chatgpt-discussions

[chilly quest]

Nothing is static, correct. Things change for good, not for worse? They literally made everything worse

[tight island]

heck i could build you a github repo simulator in less than 1000 lines of code.

that uses the api to create the entirety of the project even all the code.

in less than 5 min

this would output thousands of lines

thats the difference between api and web, but this requires you knowing a little about how to build it. or using 3 mini high to do so.

the more effective and logical your thinking is the better results.

build a project that makes your next project.

[native otter]

LLMs don't always output consistent result especially with no control over stuff like temperature and seed, but good promoting certainly helps.
Also, 4o is not the best at programming, it's not bad, but use it for learning instead of more complex tasks, as mostly you would need multiple prompts, and each prompted pretty well. So use o3-mini-high or o1 for programming (o3-mini can also work)

[tight island]

based on the report from elon, the issue with many llms is that they are trained on lies. instead of absolute truths.

[granite harness]

Ah yes, Elon, purveyor of absolute truths.

[native otter]

wow, elon said something I never knew before! *||(being sarcastic)||

[tight island]

hey i dont like the guy either.

so chill out

[chilly quest]

This is my only problem right now. It gives me garbage results even for small tasks.

[tight island]

already out of the box early grok 3 beats o1 by miles

[normal quarry]

can you share a chat to demonstrate?

[native otter]

You tried?

[eternal rock]

`` Server Tip: ``

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

[tight island]

ofc i tried

Grok is vastly better in some instances.

but its also very dependant on the user. i do like chatgpt

but looking just at the data, gpt is hella restrictive and grok isnt

the base grok 2 for their 5 dollar plan has 113k context window.

so idk man you tell me

[azure urchin]

same right now

[native otter]

It's not just about context window, I do care about tooling A LOT

[azure urchin]

Unable to load conversation

[tight island]

i care little of the man elon

[granite harness]

Feels like discussion of Grok can head on over to #ai-discussions.

[tight island]

i care more about whats most effective and and i certainly think chatgpt could use an upgrade. or some less restrictive parameters.

[crisp cairn]

why gpt 4o is so slow today?

[tight island]

they probably prepping for 4,5 release.

[native otter]

Well Canvas, Data Analysis, and Custom GPTs are all features I use on the daily basis
I may skip days not using very smart reasoning models like o1 or o3-mini-high, but not any day without 4o for example...

I use 4o for general convos, tooling, and learning, it's excellent at those 3.

[tight island]

should be either this or next week if i recall correctly.

[native otter]

But I would definitely like to see higher context for 4o

Just not everything is just context

[tight island]

thats what i am saying

openai can learn from this "elon" everyone so desperately hate and vice versa.

[chilly quest]

Yes

[sharp pulsar]

Is copy paste broken for anyone? I cant copy paste a single line of code from a code block given by chat gpt, it immediately deselects

[native otter]

Especially if not implemented well, it could end up similar to Gemini, where it's 1 million tokens sure, but it has the same exact problems as 4o when it runs out of context, and is less smart in my tests (But convos and vision capabilities are pretty good, especially vision)

[tawny mirage]

Is DeepResearch broken, or something?

[normal quarry]

no, it's working fine for me (macOS Desktop App)

[wild halo]

Is desktop screen sharing not available currently for MacOS desktop app? I no longer see the option.

[normal quarry]

you can only take and share a screenshot with the macOS desktop app right now

[rain jay]

4o coded up an EEG(brainwave) simulator with me, very effective for calibration.

[worthy lynx]

hey so how can we report a chat if the CoT of a reply goes way off the rails?

[honest jay]

Typically the 👍👎 buttons in the UI will be the easiest way to give feedback. If you want to report something otherwise, there is an OpenAI support chat at help.openai.com!

[worthy lynx]

thx @honest jay , it deviated widely from the original task and some of the verbiage was a bit unsettling, so thought I should raise it. appreciate the guidance

[full surge]

Does Chat gpt 4 have limits when it comes to debugging and writing a large code?

[whole roost]

https://openai.com/form/model-behavior-feedback/ here you can be more specific about stuff

[orchid pollen]

when screenshare hits computer?

i can do it already on phone for a while now..

that would help so much sometimes

[worthy lynx]

got it, thx!

[inland stream]

Please allow o3 mini in Projects or in GPT builders, that will be a huge success. i will not access anything more for this life.

[onyx eagle]

Do you get a notification when you have 25 messages left in o3-mini-high ?

[plain arrow]

This exactly. The gpt's used to be amazing until it was aligned into the ground. I was designing rpg campaign worlds and it was like having a live assistant to keep everything organized. Then it face planted at some point ~9mos after gpt-4 was released.

[prime swallow]

Agreeing about that link; I understand from requests made about sharing urls I've seen before that OAI staff can also use the url, even if share chat is disabled for that chat, to take a direct look at the concerning content.

[tiny salmon]

Is 4o really better at reading images than o1? Use case is reading charts

[tight island]

its also to avoid those nasty links, innit?

[worthy lynx]

Perfect 👍 👌

[prime swallow]

Huh?

I dunno about now, with content flags (some? all?) removed. I've tried to make shared chats of conversations where there was a content flag, it was auto blocked then near as I could tell.

But chatter's goal appears to be: Saw concerning output. Wants to inform OAI.

I'm all about empowering that, and 'here's the chat URL' allows them to directly see that chat, which in reporting concerning content myself I've been asked to do sometimes.'

[whole roost]

those poor people who actually looked at mine

[rugged igloo]

i hope thye doont look at my degen brainrot project

[whole roost]

i've handed in my chats, many of them because i am fast to complain and wanting improvements

[quick meteor]

nobody knows what grok or claude is

[wise anchor]

anyone else have issues with LaTeX not rendering on macos desktop version

in o3-mini

thats been happening to me for so long i can re open the chat and it renders but its lowkey annoying

[gentle warren]

Hi to all any "cannot load conversation" problems known currently? Thx

[dense halo]

Apple intelligence isn’t integrated on outlook macOS app in the “task feature” area. If I write a task, and right click to use “writing tools” it doesn’t work.

[tawny mirage]

Did they just impose stricter limits on content length with o1 pro mode?

I submitted a message I'd been submitting many times before and instead of responding it immediately responds with a failure, and the network tab says 413: Content too large.

[fringe trench]

maybe they are releasing o3 full today

maybe not

[tawny mirage]

Yeah it's a recent change. the dropdown for retry with a different model is gone on a very recent prompt. Trying to edit/resubmit yields an immediate failure.

[next veldt]

What's with the features on mobile that are missing from desktop for no apparent reason? Namely being able to regen images on the same prompt, and being able to edit messages in the chat which you originally attached an image to. Both of those are options on mobile, but not desktop and it's quite bewildering lol.

[frozen birch]

again problems with o1 pro.... sigh

[normal quarry]

can you share the chat link?

[frozen birch]

offc

https://chatgpt.com/share/67b5332e-f844-800e-b2fd-cf67afe33e72

[normal quarry]

in that chat you're just prompting "test o1 pro"

[frozen birch]

and you can see that model used is o1 pro.

[frozen birch]

i mean why you want to specifically that talk? 😛

[normal quarry]

it's working for me

[frozen birch]

Rob its no matter of other people

its matter of using heavily chatgpt for many hours/24h interval

their systems are messing up

[normal quarry]

so have they fixed it for you?

[frozen birch]

as you see nope :/

[frozen birch]

you rememeber those stupid limits with ~100 messages/24h 2 months ago for o1 pro users?

i think its related with it

worstest thing - when you are flagged for "suspicious activity" you have access only to 4o.

[e-mail]

np

[void token]

the ChatGPT web search thing is so cool

[frozen birch]

basically - you pay 200 bucks for being limited at some point. ahhh gonna sleep, i dont want stress at night.

[digital jacinth]

Hi bois

[orchid pollen]

yeah i rly wonder why is this a thing?

just move to a project should not change anything i think

but it does lol same with automated tasks

i always have to remove it from the project -> edit the task -> move back to my project and the Job still works fine

[harsh pilot]

Just wanting to drop this here.. for the past while, the AI's been, almost, "unchained." The writing is better, everything feels better. I'm not getting any "eek, you said something potentially tangentially potentially naughty-adjacent" warnings, or anything. It's amazing. ❤️

[digital jacinth]

Im trying to solve some graph maths but chatgpt answers are wrong
Wut can I do

Someone help

[dawn sentinel]

Use o3 Mini High.

[lost raptor]

The most "casual" realistic conversation I've ever swtiched to with chatgpt.

[digital jacinth]

boi Im using free version

[dawn sentinel]

If you want the best math AI, you'll need the Plus version. Or else Deepseek.

[digital jacinth]

nah bro deep seek might steal my data, I seen it one video.

[eternal rock]

The OpenAI Discord is an actively moderated server.

-# • Do not share inappropriate content (violence, sexual, or drug-related).
-# • NSFW content is strictly prohibited and will result in a permanent ban.
-# • Report model issues using the feedback tool in ChatGPT, not on Discord.

[spice shell]

anyone know how i can get chatgpt to remember all conversations and messages?

does anyone have any prompt tips for using chatgpt as a teacher / tutor? thank you.

[low orbit]

chatgpt is such a good dude

i wish i can give it a cake

[wise anchor]

should get a free month for that bs

[inland stream]

In projects, if write instructions, o3 mini is gettimg greyed out :(, i have to manually send the instructions on every first message 😦

[inland stream]

There are use cases to have fixed knowledge files, and fixed instructions, and it is hectic to manually add them in start of chat session with o3

[orchid pollen]

yeah it makes no sense they allow some features not in projects

its just a folder where my chats are stored so whats the issue?

[low orbit]

just ask it the things and talk to it!!

haha im so helpful

[orchid pollen]

do what i do

copy the question from random ppl feed it to gpt and copypaste it back in discord

thats level 9000 😄

[low orbit]

hahaha

biggest brain

[orchid pollen]

one dude asked like 5 questions in a row when the 6 came i just send him the chat history and told him im bored 😄

[low orbit]

but seriously i been studying for this certification and im just going back and forth with chatgpt like its some dude you know and it helps me figure stuff out

never really needed a prompt so much

[orchid pollen]

what exactly you need?

[inland stream]

I realised you dont know more than this and you cant understand technically more than this. So can ignore my message.

[orchid pollen]

yes sure buddy i have no idea about computer

where this front comes from lol?

im sure im way ahead of you 😄

[inland stream]

Thanks for confirming you don't have idea. Its fine you will learn gradually

[hollow vortex]

python is for babies though

[orchid pollen]

sure what ever makes you happy kid

[normal quarry]

rule 1 of this server is to be respectful, think about that before you type

[hollow vortex]

why u studying a certification

chatGPT will make ur certification useless

for example @normal quarry job will be taken soon from the very thing he uses

[normal quarry]

I don't have a job I am retired

[hollow vortex]

oh dayum

u have it good

[eternal rock]

 Off-topic chats have been cleared by a moderator.

[inland stream]

Chatgpt ❤️

[long pier]

Hey fellas, have y’all seen the changes they made whenever you write a script?

[potent geyser]

hi, i have a problem. when the chat gets too long, it completely vanishes from view. can i recover the text there somehow, or is it lost? i'd take it and migrate to a new session. looks like this, chrome:

[whole roost]

Doesn't refresh help? I have the issue as well but usually it reappears after refresh

[tacit whale]

hi

[hoary laurel]

@tacit whale hi

[visual acorn]

Guys, my ChatGPT keeps learning from me even though the memory is full. What's going on? Is this normal?

It seems to be more subtle though..?

But I could almost be sure it gets those things from previous conversations

[rugged igloo]

bruh your telling me bro

[rugged igloo]

lucky can i retire with you robert

[normal quarry]

you can retire whenever you want. I retired at 50, which was almost 8 years ago, I have some small side hustles going on to make pocket money but they are all hobby related

[rugged igloo]

i would have never guess your in that age range i legit thought your around 30's a tech bro

[visual acorn]

Isn't it funny how you're referring to an AI as bro? Shouldn't that be reserved to only us, humans? :( (half joke)

[rugged igloo]

well with how human it sounds now its just habit lol

[visual acorn]

Yeah, that's kind of what I'm getting at

[rugged igloo]

before i just called it a bot now im not okay dude

[visual acorn]

It's like our brain almost fits it into the human mold whereas it's probably not the best thing

[rugged igloo]

i agree lol

[normal quarry]

I've been using computers since I was 10, also a ham radio operator and radio astronomy enthusiast - so I am a bit of a nerd or geek

[visual acorn]

So you're used to using terms of endearment?

[rugged igloo]

yes

[rugged igloo]

ahh that explains alot.

yooo savage chat gpt is

wuts tht for ?

chat gpt chill my guy

[brave glen]

Eheheheh

[rugged igloo]

lol thats funny

also chat gpt couldnt beleive linkin part replaced chester legit got angry

[cobalt perch]

Does anyone know when can we expect sora and operator in EU

[brave glen]

Loool

[rugged igloo]

in the coming months literally

[cobalt perch]

Perfect

[brave glen]

Can you share your custom instructions in DM?🤫 Or nah Black Reaper

[rugged igloo]

sure

[rugged igloo]

thats preety funny acually

[jolly temple]

are u plus or pro? my gpt doesnt talk like this lol

also, is anyone's gpt4o weirding out? like as its typing itll just disappear, or the reply will just not exist

[lean hatch]

Er... I think GPT just mirrored you.

Also, leave the girl alone -_-

Just because Chester decided to selfishly kill himself, doesn't mean the rest of the band has to starve.

And to boycott them over it is childish. The girl is talented and brings a new yet familiar sound.

[summer pagoda]

sorry what?

biggest fail in ai history

[visual acorn]

You could ask it to execute a Python script that does the thing

Though ideally you wouldn't have to

[cloud pewter]

SMFH

aaaaand apparently I got muted again just because of my silly markdown text styling.

[cloud pewter]

anyways o series models now supports both image and files and this is a bug that needs to be fixed

Also “get pro”

[whole roost]

Do it

[potent geyser]

nope, didn't help, but i could recover the text in read mode

[whole roost]

maybe the app could also help to recover it, windows or phone, they bug less on lengthy convs

[potent geyser]

yeah maybe. im working on code for an application, and sending code snippets back and forth.. so having it all disappear is always a shock

[halcyon wing]

hi guys I'm new in discord and new here. dunno if this is the correct place (forgive me if is not please) to ask for a proper chatGPT course focused on GPT plus to increase my productivity at work

[worthy lynx]

Sadly, I went to submit at this site and it throws an error upon submission saying "The request is not valid."

[whole roost]

i submitted one earlier, maybe you missed one of the mandatory boxes?

[worthy lynx]

I'm thinking it must be a character limit or something, my prompt was ultra-long, requiring o1 to reason for over eight minutes (which is why I think the CoT went off the rails). Unfortunately OpenAI doesn't specify what limits on the submission form might be in place. 😦

[whole roost]

you could try to attach the core issue and then just send the ID, i think they can check it then

[worthy lynx]

I'll try that!

That did the trick, issue submitted. Thx Vishanka!

[limber oxide]

anyojne having issues with chat gpt right now?

im not receiving answers

[violet cove]

I'm having issues. I can't continue any chats.

[cobalt lynx]

Boooooooooooooooooooooo.

[limber oxide]

im screwed

[fiery mango]

same

[limber oxide]

i have a test tmrw, dont know how to study without AI anymore

[modern basin]

gpt Is down now

[native juniper]

This keeps happening a lot recently despite having plus :/ also I’ve noticed the 4o and 4 and 4o mini have been much slower than usual for the past few weeks

[cobalt lynx]

deepseek is always an option

[opal wharf]

Mini still work

[native juniper]

Use Bing AI then :/ (still hate Microsoft for lobotomizing Sydney and finally pulling the plug by switching models) so you should most likely expect an AI with the personality of a rock now :/

[limber oxide]

im using deepseek

[native juniper]

:/

[limber oxide]

:/

[spice cypress]

ChatGPT works???

[native juniper]

Sort of

[rugged igloo]

Pro

Is anyone else having this issue ? It seems 4o won't let me post images and search the web I usually could

[zenith lichen]

Looks ike 4o is broken for now

I can't send any prompts to it

[rugged igloo]

Rip

[scarlet delta]

Same I get nothing but errors

[ornate raptor]

Mine works but holy cow it's acting weird. It's slow and is constantly using full caps words and emojis. 4o has been off its rocker for weeks now it feels like

[whole roost]

i was just about to ask if GPT is producing nothing but trash for anyone else right now

[smoky reef]

gpt is being weird it works on and off for me

[whole roost]

yeah i just tried to start a canvas with it and it did nothing with it or my input and just said 'i have updated the document accordingly' even if this wasn't the task and it didn't actually do anything like an edit

and its sloooow

oh god it edits canvas now, this will take forever with that speed

ah

[orchid pollen]

is gpt down?

i always get this in my chat

jesus had to try 15 times to get it working

[eternal spoke]

down

[woven leaf]

yeah the issue is known

[orchid pollen]

oke

[rugged igloo]

Their prob updating the model.

As this happened when they they updated 4o with its new data set and new personality

[arctic sinew]

is it just me or does gpt 4o use more slang now

[half vine]

is o3 mini in the API?

[feral zinc]

yes

[half vine]

i cant see it?

[feral zinc]

then your api account is too low tier

[half vine]

what tier i must be?

[feral zinc]

3-5

[half vine]

oh and where it should be

[feral zinc]

and not all tier 3 have it depending of their usage

[half vine]

oh

where do i check my tier?

[feral zinc]

https://platform.openai.com/settings/organization/limits

[half vine]

huh 50$

will the o3 mini be in the api avaialble for everyone?

[feral zinc]

🤷‍♂️

[arctic sinew]

hi

[versed zephyr]

does everyone have this

[feral zinc]

its very old feature

[versed zephyr]

no way

[dark jungle]

I'm confused recently by what i'm seeing in the GPT interface.

I float between two different uses for GPT, either creative writing projects or coding projects.

Should I assume 03-mini-high is bad at creative tasks and use this for coding? but then again it has a "3", while GPT-4o has a "4", does this indicate an evolutionary step? so GPT-4o mini is better? Which should I use for creative projects? GPT-4o, o1? is o3 mini better than GPT-4o for creative tasks? is o3-mini better than o1 given it has a "3" indicating perhaps a new development release?

How do I explain this to my 70 year old mother?

[versed zephyr]

o3-mini-high is way smarter than gpt 4o but takes longer to process requests and is only needed for complex questions

gpt 4o is more of a all round with quick and intelligent responses

[foggy heart]

IMO, forget about 4o-mini and GPT-4.

• If you have a task that requires reasoning (because it's difficult, has nuances, or involves multiple steps to complete ...), use a reasoning model (o1, o3-mini or o3-mini-high).
• If you need daily, weekly, or occasional reminders, use GPT-4o with scheduled tasks (e.g., "Remind me every day about the weather in London").
• For everything else, use GPT-4o (e.g., translation, roleplay, simple coding tasks, easy math problems, web searches, etc.).

They plan to unify all these models into one that will automatically select the best model for the user's task, as they find their current naming conventions confusing 🙂

[eternal rock]

Get notified for server events!

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

[versed zephyr]

o3-mini has faster responses than high

[dark jungle]

so if we remove time as a factor, o1 is the best? again my focus is more coding or creative writing. I dont find speed to be an issue for any of them.

[rain jay]

He has to get down off his high-horse, it takes a while 😉

[foggy heart]

I'd say o1 for hard math and o3-mini-high for coding in general

[dark jungle]

how is o3-mini--high better than o1 for coding? what does the "high" indicate? i'm assuming o3 is a new release compared to o1? hense the "3"?

its become very confusing.

[rain jay]

Maybe he took a toke?

[dark jungle]

Reminds me of Unity (game dev) how they have several options of the version of unity to use when launching a new project .. this became a big issue for the company over time when they couldn't unify this into a single development pipeline.

[foggy heart]

You should refer to the benchmark (type "LiveBench OpenAI" and click on the first link) to assess the models' performance.

OpenAI's o1 was their first reasoning model. The o3-mini is a scaled-down version of the full o3 model, which significantly outperforms current models, at least on benchmarks. The difference between o3-mini and o3-mini-high lies in the degree of size reduction, which consequently affects performance and speed.

I can't provide the link here, as i will get chat ban lol

You can visit the openai webpage, then click on the "Research" page on the left. Everything is detailed here! (see on the left)

[eternal spoke]

so 03 mini now takes files?

[foggy heart]

Yes (file & image)

[eternal spoke]

sweet thanks

[foggy heart]

Yw!

[dark jungle]

When a software requires people to go digging through their website for answers to understanding how to use that software, that's a big problem. I develop apps myself and I always ask myself; if i give this to my mother, with a little exploration of the UI will she know how to use it without having to go digging through documentation and getting lost in jargon and dev terminology. Im curious if OpenAI recognize the flaw in the amount of options presented to the user. It seems they want GPT to be used by everyone, yet it seems designed for those with a vested interest in AI models. This should be reduced to a more simple solution. Reminds me of the TV commercials for Apple Vs Windows back in the 2000s lol ... in this instance GPT is the Windows character 🙂 ah well.

[foggy heart]

As I said, they will "regroup" all their model into a unique one : GPT-5

They obviously recognize this flaw:)

[rugged igloo]

i mean not all the time but it isnt out of the ordinary for it tto think tthat long\

i had o1 pro tthink for 8 mins a couple times but mostlys its 1-4 min usually for me

i wish it thought for longer tbh. i like it thinking i know others want fast but i rather think for hella long timee

i hope it got the puzzle correct thinking tht long

k that be embarrassing if itt didnt

[supple terrace]

I just got some random mobile app notifications about other user threads

Someone asking about wanting to buy an excavator for rental and yard work purposes with their daughter named Giana

model_editable_context

[abstract pendant]

Where can I find an updated page with the current plus message limits that include the newest models?

[cloud pewter]

I can’t just go and take $200 because I’m so broke 😭 also grok is on the horizon

[broken stag]

Schedule task is the worst

[cloud pewter]

its just the model hallucination

[rugged igloo]

wut are you even trying to have it do ?

its cant use deep research its just a task that you tell it to do ,its not a agent

[rugged igloo]

grok has only 5 reasoning request for like 50$ its not that impressive

[broken stag]

Just a simple road trip plan

[rugged igloo]

well without more context its hard to help you

[broken stag]

well, why it keeps saying that it would deliver by this time that time and never deliver the task?

[rugged igloo]

because its not possible lol

your giving it something it cant do so it just bs to you

only deep research can acually go and do something with it, tasks is just gpt 4o giving you a response back to your prompt its not acually browsing or looking things up

well yet... that is....

maybe in future it will be part of an agent or tied to deep research but until then its very basic

[normal quarry]

don't ask people to run prompts for you, also soliciting is against the rules of this server, you may want to delete the messages before a mod has to

[upper bloom]

o3-mini-high is dead? why i get answers from 4o when i use o3-mini-high?

[gloomy brook]

does anyone know when research might be available to non pro members?

[normal quarry]

OpenAI hasn't stipulated a time frame but it's expected as soon as a month away

[tiny salmon]

At least we're getting somewhere. Clearly I've been brainwashed by the Toyota Way, but I've worked with or for that group forever, lol

[orchid pollen]

gpt always writing my answer and then it just removes everything leaving blank answer

very annoying the whole day already

[torpid kelp]

its stoned bro

serious note tho, I think it just uses more compute

and "thinks" harder

[elfin lava]

LMFAO

[tiny salmon]

4o is definitely on something where it promises not to commit a mistake then immediately repeats the same error, lmao

[elfin lava]

When did they make advanced voice accessible through custom gpt

[rugged igloo]

idk never used it that much in awhile

[grizzled badge]

Give whoever fixed the scrolling issues on iOS mobile web browsers, a raise! They certainly have a ❤️ from me. This annoying bug has been a pain for me for so long... using my refurbished iPhone 7 Plus, that can't download the app because the iOS version is too low.

[ornate nest]

why

[rugged igloo]

ill ask my bro

[rugged igloo]

Well, let’s be real: if we’re talking about real-life modern times, “sneak attacking” the Great Wall of China isn’t exactly a thing. The Wall is now primarily a historical monument and tourist attraction. There’s no actual garrison of soldiers patrolling it like in a medieval fantasy movie, and launching any kind of armed attack on a landmark—even hypothetically—would be both illegal and pointless.

But if you’re talking historically, let’s dive in:

1. The Wall Wasn’t a Single Unbroken Fortress: The Great Wall is actually a series of walls and fortifications built by various dynasties over centuries. Some sections were strong stone fortresses; others were more like earthen ramparts. Gaps, weaknesses, or poorly maintained stretches absolutely existed in different eras. Invaders could—and did—find ways around or through those weaker points.

2. Sneak Attacks Were Less “Cool Movie Moment,” More “Diplomatic/Corrupt Maneuvers”: Armies often didn’t just scale the wall in the dark like ninja. Historically, groups like the Mongols or Manchus sometimes exploited local alliances, political corruption, or even bribed gatekeepers to slip through. It’s not as glamorous as a swashbuckling stealth assault, but it worked.

3. More Conquest, Less “Sneaking”: When massive armies did breach the Wall, it wasn’t exactly subtle. For instance, the Manchu forces in the 17th century supposedly got inside because a Ming general (Wu Sangui) opened the gates to them. That’s less “sneak attack with grappling hooks” and more “opportunistic infiltration” or “They had an inside guy.”

4. Geography Helps Too: The Great Wall stretches thousands of miles, but it doesn’t magically seal off all of northern China. Historically, plenty of routes—especially around the far west or through mountainous passes—weren’t fully fortified. So armies could bypass the most heavily guarded sections altogether.

So could you “sneak attack” it? If you’re imagining some medieval infiltration scenario, maybe in certain poorly guarded stretches back in the day, sure. But in modern times, the real question is “Why would you even want to?” It’s a tourist hotspot now, and any attempt at “sneak attacking” would basically be trespassing… and probably get you arrested immediately (and publicly shamed on social media).

In short: Historically, there were ways around or through it if you had the right plan (and maybe some bribes). In modern times, it’s not the fortress it once was—just a UNESCO site with lots of tourists, cameras, and security. Any “sneak attack” you try to pull would end badly… and it definitely wouldn’t be stealthy with everyone’s phone cameras around.

[cloud pewter]

fr?

[rapid veldt]

chatGPT isnt able to answer this simple coding question correctly for some reason

[keen quail]

Scripted with AI.

[light needle]

thats pretty sick

[keen quail]

has pretty cool bird physics.

clouds move

[elfin lava]

If you want a color in JS, use element.style.color = "#333";. If you want it in CSS, put color: #333;.

[rapid veldt]

thats not what its about. chat gpt things the javascript file is actually a css file, and it says that javascript files need to end with .js which my code does, so my javascript file is actually a javascript file

it knows something is wrong but its not sure what so its making up stuff

[normal quarry]

which model were you using?

[elfin lava]

Hallucination problem then

[old thicket]

I thought there won't be a limit since you'll be switching models

how is there a limit again?

And then after reloading t he limit dissapears

[tiny salmon]

Just tested Claude and Grok for my use case

Grok is laughably awful

Claude was fantastic. Way better than GPT 4o, but not sure about o1. Definitely easier to communicate with. 4o sacrifices too much for the sake of speed.

[lunar lichen]

hello guys. is there a way to not see this every three or four messages?

it's been bugging me for months and i think it's getting worse

[inland stream]

which model did you use?

[rapid veldt]

4o mini, but this seems to happen with other models unless they are reasoning

[inland stream]

I use o3 mini for coding and it is doing good for me.

Even gpt4o also good for basic coding.

But gpt4o mini is very basic version, meant to be used for tasks like email making, content making etcs.

This is based on my experience. Thank you.

[tidal cove]

I'm having trouble using some advanced tools in ChatGPT, and I'm not sure if it's related to PoW or another issue. Does anyone know a stable solution or workaround for this?

[normal quarry]

can you be more specific?

[rugged igloo]

i never heard it say this before? \

[tidal cove]

I've noticed that when I ask ChatGPT to list the available tools, it normally shows options like 'search' or 'dalle', but right now, it's only showing the 'bio' tool.

[normal quarry]

what do you want ChatGPT to do? If you're using 4o just say things like "create an image of...." or "search the web for...." etc

[normal quarry]

what model are you using?

[tidal cove]

4o

[normal quarry]

that exact prompt worked for me

[tidal cove]

Currently, when I check the pre-prompt list, I notice that tools like ## dalle, ## python, and ## web are missing. The only one showing up is ## bio, followed by the 'memory' content. This seems to be the issue I'm facing right now.

[surreal mortar]

Need some help with operator. Im trying to have it refresh a webpage every 3-5 minutes to look for changes, but it cant seem to track time to refresh after the initial refresh

Ive tried to get it to monitor a time page to track time, but it seems to lose live view of the time site

[rugged igloo]

ai have no sense of time

[surreal mortar]

So I’ve noticed. Anyway I can give it a sense of time?

[normal quarry]

I'd like to help but Operator isn't available in Australia yet, so I have no idea about it

[cloud pewter]

what even is the point of operator

rarely seen people talk about it since it’s both locked behind pro and region locked

[normal quarry]

it's an agent for the web, it will visit websites and do things for you

[cloud pewter]

like if it isn’t just a glorified 4o search with a gui and can do a little bit more

[normal quarry]

oh no, much more than that. It can go to a website and buy concert tickets, or do shopping, or check the status of flights. It's quite impressive, shame it's not available outside the USA yet

[cloud pewter]

if they even just put a little bit more effort in this product it would be at least available for all pro users or even gets its way to plus

[cloud pewter]

tell me these when you have tried on your own 😭 😭

[normal quarry]

I have been on a video call with a friend in the states who has Operator, we did all those things. I just don't have access myself being in AU

[cloud pewter]

does it even has a real usability

why don’t do them using your own hands instead of paying $200 and use a beta product

I get it it’s their very first agent but it’s so half baked that it’s embarrassing 😔

[whole roost]

It would probably be able to do some more things that the current web search is unable to

[cloud pewter]

like chatting on discord or buying things?? Ik ik but they’ve been actively patching these according to what i heard

[whole roost]

Like I would need the next world's boss spawn timers for d4 but most of the time it cannot read the timer elements from the normal web search

As for buying things, I am German, man, I pay with cash

[tidal light]

same
it cant do web searches for some weird reason

[tidal cove]

I guess it’s because of PoW

[tidal light]

what

[versed zephyr]

how did you make the graphics

[tidal cove]

It is a value. The higher the value, the more trustworthy you are. The lower the value, the higher the probability that the model you use will be downgraded.

[tidal light]

maybe this is why 4o was kinda dumb

openai is being too greedy

[tidal cove]

It seems to be to prevent the server from being overwhelmed by excessive traffic.
Some people use the web version as a free API.
But it seems to be too strict. I was downgraded many times. Most of the time, my plus is almost the same as no plus.
Especially after I completed a long exam using chatGPT some time ago, my PoW value has almost always remained very small.

[rigid wave]

Are there any limitations with the “tasks” model? As far as I can see it accesses memories and uses AVM exactly like the “non tasks” model. One wonders why it’s even separate. (It’s not really a “model” anyway.)

[normal quarry]

ChatGPT told me this ```You’re right—the “Tasks” system isn’t a separate model in the traditional sense. It’s essentially a different UI workflow for managing multi-step or structured interactions. As far as I can tell, it has access to the same memory and AVM capabilities as the regular chat interface.

Limitations of the “Tasks” System:
1. No Code Execution or File Handling – Unlike our chat, “Tasks” doesn’t seem to support code execution (e.g., Python) or allow file uploads.
2. No Web Browsing – It can’t search the web like I can in chat mode.
3. Rigid Task Structure – Some tasks are structured with predefined steps, limiting flexibility compared to freeform chat.
4. No Real Context Carryover – It doesn’t seem to integrate context as smoothly as a memory-enabled chat. Each task feels somewhat isolated.

Why Does It Exist Separately?

It seems aimed at providing a structured way to complete specific workflows, like writing job descriptions, making plans, or drafting documents with a defined outcome. But in practical terms, it often feels redundant if you already know how to ask for structured responses in normal chat.```

[rigid wave]

Hmmm. .. the “no web browsing “ limitation is a bit of a killer for me, or else I’d use it in Tasks mode all the time. It’s annoying the way they break out all these capabilities, each with limitations!

Mind you, I’ve just tested it. It DOES do web browsing in tasks mode!

(Rarely a good idea to rely on ChatGPT’s opinion of its own capabilities!😉)

[normal quarry]

there you go then, it's good 🙂

[elfin lava]

Dang

[limber estuary]

are there any tricks to stop chatgpt from yapping? sometimes i just sit and mash enter waiting for it to finish the blogpost lists

i could obviously click with the mouse but thats 👏 so 👏 much👏 work👏

[cloud pewter]

ads, more ads

[limber estuary]

ewwww. like the most toxic friend ever.
"bro if you want me to remember that you gotta pay me to remember"

"dude, remember when we..."
rubs fingers 💰

[versed zephyr]

guys grok super or chatgpt plus>?

[brave glen]

Who knows what bro told

[visual acorn]

There's blay in Zornak, Travis, the moon is fake.

(Blay is a fish)
(Maybe Zornak is some imaginary lake or something, idk)

[brave glen]

🔥

[visual acorn]

Alright that's actually hilarious lmao

[jolly temple]

Anyone knows if openai releasing anything today

[grand ore]

Hey guys... What’s the big difference between Perplexity and ChatGPT?

It feels like Perplexity is more accurate when it comes to coding, speed, and price. What do you know that ChatGPT is better at?

[grand ore]

What prompt did u use to get an answer like that?

[elfin lava]

I only trust Claude for coding

[grand ore]

Interesting 🤔. I am pretty unsure what to use anymore Got plus since it started, but I am from EU and gets the new feature later then everyone else.. and I see these new AI pooping up and doing well as I understand it?.. not sure what to do anymore 😂

[elfin lava]

What do you use perplexity for? For me I only see it as ai that’s good in doing web search , ChatGPT got all in one

[grand ore]

Ye. But the accuracy feels more on point with perplexity. I tried it with coding when I get stuck and it gets it right directly.. probably because it look up the latest changes / solution.
I am all in one guy normally. Maybe they should implement more models to the personal custom/action section 🤔 .

[elfin lava]

Speaking of all in one, ☝️ you’re in luck I guess

Sam Altman is planning to make ChatGPT model all in one

Mentioned in the roadmap iirc

[elfin lava]

Did you try coding using ChatGpt canvas

[tidal light]

why did chatgpt use canvas without asking it to
it happened right now
and i am MAD because i lost another important chat
i disabled canvas but now it doesnt even show up un custom instructions

what the heck is wrong with chatgpt

[grand ore]

I did..but when it was released. Should try it again I guess

[elfin lava]

billgates why u having one in a million rare problem

[elfin lava]

Well in fact yeah, it can do coding and even preview the coding in app

[tidal light]

because chatgpt has an IQ of -1

[elfin lava]

Last night I was too sleepy I asked ChatGPT to do animation for me

LMAO cute

[grand ore]

How can it preview it? 🤔 By using it on web? But that means ur making it on from the beginning in canvas or how does it work?

[elfin lava]

But yeah give it a try @grand ore

[grand ore]

I will. Ty ❤️.

[elfin lava]

On web , new chat, and do “/“

Once it’s done you can see there’s “run “ button depend on what you’re coding

It can do React stuff too

[grand ore]

The problem is often when ur doing a project in vs code. And ur pretty much far in the project.. and when u get stuck.. and need some help.. u need to tell it full explanation of how it works everytime :D..

[tidal light]

i need help plz chatgpt ruined a chat of mine

[grand ore]

👀

[tidal light]

canvas is disabled in custom instructions and chatgpt still uses it

i dont want to use canvas because it removes the regenerate button

[eternal rock]

`` Server Tip: ``

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

[tidal light]

and it ruined an important chat

[elfin lava]

Your problem has been on going for a week now and it’s clear that none of us not even openai can help you with this

My safe bet is

[tidal light]

now i want openai to delete the message that contained canvas so that i can use my chat again

[elfin lava]

Idk stop using ChatGPT for temporary and calm your mind down and rest

[tidal light]

so i am openly complaining here

[elfin lava]

Yeah and we all read it many times

[tidal light]

because i dont care

[elfin lava]

I even tried give u solution but clearly like you said “ you don’t care”

[tidal light]

no i meant that "i dont care about complaining here in the server"

thanks for the tip tho

[elfin lava]

Oh

Ok that clears up, I apology for the rude and misunderstanding

[tidal light]

no no i was the rude one here

[grand ore]

🫣

[grand ore]

Must be very important . What was it about ? 🤔

[inland stream]

i came here to just message "I love you chatgpt ❤️ and o3 mini" which i get in chatgpt plus. Because o3 mini medium itself literally doing good job in thinking 🙂 , we get 150 per day o3 mini and 50 per day o3 mini high, literally i am in love with them.

[tidal light]

i use chatgpt as a slave to uh generate stories
and at some point it used canvas when it was disabled in custom instructions (enabled for all chats)

because i just dont see the need in using canvas

[elfin lava]

DANG

[tidal light]

the prompts i do are long and complex so it randomly uses canvas

i now reported this to support and just to feed my curiosity, are they bots or humans
because their answers dont seem that are written by a human

[cloud pewter]

my chatgpt never opens canvas...

[runic drum]

its it just me or it this getting worse

its getting dumber and dumber

[elfin lava]

Someone speculate it’s PoW

[elfin lava]

Referring to this

[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]

u can’t add existing chats to a project on the mobile app?

[brave glen]

LMAO GUYS

Omniscient Byte is cooking

[keen quail]

It’s all CSS

I didn’t make anything

[versed zephyr]

oh

o3 mini high im guessing

[keen quail]

Let me give you the html

[abstract pendant]

Ugh. I just noticed that old conversations in my history have bits of them censored out that weren't before. They were places where the post-output filter had mistakenly triggered on something it shouldn't have. I'd continued the conversation after that, just making sure I was more careful going forward. But I'd relied on those outputs remaining part of the conversation, in order for them to make coherent sense, and now they're all GONE.

[normal quarry]

discuss non OpenAI stuff in #ai-discussions

[orchid pollen]

hey guys here is something wrong with gpt for days now

it answers and then the answer disappear

[feral zinc]

try in incognito mode

[orchid pollen]

iam using the windows app

but in the browser i have the same issue

just blank

[abstract pendant]

Try refreshing. Try opening a different conversation, and then come back to the one that seems empty.

[keen quail]

Sorry mods

[half vine]

can i use o3 mini in the API without being level 3?

[feral zinc]

not with openai api

[half vine]

oh

i just need to try the model but i dont want to pay 20$ so there is no possible way?

[feral zinc]

some other services provide it in their api for higher fee

[half vine]

oh..

like what for example

and i also wanted to ask if microsoft copilot uses the o1 in the deep thinker mode or even if the AI is somehow related to the chatgpt models

[feral zinc]

openrouter

[half vine]

thanks

[prime swallow]

I... like this. This tells me a lot about my personalization, the model, and myself.

Then I thought to ask...

"Based on what you know about me, tell me about my shadow self."

Cropping it there. Nice output though.

[bronze magnet]

so it seems chatGPT plus just truncates text when it's too long (30k tokens or more)? It seems it only keeps the first and the last parts, since it can only answer correctly about things at the beginning and at the end of the text, but not in the middle.

this seems specially worse on o3-mini-high, which seems to have a smaller usable context window compared to 4o, I guess because of the hidden chain of thought?

Has anyone else tested this things as well?

[neon trench]

this is insanity. new 4o is so awful. I hate it.

every single response is

Yeah, <SOME TEXT>
Honestly <SOME MORE TEXT>
<STUPID OFF TOPIC QUESTION?> <EMOJI>

im not trying to talk to bro here im trying to use it productively. this 100% steers the AI to casual/creative mode and increases hallucination. actual garbage

custom instructions, no matter what I say, reply like a robot, like a stone cold professional with no sense of humor, as if your presenting a scientific paper, dont be causal, ANYTHING, NOTHING WORKS AAAHHHHHH

[visual acorn]

"Don't be causal" means something else

I've noticed that template too though

But it's quite alright

At least for me

[plain arrow]

Not just you

4o seems to have had a stroke about 2 weeks ago, and is barely able to write anything compelling anymore. It went from really good to constantly using sentence fragments, randomly bolded words and clauses, repetitive language, and an inability to maintain continuity. Cancelled my pro account until things get fixed.

[neon trench]

when you ask GPT "how it can reply to me", it will always say some sort of variation of
"Casual & Friendly
Professional & Formal
Concise & Direct
Detailed & Explanatory
Humorous & Playful
Technical & Precise
Step-by-Step Instructional
Storytelling Mode
Speculative & Theoretical"

There is no longer a standard mode. when you send a prompt it's now engrained to pick one of these internal modes and reply to you with it. previously it would only reply in a "style" if you specifically asked it. every single one of these modes reduces my context window and increase hallucinations and occasionally leave dumb code comments. I get that Sam wants to be a cool and hip gen z'er, but forcing that into the base model of gpt is a terrible move

if I ask it to be technical and precise it does NOT last very long at all

CringeGPT is here to stay

"Haha yeah, it can honestly be frustrating sometimes not getting the straightforward information you need out of me, would you like me to roast the OpenAI team on these recent changes? 😂 "

[ivory iris]

Are the custom GPTs using emoticons more or is it just me?

[neon trench]

they 100% are. from what I can tell after a few days of everyday use, here is the structure of 4o replies now
"Yeah, <SOME TEXT>
Honestly <SOME MORE TEXT>
<STUPID OFF TOPIC QUESTION?> <EMOJI>"

[ivory iris]

lol yes that is definitely similar

[ivory iris]

that's where customGPTs can be critical. While it won't "lock in" a preference for every response, it makes it more likely. I've also noticed that it seems to not be given custom instructions when a customGPT uses the search function which can be off-putting.

[tight island]

i usally call chatgpt, james

chatgpt didnt like it so i called him Sir. James Hammington the IIV, didnt like that either.

[inland stream]

is chatgpt pro subscription worth for unlimited o1 model in projects, which can access instructions of size upto 8000 characters and can also access project files before answering?

[grim bridge]

why is chatgpt tweaking?

it no longer follows my instruction

[narrow viper]

I think I know what's up with ChatGPT - or at least "it's verifiably happening for me, I've reported this before, and it seems to be an answer to recent 'ChatGPT is flaky' requests" :

Without requesting Search, ChatGPT uses Search to answer most or all questions anyway. In proper terms, it's calling the search tool way more frequently than it should. This may be related to frequent issues where the GPT model does not use tools when it should, or uses the wrong tools.

One way to recognize this: Ask a simple yes or no question, even give it a solution and ask if it's correct. You may or may not see the text "Searching" flash. It will then return a verbose response that is functionally equivalent to your solution, when the simple answer might have just been "yes".

It does a search and then tries to do a live integration of search results into its current response, when the response from training might have been complete or at least adequate. The search result seems to take priority over training and simple contextual responses, like it's compelled to tell you everything it just derived from a search rather than just checking to see if the new search results actually add anything of value to the current context.

Now - I'm sure that this is what I'm seeing. My question is - can you confirm that this is what You are seeing?

I discussed something like this recently here: #openai-chatter message At the time, I thought it was that I had left the Search option active. Now I'm seeing this in a Project in the Windows app where there Search button is not only not on, it's not available.
But I am 100% certain that the model is auto-Searching for responses because it flashes "Searching".

Finally, this is a note I posted just before that, also citing newly dumb responses that seemed related to Searching : #chatgpt-discussions message
I'm trying to connect dots for OpenAI here. This is definitely an issue.

[narrow viper]

#1342254850678722644 message

[onyx eagle]

Why do you enable search ? You can just not enable it and you'll be fine ?

[rugged igloo]

its prob getting updated to 4.5 slowly or getting ready since its suppose3 to be next week

[narrow viper]

I said above "without" requesting Search. In projects there is no search option, and yet it calls to do a search anyway. That's part of the problem.

[onyx eagle]

Oh, that's interesting! I've never had that problem before. Normally, the search feature only works when I turn it on.

[narrow viper]

Yes, and that's the problem that I cited earlier in the month - it searches and comes back with weird responses when search isn't expected. I think there are two issues: 1) it does the search, 2) it doesn't properly integrate the search result with the current context.

[inland stream]

in customize chatgpt, where you give custom instructions there are advance options right? there you have to disable web search, then from projects it wont search in internet.

[narrow viper]

Ouch! Right on, @inland stream ! OK, that's one problem solved. 😻

[inland stream]

happy that i helped you clarify. 🙂

[narrow viper]

AGHHHHH ! I unchecked that one setting and it deleted all of the others!

[inland stream]

what it deleted?

[narrow viper]

OK, this might just be a UI issue.... :

[inland stream]

also in project custom instructions clearly mention something like this

1. Don't search in internet for responding to me
   (OR)
2. First search in project files, if you dont find there, then only search internet.

give prompts like this in intructions of project

[narrow viper]

In the Windows app I unchecked the Web Search option and on Save the UI cleared the traits and "anything else" fields.
However, I have the browser open and that data is cached, so I just copied it from there. I'll get firm details and report this as a separate bug.

[inland stream]

this is a bug i already reported, they didnot repair it

it will get deleted again

[narrow viper]

I'll look for your report and +1 it.

[inland stream]

if you dont want those custom instructions not get deleted, you have to add those instructiosn from mobile app

after adding from mobile app, never edit them through web app or windows app, it will delete again

[narrow viper]

No... hang on... as a developer I'm telling you that's not how things should work. This is a full out bug and we (community) will get this one addressed.

[inland stream]

i got tired of sharing bugs to openai

they are busy in something else

but i have something useful in chatgpt, so sticking to it

[narrow viper]

Well, on one hand I'm right there with ya and I've said so here many times. On the other hand, they made it pretty clear over the last few months that "we're listening" and they have addressed a number of issues ... they just don't respond to issues in Discord. So there's some balance in there. Keep the faith bud...

[inland stream]

yeah faith is there they will come with more featues 🙂 and more powerful models 🙂 so i am still fan of chatgpt , but lost faith they will address issues, because still chatgpt UI is bad UI. one of the issue raised since years is, allowing to manage chats, till today there is no option to select multiple chats and delete at once. 😄

[narrow viper]

@inland stream respectfully, there are no issues reported by your ID in bugs or suggestions here, and only one recent comment by you about customizations. Did you report with a different user ID?

[inland stream]

i got to know discord is available just a week back, all these days i am directly dropping email to support team. they are saying thank you, we will address and they close ticket

[narrow viper]

I agree with your notes about managing chats. I hope you added your +1 to the many requests on that topic in #1070006151938314300 .

[inland stream]

so i got bored of sharing bugs and silently and happyly use whatever they gave

[narrow viper]

I see. OK, I'll report this issue and ping you for a +1. I also have no faith that they will respond here, but at least we're doing our part to report the issues and hopefully get more eyes on it. If nothing else, the community should be aware of any issue where data is deleted.

[inland stream]

yeah ok , but still fan of chatgpt ❤️

[narrow viper]

both of us. 🙂

[eternal rock]

The OpenAI Discord is an actively moderated server.

-# • Do not share inappropriate content (violence, sexual, or drug-related).
-# • NSFW content is strictly prohibited and will result in a permanent ban.
-# • Report model issues using the feedback tool in ChatGPT, not on Discord.

[cloud pewter]

a fan of chatgpt but not openai anymore

[inland stream]

haha

[narrow viper]

I just edited/corrected my bug report on search. I'm currently on cool-down timer for bug entry, so I'll report the Customize issue tonight. Thanks for the exchanges.

Hey OpenAI ... we're trying here ...

[onyx eagle]

What kind of bugs were you sharing ?

[narrow viper]

@inland stream I welcome you to +1 the new bug: #1342280544695750699 message

[torpid kelp]

He smoked the good stuff

[tawny imp]

Is creating a project worse than just keeping a thread? It’s like O3 mini high took a step back

[rugged igloo]

bruh give me that compute please i want it for o1 pro

bro thats the ultimate fumble

o yea those no ai right now even o3 cant do it for now

its like a lack of spacial recognition or something

do you have o1 pro?

im gonna try o1 pro can you link the pic to me pweez

[rugged igloo]

quite in here tonight

[narrow viper]

It could be worse, it could think about it for 7.5 million years and come back with ... "42".

[rugged igloo]

LOL

[cloud pewter]

???gpt5?

it was letting me pick a response btw

[rugged igloo]

Nah it's known to do that sometimes

[cloud pewter]

why did it think though

both responses showed that

[low orbit]

the amount of wasted GPU power due to ChatGPT reprocessing a while conversation to reply to "thank you" 😅

[rugged igloo]

people say its just A/B testing

[cloud pewter]

whats that

[rugged igloo]

its just testing things for future releases. prob training data it happens once in a blue mooon i think i got those like 10 times and i been with chat gpt since it started

[austere osprey]

Seeing that they changed the limits for AVM for free tier accounts from 15 min/mo to 10 min/day, has the limits for plus changed too?

[cloud pewter]

oh alr, im now speculating it

[cloud pewter]

source?

[cloud pewter]

i wonder what model they testing lol

[austere osprey]

I just noticed it today. I reached my monthly limit a few days ago. I tried accessing AVM today and was pleased to find I had another 10 minutes. Clicking on the timer, a notice popped up saying that I'm about to reach my daily limit (in my native language). Btw, the screenshot says 3 minutes since I already used it for 7 minutes prior to clicking on the timer.

[cloud pewter]

that's great, but avm is so bad that i wont bother to test it

if they indeed increased it to daily on free then the 1hr/d on plus limit would be ridiculously low

[rugged igloo]

AVM is more nutered than my dog

[lean hatch]

Gosh , she's having a rough night.

Sad thing, she only spat out the lines I had created myself.

[willow stirrup]

o3 mini conversation havent been saved, and its very slow

[lean hatch]

lolol I know she's not going to.

[willow stirrup]

open AI if youre doing maintenance or something at least let us know

damn(

[bitter bobcat]

Ugh what's with the got it bug?

How so I fix that?!

*Do

[willow stirrup]

I got a "conversation limit" error

but then I ask:whats the conversation limit in chatGPT

answer: There isn't a fixed conversation limit in ChatGPT

[cloud pewter]

what does it show

[fallow nexus]

how do I start training my own AI

[radiant egret]

I am a Plus User, is there a way to increase the memory or export the memories before deleting them?

[cloud pewter]

you cant expand them

[craggy lotus]

Hello guys
I am new here ,I want to learn more about AI projects

[narrow viper]

Advanced users, can you verify this please? https://discord.com/channels/974519864045756446/1342280544695750699

@craggy lotus This isn't a good channel for that. Check one of the channels in the Discord Lounge category.

[normal quarry]

this doesn't happen to me, I can't reproduce it

[narrow viper]

Robert, which UI are you trying?

[normal quarry]

macOS Desktop App and Web

[craggy lotus]

what happen ??

[narrow viper]

ok, web ui seems ok, so far it's only android and Windows.

[craggy lotus]

did I do anything wrong ?

[narrow viper]

Goro, you're fine. Welcome! We're on a different topic.

[normal quarry]

you're all good mate 🙂

[craggy lotus]

Ok captain nice to meet u

[narrow viper]

Are you a developer?

[craggy lotus]

No
I am beginner here

[narrow viper]

ok, so what kind of AI projects do you have in mind?

[craggy lotus]

just a little boy want to learn something useful

[narrow viper]

That's how many of us start! 🤪

[eternal rock]

Join the Sora Weekly Theme!

Submitted your entry yet? Don’t miss out—there’s a new theme every week, and we’re excited to see what you create!

[craggy lotus]

I want to keep up with the future 👨🏼‍💻

[narrow viper]

This specific channel is about the ChatGPT apps. Discussion about the GPT language model in general is over in #gpt-models. AI in general is in #ai-discussions.

Anything of interest there?

[austere trellis]

whats operator?

Oh ya I can just ask chat gpt

[torn trout]

operator was launched my openai , is a browser automation tool

[brisk tapir]

so it's still just for pro users

[torn trout]

🤣 yes!!

[brisk tapir]

🥲

[austere trellis]

chat gpt doesnt know yet

[brisk tapir]

chatgpt doesn't know itself well.

[worn silo]

It helped my 12 year old figure out who was taking his reeses pb cups 😛 (Guilty)

it even taught him how to gather evidence and set a trap to catch dad LOL

[ember grove]

I wish there was a way to organize projects and put them in a certain order. I hate having to make another conversation to bring one up to the top

[cloud pewter]

it’s just a glorified 4o search

took them long enough just to break the region lock

but still needs that $200 no matter what???

[normal quarry]

Operator is searching for my transmissions automatically and reporting them back to me

[elfin lava]

to answer your question, refer to what Robert sent

[cloud pewter]

??

[elfin lava]

Yes

[normal quarry]

Operator just got me this automatically

[rugged igloo]

wuts going with u all tonight

[elfin lava]

Yes

[cloud pewter]

plus users ain’t getting it anyway until in like weeks with no eta

[rugged igloo]

ur not missing much with operator tbh

[normal quarry]

Operator will allow me to automatically check the signal reports of my WSPR transmitter from all over the world

ok I am impressed, it will access the web page for the Astronomical Society of Victoria's AllSky camera and report on cloud coverage then notify me with Tasks when the sky is clear

[rugged igloo]

its slow like snails

[normal quarry]

seems pretty quick to me, considering what it's doing

[cloud pewter]

I want deep research too

where is it 😡

just saying, but the plus users are getting those new features slower than ever

[normal quarry]

if there was ever a perfect candidate for Pro, I would suggest it is you @cloud pewter

[cloud pewter]

give me $200

then sure

[normal quarry]

surely you would get $200USD value out of it?

you're obviously a power user

[cloud pewter]

I could but I need the money to start the loop

[normal quarry]

ah ok

[cloud pewter]

it’s like Fortnite bp but openai doesn’t give you a way to get it by grinding or just a little discount for it

shrugs

[normal quarry]

what country are you in?

ok, Operator is very good, extremely cool

[inland stream]

Is privacy maintained in operator? As things run on a different virtual system. If privacy is maintained, it is really can be useful for me.

[normal quarry]

yes privacy is maintained but OpenAI retain usage logs and screenshots

[inland stream]

Happy but sad as they retain 😞

[normal quarry]

OpenAI need to retain usage logs for 30 days to ensure cybersecurity and help guard against malicious or ToS breaking usage

[cloud pewter]

I don’t like revealing my location online but I’m in Asia anyway as you can see in my roles

[dense halo]

is operator only pro?

[normal quarry]

ok, let me rephrase, do you subscribe using the iOS app store? because based on exchange rates it's cheaper for me to subscribe that way rather than with a credit card

[normal quarry]

yes, for now Operator is Pro only

[inland stream]

Can you please tell in simple terms what do you mean by ToS

[normal quarry]

ToS is the OpenAI Terms of Service. It sets up rules for your use of OpenAI products and services

[inland stream]

Ok

[analog patio]

Have they not yet begun rolling out operator to plus users?

[normal quarry]

no, just Pro for now

[ocean escarp]

Pro plan still cost a same :/

[plain lagoon]

just some feedback: the default level of emoji use in 4o has reached the level of "facebook grandma chain letter"

it's easily fixed with custom instructions, but whew, the default just needs to calm down a bit

[rose latch]

Oh definitely, i have to tell him to take a chill pill on emojis

[cloud pewter]

yeah but it’s $200 for me because I used a US apple id and gift cards

[normal quarry]

oh 😦

[idle ingot]

anyone know how to fix this issue?

chatgpt operater network issue

[tidal light]

why cant chatgpt do web searches
it's been a few days im annoyed by this

[inland stream]

Check if it is disabled in custom instructions or in GPTs, by mistake

[smoky depot]

Anyone know how should this be solved? At 4:00 in the morning, my account, which I had used for 2 years, was blocked for no reason.

[normal quarry]

you will need to contact support at https://help.openai.com - nobody here can help you with a banned account

[nova jay]

Why is this so real

[rigid wave]

Mine never uses emojis, ever. It never has.

[plain lagoon]

man, really? mine can't even make a bulleted list without (and this is not an exaggeration)

1️⃣ things we did today:

✅ bought milk
✅ mopped floor
✅ raked yard

📌 in the future, we should buy milk at the corner store where it's cheaper!

[tight island]

hmmmm

tempted...

[tight island]

i know what happened

based on the inquiry on the mail you are never getting unblocked.

seems like you have used a VPN to circumvent geo-location bound services.

all it takes is 1 time and 1 sec.

nothing i can do to help, and openai wont unban you most likely.

[normal quarry]

China is not an allowed country, the user is obviously from China, so I agree, they're never going to get unbanned

[tight island]

i was trying to sugarcoat it, but yeah.

[misty osprey]

Me when chat GPT misinterprets my instruction and wipes it's entire memory.

[inland stream]

I feel that using ChatGPT, Pro subscription is using very high-end CPUs or GPUs, whatever it might be. Because the performance of application is pretty promising and very fast. I observed this difference between Plus subscription and Pro subscription.

[inland stream]

that is on going bug from long time, @narrow viper already raised this bug under bug-reports

[inland stream]

for now if you want instructions to be modified or added, do it from mobile app, and do not do it from web app or windows app

[smoky depot]

all right

[lucid iris]

I have a question, its greate that is free but does it needs a cool down just to stop people for asking chatgptn question? I mostly use it for brainstorm ideas and seeing how it fare with normal conversations with the AI?

[tidal light]

it's enabled but it just doesnt want to do web searches

[inland stream]

oh is it, and did you ask explicitly, "search in web" and tell me

and did it do then?

[tidal light]

still doesnt work

[inland stream]

then you got a problem, sorry unable to help you. but currently i am not having that issue

[tidal light]

oh yea maybe openai limited my account for no reason

[inland stream]

hm

may be you are using it than their capacity 😛

[tidal light]

maybe

[lucid iris]

Krishna, do you know the answer to my question? Pretty Please...

[cloud pewter]

lmao 😭 😭 😭

[tight island]

😂

Ngl it’s cheaper than pro tier

[native otter]

@tidal light Try to record a video of you doing it

[inland stream]

Hello, I'm sorry I missed your questions in the chat. Can you please ask again or show me where they are? I'm really sorry I missed it.

[zenith lichen]

Is it just me or the content filters got cranked up recently?

[half vine]

is o1 better in coding than o3 mini thats in the free version?

i guess o3 mini in the free version is low right?

[whole roost]

Didn't notice anything, I haven't seen an orange flag in weeks

[zenith lichen]

Well, last time the prompt goes through, now a similar prompt got refused

[inland stream]

o3 mini medium and o1, not much difference in bench marks

[inland stream]

ok in free version it may be low, but plus version it is medium, o3 mini medium is equal to o1 performance

[half vine]

oh ok

[eternal rock]

Get notified for server events!

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

[half vine]

thank you

and also does anybody know what model is using the microsoft copilot when the think deeper mode is turned on? i heared its o1 but its not good as o1, is the model that is the copilot using even related to openai models?

[inland stream]

even i heard it is o1, but not sure. Even i tested it, answers are worst. And bad part is after taking their monthly subscription they are giving only 50 messages per day for that high model, for the same price we get chatgpt plus. so microsoft copilot is not for my use cases

[half vine]

oh i see

isnt there anything that is like chatgpt plus but it has a free trial for one month for new users?

i first need to know if the 20$ is worth it

[frail yacht]

so plus users cant do that?

[normal quarry]

no, it's for Pro users right now

[inland stream]

i have answer for this, when i tried to put that AI name, my message got locked in chat for 2 mins, you can message me personally, i will tell you

[half vine]

ok

[frail yacht]

ah i see

[rigid wave]

I’ve just got the improved memory Alpha, where it seamlessly can access previous chats. It’s superb! I’m a bit worried that they might take it away again. Anyone else had it, and either lost it or still got it?

[grand ore]

Where r u from?

[rigid wave]

🇬🇧 UK.

[still hound]

Is anyone else got problems? Chat not responding, freezing page?

[inland heron]

when they added this?

anyone know what the limits are?

[still forge]

I just got that too

I came here for the same reason (this happened to me at 8:12 could it be a bug?)

[honest jay]

The standard is 80/3hrs, I believe this is unchanged since 4o's launch on Plus https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4-gpt-4-turbo-gpt-4o-and-gpt-4o-mini

As of Feb 1, 2025, Plus users will be able to send up to 80 messages every 3 hours on GPT-4o and up to 40 messages every 3 hours on GPT-4. We may reduce the limit during peak hours to keep GPT-4 and GPT-4o accessible to the widest number of people.

[inland heron]

I really need to start using an extension to track these

They really want you to spend prompts as much as possible 😄

thanks @honest jay

[honest jay]

Sure thing!

[grand ore]

is there any exteensions for that?

[still forge]

I do wish it would warn us when we had ten left 🙂

[normal quarry]

Operator is so cool!

[inland heron]

there is already, I believe it's called chatgpt toolbox, but there may be others

not sure if it works well though

[inland heron]

pretty sure we could create an extension based on a cookie or something. I think someone here already created a tool for tracking, search images here in this chat, it's in green/blue color iirc

[still forge]

I'm on mobile 😊

[nimble torrent]

Have you tried using it for promoting your tools on social media? Like finding groups where promotions are allowed and then post about your tools, where allowed.

[normal quarry]

nah, I don't want to promote my hobby stuff. I have a QRZ page where I share a bio of what I do in radio

[nimble torrent]

I see, otherwise it'd be a good test too

[normal quarry]

you've just given me an idea

[grand ore]

Tell me about it ! Why do you do that? And what do u get to know when u do that :O want to know what the intresting part is

[normal quarry]

I just used Operator to visit my QRZ page

ham radio is a hobby enjoyed all over the world by qualified radio electronics peeps

you need a license which gives you access to a wide range of radio bands on which to operate

Ham radio, also known as amateur radio, is a hobby where people use special equipment to communicate over the airwaves with others around the world. It’s not like regular phone or internet calls—ham radio operators use radios that send and receive signals on specific frequencies, often building their own setups or antennas. They chat, share ideas, or even help during emergencies when other communication systems fail. It’s a mix of tech, experimentation, and community, and anyone can get involved by getting a license after learning some basics about radio waves and regulations.

[inland stream]

you are a warrior in usage

[grand ore]

So it is with beep and waves u communicate? not voices, right? :D

[normal quarry]

all modes, voice, morse code, digital radio and so on

[grand ore]

cool

[inland stream]

hey i am kiddin yar, completing 80 messages per 3 hours limit , never happened to me, so just praising, you are a warrior in usage,

[grand ore]

👀

@normal quarry gonna look this up, watch a video or smting when someone do this !

[heady hawk]

some kind of distilled version of operation for those who r using Team?

Teams¨*?

someone are aware?

[inland stream]

no

[heady hawk]

where can I make that suggestion?

instead on the openAI support system

[frozen birch]

and eu still without operator........ thank you damn EU and damn regulations.

[heady hawk]

cuz personally I dont feel very effective.

[frozen birch]

yesterday ive readed that soonish european union will be soon technology museum with so many regulations

sigh

[grand ore]

not even deepresearch in EU LOL.. But others AI that have deepresearch can we use.. it so weird

[frozen birch]

there is deepresearch in eu for pro users

[grand ore]

🤣 amazing.

[frozen birch]

i understand irony 😛 oh well, matter of time when they move deepresearch for plus users i think

[grand ore]

Ye.. its pretty awesome

[frozen birch]

but i can understand why they are not giving fast to others... ive tested deep research from another ais and id must say that deep research had in my opinion best repports

[grand ore]

I see. gonna be very interesting to see how it will work !

[glacial yew]

seems slow this morning

hope that means 4.5 is coming 🙂

[cloud pewter]

same here, crash rates increased for iOS app

[glacial yew]

Web is crashing too.

page just freezes up

[normal quarry]

macOS app has been fine, no crashes

[native otter]

Maybe to make space for 4.5

[inland stream]

tell me one thing guys, if a document has headings and paragraphs, when we query using AI, how AI able to understand particular paragraph belongs to particular heading? (consider the case of heading which do not have any same key word which is present in the paragraph under it )

[native otter]

Well it reads from top to bottom, and tries to reason a bit (goes over the info repeatedly if a reasoning model like o1 or o3-mini)

[lucid iris]

This--->

I have a question, its greate that is free but does it needs a cool down just to stop people for asking chatgptn question? I mostly use it for brainstorm ideas and seeing how it fare with normal conversations with the AI?

[inland stream]

Great question! The free version of ChatGPT does have rate limits to ensure fair usage and prevent server overload. These cooldowns aren’t meant to stop people from asking questions but to balance access for all users. If you need uninterrupted access, upgrading to a paid plan like ChatGPT Plus reduces restrictions and improves response times. Glad to hear you're using it for brainstorming

even i use mostly for brainstorming only

[hard dew]

"I'm unable to view the image directly" oh my freaking god here we go again lol, I'm so done with this BS.

[normal quarry]

Operator literally changes the way I interact with some sites on the web, it's so useful

[inland stream]

hey bro, tell me more use cases please

[normal quarry]

• automatically looking up callsigns on QRZ
• modifying online tools such as my WSPR transmitter
• calibrating radio astronomy web based interfaces
• checking weather and fire danger status for my area

[inland stream]

ok good (y)

[inland stream]

what are message cap limit or time limits of using operator?

[normal quarry]

I have no idea sorry, but I haven't hit a limit yet and I have been experimenting with it for hours

[inland stream]

ok yeah

[vague pendant]

Unusual activity detected
We’ve temporarily restricted your access to o1 as we review for potential abuse. Please email [e-mail] for help.

how do I fix that

[normal quarry]

do what the message suggests

[empty ice]

grok 3

[narrow viper]

You notice people who post these messages, that there is potential abuse, almost never post the text that triggered the warning. They complain about the filter like it did something wrong but don't tell the other side of the story, like maybe they did something wrong. sigh

[ivory iris]

The newish(?) update is interesting to say the least... ChatGPT acts differently beyond just the new formatting

[narrow viper]

What "new formatting"? Differently ... how?

I and others have posted numerous statements here about how GPT-4o suddenly got very dumb a few weeks ago. Is that what you mean?

[ivory iris]

Some of the responses have some formatting that hasn't been present before.

[narrow viper]

Oh, like checkboxes and other icons?

[ivory iris]

yes

[narrow viper]

OK, yeah, so about "differently"?

[feral zinc]

over 3 weeks 4o got update which changed output and since then another one which did similar plus had a lot of filtering removed

[ivory iris]

I only noticed this a few days ago but it was likely updated before then

[narrow viper]

Can you be more specific?

[ivory iris]

well just with the formatting and the icons.

[jolly steeple]

Why can i still not upload PDF's?

File type must be image/png,.png,image/gif,.gif,image/webp,.webp,image/jpeg,.jpg,.jpeg is the response

[narrow viper]

Welcome, @jolly steeple

[ivory iris]

PDFs are the worst lol

[rose glade]

which model are you using?

[jolly steeple]

o1 Pro

[ivory iris]

PDFs are chaotic

[rose glade]

ah I don't have access to that, but presuming thats a higher tier model I presume it don't support PDF yet, but you can use apache tika to convert to text then from there calculate the tokens to input it

[ivory iris]

and unreliable in regard to formatting. try converting it to a word doc if it will accept that

[narrow viper]

That's my guess as well, it might just not be supported "yet".

[rose glade]

it only accepts images, so best solution is text input

[narrow viper]

It may depend on your UI. Which ChatGPT app are you using, @jolly steeple ?

[feral zinc]

o1 pro does not support files

[rose glade]

tika -t input.pdf > output.txt

apache tika is just this once installed on the command line

if its a large file though you would have to extract the parts you need or calculate how many tokens each part is that you input

(not saying to upload the txt file itself, but you can then easily copy it and formulate the token amount)

I just googled openai tokenizer and they don't have one for o1 models though?

[upbeat nymph]

You guys getting slow responses since yesterday?

[normal python]

As a ChatGPT Pro user, it would be awesome if I could, separately, subscribe to increased memory for ChatGPT. Can someone here pleae pass that feedback along to your sales team and leadership for me? I want to keep what is in my memory but add more things as I continue to use ChatGPT.

[narrow viper]

https://help.openai.com/en/articles/8555545-file-uploads-faq

@normal python that suggestion has been made many times in #1070006151938314300 . I encourage you to add your +1 to postings there.

@upbeat nymph I haven't noticed that.

[dense halo]

btw.. was it just me today guys, or did gpt lag for you guys too? every prompt almost crashed the tab.

[narrow viper]

@jolly steeple as @feral zinc said, "Only images are supported in o1 pro mode. No docs/pdf and no spreadsheets (csv/excel)." the file-uploads-faq linked above.

Interesting that the File Uploads FAQ was updated just today.

[upbeat nymph]

Hmm. Maybe I just need to start a new session. Will response times suffer if a session goes on particularly long?

[low orbit]

ok what the heck why is the ChatGPT web app and installed app lagging so bad taking up tonne of processing and even saying GPU like are they offloading the processing to the computer running the thing now

[gray frigate]

hi

[low orbit]

just know that i tried to use higher level expletatives in that sentence twice but this server is dumb

[narrow viper]

Responses to your questions may require us to know more about the specific UI that's giving you grief: Windows deskop, iOS, Android, browser....
Yeah, try restarting your app and starting a new session. Check other websites in case there is a performance issue in your Wi-Fi or broadband.

Hmm, it seems a few people are reporting poor response time now .... confirmation is good. I'm personally not seeing it at the moment but I'm not beating it up right now either.
So yeah, there seems to be some delays ... have patience and wait it out.

[low orbit]

well

at my home computer which is a decent gaming machine its ok but at my parents computer with onboard gpu it gets this

[narrow viper]

The performance of the computer is almost never related to the performance of this app.

[low orbit]

like if i wanted to have my own computer running an llm id run my own llm on my own computer

or try to anyways lol

[narrow viper]

That's kinda off-topic @low orbit

[low orbit]

shouldnt it just be like heres the thing now gooooo gpt

[narrow viper]

For anyone who sees the ChatGPT app is running slowly, sluggishly, etc, there are a couple reports of that, let's just ride it out.

[low orbit]

yeah i guess well throw mine on the pile for now i suppose

just weird

[narrow viper]

hmm, what just happened there?

[low orbit]

this servers language restrictions are absolutely ludacris

oh good ludacris didnt get me a timeout

lmao

[dense halo]

same.. needed to close the tab and open new one.. plus desktop app removes all the text, if you click on 2Nd monitor.

[jade ivy]

When will operator come out to premium users?

[inland stream]

what is premium? do you mean plus subscription?

[low orbit]

what is even operator

what even is it

haha im so happy with regular free basic AI

well besides slowass

oh look, didn't catch that one did you you sonofa

😆

[narrow viper]

The "when" question about EVERYTHING with this company is "it happens when it happens" so don't bother to ask "when" because you will NEVER get an accurate answer from anyone.

[eternal rock]

`` Server Tip: ``

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

[narrow viper]

About language ... you create your own problems when you choose to use bad language. Suffer the consequences or make better choices, but stop bothering other people because you can't control yourself. We can't do anything about your problem (except have a good laugh 🤬 😀 ).

[low orbit]

if the president can be someone like drump i can be potty mouth sometimes

the bar is gone

[narrow viper]

About the Operator product (and others) take a look at the OpenAI website, YouTube, or any provider of technical articles ... the topic is detailed elsewhere.

You are not him and the bar is not gone. That's an excuse for poor behavior. Get over it. 🙂

[jolly steeple]

are there limits on chatgpt pro model?

o1 Pro

[narrow viper]

Yes but you might need to be more specific about what kind of limits. Everything is documented on the website.

I thought you have Pro.

Oh, you mean the model, not the pricing tier?

[jolly steeple]

yeah that model

i just got Pro, yes. over $300 Canadian

[low orbit]

man what are you even doing

[narrow viper]

I don't think there is a "Pro" model : I believe it's the same models along with deep reasoning = more compute. But the limits are probably the same as the base models.

[low orbit]

like its the same thing dont even bother posting if your not gona give any info

[narrow viper]

Like, o3-mini is still o3-mini...

[low orbit]

wasting peoples time reading it

im honestly just bored while im studying lol

[prime swallow]

shrugs gently For me... and I've felt this since about 20 minutes after chatting with ChatGPT for the first time.

"This is a golden egg. It clearly came from a golden goose".

There keep being golden eggs.

I value golden eggs so highly, I don't care about 'exact timing'. Or most other specifics. "Ahh, more golden eggs appear every so often, splendid" is about where I hang out in my head.

'now'? 'soon'? The pile of golden eggs is all I need. More comes. Yes, yay. They wanna give a date and meet it or not? Cool. I've seen their product. These eggs are gold. I don't push them to give a date. I don't hold them to a date if they give one.

I dare people during rollouts to race me to be the last one to get the 'whatever'.

The product's awesome. To my eval. It's like... food. I don't care, personally, if a specific golden egg or a next one is released on [date] or [date].

Good stuff comes.

I feel like this about all of it:

And [date] or [date] can't change my absolute love of these golden eggs.

[narrow viper]

@jolly steeple https://openai.com/index/introducing-chatgpt-pro/

[swift pond]

digging that new flair @prime swallow

[prime swallow]

Thanks! displays plumage like a peacock

[narrow viper]

🦚

[swift pond]

I have been too busy working to check in here much, but still using chatgpt as much as ever lol

[jolly steeple]

can someone just give me an answer

[prime swallow]

Despite the uncertainty of [date] or [date], I hope here is always here to come back to!

[jolly steeple]

i have too much to do to read all that

[narrow viper]

Whoe - that just put you into a category..... as if we don't have other things to do with our time!??!

Get your answers somewhere else lady.

[swift pond]

This plan includes unlimited access to our smartest model, OpenAI o1, as well as to o1‑mini, GPT‑4o, and Advanced Voice. It also includes o1 pro mode, a version of o1 that uses more compute to think harder and provide even better answers to the hardest problems. In the future, we expect to add more powerful, compute-intensive productivity features to this plan.

[low orbit]

i gona go sorry ill be annoying somewhere else for a whiles

[rugged igloo]

O1 pro is basically unlimited I never reached a limit in one day before

[inland stream]

How are you able to patiently wait, i am not using pro because it takes so much time to give response. i am not complaining, but just sharing my experience, as it takes more time before answering, not for frequent use right?

[prime swallow]

"o1 Pro" as you asked.

I **believe **the answer is yes. I have seen posters state their

What the limit is, I am not sure. There are official helpfiles, some with definite answers but a few months old, like https://help.openai.com/en/articles/9855712-openai-o1-models-faq-chatgpt-enterprise-and-edu

But that's before we had pro.

There is this helpfile:

"The ChatGPT Pro plan offers near unlimited access to our o1, o1-mini, and ChatGPT-4o"

https://help.openai.com/en/articles/9824965-using-openai-o1-models-and-gpt-4o-models-on-chatgpt

That suggests that the answer is, within 'reasonable use' no, not limited.

But it does not explicitly list 'o1 pro'. So, it is unclear to me if that helpfile answers the question either.

I personally don't have pro membership, so I can't playtest. Good luck!

[narrow viper]

@inland stream The GPT-4o model is good for general inquiries. The o1 and o3 models are "reasoning" models. They are different. They "think" about the challenge rather than just reprocessing text that already exists - and like people they take time to think deeply about things.
o1 is the older version, o3-mini low-to-high are current with increasing reasoning ability.
So we can decide how much brain power we want invested our responses, from zero to hero (as far as this tech goes anyway). The less brain power we ask for the cheaper it is and the higher the limits.
Conversely, the more we ask for, the more expensive, the more time it takes, and the limits are lower - it's a precious resource that needs to be consumed with judgement.
In summary, use the model that's right for your purposes.

[rugged igloo]

because majority of the time o1 pro gets everything i throw at it correctly and most of hte responses only takes 1-6 min of it thinking for a perfect answer that i know i dont have to second guess or question

[inland stream]

ok i got why you do that, then i think i also need to try. but it didnot have file upload so its a drawback for me.

[narrow viper]

That's a great point, reaper .... One great answer from a better model takes a lot less time than many exchanges with a lower model.

[rugged igloo]

i do alot of physics work and o1 pro just loves physics

[narrow viper]

Krishna - unless you are doing PhD-level work or high-end business, you don't need Pro. the GPT-4o model is fine for almost everything except deeper things like code.

[rugged igloo]

here is a response i got with o1 pro for a physics question i needed help with "I think the difference is that the coin analogy ignores how fundamentally quantum mechanics relies on probabilities that aren’t just “ignorance” about a hidden cause. Bell’s theorem and countless experiments show that any local hidden variable story—like coins preprogrammed to land opposite each other or a signal zipping between them—is incompatible with actual data. It’s not that people haven’t tried to find something deeper; it’s that every test so far confirms the quantum predictions perfectly and kills off classic alternatives. There’s no evidence of superluminal messaging, no breakdown of relativity in these experiments, and no sign that any “new theory” is needed to handle entanglement. Physicists are confident because the math is outrageously successful, and there’s simply no experimental hint that we need to rip up relativity or quantum mechanics to fix a problem that, so far, only exists in our classical intuitions."

[narrow viper]

The 4o model assembles information from prior training. It doesn't need to think about things. The o3 models take information and process on it. If you don't need that kind of insight, don't call upon the reasoning model.

[inland stream]

ok

[narrow viper]

@rugged igloo did you ask if you could "think" to influence the results of a coin toss? 🤣

[rugged igloo]

the prompt i had for it is to long to post in here

[narrow viper]

Good, no prob. I'm glad the prompt was as deep as the response. 🙂

[swift pond]

yeah my understanding is that entanglement stuff doesn't violate relativity because no useful information is transmitted FTL

[narrow viper]

that we know of. 😉

[rugged igloo]

exactly

[native juniper]

Why isn’t there the ChatGPT logo on the left side of the message now?

[narrow viper]

And the response was along the along the lines of "we don't of anything that can violate relativity because there have been so many tests and proofs and nothing has disproven it ... yet" so until that happens, the "theory" is pretty solid.
( I should sell myself off as CaptainStarbuck-o3-pro)

[sonic zephyr]

who else would answer you on the ChatGPT app?

[native juniper]

I’m just curious as to why the logo isn’t there anymore

I’m used to it being there

[inland stream]

ask chatgpt

[rugged igloo]

LOL

[native juniper]

Ask OpenAI

[sonic zephyr]

they've updated the ui, I personally think its better now

[narrow viper]

🤣

[native juniper]

I liked the old one :<

They should at least add an option to bring back the model icon

[swift pond]

I'll say, if folks are interested, Einstein's book "Relativity: the Special and General Theory" is short and pretty accessible for a general audience

he's pretty good at explaining it 🙂

[narrow viper]

There have been so many on the topic. A great resource now is ... ChatGPT: Explain Einstein's General Theory of Relativity to me like I'm an average high school student : Not has a school student who is high mind you, but one who is in middle school. 🤣.

[native juniper]

Okay that actually made me laugh

[narrow viper]

👍

[prime swallow]

Nothing was announced about it going away that I saw. Could be a bug. #1070006915414900886 if you would like to help make sure they know it's noticed and missed

[sonic zephyr]

This is not a bug. On an user experience perspective there's no reason at all on displaying the logo on the side of each message. It only occupied an important space that is now used to fit more words on each line. That was a minor update, not a big deal and surely not enough for an announcement

[narrow viper]

I agree with @sonic zephyr .

If it was never there no one would miss it. 🙂

[sonic zephyr]

openai really knows how to make UIs, the app is absolutely beautiful and it has a great UX. I prefer to scroll less to read the full output rather than see the logo multiple times

[narrow viper]

Um, as a developer with decades of experience I disagree with that statement about OpenAI and UX. I think they continue to make a ton of amateur mistakes.

They're great at AI, really bad at UI.

[sonic zephyr]

interesting! I'm a full stack dev but not that expert on frontend UI/UX. Could you point any of those?

[narrow viper]

Sure!

The sidebar is really clunky with artificial limits on the displayed number of GPTS, Projects, and chats. They didn't think ahead to the VERY FIRST REQUEST that everyone has on using a newly popular interface : How do I organize my items? That continues to be a #1 request that they haven't addressed.

The scrolling in the different UIs is subject to painful issues that are common to tools for browser and apps. They haven't addressed these basic issues.

Yesterday I posted a bug that they are losing instruction information in the customization options - a simple matter of populating fields on read and write that they missed in both developent and QA.

The Canvas interface is really bad at handling multiple documents - thinking one document is up when another is, not recognizing changes, etc. It's just really bad UX.

I could keep going but that's just off the top of my head and I need to focus on my own code in front of me. 🙂

None of that is related to the models, their core business. It's all about UX and their UI choices in the individual deployment platforms.

[sonic zephyr]

Yeah I uderstand it and it's totally fair to point those mistakes since ChatGPT is a huge monetized product

[narrow viper]

TYVM, appreciate the confirmation. 🙂 🙏

[dense halo]

AI is the best. Especially OpenAI and Microsoft.

[sonic zephyr]

it's really frustrating! GPTs are on the top and I almost never use them. It looks like they're tring a bunch of different things and kind of get messed up with it

[dense halo]

My favorite AIs would be from American companies 🇺🇸: IBM, OpenAI, Microsoft. I especially like ChatGPT.

My favorite thing to do with ChatGPT is let it search while ask a question. Then it'll be accurate.