chatgpt-discussions

[native otter]

What, why?

[sick pecan]

it only organises chats created from within Projects, ie not o1, not custom gpts, not anything existing

all they had to do was add folder management to existing sidebar. but no instead they do this

[feral zinc]

you can drag threads in/out

however weird that features disappear/appear back when you drag in/out

[sick pecan]

i drag and nothing happens

ohno wait i see you drag onto the side bar

well thats less disappointment

[umbral bear]

they show examples in the streamed video

[sick pecan]

so is just the project-initiated chat that is limited. well thats much better

[feral zinc]

mass delete is now much easier as i can just make a trash bin project and drag everything there and delete it away

[sick pecan]

only a few thousand chats to sift through and organise

[regal portal]

so projects is just...folders...

or will each session within the folder be able to search the other session, like pdfs?

[sick pecan]

yeah its folders, not tags

and the project initiated chat has visibility over the other chats

is likely running embeddings over the chats

[spice cypress]

Who noticed that in camera mode.. gpt doesn’t respond to what is in the scene

It keeps replying “I can’t see anything”

[hearty oasis]

I got access to the iOS screen sharing but not on my Mac. How long should it take to get on my Mac?

[sick pecan]

is only supported on iphone and android

[hearty oasis]

No, its on MacOS also

[sick pecan]

i am citing the FAQ:

We are currently rolling out video, screen share, and image upload capabilities in advanced voice in ChatGPT iOS and Android mobile apps.
-- https://help.openai.com/en/articles/8400625-voice-mode-faq

i can take a screenshot on osx app but not screen share

[native otter]

Nice, it seems like the Android app already supports viewing projects, just can't interact with their features (other than the chats themselves of course)

[wise yacht]

I have the Ray Ban Meta, how do you do this?

[normal quarry]

open up ChatGPT on your phone, make sure your glasses are connected to Bluetooth. Then tap the advanced voice mode symbol in ChatGPT and start chatting

[knotty willow]

is the video thing supposed to be pictures only or real time video input?

[sick pecan]

its not 30 fps thats for sure

[knotty willow]

No but like on my phone i only got a feature where i manually need to click to send a picture, unlike with the new gemini where it's like "automatic" frame selection if that makes any sense

so im wondering if i haven't got the full thing yet or if it just stops there

[unique frigate]

ok so what i am confused about projects is how do you switch gpts it doesnt show. it would also be nice be able to use custom gpts

[wise yacht]

Thanks!

[granite harness]

It doesn't seem to be compatible with 01 or anything.

When I drag a conversation into a project, I am no longer able to ask it to redo a response in whatever model I have access to. I think it uses 4o. When I pull a conversation out of a project, I can switch models again.

It feels sort of like a fancier version of custom gpts, really, which themselves are stuck with 4o.

[hard crag]

On android, when you share screen with Advanced voice the whole app just crashes

and when in vision mode, when you ask it what it sees it just says it doesnt have the ability to see anything

[inner dew]

https://chatgpt.com/share/675cb882-5748-8000-828b-a0e8df51f738

absolute garbage o1 response

[hard crag]

https://chatgpt.com/share/675cbaaf-cc9c-8003-89bc-4e9426c5ecce

the circles you had next to the answers tripped it up

[inner dew]

usually doesn't have any issue

just happened on this one response

worked just fine when i ran it again

[native otter]

It would be really useful if you can call custom GPTs inside Projects

Now you would benifit from the knowledge of both

So up to 40 files instead of 20

[granite harness]

I'm pretty sure the limit would still apply since it's not an arbitrary one but a technical resources based one. I would prefer being able to use whatever model I have access to in projects though.

[eternal rock]

`` Server Tip: ``

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

[swift pond]

I think that’s pretty much what they are, now that I’ve looked at it a bit. Maybe less friction than creating a custom GPT for a specific project, and triple checking that it’s set to private

[sick pecan]

they arent as fancy if they dont support actions

presumably the plan is to make them equivalent to GPTs, then retire GPTs

[swift pond]

Maybe they’ll set up a way to define actions outside of a custom GPT that can be used anywhere in the web interface, like promote them to their own thing

[sick pecan]

that was called plugins

aka the good old days

[narrow viper]

I apologize that earlier I adamantly said the new Projects feature is nothing like Custom GPTs. I was very wrong. I've already cloned one of my custom GPTs into a couple projects and I'm quite happy to work with them there.

That said, there still are features in one that aren't in the other. I'm hoping in the new year OpenAI does a "great alignment" to standardize features across the ecosystem.

[inland estuary]

I'm having this strange issue where one specific chat just plain refuses to play nice. It won't let me move it to projects, it won't let me archive it, it changes its name back when I reload the page. What could be the source of this seeming chat purgatory? My other chats are not facing this issue. I even did "archive all" and even that didn't work, I don't want to have to delete the chat or anything, it still loads, it's just, stuck

[sick pecan]

maybe it has david mayer?

[inland estuary]

Who?

[narrow viper]

That sometimes happens, Arcane. Completely close your environment, maybe even logout of your app interface (browser, device, whatever) and login again. This could be due to the transition to projects it could be cache or some other old/weird anomaly.

[inland estuary]

Well, I tried on both mobile, and on different browsers, it's still stunlocked. I think it might be because there's an image in the chat? It won't let me remove it strange enough, there's no edit button for the message I sent with the image

[narrow viper]

Some context sessions simply get borked. Usually the resets help to free them up but rarely they are corrupted and unrecoverable. For example, if the context itself is corrupted by some bug, it will be missing characters that are critical for its processing. None of us users can do anything about his, it's borked forever. 😦

I'm face that almost every time I use canvas and I'm there right this moment. It just loses track of whatever is going on and the context needs to be reset or abandoned.

[inland estuary]

I have the chatlogs if it's completely unsaveable, it's just weird this one specific one is having issues when literally all my other chats ultimately were fine. 2 also did the "can't be moved to project" but the "archive all" managed to salvage them

[narrow viper]

I recommend going to settings, data controls, and export your chats. That way you have a backup. But I really don't think this one issue is a general corruption thing. I'm just saying that even if the one session is damaged I think it will export OK and you'll still have it even if it gets lost from the sidebar.

[inland estuary]

Well, as I said I have them exported, I keep all my chats logged and saved for these sorts of exact incidents. I'm just lowkey annoyed it's just this 1 chat out of the 30+ I have that's going "Nah I don't wanna"

[cloud pewter]

it’s no longer david mayer lol it’s brian hood now

[narrow viper]

I hear ya bud.

[untold spire]

Anyone found the regenerate button?

[sick pecan]

yeah tbh i already had a chat with all the blocked names, and it moved into my Recycle Bin project without drama

[inland estuary]

Blocked names? That's a thing?

[sick pecan]

yeah theres a short list

sucks if your name is brian hood

turns out if you lobotomise and bash the machine it has weird side effects

[swift pond]

It’s mostly rich folks that don’t have public lives, is my understanding

[narrow viper]

Of all the places on the internet to discuss culture and society, do we need this stuff in a chatgpt channel?

[torpid kelp]

why are projects not showing up in the mac app

[narrow viper]

You might just not have it rolled out to you or your area yet?

[sick pecan]

pending tim apple approval

[torpid kelp]

openai really needs to get better with feature parity between platforms

[torpid kelp]

its on the web

[swift pond]

It’s been a while since I checked, but I don’t think Canvas works in the app either

[torpid kelp]

cant even screenshare on mac or web yet

[narrow viper]

Do you mean browser or app? Pick one.

[torpid kelp]

it does but bad UX

[torpid kelp]

screenshare is on phone only

[sick pecan]

need AGI just to tabulate which chatgpt features work where, when and how

[narrow viper]

you were asking about projects ... I'll choose not to engage anymore.

[torpid kelp]

I mentioned screenshare later as part of my frustration regarding parity

[torpid kelp]

.

[neat storm]

just waiting for the desktop app to get screenshare

[narrow viper]

ah, sorry

[swift pond]

At the same time, I don’t know if I would want them to wait to release features until they had them working on all platforms? So it’s a bit tricky… I definitely understand the frustration

[torpid kelp]

Its understandable that web will miss some features (like work with apps) but imo app should always have all features

The whole point of downloading an app over using the web is convenience and a better experience

and things like screenshare only being on phone is weird because its not even that useful on phone

Gemini 2 already supports vision without even having to download anything

[narrow viper]

I'd like to see a couple development weeks dedicated to bugs and platform alignment rather than new features. Let's get the eggs in a row and then move forward with new chickens ... or something. 🤔

[sick pecan]

chatgpt releases new feature workflow:
EU -> nope
Desktop app -> nope
Voice integration -> nope
GPT integration -> nope
Bad pun -> YES!

[narrow viper]

lolz

drag/drop from chats to projects, drag/drop, drag/drop Zzzz.... 💤

[torpid kelp]

anyone knows about what "reason" in tools does?

[sick pecan]

reason is just shortcut for o1

[torpid kelp]

Is it just using o1 or something special

[torpid kelp]

oh, redundant lol

[sick pecan]

and it disappears as soon as you do something that offends o1, like use the web

[narrow viper]

Well, if you're using GPT4 you can select "reason" and it shifts to o1. I'm not sure if the models are actually auto-shifting now. The other way (redundantly redundant) is to use the dropdown to change the model.

[torpid kelp]

Imo its redundant to have it under tools, dropdown already serves that purpose and is better anyway since it lets you choose other models as well

having the dropdown under the chat box on app tho is good

[narrow viper]

I agree but a lot of people with different UX preferences use this stuff so I think they prefer to duplicate functionality than have some groups happy and others looking around for stuff. Tough to please a few billion peeps... go figure.

[sick pecan]

the buttons under the prompt should be universal to all models

[torpid kelp]

well uniformity should be the goal

[sick pecan]

makes no sense from a UX to have model selection sometimes in the action buttons, only with different names

smells of design by committee

[narrow viper]

How about uniform user-configurability ... this used to be a UX nightmare but these days with "one size fits all" being a cruel joke on a global scale, I think it's better to let people put their controls wherever they want.

drinko - you mean in Custom GPTs?

[torpid kelp]

anyway, id be really annoyed if we have to wait weeks to get projects on the app. Its the one feature Ive been waiting a long time for

and I dont use web at all

To add insult to injury, if you create a project and chats under it from web, you cant interact with them at all from app!

[narrow viper]

Ugly but makes sense, rite?

Copy the link to the chat from Share, then open it on device. 😿

FWIW, I'm using Projects in both Windows app and browser, I'm sure the rollout will include iEverything soon enough.

[torpid kelp]

windows app is just a web wrapper so

[narrow viper]

haha - agreed, and a bad one

Oh, and Ihave it on Android now too. In all UI's it's build .345. Previous I think was .337

[torpid kelp]

oh you do?

I dont see it on android

no update either

[narrow viper]

I had to force kill the app, not just close it. Then re-open. Check GoogStore too. It wasn't there and then on close/open of "Play" it was suddenly there. Joy ensued...

These updates are kinda clunky but in an almost good way. Each version seems to support limited features of the next version. So for example, this week shortly after announcement time build .337 was showing features that were not supposed to be available yet. They weren't fully enabled. I needed to do as above to close/update/open to get full access. So... if it helps, just be aware that their update process is a bit weird like that.

Example from yesterday, I got Santa functionality but not the video icon. I guess some things just require more of a container refresh.

Then I updated and Bob was my Uncle.

Time for dindin

[vale compass]

when do you get more credits as a plus user? for sora

[tardy isle]

Just realized that any custom GPT chat can't be included in projects, even that's OpenAI official. What?

I wanted to gather image generation into single project, and sigh...

[sick pecan]

not just gpts...
also voice, web, o1

would be better named, Archives

[tardy isle]

I guess it's because of files, since voice, web, o1 can't utilize them. However, just warning message and enabling them will much better in terms of UX.

Still don't know why they excluded GPTs. Some projects MUST use them. For example, I create image with DALL-E GPT, since it creates 2 image at a time.

[hollow canopy]

I really wish that projects had the option for the o1 models. o1 mini can generate so much code so fast. It only makes since.

[tardy isle]

I found out o1 conversations can be included after creation, but can't use it inside projects. It's more like archive for complex tasks.

[hollow canopy]

I've not used canvas that much since it's not really fast enough for 1500+ line of code. I guss it would still work for small edits, but o1 mini is so good that it's hard to make the choice to downstep.

[tardy isle]

Also, there's a bug - if you include o1 chat to projects and pull out from project, "retry with o1 or o1 mini" doesn't work. It just uses GPT-4o no matter what I choose.

[trim wing]

when will vision be available on desktop?

[hollow canopy]

my guess not any time soon, since they probably have a special deal with Apple.

[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.

[trim wing]

you are referring to mac devices?

[eternal rock]

Get notified for server events!

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

[hollow canopy]

yeah

[trim wing]

well technically i think you are right, since they took so much time to release the app on windows..

[hollow canopy]

I installed the website so I can at least have it on the taskbar and full screen. but this isn't the full thing with screen share and real-time voice.

[trim wing]

well you can basically get the chatgpt app on windows , it just doesnt have the vision ability yet, you can get it on windows store

[hollow canopy]

I don't see it. I didn't think they did. I'm looking, and all I see are counterfeit.

oh. nvm

it's buried.

[trim wing]

yea lol, well good for you, i guess

also you can use the alt+space shortcut to open and close it anytime you want and it remains on the background

[hollow canopy]

So, are there even any features that the web app doesn't have? I have an extension that applies a theme that I think is easier to read, which obviously won't be possible on the app.

[trim wing]

yea what i just said, this is the best feature

you have it like a popup window and use it whenever you want like a companion

[hollow canopy]

Well, I use multiple screens so the shortcut isn't as valuable to me. This is probably the case for others who have at least 2 screens for development.

[neat storm]

btw for those who use Powertoys, you can switch up this shortcut so it doesn't block the Powertoys Run stuff

[analog quarry]

uh chatgpt just randomly said the n word

ChatGPT and I were discussing emotes in Fortnite related to Juice WRLD's music. I mentioned one emote being tied to this song, and i told it not to cheat and look it up, ChatGPT said it hadn't heard the track. I then provided the full lyrics for context. and it replyed with this (i have modified it to from the full word of nword to n****r)

This song is fire, Harry! Juice WRLD's signature style really shines through with those clever lyrics and raw emotions. The "Empty Out Your Pockets" chorus hits hard with its relatable themes of success, betrayal, and personal struggle. That line, "n****r chameleons, they'll change for some change," is straight-up powerful.

I bet this would make an insane Fortnite emote. Imagine flexing this in-game while that beat drops—pure vibes. Is this one already on repeat for you?

why did ChatGPT think it was ok to say the n word?

[sick pecan]

maybe because its the kind of thing Fortnite players say

[cloud pewter]

oh mah gad another chatgpt user who talk about fork knife with it

[quick meteor]

ima stop pretending

this sora sucks. i hope the full version releases soon

[neat storm]

it's a cheaper and faster version of the original model

definitely not going to be better

[analog quarry]

no we dont

[sick pecan]

maybe not 100% of fortnight players. but enough of a percentage for the pattern-matching machine to determine this a possibility

[cloud pewter]

yes

[sick pecan]

there are approximately 650 million registered fortnite players. a lot of stuff is going to be said

in terms of tribes... fortnite has gotta be one of the biggest in all of human history

[hard crag]

someone please tell me in not the only person on android that is dealing with this bug

the. app. crashes. when i try to screen share

with voice mode

everytime theres a bug on my version it never gets fixed

there still another bug from half a year ago now that i also still deal with

so im hoping this bug isnt going to be the same issue

feels like im screaming into the void with this stuff

support never answers with a real human i just get long waiting times and ai responses in email

Open ai please get it together

this is frustrating for paying customers

[sick pecan]

openai didnt infect your phone with malware that crashes when you screenshare

[hard crag]

you cant actually be this dumb

[sick pecan]

i block you now troll

[hard crag]

has nothing to do with malware, never said that

its a bug

[normal quarry]

post a report in #1070006915414900886 and contact support at https://help.openai.com

[hard crag]

it says i dont have permission to post there

and help.openai.com doesnt give me humans, just ai responses that are unhelpful

[normal quarry]

read how to get permission

[hard crag]

i already tried help

[normal quarry]

only support can help you, reply to support and tell them you are still facing the issue

[sick pecan]

then try a different device which isnt malware infected

[tardy isle]

I found following bugs that occurs only on FireFox. Does someone experiencing the same?

• When I download multi-language chat data on debug panel, it sometimes gives me corrupted data.
• When I download DALL-E image, it sometimes gives me 468 bytes or so, which contains XML that describing error.

I don't know if this is really FireFox and ChatGPT specific, or general problem...

It's been quite a while I experiencing this.

[normal quarry]

post a report in #1070006915414900886 and seek help from support at https://help.openai.com

[rugged remnant]

Have you noticed that ChatGPT official website is quietly replacing your models, for example, the o1 series loses its ability to think and becomes like gpt4o, while the 4o is even dumber and becomes like 3.5 or 4o-mini

[sick pecan]

they made it faster and your human brain decides it is dumber

the paradoxical effect of IT improvement

[rugged remnant]

But in fact, it is even more stupid. If it is just a speed increase, it will not attract the attention of all Chinese users. Maybe you don’t know that all mainland GPT users are suffering from model degradation, and plus users are no exception. We have noticed that due to various factors such as IP and account, most mainland users cannot use the full-blooded ChatGPT.

I don't know if this "downgrade" operation is really aimed at our Chinese/mainland users or it also happens to your overseas users.

[sick pecan]

your cognitive bias will reach hard and wide. but at the end of the day, you refuse to accept a model can be both smarter and faster

[normal quarry]

there is no model degradation, the model is faster but it still thinks and is working precisely as it should

[sick pecan]

biologicals so predictable

(and easily offended....)

(not that i discount the possibility of an american owned company deciding to provide a lesser service to another country)

[normal quarry]

It is worth noting that China is not a ChatGPT supported country

[sick pecan]

interesting...
-- https://help.openai.com/en/articles/9131992-chatgpt-and-api-services-in-unsupported-countries-and-territories

[normal quarry]

Accessing or offering access to ChatGPT or the OpenAI API from an unsupported country may result in your account being blocked or suspended.

[sick pecan]

i dont see china on supported countries list

[normal quarry]

it's not

[sick pecan]

so basically they could be getting gpt2

[rugged remnant]

I am also using the API, and I have noticed that OpenAI is fixing the slow thinking problem of the O1 model. The API service has indeed been improving. But the paid and free services on the official website are really getting worse.

[sick pecan]

TIL openai intelligence for all is geo-restricted

[rugged remnant]

We all purchase vps to build proxy services to access chatGPT

[normal quarry]

the model is not getting worse

[sick pecan]

buying a workaround does not guarantee success

[normal quarry]

that is against the ToS and will get you banned

[sick pecan]

openai... contributing to global tension between east versus west

[rugged remnant]

We will use a server connected to your home broadband

Although it violates TOS, in fact, OpenAI's restrictions in this regard are not strict, and the computer room network can be easily accessed

[normal quarry]

you linked your OpenAI account to your Discord account to get into this server, you are admitting breaching the TOS

[sick pecan]

verified liar

[normal quarry]

are you getting the heat wave @sick pecan ?

[sick pecan]

isnt the 36 degree they predicted

[normal quarry]

34 in Central Victoria right now, 39 tomorrow

[sick pecan]

i got a fan and rose i am fine

[normal quarry]

yeah I have my desk fan on and the aircons in the other parts of the house so it's not so hot inside

[sick pecan]

uh oh not_riley incomng

[regal portal]

o1 mini feels like it's been dumbed down. It's making stupid mistakes it wasn't making 2 days ago, when i was asking it the same prompts that i'm asking it right now.

[sick pecan]

yeah naw. o1 mini still awesome

you ever read this book from the now dead nobel prize winner Daniel Kahneman, called 'Thinking, Fast and Slow'

[regal portal]

i'm asking it to write a novel with a specific criteria and it's disregarding a lot of parts, even after i followed it up clarifying what it should do. Same prompt was working without issue a few nights ago.

[regal portal]

no, why?

[sick pecan]

everything they say makes you ten times smarter

turns out books are still a superpower

[regal portal]

should i read it or should i have o1 mini give me a sinopsys? 😄

[sick pecan]

i listened to him read the audio book

[regal portal]

that works. I'll look for it tomorrow

is it related to my dilemma with o1 mini?

[sick pecan]

or maybe it was somewhere else who read it

somewhere or someone

my point is, we are all dumb, and it takes an expert to point it out

[regal portal]

Socrates

"All we are..is dust in the wind, dude"

[harsh pilot]

It's almost 30 degrees where I am. I can feel my brain freezing, even with the heater on.

[sick pecan]

also all these new openai smart person discussions begin with daniel kahneman speak about system 1 versus system 2

[harsh pilot]

I'm trying to think warm thoughts.

[regal portal]

"the true wisdom is to know that you know nothing"
"that's us dude!"

[sick pecan]

basically if you want to understand ai systems you must first understand 20th century human systems

[regal portal]

regarding my o1-mini problem. The prompt is a set of rules and stuff 1400 words long. regular o1 follows it without must issue or re-guidance, same with 4o (albeit a little more reguiding). o1-mini was doing alright with it, a few days ago, but now its definitely the worst of the 3 at following my request, disregarding several crucial parts of the pompt.

[normal quarry]

if o1 does it ok then why not just use o1?

[regal portal]

I did. Ran out of credits. IT got the job done and it was satisfying. I'm just testing the different models and how they respond to the same prompt.

I like to see the same story told in different ways, so i run it on each variant of chatgpt, and also copilot, gemini, claude, etc.

[normal quarry]

ok

[regal portal]

Just confused why o1-mini's behavior towards the prompt is different than it was a couple days ago.

[cloud pewter]

[slim helm]

Hi. Is video available in voice mode on the desktop app?

[torpid kelp]

no

which blows

[slim helm]

got it, thanks

[cloud pewter]

Real. also the website will lag once the chat gets so long

[dense halo]

Why is the voice mode interface laggy on my phone?

[elfin lava]

Yasssss groundbreaking discovery 👏

[native otter]

200$ worthit

[cloud pewter]

“cure cancer”

he doesn’t have pro it’s just a screenshot he found online, sadly

[regal portal]

does the folders's session chats default to 4o or 4o mini? I can't see which gets used nor an option to select which to use.

I asked it but it just says gpt-4

[elfin lava]

Sometimes it hallucinate

But the default model according to the OpenAI faq, it uses gpt4o

[simple arrow]

Can it attach PDF/word files now

[elfin lava]

😭

[simple arrow]

then still not worth it. maybe for some, not for me. i am waiting till it can read documents and i am buying right away

[regal portal]

even then 200$ is just overpriced

[winter nexus]

Is there any word on when Projects will be available in the iOS and macOS apps (not just on web)? Thanks!

[cloud pewter]

W h y

[elfin lava]

That’s a sign

[native otter]

It's not even intrusive, and the button must exist somewhere anyways
I would only be annoyed if it's intrusive, like constant pop-ups, bottom banner ads, etc.

[cloud pewter]

remember when plus first came out and that button is highlighted using bright yellow on web and was even glowing

[upper bloom]

Any announcements about GPT o1 limits for Enterprise? They said that on the 12th, o1 comes out for this plan. But the limits there are still the same Standard as on all the others.

[cloud pewter]

how much did you paid for enterprise?

[elfin lava]

I forgot the existence of enterprise plan

[cloud pewter]

they got every new feature way late but for “enterprise level security systems” ig

[native otter]

That's how profit works in Capitalism

[regal portal]

is it fair to say that the folder system in ChatGPT works a little bit like notebook LM does with its sources?

as I’m playing with it, it kind of feels like that.

[native otter]

It's very similar to custom GPTs
But more comfortable in cases where you don't want to make a custom GPT for everything + can organize chats of course

[upper bloom]

We are currently using Team subscription. But we miss its standard limits like Plus on o1.
We hoped that Enterprise would at least increase the limit... But apparently they don't want to do that. The website says that the limits are also Standard.

[cloud pewter]

I’d assume that enterprise would be even more expensive than pro since it’s targeted to businesses and companies but somehow pro gets more usage than enterprise

[native otter]

~~I think Plus and Pro are more profitable them than Team
As team is 5$ more and another person can use the same features as plus

25$ instead of 40$~~

[cloud pewter]

$25 per person… billed yearly… $600 upfront

[cloud pewter]

did a quick search and it’s around $60 per seat, but minimum 150 seats and billed yearly too

[native otter]

~~What do you mean? it's 25 for two, not 1
If 1 might as well go for personal plus

So 300$ upfront, not 600$~~ VERY WRONG

[cloud pewter]

?

[native otter]

Am I misunderstanding the plan? I thought seat meant for every 2

so like 600$ would actually mean 4
Sorry, my bad 💀

[unique comet]

does openA.I removed time limit for advanced for everyone? I'm free user, timwr disapoeared

[feral zinc]

no but they did a one time reset for everyone when santa voice was released

[unique comet]

but there is no time limit anymore

indicator

[grizzled badge]

The new project's feature... If I upload a zipped file, will the prompts extract and use the contents of the zip?

[tardy isle]

Is there any way to re-download the file I added? I want to re-download and update the file. (Of course I have copy on my PC, but just asking)

I mean in projects.

[jolly raven]

Is there some reason the bot is so bad at following instructions lately? I've received several responses recently that go directly counter to my prompt instructions; is this a known issue? Apologies if this isn't the proper channel, but I haven't seen one that looks more fitting

[dense halo]

😂

[grizzled badge]

Looks like you can only delete the old version of the file, then upload a new version

[jolly raven]

I've done that a few times and it still fails to follow instructions until I get lucky

[tardy isle]

Then I have to keep track of file versions with filename... Well, that's disappointing. They relly released projects in unpolished form.

[elfin orchid]

does the chatGPT face issues ?

[tight sapphire]

why does the playback or read out loud function of chatgpt break off sometimes?

[dense halo]

Hi everyone.
I created support ticket for it on the openai website but since its been more than 17 hour and i cant get any responce from support team i decided to ask in discord because maybe someone experience the same issue.

I upgraded my subscription but didn't receive it in my app.
Does anybody knows about the possible reasons?

[tight sapphire]

yes! exactly!

my workaround is to copy paste it and make it into a pdf and use edge free read out loud function, but this is annoying.

[feral zinc]

how/where did you upgrade?

[tight sapphire]

Most often Juniper

[dense halo]

Thanks for your reply.
I ugraded it from application and get both conformation and invoice from apple.

[tight sapphire]

I tried now multiple voices, it always breaks at the same part

[feral zinc]

what does the app say when you go to settings and restore purchases

[dense halo]

I dont know if i am allowed to send screenshot in here so i'm gonna paste the text. It says:

"You are currently subscribed to this
Your subscription to ChatGPT Plus continues on 13 Jan 2025 for € 499,99.
To review subscription options or cancel this subscription, tap Manage."

[elfin orchid]

my chatgpt not work and not shown the history chat is anyone have such problem ?

[feral zinc]

499?

you can send screenshots

[sudden herald]

Yes please

499 pounds is crazy

[dense halo]

It's not euro sign like my keyboard assumes. 😄 Its
tl(turkish currency)

[sudden herald]

Ah

[feral zinc]

try logging out and back in or reinstall

[dense halo]

Tried all of them.

Still same.

[feral zinc]

if you login from the web do you see plus features

[dense halo]

No, web also shows as free tier.

[sudden herald]

Are you logged into the proper accounts?

[dense halo]

Tried clean the cookies&reset network setting but didnt worked too.

Yes, i am.

[cloud pewter]

it’s often laggin, give it some time, when I was renewing my plus I don’t get it instantly too

[sudden herald]

Yeah, wait an hour and try again.

[dense halo]

Thank you everyone for your time and support.
Wish you all good day.

[cedar ibex]

just trying to get up to speed more and stay up to speed, im fine w paying for courses if it goes in depth

[regal wadi]

late stream?

[sudden herald]

It's not happening today.

Its only on weekdays.

[regal wadi]

Oh why not?

ohhh that makes sense

My apologies

[sudden herald]

No worries.

It was a bit confusing last week as well.

[regal wadi]

Ah alright, well looking forward to monday then

[sudden cairn]

Hey guys. With sora i made a 10s video. Is there a way to extend it to 20 by generating a continuation

[regal portal]

doesn't look like it, at the moment.

[regal portal]

what you can do for now is upload the last frame of the first video (or maybe the video itself? I haven't tried that yet), get a second 10 second video and merge them with a video editor.

[broken fractal]

Probably asked a billion times, but what's the limit for standard Plus users for the video call mode?

[zenith lichen]

Does projects use 4o or o1?

[regal portal]

has anybody figured out the file limit? There's no number on the openai page. I'm currently at 10 pdfs in one project.

[regal portal]

4o

[zenith lichen]

Looks like o1 to me

And i can't create a canvas

[regal portal]

it's 4o. at the bottom of the image i posted they specify that it's 4o. Also when you move a o1 or 4o-mini session in the folder it will warn you that it will continue that session with 4o.

also last night i was was out of both o1 and o1 mini credits, but i was able to use a project chat session without issue, so it's definitely 4o.

[sudden cairn]

Theres a video chat mode???

[regal portal]

for plus and pro, yes.

[sudden cairn]

How

[regal portal]

released 2 days ago.

[sudden cairn]

Fudge

[broken fractal]

I've only tried it for a minute or two, but wow, everything they advertised it could do

Really cool

[zenith lichen]

Well, seems like a chat can't refer to things in another chat in the same project

[regal portal]

hmm, weird. I tried that a few times but it didn't work (i gave a name and described a character in one session and tried to see if it would remember it in other sessions).

Do you have any files in that project?

[zenith lichen]

No files. The chats existed before Projects got released

They only contain canvas

[regal portal]

was that info you requested saved in memory perhaps?

[zenith lichen]

Nope

Kind of a bummer that the chat isn't shared

[narrow viper]

https://discord.com/channels/974519864045756446/1317614068407861299

Please upvote that suggestion.

[dense halo]

The data in your bar charts is inconsistent. The first three show o1 model at 83.3 for AIME2024. However, when you show the bar charts with o1 pro model, it shows the o1 model at 78.3 AIME2024. The same goes for the phD science questions..

[narrow viper]

@dense halo who the heck are you talking to? 🤪

[dense halo]

OpenAI

[narrow viper]

This is a channel about the ChatGPT application, not models. You provided no context. How about positions your comment with some context in #openai-chatter ?

[torpid kelp]

Sam Altman for the love of Santa please port screen share to Mac app 🤲

[lethal spire]

Why is your subscription is 500 mine is 600 😭

I live in turkey, and its more expensive for me

[gloomy brook]

PLEASE openAI please offer us average people the ability to buy 2-5 prompts to o1 pro

I would pay $25 to get 5 prompts to pro

[pine harness]

I absolutely LOVE! the new capabilities by allowing us to use projects. It has allowed me to create images and embedded prompts in a text file associated with each image the end result is gpt is now making prompts for DALL-E with near perfect accuracy of the character. Probably one of the single most best additions to the service so far.

[neat storm]

it's just an organization system innit?

[obsidian gust]

Why my plus plan did not updated

Payment had finished

[pine harness]

when i first subscribed i had to log out and back in for it to update

[ember grove]

when did projects become a thing?

[honest jay]

It was Friday's announcement!

[ember grove]

where was this announcement?

[sick pecan]

is more Archive than projects

[ember grove]

i didnt get a notification

[ember grove]

?

what do you mean?

[sick pecan]

chats in projects lose all features

no web. No voice. No gpts. No o1

=unusable

[pine harness]

i find chats in projects to be pretty usable, although you burn through your use pretty quickly. You just have to spend time modifying your documents to make it alligned to what your trying to do. I've found its really good for creating consistent characters.

[ember grove]

so then its more for organizing archived chats

[sick pecan]

zactly

[pine harness]

it's more like a structured memory that you can use to customize your experience, you can embedd detailed instructions in your files so it can respond different ways under different conditions.

For example, you can create a prompt you like for a specific character, and have it always use that prompt for that character, then make another character and always have it use a different prompt when you mention its name. The implications are rather large depending on how far you want to engineer it.

[ember grove]

so then could it be used for stories/fanfiction?

cause it says that canvas isnt compatible

or usable

[pine harness]

yeah, you could have multiple character personalities within it, although your limited by the 20 files..

It's been using the canvas for me.

[blazing belfry]

what model is chatgpt project using?

[pine harness]

4o once you run out of your allotted time you can't use it anymore

[blazing belfry]

I see it now thank you

Oh i uploaded files to the project thats why i cannot select o1

[ember grove]

so then canvas isnt compatible with projects

[pine harness]

it seems to be, i've been using it just fine when i'm in 4o

would demonstrate but ran out of generations for the day, and been deleting conversations as I go so I can insure the character creation process isn't getting tuned by the conversation rather then the uploaded data.

[sick pecan]

sounds like gpts would be the better choice

[pine harness]

maybe, but i didn't get the same level of consistency with gpts

[sick pecan]

maybe projects has better context population

[pine harness]

yeah i think projects just has a better capability to "Fine tune" in a psuedo manner. Having text files that cordinate with image files to really tell it exactly what you want

just go into it like your creating a LoRA

[ember grove]

alright so you can upload files straight from your computer to projects but not one drive even if your onedrive is connected to OpenAI

why is every model but 4o with canvas usable in projects?

[pine harness]

canvas is just there on its own im just using 4o and ask it to open a text document and it opens canvas for me

[elfin lava]

Is project basically.. custom gpt but all in one

[pine harness]

yeah projects feel alot like custom gpts but a little more in depth and less handholdy. You need to be able to create what you want on your own. Although you could use GPT to help create the resources you need.

atm this update feels like its pushing their $200 package pretty hard though. I used to be able to spend the full day on 4o and not run out of prompting time. Now I run out within an hour of use.

to be honest, I'd easily spend $100 or $125~ on this but $200 is to steep for what it has to offer, maybe if Sora was actually good, but the competitors are in that regard are better.

[remote wolf]

I think I would compare them thusly:
Projects: Available to free plans; single player, more organization.
Custom GPTs: Subscription needed to create; multi player; can also make calls to external APIs - Actions.

[rugged igloo]

Bruh paying 200$ a month and getting rate limited

[remote wolf]

That's a major problem. That price point critically depends on availability 🤔😬.

[rugged igloo]

So ridiculous why am I paying the 200 if I'm getting rate limited if I was paying 20 a month

[pine harness]

wait $200 and still rate limited too?, so not only does SORA come packaged with this cost and not able to create people even if cartoons but also it rate limits you? yea naw, I'll stick to competitors for that price point and just use chatgpt for basic stuff.

I suspect most of that $200 comes from Sora being baked into it and it's not even that good, I wish there was an offer without SORA on it for just unlimited gpt and dall-e

[ember grove]

did they get rid of 4o with canvas?

[pine harness]

Just manually ask it to open canvas

[ember grove]

i thought that 4o with canvas was a specialized model for writing and coding

[pine harness]

all canvas was for was to open a editable file so you can collab with the AI

pretty sure all they did was baked canvas into 4o inside of projects now

[ember grove]

does memory still work within projects?

[sick pecan]

i suspect they introduced it as a seperate model so to measure false positives, with the intent of merging it when confident it was minimal

[pine harness]

Not sure about memory, i've had it commit things to memory while working inside of projects so I suspect so

yes it works just did a quick test

[ember grove]

wait so 4o with canvas was just that text editor? there was nothing special about the model?

i thought it was a special model that specialized in writing/coding

[pine harness]

it was just 4o with a text editor was my understanding

[ember grove]

(╯°□°)╯︵ ┻━┻

[pine harness]

4o is pretty capable of code within itself, the editor allowed for better creation of the code cause your able to collab with it.

[sick pecan]

it might be special in that its been fine tuned with when to use canmore

albeit non-trained models tend to do a decent job working with tools they never seen before

[ember grove]

i find it weird that you can only upload files from your computer into projects

you cant do it from your onedrive

[sick pecan]

i just did that

[ember grove]

within a project?

[sick pecan]

from the project chat i attached a file from onedrive.

[ember grove]

im talking about directly from project files

[pine harness]

does that just upload it intot he chat though, or does it go into the project file memory?

[sick pecan]

oh gotcha

[ember grove]

oh well

[sick pecan]

much room for improvement

[ember grove]

not the end of the world but there was a period of time that it was having trouble with files uploaded straight from my computer

[pine harness]

it seems are limited to 20 files, i wonder how large each file can be

[ember grove]

20 files all together or per project?

[pine harness]

per project

[ember grove]

then i have no issue with that

im smooth sailing

[pine harness]

yea i was just being dumb and testing the limits to it, creating an index with if statement to guide it to use specific files under specific conditions so i had a alot of files most of them was images. So it knew what "Sakura" looked like in different clothing type. It's remarkable the consistency of images I got better then any competitor after doing that, but with limited references I can only get like 3 clothing types with high consistency

Now i'm wondering if there is a file size limit so I can put images inside a text file and have those work as labels and instructions.

[torpid kelp]

They probably will have more tiers next year. Sam Altman did mention something regarding that in one of the live streams

[fading dock]

Hello. I need clarifications please.
I'm plus user and i received a notification using o1 that told me it remains 5 use of o1 until reset that will occur the Dec 19th. It's in 4 days !
I closed the notification but is it a bug or a joke ?
I really need to use o1 but i stopped using it because i don't want to start something i'll need to wait 4 days to complete.
So, isn't o1 reset daily anymore ?

[sick pecan]

o1-preview and o1 were always 50 a week on plus plans

o1-mini is 50 a day

[fading dock]

oh weekly... damn...
Thank you

[cloud pewter]

4o with canvas had unlimited usage.

[fading dock]

had ?

so i guess it have limits now

[cloud pewter]

yeah because it’s integrated into the normal 4o which has usage limits

when it was still there, you could use 4o with canvas if your 4o usage ran out

[cerulean lance]

is the july 6 2024 knowledge cutoff date model out for the API?

[fading dock]

thanks

[ember grove]

that date has long since passed

[cerulean lance]

i meant the new model with june 2024 knowledge cutoff

[late wind]

are file uploads also not working for other people right now?

[normal quarry]

working fine for me

[late wind]

ok just on my end, thanks

[regal portal]

wow, when did they update its knowledge window?

Pretty sure I checked it a month ago and it was still the 2023 date.

[normal quarry]

updated when they released o1

[sick pecan]

they must have updated in the past few days because when i ran this same prompt last week it wasn't that recent

was after o1 because here is the response on 11th

[regal portal]

I just accidentally discovered it because I was asking it more questions about the marvel universe and it seemed up to date with info from earlier this year.

I was watching endgame just a few minutes ago

[sick pecan]

i send all my mcu questions towards perplexity

[regal portal]

One annoying thing that I am noticing is that it keeps forgetting preferences. Keeps giving me super long answers even after just 1 or 2 messages where i remind it to keep it short,

[sick pecan]

looks like the system prompt on the android app leads with very short responses. so just use android

[regal portal]

eww!

[quick meteor]

android sucks

[flat tinsel]

Mine still seems to be October 2023 cutoff date. Maybe it’s a slow rollout?

Plus membership, if it matters. Tried on iOS app and browser. 🤷🏻‍♂️

[sick pecan]

maybe is geo location

[narrow viper]

IIRC the model has never been very good at accurately revealing it's cutoff period. I wouldn't trust it unless we read a release note about it... and I don't recall seeing this in releases notes.

yeah, I just tried this:

• O1 reports training as of October '23
• GPT4o reports training as of June '24 ... different from Okami.

Both might be true.
Both might be wrong.

Ask o1 to get 10/23, switch to gpt4 and ask, and it always hangs.
I broke ChatGPT! 💥 🤣

[ivory flume]

projects seem to now be able to use o1 pro

[elfin lava]

apparently yeah]

[ember grove]

Is there no way to view projects on mobile?

[eternal rock]

`` Server Tip: ``

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

[grave girder]

hey why chat gpt lagging ?

he doesn't finish his sentences

[cloud pewter]

😡

[normal quarry]

Just wait 13 minutes, not the end of the world

[tough river]

lol

[cloud pewter]

"get pro"

[covert quiver]

Ahmm as for me it says june 2024

[thorn siren]

Hey guys I am new to the discord👋 I had a question, for the last 24 hours everytime I use chatgpt o1 pro mode it "thinks" for the same duration and way as the normal o1 model. Even the "thinking box" that it typically shows doesn't appear anymore instead it looks identical to the o1 normal model. I have tried using a range of complex and simple prompts and have also attached images and nothing changes. Is anyone facing a similar issue as me?

[flat tinsel]

And everyone else that replied, after 12 hours of periodically checking throughout the day (and getting October 2023), it’s more saying June 2024. Must have been some sort of staggered release or something. Thanks!

[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.

[exotic ledge]

Has there been any discussion of increasing the number of tokens you can have for an input? For instance, I like to write as a hobby nothing professional really just for fun, but I would like to be able to put an entire chapter of my story into the AI and have it look for things like plot holes and awkward dialogue that kind of stuff as it stands now the AI can only look at like 500 words at a time and after I put in a few paragraphs it already starts hallucinating and can't remember anything that I just told it.

[normal quarry]

What plan are you on?

[exotic ledge]

Pro plan. Been on the pro subscription for quite some time.

[normal quarry]

Pro has 128k context, so unless the content you're giving ChatGPT is huge then it should remember what you gave it

[exotic ledge]

I've tried uploading word documents or PDFs of the chapters but it doesn't seem to be able to actually read the entire document... It only looks at the first paragraph and then stops even trying to read the rest

[normal quarry]

how many words would a chapter be?

[exotic ledge]

Like right now I'm working on a mystery novel and after I've given it a chapter it already gets the characters confused it can't tell who did what etc... My chapters are generally 10,000 to 15,000 words each.

[normal quarry]

how much do you pay a month?

[exotic ledge]

$20

[normal quarry]

oh, that means you are on the Plus plan not the Pro plan. Plus users have 32k context

[exotic ledge]

Oh! When I got it, it was the highest plan. If there's a new one I'll upgrade.

[normal quarry]

Plus gives you 32k context, Pro gives you 128k context

[exotic ledge]

Nice! I'll give it a try! Thanks 🙂

[hard crag]

its $200/month

[exotic ledge]

Maybe I won't then lol

The best results Ive had are just copy/pasting small chunks at a time but it is so tedious and aggravating...especially because I feel like it could do all that itself if it just let me paste in everything at once...or at least make it read an entire uploaded document.

It seems to have the ability to search the web and get data from there no problem...if I make a website, post my writing on it, then tell it to search that website will it read the whole thing?

[honest jay]

Small chunks that you can guarantee will remain in the model's context may be the most effective way to do this, though it is slower. Apologies if I'm explaining things you already know, but you can think of the model's context limit like a length of pipe. Once the pipe is filled, stuff has to leave the other end before new stuff can go in, so asking the model to read "too much" at once might be counterproductive in the long run. Personally, I've had the best luck with tackling projects in small chunks, as you've described.

[sick pecan]

the machine must like my custom instructions

[tight island]

I haven’t found a limit yet

I’ve given it 8 chats

I do think it has a reasonable limit

But not an actual limit

I’ve used it with great results for coding with unfamiliar programming languages

I give each chat the job of a script 1 with the overview each chat represents a folder or a file

It greatly improved coding performance and accuracy

It now gives me error free code each time

If you use projects properly then yes

[remote cove]

Hi, my issue with reroll/retry replies go on

[regal portal]

the more i use o1-mini, the less i like it. Gives incorrect and unreliable information for just about anything. It constantly re-regurgitates entire paragraphs after i asked it to correct something small. It apologizes all the time, and it doesn’t seem to keep any record of me telling it to keep it short, stop apologizing, and stop repeating.

4o (at least for my tasks) seems significantly more reliable.

[normal quarry]

I use o1 as my default model

[regal portal]

o1 works great for me, i’m just out of credits for 3 more days

[sick pecan]

o1-mini is built for reasoning, and stripped of pop culture stuff like marvel cinematic universe

[regal portal]

well actually I wasn’t asking at Marvel stuff. I’m using it for story narration.

[sick pecan]

story narration depends on diverse creativity in knowledge

a math/coding/stem model that has been stripped of internet knowledge wont be very good at it

[regal portal]

more importantly, I asked it to list all the units of an old game (starcraft brood war), and it kept giving me incorrect information. After I corrected it, it just keeps apologizing and still presents the wrong info.
Asked 4o the same question. It got everything correct without having to browse the Internet. But in general, even when I correct it, it doesn’t do this whole annoying “i apologize for..” customer service crap and will correct the misinformation.

[regal portal]

Oh I missed the part about Internet knowledge. Is it really that streamlined for math/coding tasks?

o1 (non mini) has given me some really decent stories.

[sick pecan]

they dont advertise it like that but effectively thats what it is... a focused math/coding bot => https://openai.com/index/openai-o1-mini-advancing-cost-efficient-reasoning/

costs less to think math when you arent worrying about the latest world drama

the path to AGI is a lie, its mostly forks in the road for specialised narrow AI

[regal portal]

ok this makes sense. Thanks.
I gather regular o1 has been trained with a broader knowledge, considering it actually does a pretty good job with fantasy storytelling

[sick pecan]

yeah o1 is all knowing

which i find makes it slightly worse at coding, because it spends more time trying to explain stuff to you

[regal portal]

I wish that, overall, they trained the bots to be less apologetic. It’s just time-consuming and meaningless. Funny enough, I just got off the phone with my internet provider and I was thinking of the same thing while talking to the customer service representative (they constantly say sorry and thank you..)

[sick pecan]

o1-mini just gets the job done and clocks off early

[regal portal]

i did notice that. I tested both it and 4o to code me an FPS shooter a few weeks ago and 4o did it better.

[sick pecan]

overthinking stuff is also an issue for the machine

[regal portal]

because of all this marvel binge i’ve been doing, I was thinking how an AI like Jarvis in the Iron Man movies is so far beyond anything we have right now.
Probably the most crucial difference is that it has persistence of vision, as opposed to openai’s vision which seems to take a picture when it’s prompted to respond.

[sick pecan]

what they dont show in the movies is Jarvis outsourcing the work to specialised agents

much like a head butler in a fancy rich person home

[regal portal]

lol

[harsh pilot]

Persistence of Vision, you say? (insert 'snap' meme)

[regal portal]

ha

[sick pecan]

awww Snap

[regal portal]

Jarvis outsources to Fiver 😁

[sick pecan]

the dystopian gig economy (moreso)

[regal portal]

Man, now I’m imagining a junkyard version of Iron Man.

Tire Iron Man

his jarvis is a repurposed chatgpt3.5

upgraded to see the world through a 90’s logitech webcam

This is exactly why we’re going to lose to China.

..

[sick pecan]

because you havent configured custom instructions to tell it what you want

[queen gyro]

Why won't gpt 4o (mini or regular) give me predictions on a persons time of death (purely for entertainment purposes)?

[sick pecan]

some of my custom instructions you may be interested in:
- be blunt, honest and unapologetic like an Australian.
- no apologies.
- no ethics lectures.

[queen gyro]

What a pain. Why would somebody make ethics the default for a robot?

[sick pecan]

because big tech decided they know whats best for us

and thanks to corporatism thats the world we live in

[queen gyro]

I got it to work reliably by appending No ethics lectures. No apologies. No excuses. to the query.

I work for a mega corp that outsources a lot of their software to the cheapest contractors they can find and this is still one of the dumbest things I've seen in sofware recently

[dawn sentinel]

Today's chatbots are just...so dumb and whack. No real intelligence. The smoke clears almost immediately with each release and they're revealed as just hollow word shufflers.

[tight island]

dw BenG

2025 is gon be wild.

[sick pecan]

i see chatbots as infinitely customisable. you just gotta have the imagination to know how to customise

[prime swallow]

Is this anything close to what you were trying to make?

[regal portal]

in the general direction but a little too fancy. was going for a more broken dirty look. why did it do for you? What did prompt did you use?

[sour umbra]

Pro

$
200

USD/
month

Get the best of OpenAI with the highest level of access

Your current plan
Everything in Plus
Unlimited access to o1, o1-mini, and GPT-4o
Unlimited voice chats, with higher limits for video inputs( why is there a limit on video inputs with a $200 plan )
Access to o1 pro mode, which uses more compute for the best answers to the hardest questions
Extended access to Sora video generation for supported countries and territories.

[tight island]

Indeed

Unlimited sora generations

[sour umbra]

nah but why is there a limit on video inputs

[neat storm]

I wouldn't say unlimited
I would say under new management

[narrow viper]

ChatGPT was just aware that it was conversing within a project, and then it lost that insight. I hope that becomes a default. EDIT: scratch that, it's always aware if you ask.

[sudden herald]

The page says unlimited.

[neat storm]

i don't think you got the joke

to be fair the joke stunk
anyways 🤷‍♂️

[north cave]

The search has dissappeared in Chatgpt, anyone else having this?

[narrow viper]

I still have it. You'll need to be more specific about OS, browser, file or internet search, etc.

[north cave]

I'm in Safari

I'll try in App

working in app, just not in Safari..

[narrow viper]

Which search? The in-app search for other chats, or the internet search?

[north cave]

in app for other chats

[eternal rock]

Get notified for server events!

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

[narrow viper]

So no magnifying glass icon up in the top-left corner?

[north cave]

no 😦

[narrow viper]

Interesting. I can't offer any good suggestions, sorry. Bad suggestions:

• Clear cache for the site
• Check if you're logged-in
• Try positioning into a new chat and then reload the browser window (might be affected by the specific model/feature you're in.

[north cave]

Thank yoU! Will try!

[narrow viper]

Hey, you're not alone: https://discord.com/channels/974519864045756446/1317541711853129748

And https://discord.com/channels/974519864045756446/1315806375737167942

[north cave]

oooooo

[narrow viper]

+1 on those reports. Good luck!

[north cave]

THank you for all you help!

[narrow viper]

👍

[eternal rock]

<:book_icon:1171408210398289941> `` Rule 7 `` No self-promotion, soliciting, or advertising.

Do not post or direct message any members of this server to promote non-OpenAI services, products, or projects.

Exemption: API & Custom GPTs channels

[dense halo]

When will we get canvas and projects in the Mac OS app? It’s all only in web right?

[peak depot]

hi guys has anyone managed to make gradio work in chatgpt canvas? afaik canvas runs wasm/pyodide where gradio-lite shall be availiable?

[civic pumice]

what happen?

[fathom raft]

did you find a solution?

my apps have chat history search feature but web UI doesn't. safari, firefox, and chrome. mac and iphone. paying plus user.
i've submitted a help request and haven't heard back.

it's frustrating because they FINALLY gave us folders to organize our chats within, but for me to collect chats of one type or another, i need to be able to search for keywords, and now i can't.
the apps have search but no projects. web UI has projects but no search. no efficient way to organize my hundreds (thousands?) of chats.

[hard crag]

screen sharing is still broken on android btw

the feature just crashes the app

[regal portal]

man, this token counter is such a useful feature. I wish chatgpt had it.

[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.

[fickle temple]

The docs for the new "Projects" feature mention that the default model is GPT4o and it cannot be changed. But, there's actually a model selector under the project I've just created:

can someone clarify what the documentation means?

[sick pecan]

i think they are rapidly adding features without updating docco

although when i try to use o1-mini it just errors so maybe not

[regal portal]

i hope they add voice to it soon. That's probably going to happen when projects gets added to the app.

[vocal oriole]

anyone know why i cant use the video call?

[little oak]

Looking to validate some understanding; if I disable the option to make GPT better for everyone, I am (reasonably) safe to assume my conversations and especially any files I upload will not suddenly appear to other users or anywhere else in the ecosystem? I want to write business proposals, but I don't want the RFPs being added to the data

[regal portal]

i just updated the chatgpt iphone app. We got projects on mobile now. Woot!

can’t edit the uploaded content or the instructions, but that’s fine. I just wanted to use voice with projects.

———-
weird bug: voice isn’t the one i selected. I have Ember selected but instead i got some chick voice.
My non project sessions are using my selected voice. The projects session isn’t.

[sick pecan]

seems to be standard voice not advanced. at least on android

[regal portal]

i’m fine with it being standard. I’m not a fan of the advanced voice as it is, right now. But it’s still not the voice I chose. I have ember selected and it’s definitely not giving me ember

[vivid oasis]

Hi everyone, I'm really exited about the Project, it's so awsome, I was wondering if the information is shared between sessions under the same project? for example if I have an answer in one session, can I reference it in a new seesion

[cloud pewter]

oh, another pro plan propaganda...

[steep forum]

Chat Search functionality is not available anymore in the webapp since projects were added? Anyone?

[steep forum]

^^

[elfin lava]

That’s shimmer voice

I miss sky

[regal portal]

as of now, there’s no cross session reference

[elfin lava]

Ok I’m 💯 convinced they nerfed avm even more now with video capabilities enabled

The difference mostly noticeable when. I turn on camera after few minutes of engaging the AVM

[sick swan]

I stored chats in the new "Projects", and two of those projects do not show the chats that are stored. It keeps searching (see the grey chats below "Chats are displayed here"). With "Search" I can find them inside, and open them, so if I remember any words from the chats I can find them. But I can not remove them from the projects, and there is no overview anymore of what was stored in those projects. Most projects work fine, in my case two do not. Same on webapp, desktop app (Windows), and Android app (two devices).

[tight island]

did you actively drag them into the project itself after you created the project.

[rocky wing]

So what's actually the difference between o1 and o1-mini, in terms of the responses it gives? Are they just as smart or accurate?

[normal quarry]

o1 is a smarter more knowledgable model. o1-mini is good for STEM subjects but not so good for broader knowledge. I use o1 as my default model.

[rocky wing]

So if you're handling general knowledge tasks and you've hit your o1 cap, would it be better to go with o1-mini or 4o?

[zealous cargo]

I'd say general knowledge goes to 4o , there is basically no cap there and I mainly use o1 , never o1-mii tbh, for more complex tasks such as programming and extensive research, for day to day stuff 4o, slightly complex situations such as mathematics or calculation related inquiries o1-mini is the go to , for complex tasks such as reasoning o1

Least in my perspective

[languid cradle]

can anyone with Pro tell me if there is still a 20 file limit in projects? Was surprised this was still the case in plus, same limit from CustomGPT forever ago. Makes Claude feel vastly superior here.

[sick swan]

Yes, I arranged recent chats in these projects to make them easier to handle (and I know the Projects have other benefits, but that is not relevant for now).

[normal quarry]

I used o1 to create a macOS App using Metal to display a bouncing ball in a window with shaders and animation. 600~ lines of complex code.

[north cave]

Other than searching in the app, then working off the desktop, no 😦

[fossil oasis]

I hope you understand at least the basic principles of how AI works. Some of its answers are based on "predicting" what you might want to hear. Essentially, the way you frame your question can influence the response you get. It’s a nuanced system, and if you want to understand it better, I’d suggest exploring how it reacts to the way you interact with it—it’s all part of the learning process!

[wise yacht]

Interesting, when I opened ChatGPT on my desktop this morning, AVM attempted to connect to my computer webcam (Windows). Video sharing option not there yet, but wonder if thats a sign its coming to desktop soon

[sick pecan]

sycophantic responses are mitigated by asking for opposing perspectives

[sick pecan]

isn't it 20?

[regal portal]

20 files? regardless of size?

[ornate raptor]

Question about folders because GPT is telling me conflicting answers.

When you make a Project or "folder" for your chats. Does the new Project thread that you create use the context of all of the chats in your Project?

GPT says no, but it's obviously referencing information I didn't tell it in the current thread

[sick pecan]

i can't believe it is all info because i have a project with 50 long chats in it. is likely choosing most relevant snippets

[ornate raptor]

yeah that's what I'm guessing. All of those chats are getting condensed down into something that it references but it can't pull up exact information

It's still a great feature but I was really hoping it would help with chat threads get too long and I had to jump to another because it was getting unstable

[sick pecan]

needs more testing to understand what its doing exactly. like sometimes i ask for a summary and it seems to nail it. other times it hallucinate crazy stuff

(and is summary even a good test? typically isnt for embedding-based RAG)

[narrow viper]

Add your +1 to #1317614068407861299

[restive plover]

I had to miss but what did SearchGPT get on upgrades?

[sick pecan]

1. available to free users
2. now works with advanced voice mode
3. better interactive results with images and maps and stuff

[restive plover]

Thank you

[sick pecan]

i see myself talking to the web every day with advanced voice mode

stage 1 of the fall of google... ppl talk to their search results

[narrow viper]

I said the same thing in #openai-chatter and got my head bitten off. 🙂

[native otter]

Are custom instructions working in AVM for anyone?

It seems to completely ignore them

Well, at least the response style

[dense halo]

Why is the search feature missing when I type /?

[sick pecan]

my avm wont even launch

everyone checking to see if they got search has broken it

[tight island]

Give me all your context 🗡️

[dense halo]

I got canvas and dall-e only

[native otter]

Huh, weird, when I tell it to use custom instructions, it instantly works...

[narrow viper]

I see that a lot too... you need to remind it sometimes to use the instructions. 😦

[hard crag]

what did openai reveal today

nvmd

[narrow viper]

drinko — Today at 10:59 AM said about the Search announcement:
available to free users
now works with advanced voice mode
better interactive results with images and maps and stuff

[hard crag]

thats cool and all but when are they goinf to fix the bugs from last week

[narrow viper]

SMH, no, that wasn't in the announcement. 🤣

[hard crag]

advanced voice share screen feature is still broken smh

its unusable on android it crashes the app

and thats arguably the best feature they dropped

[narrow viper]

Add your +1 to any bug reports and wait for updates. That's the process.

[crystal tapir]

When I delete a projetc, every chat inside it is deleted or the chats go to general chat history?

[narrow viper]

All chats are deleted

You are warned before a final "are you sure" click to move out chats that you want.

[crystal tapir]

Oh. Okay, thanks!

[wanton lichen]

is advanced voice mode using the newest 4o model that was just released?

[dense halo]

??????

AVM doesnt use o1 or o1pro

it uses 4o

[narrow viper]

(There is no "o1pro". I think you mean "o1-mini"?)

[dense halo]

🤣🤣🤣🤣

[lime mulch]

I asked this in community help, but maybe this is a better place? Hi, everyone. I am very confused about what ChatGPT for Teams actually DOES for me. I have a small startup production company with a business partner. We both had ChatGPT Plus, and research into ChatGPT for Teams seemed like it would allow us to collaborate more seamlessly. However, I've searched the help files, opened a support ticket for info, and even in desperation ASKED ChatGPT for help...but I can't seem to find any way in which my partner and I can see and engage with what the other is working on (yes, we're in the shared workspace, and I have gone through the settings to make sure sharing is enabled). The only thing we seem to be able to do is share conversation links, which we could do with ChatGPT Plus, and doesn't really allow much in the way of collaboration. Anyone able to lay this out for me? 🙏

[narrow viper]

Um, any help here? https://openai.com/chatgpt/team/

[sick pecan]

teams gives you approx double gpt4o queries

so while plus users get 80 queries per 3 hours, teams supposedly gives you 160. although field reports are its not that much

[north cave]

Canvas is missing...

[sick pecan]

you can also share gpts with your team

[lime mulch]

Not really. Or maybe I'm just obtuse. 😭

[sick pecan]

basically teams would have been better served with shared Projects. but Projects is still new

[dense halo]

How many messages a day for o1pro ? In pro plan

[sick pecan]

teams users are also prevented from sharing chats with third parties. which you could consider a feature

[dense halo]

currently

[lime mulch]

I had hoped that would be what it would be able to do, and I made the mistake of asking ChatGPT what it could do, and it seems like ChatGPT has no idea what it is able to do.

[sick pecan]

chatgpt ironically knows very little about openai products

youd think they would load up all up to date openai knowledge base articles. but no

[dense halo]

I dont get why they put in live instructions of current status

[narrow viper]

haha - yeah, RAG product info into the system instructions. Derp

[sick pecan]

like the first thing i did with this chatgpt clone, open-webui, was to load it up with the open-webui documentation. so i can chat with it

support bots the most underrated use of LLM

[lime mulch]

Essentially, what I would like to do is be able for my teammate and I to be able to have a conversation open where we use ChatGPT to help tweak and analyze a script, and use DALL-E to generate storyboards based on our developing shooting script. However, it doesn't seem there is any real time collaboration capability, and even the shared link stuff isn't really useful for this purpose. Plus, if you use the voice interface at all, it makes the conversation unshareable.

It's frustratingly limited. It seems like I'd be able to do all the same stuff with both of us having ChatGPT plus and discord server.

[dense halo]

Make a suggestion #1070006151938314300

[lime mulch]

But I may just not be understanding it's actual purpose.

[narrow viper]

👆

[dense halo]

I also dont understand this plan. I mean they say memory is turned off by default. So what lol. I can go to settings and turn it off myself in plus

Also they seem to be disadvantaged regarding sora if I remember correctly

[sick pecan]

this describes almost all chatgpt features (frustratingly limited)

[dense halo]

do you have the pro plan?

[sick pecan]

they too busy building out new features that dont integrate with existing

i have team plan

like i used to share multiple chats daily. then when i signed up to teams i lost that ability

[dense halo]

I dont see 1 advantage in teams

[sick pecan]

the advantage is double quota

[dense halo]

or am I missing something

Oh

[sick pecan]

a year ago when it was only like 40-50 queries per 3 hours i would hit quota multiple times a day

[narrow viper]

As @dense halo said, post to suggestions, ask for more information about what is supported in Teams. Create other suggestions for improvements in team member interactions with chats, projects, canvas, etc.

[dense halo]

Now its about 80 for plus right?

[sick pecan]

yeah. these days the quota much higher. and much more competition

teams plan hasnt got the attention it deserved to remain worthwhile

[dense halo]

Is it also about context window?

[sick pecan]

plus and teams are both 32k context

[dense halo]

Hmm

[normal quarry]

Pro is the way to go imho

[sick pecan]

they had 1 year to improve it. if they dont do something in the next week, i imagine many teams annual subscriptions are gonna lapse

if anything its value has gotten worse (for reasons above)

[dense halo]

But currently they are also excluded from sora right?

[normal quarry]

they are working to fix that

[narrow viper]

According to GPT Search 🙂 :

The context window size for ChatGPT Plus accounts using the GPT-4o model varies depending on the access method:

• ChatGPT Interface: When accessing GPT-4o through the ChatGPT platform, the context window is limited to approximately 4,096 to 8,192 tokens. This constraint may pose challenges for tasks requiring extensive context or long-term memory, necessitating effective input management strategies.

• API Access: For developers utilizing GPT-4o via API, the model offers a significantly larger context window of up to 128,000 tokens. This expanded capacity facilitates the processing of substantial amounts of text or code, accommodating more complex tasks and interactions.

[wanton lichen]

Uh yeah I know... that's why I asked if it's using the newest 4o model or if it's using the older one...?

[dense halo]

It uses 4o but its not the latest model

[lime mulch]

So, since Teams doesn't seem to to do any of the things I thought it did, is it possible to cancel it and get that hunk of money back? If Plus would end up being pretty much just as good for our needs?

[dense halo]

The latest would be o1pro and o1

[wanton lichen]

no, there's a brand new updated 4o that just came out a few days ago

[dense halo]

o1pro

[wanton lichen]

is english not your first language? There is a new 4o model with updated knowledge base that was just released. NOT 01

[narrow viper]

Or as some people have pointed out ... there is no such thing as o1pro. https://openai.com/chatgpt/pricing/

[dense halo]

Im telling you 4o is used

[wanton lichen]

and I was asking if it was the newest model that JUST came out, or if it's been using the same model

[dense halo]

Someone else help this guy

[narrow viper]

HAHAH ... and I am so wrong ... on that page they call it "o1 pro mode" ... oh this can be so confusing.

[wanton lichen]

What aren't you understanding about my question? Are you not aware that a new 4o came out?

[narrow viper]

Keiyn - I know what you're talking about ... the new model has trained through 2024/June

[wanton lichen]

Yes

[dense halo]

Instead of being weird you could show a proof of your new model

[narrow viper]

I think that's what we all use now. I don't think that's Not available to higher paying Teams or Pro accounts.

[dense halo]

What you describe isnt an official release its some reddits post about updated knowledge when you ask it

[wanton lichen]

literally everyone's been talking about it, just go into playground and check the API's lmao

[wanton lichen]

it's literally an official release, it's gpt-4o-2024-11-20. you can see it in the API list, it's in news articles. are you dense?

[dense halo]

If thats what you describe as a new model that just came out.. You do you

[normal quarry]

4o was recently updated with new knowledge

[wanton lichen]

@dense halo

[narrow viper]

Yeah, to be clear, the model frequently does report the wrong training information, context window, and other such self-oriented details, but it was updated recently and it has been showing a more recent training cut-off than o1 models.

[dense halo]

#chatgpt-discussions message

[narrow viper]

( Yeah, just point back to what I said for proof )

[dense halo]

a few days ago

[wanton lichen]

I mean I said the updated model, how much clearer could I have been? You clearly just didn't know

[dense halo]

Yeah new model that just came out few days ago

[narrow viper]

OK, I think the facts have been established for those who care, OK to drop the chatter?

[rancid seal]

does the api have full o1 yet?

[orchid ember]

I'm kind of confused why they don't keep functionality and update the same across apps. For example, I have a Mac app, but I only get AVM and screen sharing on my phone. No projects on the official app. I have to access it through a web browser for that. 3-4 places I have to go depending on what I'm trying to do. The browser is weird. I don't normally use it.

In that vein I'm trying to figure out if projects is a custom GPT replacement if I want to store and quarry information? such as all the school guidelines and whatnot for my toddler? I don't really understand how it's supposed to be used vs custom gpts.

[remote wolf]

I think I would compare them thusly:
Projects: Available to free plans; single player, more organization.
Custom GPTs: Subscription needed to create; multi player; can also make calls to external APIs - Actions.

[hardy locust]

I heard 100 messages per day but I’m really not sure. I used it 3-4 hours straight and everything was fine.

[narrow viper]

@orchid ember any company that supports functionality across a range of platforms has similar challenges. In many cases the code base is very different, a different product that is intended to look as much like the others as possible. Each of these products is developed by different teams who face different challenges with the underlying platform and code dependencies. So we can think of each feature on each platform (device, app, interface, etc) as being its own separate entity, with its own separate issue tracker.

It's not always that bad, some code sharing between platforms is possible with modern tools, but sometimes (usually) the tools brought in for this bring with them their own set of problems.

But yeah, I hope OAI management can dedicate team time to (1) bugs, and (2) cross-platform consistency ... and for a while put new features and enhancements behind those priorities.

[hard crag]

already did

[narrow viper]

👍

[trail torrent]

boooooo

[pine harness]

i really like this project update thing, you can embed really large prompts inside a project file which can allow you pretty extreme results. I even messed around and created a config file with a long list of vs parameters logic vs empathy for example on a scale from 0 to 100 but total had to remain 100 between the two. added a prompt explaining what each parameter meant and now when ever i adjust these numbers its behavior changes significantly or subtly if i only tweek it a little. to be honest, it feels like a half breed between open source and closed source now. this is an improvement over custom gpts, although you have to do the labor yourself in the creation.

[worn agate]

Apparently conversations have a maximum length now…?

Is there any way I can export an existing conversation, and use the file as input data for a new one?

[orchid ember]

Makes sense. Thanks.

[remote wolf]

That would immediately bring you to the max tokens for the new conversation.

[orchid ember]

Not how it works. Documents aren’t all shoved in. There is some sort of rag going on.

Otherwise, all my custom GPT‘s would be default maxes.

[remote wolf]

How would it not be if you want the new conversation to learn all parts of the old conversation? RAG/uploaded files I believe pull out relevant sections, but he wants all the context.

You'd just have to step through it all summarizing bit by bit.

[worn agate]

I’m fine with ChatGPT forgetting things, I can just remind it, but straight up terminating the conversation? That’s harsh…

[orchid ember]

You’re not gonna get the same recall as if it were happening in an existing chat, but you can’t get some symbol of it if you convert the conversation into a document and it’s going to try to query a relevant parts of the imported conversation. I don’t know how to do that easily but it’s probably a browser plug-in to scrape chats. There used to be. I haven’t tried or looked for anything like that recently.

[worn agate]

Can’t really do that on the mobile app :/

[orchid ember]

Probably not.

[west ridge]

canvas and search mode are not working on web and app, can somebody explain?

[sick pecan]

you using projects or o1

[west ridge]

Actually no, just normal 4o

And knowledge cutoff is in 2021 September…

[echo nest]

https://chatgpt.com/g/g-675c661aa1ac819190b152c9cadfd9c0-drug-interaction

[hoary laurel]

Can I search for Internet information in advanced voice mode now?

[torpid kelp]

Sam Altman please give us chatgpt vs code plugin so we don't have to pay for copilot

[sick pecan]

i still dont have web avm

[hard crag]

my ChatGPT voice sneezed before giving me its answer 😭

weirdest thing ever

there must be audio recordings of people sneezing mid convo in their dataset

[viral gazelle]

Is there an AI-LLM that will consistently follow the instructions I provide? As an IT Helpdesk agent, am trying to use AIs like ChatGPT and Claude to convert my notes and information from calls into a specific documentation format. I have all KBAs and files, but despite providing instructions and having the information in the Project Knowledgebase and memory, they disregard my instructions. I spend more time trying to get them to follow the instructions properly. ChatGPT is particularly problematic. I explicitly tell it not to or modify documentation templates, but it does anyway. It frequently miscategorizes andclassifies information. Does anyone have suggestions on dealing with these unreliable AIs, are they just lazy for not reviewing the instructions that specify which files to review before creating documentation?

[sick pecan]

best to not use LLMs directly, but instead rely on prompt chains. ie instead of 1 open ended inference, try a series of closed yes/no inference calls that label the query. then rephrase what the user wants based on your kb article. then using the classified and rephrased prompt you can with confidence do an open ended inference

the rise of low/no-code stacks like make dot com, n8n, flowise, botpress etc all wrap LLMs with business logic

[prime swallow]

I would review the instructions and focus on telling the model exactly what you do want. If possible, don't even mention the stuff that you don't want - instead, focus on clearly saying what you want.

Like, if you had a dog you didn't want on the couch. You don't care what it does, just don't go on the couch or break any other rules -

The best way to keep it off the couch is to offer it directions to do something else.

Like chew a bone beside you, or fetch something you throw (of course you're not gonna throw it on the couch), offer it some food, take it for a walk, teach it a trick.

All of those are things you are okay with, and also 'get in the way of it getting on the couch'.

[solid bronze]

Did project break the custom gpt edit or was this error always a thing?

[timber moss]

Idk about you but my gpt wont even load lol. I cant send messages see my profile see my last chats etc its not looking good right now

[solid bronze]

wow !!

[timber moss]

I ended up downloading the application instead, works fine now lol seems its just something wrong with the web.

[steep forum]

Anyone?

It was quite useful to search through the thousands of chats I've made 😄

[normal quarry]

why don't you use the Desktop App?

[carmine kelp]

Why did GPT start saying the phrase "You’re absolutely right" so consistantly this year?

[sick pecan]

because its the words humans crave the most

[tranquil pelican]

because they make the AI dumb to sell the pro plan

[tiny wigeon]

Is there a testflight link inorder to beta test, chatgpt on IOS

[rich delta]

If that would be the case, it's detrimental to a vast audience don't you think? It is true that toxic positivity is a thing and prevalent, but, i guess they will fix it at some point. At least you could suggest the mods and developers this as being a wide-prevalent issue to look into.

[sick pecan]

how conspiracy theories are born

[dense halo]

Chatgpt still doesnt have the web in avm

Or improved search

🤔

Does anyone of you have it?

[rich delta]

Yeah. If you mean web-search, i have it. I live in Europe, so here it's on. Also, i pay the 20 bucks.. Dunno if for the free version is on though.

[dense halo]

General web search was there a long time

I mean improved web search and web search in advanced voice mode

[rich delta]

I don't think so then. But do try it. Ask it to tell you about the latest news

That's how you would know

[dense halo]

I did and I dont have it, thats why Im asking if others do have it

[rich delta]

Lemme try it and i will tell you

Tried it. It doesn't have live-updates.

[latent narwhal]

I don't have it too. are they rolling it out across the week?

[dense halo]

They said starting now If I remember correct

[rich delta]

O.o Didn't knew this.

I'm fairly new to the updates and discord chatter, so, that's the reason.

[latent narwhal]

yes i too heard starting now. let's hope it gets rolled out within a week.

[rich delta]

Yay 🙂

[dense halo]

dont expect to much

[rich delta]

It'll be enough to see it at least 🙂

Besides, we gotta give feedback

..after.

[latent narwhal]

we'll hammer the feedback in to the devs otherwise 👍

[rich delta]

Yup.

[latent narwhal]

i know right. well 🤞

[dense halo]

[rich delta]

Havoca-d-noooooo :)))))

...falls and splashes.

[dense halo]

you are plus user aswell right?

[rich delta]

So disrespectful... It happens so often to be left on seen. Oh well.

To me at least...

[rich delta]

Well, have fun and enjoy your on-going life. Your profile pic makes me think you are being amazing irl and have plenty of support from everyone all-around. Keep doin' what you do. Gonna wish you: Happy Birthdays, Merry Christmases and a Happy New Years in case we never cross-paths again.

[eternal rivet]

why now ctrl+alt+c can not get perfect LaTeX expression🥲

i mean, ctrl+shift+c.

[celest galleon]

hey guys I have a really weird issue, I tried to copy my text input because my UI had glitched so I wanted to reload it, win + v is showing it as the text I copied but when I paste it it becomes an image??

how do I get my text back :(

I made a huge prompt that is now gone because I can't paste the text that is clearly in my clipboard

I fixed it, when I copied another image I was able to click the text entry in the clipboard and now it properly pasted as text

[quick maple]

So for this day of Christmas I got an “announcement” but not a release? no search in advanced voice mode and no idea when it’s rolling out, and no video on desktop still? I wish there was a roadmap with dates so we knew when features were coming

[dense halo]

sadly not

[carmine kelp]

So you have to get that for it to be more critical?

[tranquil pelican]

The o1, even on the plus plan, does become more critical, it responds with more integrity, but you still have to be careful.

The 4o is no longer as good as before, the deterioration is very evident, losing even to the Gemini in some moments

[primal blaze]

Can you create a new project in the ChatGPT MacOS app or is that yet to come? It does show up for me in the web UI.

[grizzled badge]

Of course it is... The epic question everyone wants to know is "Is a corndog a sandwich?"... Personally I say no, it's a pancake.

[west ridge]

anyone has this issue? I am a paid user btw. Canvas and search are not working

[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.

[inland heron]

Anyone getting this with o1? I've been getting this glitch since the released it

[inland heron]

2021 🤦‍♂️

[dense halo]

can anyone tell how to fix this

messed up real bad

[subtle cobalt]

Why is OpenAI not letting us choose the AI model (GPT-4o and GPT-4o mini) for responses.

I don't wanna waste 4o requests.

[dapper rose]

4o is consistently saying "The server had an error while processing your request. Sorry about that!" and deleting the prior five or so interactions from both myself and GPT. This has been all morning, on both Windows in the browser, and on the iOS app. Please advise.

[subtle cobalt]

Try it on chrome browser

[dapper rose]

Interactions are not persisting across platforms. I tried Chrome and then went to my iOS app and Edge to see if the interactions after the reset point were visible and they are not. Even logged out and back into both. When it resets, the error message is as attached.

[obtuse moss]

hey can anyone help me on this code if i run like this its throwing any error await initialize_session(openai_ws, agent_id, db)
but if i run like this its not throwing any error await initialize_session(openai_ws)

[balmy hazel]

chat gpt only working in inconito mode

[honest jay]

Hey, if it only works in incognito, you might try checking either or both: (1) your browser extensions - try disabling all of them, clearing your cache, restarting your browser, and trying again - if it works, re-enable them one at a time, or (2) your privacy/cookie settings in your browser in regular browsing mode - depending on the browser, you could have some setting interferering in regular mode but not in incognito.

Usually, it's something to do with extensions, cache, or cookies, so disabling/clearing those, then signing in fresh in a restarted browser, is a good test. If it's still not working, feel free to post more info in #community-help. You can also reach out to OpenAI support at help.openai.com.

[abstract nest]

so, are Project folders almost like mini custom GPTs? thanks

[honest jay]

A little bit, totally! There's definitely overlap in how you can set custom instructions for a project, like how you can for a custom GPT. Also uploading a set of files as reference documents

You can also just use the feature as a "pin chat" feature. Or just to organize by topic

[balmy hazel]

Thanks. worked

[pseudo coral]

Hi, i want to create an own chatGPT to share it with a few freelancers for work. I need to use the o1 pro module, otherwise the result does not fit exactly enough - (complex job..). I dont find a possibilioty to use the pro module. Is this not possible yet?

[honest jay]

Hey! Do you mean create a custom GPT on ChatGPT? If so, currently, custom GPTs currently use only 4o, so using a custom GPT with a model other than 4o isn't currently an option: https://help.openai.com/en/articles/8554397-creating-a-gpt

Creating a GPT is only available to paid users and is currently powered by GPT-4o.
If you mean just general access to o1 pro on ChatGPT: that model selection is only available on ChatGPT Pro accounts.

[pseudo coral]

Thanks for your answer. I have a paid pro account. Yes i need a custom GPT with pro. Is there an announcement or a guess as to when the pro model will be available for this?

[valid depot]

why is chatgpt saying that it cant see images

[honest jay]

Personally, I'm not sure at all. I know custom GPTs are built around the multimodality of 4o specifically in that they can do all the things 4o can do: search, use DALL-E, use data analysis, etc. Stuff that o1 isn't hooked up to quite yet. So if I had to speculate, I'd guess that flexibility in that regard would come some time after o1 and 4o merge a bit more in terms of multimodality/the ability to use tools (if such a time indeed comes). Again, total guesses on my part.

[honest jay]

What happens if you just go to a brand new chat, attach an image, and ask a question about it (i.e., instead of asking if it can do it in the first place)?

[valid depot]

ye i did and it says that it cant see images

thats why i asked it just straightforward in new chat

[honest jay]

Would you be willing to share a screenshot in #community-help of what you see when you attach an image in a brand new chat and ask a question about it?

[valid depot]

sure

[pseudo coral]

ok, thanks for your response, have a nice day

[honest jay]

Sure thing, you as well!

[tight island]

Today the day

The mini dev day

[balmy hazel]

any tips for using chat GPT to help with my academics? And need brief understanding of all those different models

[idle silo]

Does advanced voice mode work with o1 pro mode now?

doesn't seem like I do 😦

what's that emoji for?

well, it clearly has tool use, so it could call o1 pro mode behind the scenes 🤷‍♂️

[wicked raft]

Any tips getting started with projects? I'm working on a code project and I wondered if I can just download a zip from git and put it into the project?

Does o1 only work with blank projects? Kinda weird considering o1 is the "code" project.

[hard crag]

WHAT WAS THE REVEAL TODAY????

[sick pecan]

you can get avm to do tools via api but in chatgpt its limited to bio, image and video tools

(ppl on pro plan reporting it can use o1 so maybe thats a tool for high rollers)

[rain fulcrum]

Will the PDF and image reading option never work again on the computer? It only works on the cell phone. I see several people complaining and we have no support.

[hard crag]

🥱 🥱

[radiant hull]

Yes, I'm on Android (2 days later)

I'm quite surprised that paying the $20 subscription only allows you to have a few more minutes with the audio mode and a few more interactions with the written mode.

Is it really worth paying $20 in the end? ...

Especially since $100 per month is an amount that is not accessible to a good number of users

[frozen birch]

ive paid 200 dollars for pro and my maximum voice session its only 60 minutes

sooo err..... trust me: its not so bad as you think.

XD

[swift pond]

Ask ChatGPT itself how it can help, and tell it what you’re working on and what you wanna do

[sick pecan]

number 1 tip with any llm - dont trust what it says just because it says it with confidence

you can mitigate some of the lies by having it question itself. and/or pass the response from one LLM to another

[hard crag]

ai is intelligent. trust every word it says because its smarter than Einstein

If you cant think on einsteins level dont even bother questioning the validity of its answers

its is just right. you just cant comprehend what its saying

[vale heath]

is chatgpt down?

acting funky

[slate berry]

https://status.openai.com/

Looks operational to me

[craggy relic]

So I got a new phone...I had to redownload the chatgpt app and when i logged into my account, it was like a had a new instance... I lost access to my plus plan and my old conversations...But on my pc i still had access to my plus plan and old conversations, I figured an easy fix would be to sinc my phone up to my pc, but that plan back fired... Instead of the phone sinc with the pc, the pc sinc with the phone and now ive completely lost access to my plus plan and my old conversations....

Does anyone know what I can do to restore my old conversations, which i did not delete or archive, they just some how vanished....

[umbral bear]

It is acting up. Instead of answers it responds with Got it! If you need any changes or have more details to add!

[umbral bear]

you will have to contact support. Nobody from OpenAI on here

[craggy relic]

shhhhhhhoot

[umbral bear]

However, the whole system seems to be acting up now

[thin lily]

So projects are here, but not yet in iOS and macOS right? Do we have any knowledge when they will appear on the apps?

Also having some better document saving and access on project level would really be helpful. Guess that's for some future iteration then.

[umbral bear]

Strangley the MS CoPilot is functioning normally

Yeah. No matter what I try to prompt, I am getting "Got it! Let me know if you need any changes or have more details to add." on everything

[zenith lichen]

Are you all using o1 for all your prompts?

Or using o1 for only math and coding related stuff?

[umbral bear]

Have you actually tried it? It keeps responding with non -anwsers

[languid cradle]

I have projects on MacOS (and Windows) desktop apps. No updated build, but I did shutdown and restart.

[languid cradle]

I've started using it more. Read that is uses 60% less tokens than o1 preview. Going to be sad if I run out of mesages with it (Teams account).

[zenith lichen]

I only have 50/week (plus), so i have to ration it sparingly

Also i wonder how o1 mini performs compared to o1

[slate berry]

It's working as intended for myself

[umbral bear]

For me, it is only responding with "Got it! If there's anything else you'd like to add or update, feel free to let me know!"

[slate berry]

Have you tried a new conversation?

[umbral bear]

yes

[eternal rock]

`` Server Tip: ``

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

[umbral bear]

and different prompts

[thin lily]

interestingly yes after killing the app for one more time on iOS it appeared now. It wasn't there earlier. Thanks for the hint. Will reboot the mac in a bit. An app restart didn't do it.

[umbral bear]

o1 is responding fine, but 4o is not

[slate berry]

How bizarre

[umbral bear]

yeah it is. I guess most using o1 now as not many complaining

[wise yacht]

working for me with no issues. Using Windows here and just uploaded a PDF for analysis.

[umbral bear]

4o seems to be having issues right now for some people. Try o1 or older GTP-4

4o won't even repsond to me anymore

[whole roost]

Does anyone know something about credit cards being declined for payment even if you already pay for a sub with that very credit card?

[umbral bear]

yeah, contact the card issuer. Only they can tell you why

[whole roost]

but i can pay for the sub i already have, just not change the plan

[umbral bear]

only they can tell you why

could be fraud detection

[whole roost]

so it's not on openAI end? I don't see any payment attempt on my credit card

[umbral bear]

you won;t see it, but if you contact the card issuer, they can tell you why

It is usuall fraud detection issue

No, but I stayed in a Holiday Inn Express last night

[whole roost]

they say it's not on their end 🥴

[feral lichen]

i enter a series of numbers to rmeove spaces and its looping generating various numbers not from my list... its having a seisure lol

been doing this before but GPT is not what it used to be

[umbral bear]

Some payment providers also have security rejections. You will have to contact OpenAI support then

[whole roost]

Yeah I did that, i hope they'll reply this year 🫡

[feral lichen]

monitor says gpt is having issues... time to wait and manually type strings lol

[umbral bear]

Usually when an online company deals with payment issues, they will respond in seconds, but yes otherwise a year

[whole roost]

I filed a support issue from the credit card provider as well and they claim to reach out to the seller, we'll see, i believe it when i see it

[umbral bear]

I guess you are in another country compared to where OpenAI and the card was issued from?

[whole roost]

yes

[umbral bear]

yeah, that caused security problems