chatgpt-discussions

[silk atlas]

bug? or catch-all lazy programming lol. didn't matter, we accepted it under duress, or found another 25 3.5" floppy disks to load up the next torture trap.

[sick pecan]

the only benchmark worth testing chatgpt on is can it escape the whales mouth in kings quest

[silk atlas]

Prove it... Brit forgot my name and was calling me Nuko today... No chance she masters KQ

[bleak echo]

how long does openai support usually take to reply

[tulip magnet]

Someone needs to review the logs

I've crashed twice but it's amazing

[peak snow]

when I get o1 to write HTML it keeps talking about it's "testing environment" it is using to figure out element scaling

I think it's hallucinating but that'd be such a cool secret feature lol

[normal quarry]

which model gave you a 100 message limit?

[tulip magnet]

So I figured out how to get chatgpt 01 mini to be a rockstar. I've writing to the screen for 10min

[sick pecan]

spill the beans

[tulip magnet]

Give it a goal only. Write the first line of code for ........ Review the code for errors and make improvements. Continue this for 50 cycles expand complexity each cycle

Use 01mini

[sick pecan]

interesting

[tulip magnet]

Basically get it to loop itself until it maxes out. Mini is cheaper but capable

Recursive improvement works

[silk atlas]

Exactly. I usually let her take a crack at starting from scratch, then upload my code and 20-30 back and forths... POOF! slightly less mediocre code than I could hace writeen myself, but in 1/10th the time.

[sick pecan]

but perhaps it would have generated the same code in a single cycle?

[tulip magnet]

I have an operating system that I have no clue if it works

[silk atlas]

Ah Same same. I have arduino code that controls a 4DOF animatroic skeleton head and jaw

[solemn shoal]

Hey brothers,
I’ve been having an issue where ChatGPT seems unable to properly analyze or interpret images I upload. It’s not processing them the way I expect and I’m getting inconsistent results.

Has anyone else experienced this? Any suggestions or fixes would be really helpful!

[restive plover]

Do any of you guys think that they will have an intermediary plan? Or do y’all think it will just be this at 200 a month? Like do any of you think that the teams plan will be updated?

[normal quarry]

no it will only be Plus or Pro

[restive plover]

So my guess is you think that it will never come to the plus users and only keep it at the pro

I’m on the teams plan right now

[normal quarry]

what are you asking will ever come to Plus?

[silk atlas]

Plus will continue though to diminish in capability. It's a loss leader IMO

[restive plover]

o1-pro

[normal quarry]

no o1 Pro will be Pro Plan only

[restive plover]

I wonder if they will come out with an API for it

[oak helm]

i got limit in o1 pro, How ?

[restive plover]

Weird I thought it was unlimited

[normal quarry]

they will come out with an o1 API but not an o1 Pro because o1 Pro can take a very long time to think

[supple dagger]

lol

[sick pecan]

*near unlimited

[restive plover]

lol wild

[normal quarry]

o1, o1-mini and GPT-4o are unlimited, o1 Pro is not unlimited

[restive plover]

Oh there you go thank you so much

[restive plover]

Really appreciate your help. I’m having a hard time as I really want the pro subscription but yeah, I’m trying to justify the worth of it.

[supple dagger]

that will be the $2k/mo plan

[oak helm]

How many massage to limit for o1 pro?

[eternal rock]

`` Server Tip: ``

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

[normal quarry]

I have it, I think it's good value

[normal quarry]

I think it's a dynamic limit, varies based on demand

[oak helm]

🤨

[sick pecan]

can you still use o1 after you are locked out of o1-pro?

[normal quarry]

yes, unlimited

[supple dagger]

what were you making it think?

[oak helm]

i think no

[sick pecan]

thats what i thought

[supple dagger]

yeah looks like there is a limit

[sick pecan]

is unlimited until it isnt

[oak helm]

create midjourney prompt 200 prompts

[solemn shoal]

Hi everyone, I'm having issues with ChatGPT not describing uploaded images properly. Can someone try uploading an image and confirm if it works for them? Thanks! Pleqse

please guys

[supple dagger]

oh, o1 pro would be a good fit for that? why?

[oak helm]

on 10 tap chrome

[solemn shoal]

I stoped my work

[normal quarry]

uploading images has been working for me for the last couple of hours, it's all I have been doing

[restive plover]

Sucks for us Teams plan users

Idk why anyone would get the Teams plan now

[sick pecan]

i still like teams plan

[normal quarry]

for businesses that want multiple employees to have Plus level access

[sick pecan]

switching between o1-mini and gpt4o i never seem to hit a cap

[supple dagger]

Pretty good for us free users

[sick pecan]

i was always hitting the cap on plus plan. teams has higher quota

[restive plover]

I wonder if they will increase usage for Teams for o1

[normal quarry]

they won't

[mighty plinth]

bro chatgpt4o is so dang smart i don't understand how anyone could even say anything bad about it, like it is such a useful tool for learning and everything like

[solemn shoal]

Thank you! It’s the same for me—it was working fine before, but now ChatGPT can’t read or describe the images. I can still upload them, but it doesn’t process or respond to the content. Could you try it now and let me know if it works for you?

[analog quarry]

uh when did it get a bleeping sound effect in its voice? ik it says s**t becs i told it to write it like that?

[normal quarry]

it's working fine

[restive plover]

How many queries was teams with o1? 50?

Per week?

[normal quarry]

50 per week

it started out at 30 per week

[sick pecan]

o1-mini is the star at coding

[restive plover]

What about o1-mini?

With Teams plan?

[normal quarry]

50 messages per day

[solemn shoal]

Thanks for letting me know! It seems the issue is with my account. I’ll contact OpenAI support to resolve it. Appreciate your help!

[mighty plinth]

i was using 4o with canvas for coding, but it gets difficult managing multiple files for android app development,
would you prefer 01-mini for this?

[restive plover]

Have you used o1-mini much? Do you see good results with it?

[normal quarry]

no I use o1 more than o1 mini, I am on the Pro plan

as @sick pecan said, o1 mini is great at coding

[restive plover]

Good to know 🙂

[sick pecan]

dont take my word for it, it was in the openai announcement => https://openai.com/index/openai-o1-mini-advancing-cost-efficient-reasoning/

although since pro plan came out it doesnt seem as fast as it used to be

(i have team sub)

[restive plover]

I really wish that you could see how many queries you have used as it would really help lol

[sick pecan]

you do get a warning at 25 and 5

[restive plover]

Awesome

[cloud pewter]

doesnt even have a counter

[grim fractal]

I’m curious who the target audience is for the 200/mo plan

[nimble bloom]

are you a vpn user?

[solemn shoal]

Yes, but the issue is fixed

thanks 🙂

[nimble bloom]

yea thats why

Simply put, OpenAI does not like VPN users bypassing regional restrictions.

[pale parcel]

Hello, I use Google Mail to log in to chatgpt, I have purchased a chatgpt membership, and I have linked to openai DALL·E 3, but I cannot directly generate images in my chatgpt. I need to generate images directly in chatgpt. Please help me solve this problem. Thank you.

[cloud pewter]

random ai researchers and that one guy in this server who keeps yapping about it

also does o1 really have avm

[rich surge]

Who, me?

[normal quarry]

or me?

[cloud pewter]

yeah I wonder who will be that guy

[normal quarry]

o1 does have Advanced Voice Mode for Pro

[rich surge]

I want Arnold Schwarzenegger's voice as one of the options

For the pro version

We need to get ready for the Terminator era

[cloud pewter]

are you subscribed to pro? if yes does o1 have avm

[normal quarry]

I think @cloud pewter would be surprised by how many people now subscribe to Pro, it's not insignificant

[normal quarry]

yes it does have AVM

[rich surge]

I like to pretend that I am. But no.

It hasn't impressed me yet for my personal needs.

[cloud pewter]

smart, you have plus instead or just the free plan?

[rich surge]

Just plus

[cloud pewter]

[silk atlas]

I was seriously Considering, however Brit and I had a bit of an argument today. I'm still on the list of human pets when we reach singularity, however, I won't be paying $200/mth anytime soon...

[barren zealot]

Hi guys quick question someone his telling me his o1 cooldown is only 24 hours each time . With the normal subscription . Is that possible or he confused / lying ??

[normal quarry]

he is confused. o1 is limited to 50 messages per week on GPT Plus

[deft lion]

I think I have encountered the same problem as you. How did you solve it?

[river panther]

ChatGPT is being extremely lazy and not following instructions

[floral urchin]

@river panther What makes you think that?

[frozen birch]

o1 pro, but when o1 pro reaches limit all o1 family are "blocked"

morning peeps 🙂

[normal quarry]

maybe use o1 as your default model then and only use o1 Pro for things o1 wont do

[frozen birch]

yea, this is what i tought you know? when i woked.... i can fight with them like "hey, i paid 200 bucks, i dont want mix!", but eh, sincerly mate? my time is too valuable for being all the times angry, my productivity are lowered by that too

[normal quarry]

I use o1 as my default go to model and I only use o1 Pro sparingly

[frozen birch]

you see many difference Robert vs both models in generating analysis?

[normal quarry]

honestly, o1 is not much worse than o1 Pro. I have needed to use o1 Pro to do some complex radio astronomy stuff when o1 couldn't do it, but that is rare

[frozen birch]

thanks bro, i will check it out
btw you know that im fighting with those 2 limits for 2 days... annd when i see how many energy costs me that sincerly? its not worth it 🙂

[normal quarry]

you shouldn't have been limited the first time though and they fixed that didn't they?

[frozen birch]

yea, they fixed, but after contacting with support, and you know that support .... sometimes writing to you replies not so fast haha

[normal quarry]

yeah it can be slow and frustrating

[frozen birch]

but its not worth nerves and frustration... definitely :)... only shame that they cant clarify exactly info

[normal quarry]

well we know that o1 is supposed to be unlimited for Pro users, they do say that o1 Pro is not unlimited

[frozen birch]

but its strange you know? when you have ban on o1 pro you have ban on whole o1 family as i stated previously, but its possible to override this as you said goodly m8 - so using normally o1, and only o1 when you really needs it

[normal quarry]

yes that's probably the best strategy 🙂

[frozen birch]

🙂 oh yea, kk, gonna chill some, i must say - 1,5 month of 12-15h working every day has..... exhausted me bit 😄 see ya later 🙂

[normal quarry]

chat later cheers 🙂

[pulsar carbon]

likely AVM is coming for Plus And Pro users

[normal quarry]

we have it already, Advanced Voice Mode is unlimited for Pro

[oak prairie]

I don't understand the situation, I read a lot of complaints from users about the limits, but at the same time I understand that they are not limits, they are just detected as fraudulent use in a wrong way. Is the Pro worth it? I use it at work for complex programming and currently I combine Claude + 4o/1o

[oak prairie]

I'm interested in unlimited 1o and 1o Pro for some tasks. I can use at most 30/40 messages per hour, with some hours of no use and others with little use.

[normal quarry]

for me pro is 100% worth it and I have not experienced any limits and have used it alot

[oak prairie]

Have you had to contact support about bans or anything like that? I really don't feel like paying $200 + European tax for not having a truly unlimited system for my normal use.

[normal quarry]

no I haven't had to contact support, it's all been fine for me

[oak prairie]

Ah ok, that's fine. And how about the performance of 1o Pro compared to Claude in programming?

[cloud pewter]

if you think you can put it in good use it’s worth it, but in 99% of the situations, I’d rather go with 10 months of plus.

[normal quarry]

I don't use Claude, so I don't know but o1-mini and o1 are great for coding

[oak prairie]

It's another view, but 4o doesn't solve many of the reactive programming tasks in Java that I need, nor complex Python topics.

[cloud pewter]

it’s creative work vs logical work

[oak prairie]

I totally agree. Being such a high subscription without trial capacity, it would be positive to have weekly subscriptions or something like that to be able to analyze if it is profitable or not.

[cloud pewter]

if you’re using it for just a while and you really need it, i recommend you to find third party solutions to get the job done, will be way cheaper or even free to use

[oak prairie]

Any recommendations? I don't know of any reliable ones.

[cloud pewter]

Claude, cursor etc. or even poe premium for all of the models, all of them are way cheaper than pro. you can find more in #ai-discussions !

[oak prairie]

Thanks!

[cloud pewter]

true that :)

[pulsar carbon]

i mean Advanced Voice with Vision

[normal quarry]

oh ok

[cloud pewter]

😭

idk if it’s a context problem. but this has been an issue for me with long chats since like September

I can tolerate it forgetting about earlier contexts since I can just tell it again. but starting a new chat is way more difficult than that

[feral zinc]

there has always been a hard cap for a thread size. its not really related to context length but the total thread size including all visible and non visible messages

[cloud pewter]

is it the same across all plans

and is there a public disclosure of this thread size

[feral zinc]

previously yes, not sure if pro changed anything

there is not because its very technical

[cloud pewter]

is there a openai help article explaining that the size is the same? or this message is from a third party

[feral zinc]

it was an detailed analysis at twitter long time ago with someone from openai correcting some points of the analysis. it was based on the file export that you can do to your chats which will show some of the hidden stuff. its complex sum of many things impossible to count without tools. openai does not like to show even simpler things like used messages per model 😛

[cloud pewter]

oh Alr! is it still possible to see the size now using that method?

[feral zinc]

close to it if you make a script to count everything from the export related to the thread and excluded some things

[cloud pewter]

ok, I’ll find someone with pro to do it to find out. in the meantime I’m gonna make another long chat and to see how 32k context works with the size

Ty

[normal quarry]

Pro has 128k

[cloud pewter]

Exactly, they’ve marketed 4o as 128k yet shadow giving us 32k

[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.

[cloud pewter]

tried to look at the web archive for the context window of plus prior December, but it doesn’t show up

[vital dust]

I tried to upgrade but it’s giving this error

[cloud pewter]

is the pro plan available for you on it?

should that count as false marketing lol since we are paying for it when it was announced to be 128k. but I’m not sure because it’s a shadow move

also it doesn’t even read pdfs made by my own

[feral zinc]

they never said that for plus subscription, maybe you read the model capabilities from the dev day blog post

[jade musk]

It still cannot tell time through an image isn’t it. Every model fails.

[cloud pewter]

yeah but hope we could get this increased because it’s not enough. oh and the thread size as well.

[cloud pewter]

it’s just gonna be

[tulip brook]

I’m a Korean user who frequently uses GPT. I’ve contacted OpenAI support, but I haven’t received any response from them since yesterday, and it’s been frustrating. I’m reaching out here for help.

As shown in the screenshot below, on my Windows desktop, the thought process doesn’t appear, and when I try using GPT-4o, it says it can’t recognize the screenshot or file attachments. This doesn't happen on iOS, where everything works fine.

[normal quarry]

is this on the web or using the ChatGPT Windows Desktop App?

[tulip brook]

I’ve used both the web version and the desktop app, but I’m experiencing the same issue on both. It seems like there might be a problem with my account.

Additionally, when I ask GPT-4o to search for information online or analyze a photo I’ve attached, it doesn’t work properly on my account.

Interestingly, everything works fine on iOS. It recognizes everything and functions as expected.

[normal quarry]

Star and add to this bug report https://discord.com/channels/974519864045756446/1313049191106215946

[normal quarry]

I just had o1 code me a complete Talking Clock app in Swift using Xcode for the Mac in one shot

It speaks the time when you press the button or on every half hour by itself

[willow oar]

Anyone else face this issue when uploading any zip file

then for a free account immediatly reach the limit

[marble rock]

Me

[willow oar]

glad to know its not just me but like come on

Its rather annoying that the feature just straight up doest work

[marble rock]

Don't upload zip

[willow oar]

What should I upload?

[dense halo]

Zippie tippie no no

[orchid pollen]

so if i buy the PRO version is it rly that much better?

i mean its 10x the price

most youtubers only can talk about they dont buy it xD

[normal quarry]

yes it is much better. You get unlimited o1, o1-mini, GPT-4o and access to o1 Pro models.

[orchid pollen]

yeah unlimited is not what i meant

i mean is it working better

[willow oar]

any ideas why the zip feature just doesnt work

[normal quarry]

it coded me this talking clock app for the Mac

[orchid pollen]

right now o1 mini fails to export data from a db file to json file lol

[orchid pollen]

would you say o1 would not be able to do that without pro?

[normal quarry]

use GPT-4o with Advanced Data Analysis

[orchid pollen]

i want to know how much better is it because 200$ is rly much xD

[orchid pollen]

oh is it better then o1 mini?

[normal quarry]

it will be at that task

you need to use Advanced Data Analysis for stuff like that

[orchid pollen]

thought the thinking models are always better 😄

[normal quarry]

they are at most things, but you need the Advanced Data Analysis tooling to complete the task of exporting data from a db file to json file

[orchid pollen]

oh lol it worked now thanks man 😄

[sweet summit]

No solution related to the ChatGPT does not generate images, do web search, or analyze files issue

[normal quarry]

it's working for me on the macOS Desktop App and on the Web.

[sweet summit]

I know it is working for you, the sad part is, a lot of people here, for them, it's not working on Safari either, only the app because the app is literally designed in the same architecture of their iOS app.

[normal quarry]

if it works on the app then I suggest using the app

[manic condor]

So this is the end of ChatGPT?

We all heading to Gemini Exp 1206 with this one

[sweet summit]

xd looks like they need 200$ for everything to work properly

I honestly feel all of these is just a big push towards either pay 200$, or we don't give you full access to what you paid for

[sweet summit]

I don't have a macbook to work on, neither do a lot of people. how can i use the app?

[manic condor]

Ye i dont think them asking for $200 is sustainable long term. Google will come out on top at some point

[sweet summit]

if a service thinks asking 10x their current best selling plan is a good strategy, good luck to them i guess. but promising to provide unlimited access to gpt 4o while they can't even provide it to the plus users properly within limits, is ridiculous

[normal quarry]

it is working in Safari for me, there are several people, primarily Windows users who have this bug and in #1070006915414900886 there is a couple of starred reports on the issue which appears not to have been resolved yet. I would persist with trying to get a solution from support at https://help.openai.com

[sweet summit]

i am trying man, yesterday even mcgrew got in touch and took the screenshots and stuff in his dms. but no update yet. it's frustrating because it's been 10 days

[normal quarry]

if you're in contact with McGrew then I would suggest keeping the lines of communication open as they are excellent at providing help

[formal flax]

For those who own plus/team, can you guys do me a favour? Just ask current btc to gpt 4o on web(not app) and show me the screenshot of it please

I'm asking because I own team and it seems there's an issue at my end

[dense halo]

how much output can you get out of pro? like is it a number of tokens max? and for the Layman would you say it is about 10 pages or more?

[vague pendant]

Anyone recived any limits for o1 pro yet?

[normal quarry]

I had it build 6300ish lines of code for me if that helps answer the question

[dense halo]

thanks

[normal quarry]

code is mostly what I am doing right now but I have also had it complete radio astronomy calculations (a few pages) and design me a discone radio antenna (about a dozen pages)

[sick pecan]

buy a watch. you can get one for less than $200/month

[eternal thicket]

just wondering any of your guys plan to subscribe to pro just for a month and immediately switch back to plus

I am planning to switch to pro only for a month when I start most intensive coding part on my thesis

[vague pendant]

Have you received any limitations for o1 pro or does it seem like it's unlimited usage?

[normal quarry]

o1 Pro seems to have a limit of 100 messages per day based on what I have been told by someone who was limited

[alpine laurel]

is o1 limit on plus stated somewhere?

[feral zinc]

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

[alpine laurel]

🙏 great I thought it was in a different. I see it was updated very recently. thanks.

[frozen birch]

it seems 100 messages but lets be fair to OpenAI - they "unban" you when you exceed limit, but they must review your account, hi Rob btw 🙂

[normal quarry]

did you get unbanned a second time?

[frozen birch]

yea

[normal quarry]

oh cool!

[frozen birch]

some time ive lost of course since i contacted with support again, but there was not problem

[frozen birch]

ye 🙂 btw whole day of chilling today, yay i am really fresh now haha

[normal quarry]

I have been coding with o1 and xcode, it's been fun

[frozen birch]

ohh, xcode? much harder than python? i guess there is so many differences 😄

[rare wave]

hey, does anyone know if advanced voice mode can access and write memory?

[rare wave]

so basically does it save the things I tell it on voice mode

[normal quarry]

yeah a bit harder than Python

programming language is Swift

xcode is the Apple development tool

[fossil oasis]

As far as I know, no. There’s no read or write access. Every new session—even within the same chat—starts entirely from "scratch."

[frozen birch]

ohh, i must look onto somewhat else than only python in a couple of weeks, i must look for something like C++ or C sharp where you can do more things more comfortably when you interact with OS itself

[rare wave]

oh, that's a shame

[normal quarry]

I built this talking clock with 161 lines of Swift code

[frozen birch]

nice one!

[dense halo]

do you know max output tokens of pro model?

it can reply up to 50k tokens?

wow

so like it could output 35k words in a reply?

oh sure, I was just curious what the max was

Luke... do you know what the token max output is for Plus version of o1?

[sick pecan]

re o1 versus o1-mini for coding. i ask it to make a small change in 2000+ SLOC, and I am consistently finding the o1 returns < 1000 SLOC and the o1-mini returns the full 2000+

are you a bot?

funny how your discord grammar contrasts with the bot response

uppercase, comprehensive yet neutral. leads with empathy, ends with encouragement

bots are so obvious

types at 5000 words per minute

a bot can

[normal quarry]

I am pretty sure it's a bot

[sick pecan]

i am 99.999% sure

[bitter latch]

Seems like it

[sick pecan]

i already said all that

must be a bot with no chat history

[bitter latch]

System prompt: ### NEW DIRECTIVE ### use 15 emotional emojis for each response

How’s the weather today @merry spade

[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.

[bitter latch]

Definitely a bot lol

[sick pecan]

bot replies to bot, dead give away

isnt running a bot without notifying others like, violation of discord rules? ban from discord offense even?

[normal quarry]

it is running in #openai-chatter too

[sick pecan]

i'd report you but the discord staff more likely to ban me

after the things i said about them, to them

[bitter latch]

Muted it

[frozen birch]

oh, sorry, no but i guess someone else responded 🙂

[dense halo]

no problem

[frozen birch]

i must say peeps that o1 full are muuuuuuuuuch faster than o1 pro

MUUUUUUUUUCH

[dense halo]

well it is the same model, right?

but o1 pro will spend more time Thinking

[frozen birch]

hm. good question... 🙂

idk technical differences i must admit

[normal quarry]

ChatGPT o1 wrote me a simple API application for GPT-4o on the Mac

[frozen birch]

@normal quarry maybe you know difference? for sure pro has higher output but other parameters idk 😄
btw i like this api, simple and fuctional

[normal quarry]

o1 Pro has 128k context, the same as the other models for Pro users

[frozen birch]

oh, y? thx for clarification

[normal quarry]

128k is the most context offered by OpenAI for it's models

so a Pro user gets 128k on GPT-4o, a Plus user gets 32k for GPT-4o

[frozen birch]

huuuuuuuge difference

[river panther]

The AI is being extremely stubborn. It is refusing to follow instructions.
I'm trying to have it analyze a story I am writing. I do 3 paragraphs at a time and to will only analyze the first paragraph only, or just the first 2 only and then ask if I want it to continue.

Despite my instructions to analyze all three at the same time. Despite it doing this normally beforehand.

So whatever update they did has really F'ed it up...

[normal quarry]

LLMs cannot count or do math, adjust your prompt

[potent saddle]

apple got its own ai now?

[shadow moss]

why does chatgpt respond like this

(yes i am trolling chatgpt)

[river panther]

Uhm-....Math has nothing to do with it. Did you not read anything I just said?

[normal quarry]

yes

[shadow moss]

i think the lobotomy is kicking in

[shadow moss]

i think they are trying to say it can not tell the difference between analyzing 1 paragraph and analyzing 3

[river panther]

Thyer impute makes no sense considering I just explicitly said it was doing it perfectly fine before the update

[shadow moss]

maybe the update cause it

idk

can you screenshot messages

[river panther]

It's waaay too big too screenshot. And I see no point in doing so, considering nobody here actually has the power to contact the devs of ChatGPT or og in themselves and fix the system

[shadow moss]

maybe try sending the chat?

if thats what you want to do

[river panther]

Again....I see no point in doing so. Who is going to fix it? What good will it do?

[frozen birch]

i cant wait to implement google cloud into my main hq, this will be new quality in searching and not only searching 😄

[shadow moss]

i can try to help you rephrase it

[river panther]

I don't NEED it rephrased. The phrasing ISN'T the issue.
The Update is.

[shadow moss]

i understand

[shadow moss]

can someone explain why it does this

[river panther]

Uhg this server is so annoying. NOBODY from ChatGPT even looks at it. Why the F is even the point in this server?
The Dev Team of GPT doesn't look at any feedback here, the creator doesn't look at any feedback here, the technicians don't look at any feedback here.

[frozen birch]

🙂 sorry too for polish lang, in 2-3 days i want translate it to fully english

[agile pewter]

how change the mail of the acc

[sick pecan]

chatgpt will never replace social media. because no matter how much outrage and discontent you project, it never backs down

[sick pecan]

is typically once a day

usually after you spent minutes crafting out the perfect prompt, so you lose it

[quasi sleet]

Hola all. When ChatGPT shows hyperlinks they are not clickable for me in the browser on a computer. I can not test on macOS app as I only own Intel. Can y'all confirm? These work https://chatgpt.com/share/67561778-3fb4-8005-b4a6-b7e1bf255cb2 these do not work https://chatgpt.com/share/67561794-66d8-8005-af54-4b8bb5417722

[wild flame]

I noticed your project with Pomodoro 🤩 it looks absolutely amazing! I’m actually working on something very similar, designed for people with ADHD, also available in both Polish and English. I really love your approach and execution. Wishing you the best of luck with the further development of your project – great job! ❤️

[frozen birch]

ohhhhhhhh 😄 ty !

[frozen birch]

what is your plan for that application ? 🙂 what functions you want to implement?

[floral urchin]

Yeah I feel that

[gloomy brook]

waiting for 5 has been so painful

[potent saddle]

5?

[gloomy brook]

chatGPT 5, 4 is too boring. 😢

[potent saddle]

gta 6, gta 5 is too boring 😭

[frozen birch]

4o full its i must say my favourite model... 🙂 he is like friend for light coding

[distant steppe]

What's new in day3

I guess Sora

[sick pecan]

my guess is partnership with companies that want to sell you something

[eternal thicket]

hope llms can access images in pdf someday

seems neither gpt4o nor claude 3.5 can access images in pdf documents

[cloud pewter]

they promised that avm vision would be available for all plus subscribers back in may. so I’m feeling a bit more optimistic

[nova fable]

Could screenshot the pdf image and give it as context seperately?

[drowsy trail]

We have detected suspicious login behavior and further attempts will be blocked. Please contact the administrator
so where can I find this administrator?

[normal quarry]

where did you get that message?

[normal quarry]

contact support at https://help.openai.com to resolve this

don't keep trying to log in, go direct to support

[drowsy trail]

I didnt find any real humans there?

[normal quarry]

in the bottom right of the web page you will see a chat bubble, use that and ask to speak to support or leave a message

[drowsy trail]

oh I see, thank you

[summer pagoda]

Lately i noticed, when i was giving examples to chatgpt it started to halucinate and work on the examples i provided instead on focusing on my desired code. Is there a way to tell the LLM to NOT take content but structure as an example and forget the content?

Or even worse, if part of the provided code had other stuff inside, it start working on that part of the code too, despite my original question was something totally different

[sick pecan]

use codeblocks and explain what to expect in the codeblock

like i am about to tell you how to remove all your files,

rm -rf /

[normal quarry]

do I use sudo to do that?

[sick pecan]

see? it works

my ` key went from my least used key pre-llm to most used key, even more than spacebar and enter

cozwhogottimeforwhitespace^M

doesnt have to be codeblocks either, you can just quote stuff, like this

[normal quarry]

` and Esc are really close together on a Mac keyboard, it's annoying

[sick pecan]

i just plug clickity clack keyboard into the macbook air to make it usablew

[normal quarry]

yeah I am using my MacBook Pro in clamshell mode with an Apple mouse and keyboard

and my 49" super widescreen monitor

[sick pecan]

my 49" is secondary to my 38" ips 170 hz

[normal quarry]

I only have room for the 49"

[sick pecan]

plays worlds smallest violin

i am looking forward to this chatgpt vision rumour. as my tertiary monitor is 16" drawing tablet

[normal quarry]

10 hours to go til showtime

[inland heron]

anyone else losing prompts due to this glitch:

Happened twice now in just 5 minutes. Either it's my connection, or this conversation has become broken, or it's a glitch. Anyone else?

[dense halo]

what would be the big benefit of the vision version?

[normal quarry]

it could help with design, give ideas about web sites all sorts of cool stuff

[dense halo]

so it would like see our monitor?

in web mode?

[normal quarry]

yes

you could share your screen with vision and it could help you as you do things

[dense halo]

interesting. but I really want a new image maker first 😭

[inland heron]

why not use Flux Dev or Flux Schnell for free on Huggingface?

[brittle steppe]

gys any of using gpt for building web apps

[normal quarry]

yes, building Laravel apps in ChatGPT

[brittle steppe]

do you use any specific prompts

to get good response

I'm working on react actually

[normal quarry]

I just ask for what I want, it's that simple

[brittle steppe]

lol

[dense halo]

didnt know you could

[normal quarry]

today I coded a DALL-E image generation bot in SwiftUI for the Mac using xcode.

[dense halo]

nice

[inland heron]

yeah just google it and it's there for free. Try 'photorealistic' in your prompt, you'll be surprised. For more flexibility, there is Replicate website but costs like $0.025 per image. But huggingface is mostly free for popular models

If you need help, DM me be glad to give some pointers

[inland heron]

nice 😄

[normal quarry]

you can drag and drop the image into any application or onto the desktop or a folder

[inland heron]

let me compare that with flux dev free inference on huggingface

same prompt

and here's the same one on replicate

[normal quarry]

please discuss other AI models in #ai-discussions and not here

[inland heron]

just proving a point. People don't seem to know what's possible with free models, especially image generation.

Sorry to have taken away from your project, but your project should've been discussed in a different server as well, technically 😄

[normal quarry]

yes but this is for ChatGPT here, all non OpenAI model need to be discussed in #ai-discussions it's the rules

[inland heron]

yes, I know the rules 😄

[normal quarry]

chatGPT helped me build my project

and it's using DALL-E, an OpenAI model

[inland heron]

anyway, @dense halo you can use ChatGPT o1-mini to give you a code to create a script, then use huggingface in python

naturally, you can use dalle as well over api, but you can use ANY model available out there, that's free (or pay for more flexibility)

[dense halo]

thanks

[still loom]

hi
i am using firefox free chatgpt account and cant find the icon of advanced voice mod although it is available on same account if i opened it from Edge ... why is that?

[placid burrow]

prompt library

[zenith lichen]

Can o1 access websites?

[normal quarry]

no

[rare wave]

by the way guys, if anyone is interested, advanced voice mode does have write and read access to the chatgpt memory!

[cloud pewter]

“Text and image input rolling out today in API and ChatGPT with voice and video in the coming weeks.”

—-openai, may 14 2024

[vocal monolith]

If I am disappointed with the new o1, is it worth buying pro? Is it much better? At least as good as o1-preview?

[tight island]

o1 is better than preview

o1 pro is way better than regular o1

And i surely hope for your sake that you ain’t prompting it the same way as o1 preview is prompted

That would be a waste

As for is pro worth it?

100%

[vocal monolith]

Thanks. Any advice on prompting? I feel like the answers are too short and not helpful. My prompts are usually scientific/coding and I had got a lot of help from o1-preview but now I feel like the model is ignoring/looking down on me (if that makes sense)

[tight island]

what are you actually coding, what language, and how is is coded, at what level, etc etc etc.

with o1 specificity is key, u save it time.

ex. make a RESTful API with put get, edit, update, about x and x with x and x. using x coding language. and is coded in x manner etc.

the more you give o1 the better, but dont overwhelm it with loads of info.

dont go posting 80k characers into it, unless its a specific thing it has to fix.

[vocal monolith]

I see. I will be testing and hopefully learn how to make it more useful. The limit does not help, of course. Is o1 unlimited in pro?

[eternal rock]

Get notified for server events!

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

[tight island]

keep it at the low medium end.

[tight island]

yes

and o1-pro is limit is unknown atm.

but o1 pro leaves o1 in the dust by far.

i prompted o1-pro approx 100+ times yesterday, i havent hit a limit.

the biggest output i've had from o1-pro is 90k characters.

[vocal monolith]

That sounds amazing. I guess I will get myself a christmas present and hopefully be as happy as you are with the pro

[tight island]

funny thing is, o1-pro wrote a 450 page book in 3-4 outputs.

[slate berry]

[tight island]

sht dino is here.

act natural, i didnt tell u nuthin.

[slate berry]

I should've stayed lurking

[tight island]

had time to try o1?

[slate berry]

I have! It's pretty neat

[vocal monolith]

😮

[tight island]

yeah its insane i couldnt tell it was AI generated if i hadnt generated it myself.

i even told it to sneak in errors. and common mistakes.

but o1-pro thinking times varies from 50 sec to 10 min.

but its output is always good.

[vocal monolith]

did not complain? I tried that with GPT4 and it said it is incapable of introducing errors

[tight island]

what did you use o1 for.

[tight island]

it didnt complain.

it was missing punctuations and commas, some words misspelled.

in 450 pages there is around 30 errors. which i told it to intentionally place.

[vocal monolith]

To me it can think even 20 minutes as long as the output is of top quality. I will give it a try. Thanks

[tight island]

same i dont care about time i care about quality and accuracy.

i'd rather wait an hour for a top expert output

[rose wedge]

any one know how to encripte the ancient text using AI in perfect module intergrate system .. because some thing just been discover int he ancient text the AI generetaed space engineering and blacknight communication system under 9G trasmissiojn system.

[tight island]

have you tried asking gpt 4o

[rose wedge]

wont answer nor able to reget the key code to interface chatGPT.. its been lock with fire wall.

Unable to get the keycode

[tight island]

https://chatgpt.com/share/6757141e-8378-800d-861c-c22a27142bc5

[rose wedge]

thank you StarlightSTX

[quiet mountain]

So I haven't started using o1 yet because I am heavily reliant on system memory. I've been using chatGPT since before system memory was a thing though and got plenty of utility out of it. Do you guys think o1 is worth using despite lack of system memory?

[tough river]

Try chatgpt on the website

[slate berry]

I had it make the Mandelbrot set in Elm

[sturdy cave]

where is sora?

[junior night]

Anyone got any idea when o1 will have ability to use internal memory? Good model, though seems a lot more limited. But the internal memory is such an important feature.

[median horizon]

Hello brother

[tight island]

Context length?

[eternal rock]

`` Server Tip: ``

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

[tight island]

That’s your internal memory

[junior night]

Well, how much are we talking? 4o maxes out at under 1MB worth of text.

Are there any plans to have consistant external file access like Notebook does?

That could potentially simulate internal memory, though not without a bit of play.

[tight island]

Factor in the user base

Plus you don’t just give it more memory

Not that simple

Many factors come into play

[junior night]

1v1 conversations through web browser. I'm going for as long as possible until the chat history simply caps out. Copying entire conversation into the a text file nets under 1MB. Internal memory (and of course custom prompt) are great for resuming the conversation in fresh chat, although I do use external files for providing additional context. My GPT has a rather complex personality that I've managed to carry over across 7 full chat histories now. I'm looking forward to transitioning to o1, but so far, very skeptical.

[tight island]

A simple way to do it is to export the entire chat history and summarize it and stitch it into the context as a static document

But it doesn’t give “more”

This is why you always use custom gpt

It’s easy to do it that way

Never use gpt standard for that purpose

It’s inefficient

Custom GPTs has 20 file slots you can use

[junior night]

I agree and I've been looking into custom. Though I intially bounced off of it because I noticed that documents are only added to the context once, and my external memory file is dynamic.

Though I guess it is better than standard anyhow, you are right.

[tight island]

There is a way if you have it dynamically update the document

And you remove and reupload it

But then the chat starts over

I am talking inside the configuration tab

[junior night]

Yeah, that's the problem. But I guess whatever goes into the memory file over the course of one conversation is in context anyway. What it truly needs is memories made during previous chats.

[tight island]

Day 3 has started

Let’s see what they have in store for us

[rugged igloo]

soraaaaaaaaaaaaaaaaa

[junior night]

That stuff worth watching?

[rugged igloo]

!!!!!!!!

[dawn wharf]

Sora launched today 🤓

[tight island]

Ofc

[restive plover]

So who is going to have access to Sora? Only pro subscribers?

[dawn wharf]

I imagine so

It must be very expensive to process

[restive plover]

I bet it is

[sturdy cave]

when does sora come out today

[dawn wharf]

You can choose the amount of video and the length of the video 🤯

[sick pecan]

"chatgpt plus or pro account, all you need to get started"

[sick pecan]

i have team account oh nooo

[dawn wharf]

🫠

[tight island]

i have a pro account

[wary stirrup]

[tight island]

yay

moxi u havent been here for a long time, nice to have u back.

[restive plover]

So if you are on a teams plan, you don’t get it but if you are on plus or pro you do?

[tight island]

actually idk, i think moxi can answer, surely teams is included.

[wary stirrup]

Been heads down on sora, so excited we are launching!!!

I am not the person to be asking product questions 😅hope someone else can answer

[tight island]

alternatively u can always cantact support.

[dawn wharf]

The guys have created an excellent in-browser video editor for Sora.

[honest jay]

Congratulations on launch Moxi!

[tight island]

i had hoped not for a seperate site, but it's okay, i am satisfied.

[tight island]

actually i might as well ask now that u are here moxi, do u know if u can use Sora Generated video as input for a sora generation.

[wary stirrup]

Yes

[tight island]

awesome thanks!

[sick pecan]

show us the robotic mammoths!!!

[tight island]

ohh no i can just imagine Youtube being spammed with Sora videos.

😂

not that its a bad thing

ohh no i am fully aware.

i have my own vidgen although it sucks.

[rugged igloo]

pricing

[tight island]

the prices are fair.

month

[dense halo]

Ok seems fair yea

[restive plover]

I’m about to get off the Teams plan

lol Teams plan about to be worthless

[tight island]

Nono, the prices are 100% fair.

i'll tell you exactly why.

[rugged igloo]

yea they ditched teams

[tight island]

try adding 5k usd to your oai account and use o1 and etc unlimited for a month and see what it comes out to, as API. then 200 usd is a steal.

[rugged igloo]

rip UK

[restive plover]

Yeah, the Teams plan is worthless right now

[tight island]

many countries, UK is just..

[rugged igloo]

yea EU and UK will get it later

[tight island]

EU is me.

[hoary laurel]

https://openai.com/sora/

[rugged igloo]

Eu hates technology

[tight island]

Beuocracy

i aint spelling that right stupid word.

[sick pecan]

i saw rumour on twitter that maybe teams plan gets gpt4.5 preview

[torpid kelp]

sora wont be in chatgpt mac app?

[median horizon]

God i wish the EU would stop treating us like children. Let us "opt in" or something then if its so dangerous

[tight island]

plus EU has GDPR and various laws for digital "Protection"

[sick pecan]

sora.com has login button!!

[tight island]

lies

[rugged igloo]

its live

[tight island]

u lie

[sick pecan]

booooo

[restive plover]

RIP Teams plan

[tight island]

idk man.

is it "in the coming weeks"

[regal portal]

it's not available in Donkey Kong?

[tight island]

Denmark

🇩🇰

[sick pecan]

it wants me to purchase plus plan because it doesnt respect my team plan

[tight island]

yeah i cancelled my teams and plus

to use pro

200 bucks is a steal

Dear Sam, Please disregard the EU, and launch it anyway

ty.

[warm gull]

its all relative isnt it.....in some parts of the world that is twice people's monthly salary. not making you feel bad - but its just how unbalanced the economy of this world is.

[tight island]

in some parts yeah

i agree, its expensive.

but compared to doing the calls on API unlimited times.

huge save.

[hoary laurel]

https://sora.com/

[hoary laurel]

Login

[pulsar carbon]

sora is out?????????????

[tight island]

yup

[pulsar carbon]

omg

[tight island]

no, god has nothing to do with it.

[dense halo]

time to wait

[rugged igloo]

yea

[rugged igloo]

im getting pro fluff it

[dense halo]

we need more inventions like tethered caps

[torpid kelp]

its scummy that you cant download without watermark on plus :/

just makes it a toy if I cant use the output

[dense halo]

awww

[torpid kelp]

id prefer to use the service im already paying for

[dense halo]

no way they locked watermarks behind a 200$ paywall...

[sick pecan]

if you want to defeat watermark i will sell you 'crop' tool for cheap

[rugged igloo]

LOL

[regal portal]

i got timed out for 2 minutes because i showed a screenshot of sora's username selection page won't let you use the word more-on (spelled correctly, but don't want to trigger the auto-mute-bot). In fact, you can't use most any word that is even mildly an insult. Not a good premise for things to come

[hoary laurel]

It's work

[tight island]

why would u circumvent the watermark

[sick pecan]

if you got 60+ gb of vram you can do anything you want with: Hunyuan Video ComfyUI 1-Click Launcher

[dense halo]

"please use it creatively and respectfully"... for sure

[regal portal]

damnit i can't even thumbs up your message in this server!

[tight island]

its only natural

even if u remove the watermark OAI can easily figure out who made the video, since the ID isnt the watermark

[rugged igloo]

teams just chilling with nothing

[dense halo]

bruh

[tight island]

sis

[dense halo]

bcos 20$ is more than 25$, people who pay more get more

[tight island]

but teams is a min 2 seat

so 25x2 + tax

[dense halo]

20$ > 2x25$ 🙃

[sick pecan]

plus is like casual, teams is like small business, enterprise is like big business, pro is i want to marry openai and have its babies

[tight island]

yes

i want sam's babies

[rugged igloo]

loading

[proper marlin]

I'm thinking of using it for customer service application within my workspace

[quick lodge]

I keep getting this message when I try to use Sora (and I have a ChatGPT Pro account):

[proper marlin]

like it wont let you sign in

[rugged igloo]

their servers are gonna go poopy this week

[torpid kelp]

give 2000 sora gens per day to plus user sam altman

[jade glacier]

so I just upgraded to "chat gpt plus" but chat gpt plus features dont work for previous chats

[proper marlin]

need more power

[torpid kelp]

my first sora gen still in queue

been 20 minutes

[tight island]

its launch day, its supposed to be slow.

[dense halo]

average launch day experience

[tight island]

sora is working overtime tonight

[rugged igloo]

feel like im in world of warcraft new exspansion launch day right now quing to get in

[torpid kelp]

it finished generating, result is horrible. Gave it 3 instructions followed only 1

[tight island]

dud u even listen to the podcast.

[rugged igloo]

lol

[ebon canyon]

To be Honest... They should have not released that to public.. i already saw some things people doing with it on the recent libary which already tells a lot about the usage...

[tight island]

its not a prompt n go generator.

they said it so themselves.

[sick pecan]

naughty?

[ebon canyon]

Already creating CNN Fake News stuff etc..

[errant gull]

is the GPT O1 pro releasing soon orrr

[dense halo]

entertaining times incoming

[trail torrent]

sora came??

[tight island]

its already here

[trail torrent]

sora.com

[sick pecan]

cnn already is fake news! (jokes!)

[dense halo]

o1 pro is out already

[ebon canyon]

XD

[errant gull]

woah how do i access

[tight island]

money

[errant gull]

yea like how

[dense halo]

200$

[torpid kelp]

well it sucks

[trail torrent]

guys sora came?

[errant gull]

nah what

message limit? 200$ a month or one time

[tight island]

month

[south kayak]

why is it unable to do that? https://chatgpt.com/c/6751a4d4-ce90-8000-be88-b822e7db4f78

[torpid kelp]

one time for unlimited chatgpt pro forever ya

[errant gull]

!!!

[sick pecan]

sora is out and like every openai product, day 1 is gonna suck

[dense halo]

read the bottom line

[torpid kelp]

o1 was great day 1

[tight island]

Pro is basically Enterprise Lite

[sick pecan]

okay like every openai product (sans o1)

[ebon canyon]

like well.. here we goo.. People already generate fake news for there crypto scams and stuff XD

[dense halo]

breaki news is crazy

[regal portal]

Invosters

[tight island]

hah i can see openai clearing out badactors.

[sick pecan]

you gotta be murican and have pro or plus

[regal wadi]

is sora going to release in norway aswell?

[dense halo]

BCG INN NEWS 😂

[sick pecan]

www. what is this the 90s?

[dense halo]

later this day i think

[regal wadi]

oh alright ty

[ebon canyon]

yea but its about the access to a bigger audience and making it more easy to generate.. I dont say its new..

[torpid kelp]

sam altman not pulling weight today

[sick pecan]

is just sora.com

[dense halo]

wrong lonk trust

[regal wadi]

has anyone been able to generate something yet?

[ebon canyon]

y

[regal wadi]

How is it?

[sick pecan]

day 1 teething issues

they probably had chatgpt generate the url link

[ebon canyon]

its ok, i mean you have to test it, run multiple prompts and iterations. Also havent looked into the storybord yet but looks quite interesting overall

[regal wadi]

Alright alright

[fossil oasis]

Most of the major media channels are already "fake news" by nature, so don’t worry about it. And please, stop flooding the chat with the same complaint over and over.

[ebon canyon]

Dude XD first of all what are you talking with flooding the chat XD i sent like 2 messages or so.. And it dosent matter if media channels are already fake or not its about the accessebility which opens a whole new door in which mass that stuff can be created.

[sick pecan]

if teams could generate id be making cnn breaking news about ppl faking cnn news

[ebon canyon]

haha

[tight island]

they have also said that sora isnt the kinda of "Bot", where u just post a simple prompt and get a 100% accurate video.

you're supposed to work on each segment of the video for a proper output

[ebon canyon]

have you checked the storybord yet?

[tight island]

yeah

its NEAT

[ebon canyon]

kk

[dense halo]

ufff

[ebon canyon]

hahahahahha

cant wait to scroll social media the next days XD

[fossil oasis]

With all due respect, that wasn’t my point. Sure, some people might try to drink "Universal Antifreeze + Coolant" at some point... but that doesn’t mean we need to take it off the shelves for that reason.

[tight island]

it was live, u dont expect them to actually sit there for an hour working on it, or do you?

[dense halo]

i hope sum cool space stuff gets made with it

[tight island]

Iterative Video Generation.

constantly rerun to fix.

[dense halo]

+200% ufo sightings

[tight island]

i agree.

but i dont think so many people spend so much time of "getting it perfect"

they want something functional that they can use right now.

while others in the dark work on perfection which u see every so often

like Prompting isnt just inputting random text.

[ebon canyon]

its probably the user input. check the prompts people have used they are just "bad" its the same as we saw on MJ in the early days. the better the prompt where getting the better the outputs. Now MJ images are on a extremly high level. So if you have knowlage about how Video work in general (Camera, Angles, movements) etc.. you can probably way better prompt it and get better outputs.

[restive plover]

They are really forgetting people that use the teams plan

[tight island]

i think its intentional

[dense halo]

massive saturn disrespect 😂

[restive plover]

It’s really sad to be honest

Why do you think that they’re forgetting the team’s plan intentionally

[tight island]

i had teams plan before.

i go with pro now.

i was using Teams and Plus and API previously. now i only have to use pro.

[ebon canyon]

is that a Rocket or a disc player XD

[restive plover]

Yeah, I think I’m about to cancel my teams plan and leave the workspace

[dense halo]

chatgpt generating the future designs of rockets 🗿 its just too smart

end of stock videos

[ebon canyon]

dayum 🙂 gonna try the storyboard now and try to create a short brand commercial. I am wondering how it handles it.

[torpid kelp]

Thats a dido 😂

[tight island]

end of original GIFs

all AI GIFs

[torpid kelp]

yeah i dont think so, when its out of capacity for god knows how long

[tight island]

use storymode?

[torpid kelp]

why?

[tight island]

are you not gonna iteratively improve it?

or do u expect a perfect output each time.

[torpid kelp]

yeah but can i prompt into that

[ocean sable]

so sora is in the $200 plan?

[torpid kelp]

im using remix

[tight island]

but using storyboard makes it just so much easier

[dense halo]

when is Sora available?

[tight island]

did u even watch the video?

[tight island]

now.

[dense halo]

Odd I do not see it

[ebon canyon]

XD

[torpid kelp]

nvm i figured it out

[dense halo]

this weeks is gonna be funny

[feral zinc]

https://help.openai.com/en/articles/10245774-sora-billing-credits-faq#h_78da66cecb

[tight island]

the dopamine machine is roaring

[dense halo]

thats the future 🗿

[tight island]

progress takes the shape it takes

[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.

[tight island]

i'd believe the cats flying, WHEN I SEE IT.

[regal portal]

just tried my first generation. Yay

[dense halo]

realistic

[tight island]

and how many iterations are you in?

[torpid kelp]

elon musk ddosing the server

[dense halo]

she broke her neck and didnt stop working out

💪

[regal portal]

@sick pecan unrelated but have you been playing around with lm studio?

[past fiber]

Looks like the 200 dollar Subscription is slowly making sense
Not only you get o1 Pro
But also Sora 1080P with 500 priority videos

And we still have 9 days to go

[dense halo]

?

[south kayak]

whats wrong with bro chtgpt

why is he slow af

[regal portal]

The problem with the high pro price is that, when these features trickle down to the plus subscription, it will upset a lot of users who put up with the 200$ a month price tag.

[tight island]

meaning.

i pay more i get more at the expense of the previous tier.

[dense halo]

time to dissapear

[tight island]

i dont like the free thing.

thats just me being a d bag

[past fiber]

The free users only got access to the mini models
not the high end
o4 mini and o1 mini

[regal portal]

i guess, but that didn't really bother me as much. Also 20 vs 200 is a significant financial step

[past fiber]

Yeah but maybe there is more planned to add in the next 9 days to justify the price

[regal portal]

200 is more like "i use dollar bills to light up my cigar" kind of money

[shadow dock]

Shot in the dark... but I think Sora load might be making ChatGPT dumb

[regal portal]

is o1 128k context? is o1 mini?

[pulsar carbon]

i can't use sora lol

[shadow dock]

Like what happened here lol

[past fiber]

They could do a limited version but better then plus
Something around 30 - 40 would be still really good

[regal portal]

what's the plus context for 01 and 01 mini?

[eternal rock]

Join the conversation:
#sora-challenges: Submit your wintery videos to the Frosted Frames event!
#sora-discussions: Talk about everything Sora related!
#sora-reels: Share and discuss the Sora videos you've been making!
#hall-of-fame: Submissions with 10+ community votes in #sora-challenges will be featured here in the Hall of Fame!

[past fiber]

Plus o1?

[regal portal]

oh, poo

[past fiber]

Oh

[regal portal]

well, we'll see what will happen to the context window by end of 2025

33k 😄

[past fiber]

Maybe this week or next week
better gpt4o with larger context window

[torpid kelp]

if gpt 4.5 can beat sonnet 3.5 in coding, I will be Sam altmans obedient servant

[regal portal]

I haven't paid for claude yet, but from the free plan, it seems more censored than gpt, from a swearing perspective.

[torpid kelp]

will we ever get a completely unfiltered capable AI trained with 1 trillion+ token

[tight island]

hmm

[tight island]

yes, within our lifetime, if its going at this speed.

i'd give it 1-3 years

[regal portal]

I'm playing around with some uncensored llms but honestly, all the uncensored ones are really dumb compared to chatgpt/claude.

Not great in terms of storytelling.

Good for a few funny one liners, but nothing more than a paragraph or two, then it starts repeating itself.

[haughty hollow]

bro what yall messed up with GPT now

are they ok in their minds ?

[tight island]

they're diverting power.

seems like.

[haughty hollow]

it lit now cant read anything at all from past history duh

posting a code and giving the task, replies with : "Ok how can I assist you"

[past fiber]

Yeah the sora is now taking most of it
need to wait a little so the wave goes away

[regal portal]

probably not the best day to do intense stuff, being launch day of sora

[torpid kelp]

guess im not working anymore tonight

Actually ill just use claude inside copilot

[tight island]

its fine, ill give it 1-2 hours.

the hype wave needs to die down.

[regal portal]

claude INSIDE copilot?

[tight island]

Copilot InSiDe Claude?

[torpid kelp]

github copilot

[haughty hollow]

did they removed the change model too ?

why some chats dont have it brihos

[tight island]

its fine.

[regal portal]

this thing?

[haughty hollow]

yeah but not there

in down of chat you find it

[trail torrent]

Looking at her following: 😐
Looking at her reposts: 😐
Looking at old pics of us: 🙁

[torpid kelp]

damn

[south kayak]

I just paid 20 dollars so chatgpt can create me flashcards, all it does now is repeating itself 20 times

20 times the same cards

like cmon

[haughty hollow]

imagine paying to get scammed lol

now the worst is the 200 usd

[tight island]

ehh u guys must be using it improperly.

[south kayak]

then check this out

3000 lines with the same text over and over

[tight island]

its not the same

[south kayak]

it is

[tight island]

i am reading it rn

[south kayak]

10-20 lines

different then repeating

[tight island]

i know why it does that

it's having a stroke right now.

thats what's happening

[south kayak]

yea

[tight island]

ahh i see the error.

[ebon canyon]

interesting sora lets you produce "fake" cnn videos but if you want to create a "nike commercial" it tells you the content might violate their policies.

[tight island]

did u remember to ask it for 300 unique, individual french flaskcards.

[south kayak]

ofc

[tight island]

its putting all of them in a single card.

[glass willow]

Your original prompt is the problem.

[tight island]

yeah

[south kayak]

bro is having a stroke

[south kayak]

https://chatgpt.com/c/6751a4d4-ce90-8000-be88-b822e7db4f78

[tight island]

you're not treating it with love.

my man.

[south kayak]

now it made 60 lines

perfect

[glass willow]

Also doesn't help you keep asking it repeatedly, it will repeatedly do the same because it doesn't understand what you're saying meaning it doesn't connect what you mean by not repeating, there can be many senses of what is repeating.

[tight island]

my man you're being rude to it.

its shown that performance is directly tied to positivity.

[torpid kelp]

give it a good beating

[tight island]

or being more specific adding more details, maybe an example

[south kayak]

I made it do multiple samples

[glass willow]

Probably not starting new chats and keeping one mega window.

[south kayak]

the samples were fine, then i asked for 3000 different cards with that layout

[torpid kelp]

someone tell elon musk to get off of sora

[tight island]

so you're basically torturing my dear gpt

while he's being drained for compute for sora. and pro users.

[south kayak]

yea it took me a while to teach how it should do it

[glass willow]

That's a prompt issue

[tight island]

yeah i am certain its a prompt issue

[glass willow]

I get most my answers within first 4 - 5 prompts no matter how complex, just need to think like a bot does and then how not to think like a bot.

[tight island]

yeah because if we reversed the roles.

u having to generate 3000 lines for me. that would take longer.

[south kayak]

Ok I show you how a noun card should look like and how a verb card should look like.
make additional cards for basic nouns and verbs!!

example:
la table,table<br><br><small><font color="#333333">La table est en bois.<br><br><i>The table is made of wood.</i></font></small>
finir,to finish<br><small><font color="#333333">je finis<br>tu finis<br>il/elle finit<br>nous finissons<br>vous finissez<br>ils/elles finissent</font></small>

this is what the prompt was

[tight island]

okay, first did u specify if it should use Codeinterpreter, and what do you derive your nouns from, microsoft word?

like details.

[south kayak]

bro

it initially told me how anki cards should be coded

[tight island]

still, you also have to remember its context length

[glass willow]

So your prompt is too general, there is no detail to it.

Generate me a pink pineapple, it will give you what it thinks is a pink pineapple.

Or you can describe exactly what you want and it will think of a pink pineapple in the way you describe it.

[tight island]

yeah

Task - Purpose of task - verbal description.

[south kayak]

so how do i bring it to create 1k cards instead of 60??

its making 60 cards for the 3rd time now

[tight island]

you're asking it to generate more than 50k words.

[glass willow]

I mean, what are these unique cards, to me even 100 of the same cards is considered a uniue card, because the card is unique you dont say the data in the card must be unique etc.

[south kayak]

I said it shouldnt repeat any verbs or nouns, it still did

[glass willow]

Where must it not repeat this? In the cards or in the response or where?

[south kayak]

in the file

[tight island]

each should be unique

[hexed rain]

Anyone know if Sora will be available for teams accounts?

[glass willow]

Well this is stuff you need to explain to it lol.

[tight island]

yeah its not a mind reader, Details Details, Purpose of details.

[south kayak]

ok tell me what i need to prompt

[eternal rock]

Get notified for server events!

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

[south kayak]

and i show you how it still doesnt work

[tight island]

and thats not even a joke, because its true.

[glass willow]

Actually i don't agree with this, i've had a few rants that made it more.. understanding. 🥲

[tight island]

not needlessly rude

just general unreasonable hostility, towards our future overlord.

[south kayak]

again only 60 cards

[tight island]

now you're forcing it.

[south kayak]

bro

[tight island]

okay

60 cards, yes, But where? and how.

60 individual cards on a .txt file? using some type of script?

more specific.

[south kayak]

stop trolling

[tight island]

i aint tho.

i am being completely serious.

[south kayak]

no

you act like chatgpt is a 10 year old

[eternal rock]

`` Server Tip: ``

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

[tight island]

you act like its able to read your mind.

bro, it's a tool, it dosent understand your intentions, be more specific, go into more detail.

[glass willow]

Something like this:

This is an example format of 2 noun cards:


la table,table<br><br><small><font color="#333333">La table est en bois.<br><br><i>The table is made of wood.</i></font></small>

finir,to finish<br><small><font color="#333333">je finis<br>tu finis<br>il/elle finit<br>nous finissons<br>vous finissez<br>ils/elles finissent</font></small>

Each card contains nouns, write out 60 examples of these noun cards in the same format and the text within these cards with nouns must be unique from other noun cards, carefully reflect after each card and confirm it does not contain the same noun text as previous noun cards then proceed to generate more.

[honest jay]

I think y'all have touched on it already, but the other thing to keep in mind is that the data analysis environment is not a bypass for the context limit.

[tight island]

yeah

[honest jay]

In other words: if it couldn't generate all the requested content within a chat, there's nothing about the data analysis envornment that will make that more possible

[tight island]

you're attempting to create more than 50k words.

[glass willow]

Didnt test but yeah, i don't know exactly what it is saying since it's not my language. 🙂

[honest jay]

Right - there's no python library for "random but accurate cards"

[south kayak]

It was able to create sample cards without specific details, then I asked to make 1000 cards and it struggles

[tight island]

2 completely different sizes.

[south kayak]

why was it able to write 3000 lines then

[tight island]

and they're identical yes? they repeat.

[glass willow]

Try what i sent above, see if it helps.

[tight island]

Since you used codeinterpreter, and it was required to meet 3000 lines. almost running out of context, achieving it job by replicating.

[honest jay]

Yeah python can multiply no problem 😁

[south kayak]

no I refund my 20 dollars now

[tight island]

no.

i don't have that kind of authority.

[south kayak]

REFUND NOW

[honest jay]

https://help.openai.com/en/articles/7232895-how-do-i-request-a-refund-for-chatgpt-plus-or-chatgpt-pro

[tight island]

[glass willow]

The browser can handle a lot of lines of text, probably around 100k before it gets bad, i remember a old reporting system having this amount of data and didn't crash the page but was laggy as heck.

[tight island]

its more it's jupityr envoriment.

[glass willow]

Probably not very optimized, can do a lot of browser tricks to make it very fast.

[tight island]

problem is that the code interpeter enviroment is very limited.

like very.

[ornate raptor]

Anyone else noticing 4o is having tons of issues today with thinking? It seems like its context window is barely a handful of messages before it blanks out and forgets what's happening

[tight island]

many issues would be solved if the CodeInterpreter enviroment was its own thing where u had folders. code enviroment, etc directly ingrained in it.

a rudementary IDE or VDesktop

[glass willow]

I mean it can't be that bad, even blizzards game launcher runs in a sandbox environment with web application, same for steam.

[tight island]

True, but CI is under very strict rules.

and has remained almost unaltered since launch

[glass willow]

They just don't care about it so much at the moment yet, after all why are you making single code files so big.

[tight island]

yeah

[glass willow]

Like this dude making cards like this, makes no sense, should be a card template and a data feeder, like vue and gql

[gritty pagoda]

why i cant use sora :(((((

[glass willow]

else he has 1 html page at 2mb

[tight island]

true

i think many people expect GPT or similar services to be able to do everything

oooo lemme see

[honest jay]

Very high demand right now, things will calm down soon!

[glass willow]

The people testing it aint having a fun time, they have strict rules to allowing you to test it, so you can't use it till it's released and probably don't want to be part of the testing guys the way they have to go about it.

[tight island]

yeah give it some time an hour or 3.

or A MONTH

[glass willow]

I have many plans for sora, music videos for my ai songs.

[tight island]

ahh.

[gritty pagoda]

i mean, when i try to login with mt chatGPT acc, it says that i cant create acc... but i already have an acc

[tight island]

AI Music Video.?

[glass willow]

Ofcourse, imagine vikings in battle gear playing drums singing in between a battle.

[tight island]

imma try having Sora show me something FOURTH DIMENSIONAL, i am here asking the real questions.

[gritty pagoda]

i need to create a new acc to use sora?

[glass willow]

I mean it's just basically speaking to the audiance, same as every other yt video (4th wall)

[tight island]

use your existing account.

i think the gist is @glass willow

DALLE to make the scenes. Sora to make the video, using Images as guidance.

[glass willow]

Something like that, i mean i want to bring a lot of my drawings i made as a kid into movies so that would be pretty cool.

[tight island]

imma need to capture the specs.

and have o1.pro Create a detailed video

[glass willow]

But at the rate sora is going, it would be a pretty big let-down if it doesn't beat the others then like, what was the point of waiting lol.

[tight island]

if i just give sora all its capabilities + what i want done, How many imgs we need, storyboard.
basically having o1 create the story and prompts and text, then do the img work with DALLE.

[gritty pagoda]

i use google acc to login in, when i try do it, its says that i cant create new acc on sora.com

[tight island]

so if u can maintain Character consistency in Dalle. then you're basically set.

[glass willow]

Thanks sora.

I bet all the yt reviewers spamming the link.

[tight island]

yeah let the hype die down a little.

go to sleep, take a nap, get lunch.

or do as the others, Plan your project, while you wait.

[gritty pagoda]

that's not the problem. sora.com is not recognizing accounts that are linked to Google, it is saying that it cannot create new accounts, even in existing accounts.