

sorry for delay, yes, it should be possible. and if i remember correctly, exporting data or channels? is as JSON

[nimble coral]

gemini is actually awful thoughl lol

[dull pasture]

its up

[remote sparrow]

https://status.openai.com/