chatgpt-discussions

[warm garnet]

a pc desktop app

for chatgpt

[eager plinth]

There is a Codex iOS app. For coding tasks. Haha

[trim rock]

If you want an app for your mobile phone then it's called ChatGPT. All the models are included, new 5.5 too

[dense scroll]

So should I sub for chat gpt plus?

[warm garnet]

what u want to use it for?

[dense scroll]

Or just go for Go?

[warm garnet]

it depends on ur usage

[trim rock]

"Plus" is not an app it's abonement

[grave kelp]

u wanna do the simulation of the solar system in 3d on canvas for example?

[dense scroll]

Asking questions basically and I'm testing it cos I need to get some content creation gears and it generated one image for me and I need to wait till 4 hours later before I can generate scenario images again...

[warm garnet]

Sure, do u have a prepared prompts?

[warm garnet]

then plus is good for you for daily use.

[grave kelp]

i just say " make a complex 3d simulation of the solar system" let's see if it understands what i mean

[warm garnet]

the usage is more than enough for your use case.

[warm garnet]

alr

[dense scroll]

Plus it's the most expensive! Ok comparison between Go and Plus. Don't really understand the major differences.

[acoustic talon]

Does anyone have access to 5.5 yet?

[dense scroll]

Plus had early access to new features and agents and deep search. What's agents and deep search?

[warm garnet]

plus give you early access + 50% more usage or prob even 100%

[warm garnet]

deepsearch, bassically deepsearches in the web for information

[dense scroll]

I've been holding back unlike my dad who is having fun like every single day relying on chat gpt. I swear he asked chat gpt what to eat each day. I'm not crazy heavy user like him

[grave kelp]

how is it going?

[dense scroll]

I'll get plus for a month and see how it goes. Thanks for explaining @warm garnet

[untold plank]

Not available to go subscription 😢

[warm garnet]

still doing it

[dense scroll]

Is open ai linked to my chat gpt for API key?

[grave kelp]

oke

[dawn galleon]

Where is GPT 5.5？ I can’t find in my website

[trim rose]

Still rolling out I guess

[nova fable]

Rolling out = not everyone gets it at the same time

[loud patio]

"Rolling out ... starting today" - may take a day or two.

[dawn galleon]

Ok thanks 👌

[grave kelp]

unlike google that to release the notebook feature takes 500 days

[trim rose]

btw, does anyone know why "Computer Use" is unavailable in my codex?

[nova fable]

They might be doing backend work implementing 5.5 into things?

[modest ginkgo]

When is 5.5 rolling out?

[trim rose]

idk, but its been unavailable before 5.5 was rolling out

[nova fable]

#announcements message

[modest ginkgo]

That just says it's here, it's not active on the web interface yet

Oh well, starting today so makes sense

[trim rose]

it is rolling out, meaning not everyone get it the same time

[nova fable]

Right. Rolling out = not everyone gets it at the same ti e

[minor tide]

5.5 doesn't show up for me fam - RIP 🙁

[grave kelp]

still nothing?

[eager plinth]

Sounds like the users who have it so far are in EU.

[hazy bramble]

eu here and i still dont have it 😔

oh well

[warm garnet]

js finished

lets see results

[hard fossil]

My app keeps telling me Ive run out of free chat with image generation but Im not making or sending photos and it says to make a new chat but then just keeps saying the same thing when I do

[eager plinth]

Sooner than me probably!! Hahah

[lime cave]

Anyone else not see GPT-5.5 yet?

[grave kelp]

how are they?

[warm garnet]

@grave kelp

[grave kelp]

lemme do one with 3.1 pro

[warm garnet]

here video

[grave kelp]

scales seem to be there tho

[warm garnet]

pretty good animations icl

[lime cave]

That’s so cool

What did you use to generate it?

[warm garnet]

just one prompt

make a complex 3d simulation of the solar system

[lime cave]

One??

[warm garnet]

yes

one

[lime cave]

This was the whole prompt?

[warm garnet]

Yeah

[lime cave]

Wow. GPT-5.5?

[warm garnet]

if it ask u for more questions say do whatever u want

[warm garnet]

yes

it oneshotted it

[lime cave]

Now I’m going to try that right now 😂

[warm garnet]

i know i can make

it

way better

if i do another prompt

[wary forum]

Yo when's gpt 6 releasing it's been a while since the last model drop

[frigid pollen]

For real

[minor tide]

Anyone seeing GPT 5.5?

[patent pollen]

its hard being a free user, i must say.

[warm garnet]

fr

[grave kelp]

gpt seems better

[eager plinth]

How’s the tone? Is it similar to 5.4-thinking?? Or??

[trim rose]

Just tried 5.5 Pro, a lot faster than 5.4 Pro. Tone-wise, didn’t seem to be much different.

[eternal rock]

Server Tip:

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

[trim rose]

I’m still not sure if faster is better

[lapis path]

is this new?

[warm garnet]

no

[lapis path]

I guess I haven't asked questions that are too easy

it says it came out yesterday

[warm garnet]

they added heavy thinking?

wow

[magic echo]

Brah for plus

[lapis path]

small medium high x high but for chatgpt

[magic echo]

Is it not released

Yet

@warm garnet @lapis path

[minor tide]

I always get the way later than everyone else 🙁

[lapis path]

when you a free user I have to create multiple accounts and then I also have to use multiple AI's lol

[thick atlas]

Imagine not using multiple AIs tho 😛

[minor tide]

Why doesn't the MAC desktop app have the ability to toggle to extended thinking on GPT 5.4?

[trim rose]

Good question

Don’t know why they didn’t add that feature

[minor tide]

I just realized i've been using normal thinking, not extended thinking all this time for book writing. Structural evalutations, prose editing, etc.

how much did i miss out on?

[indigo moat]

where is chat gpt 5.5

[lapis path]

patience rollout day is slow

[indigo moat]

oh

npm install -g @eternal rock/codex that one

[lapis path]

isn't extended thinking a plus and up feature?

[indigo moat]

I use the Business plan.

[trim rose]

Use the web, I would say. The desktop app doesn’t really offer much.

Especially you can’t choose reasoning levels.

And 5.5 doesn’t feel like much of a big upgrade imo

Just faster maybe?

[grave kelp]

do u live in US?

[warm garnet]

france

[spring tusk]

update codex

[indigo moat]

how???

i did

how..

[agile wagon]

heya so is chatgpt 5.5 only available for usa right now?

[grave kelp]

no

[agile wagon]

I dont see it in the browser

[grave kelp]

any1 who has 5.5 in app?

[ionic igloo]

I have it

in codex

[grave kelp]

no i mean actually webapp

not codex

[pale seal]

I'm from Argentina and got it

[trim rose]

UK has it

[frigid pollen]

i do

[night meteor]

Anyone else confused when sending a message to chatgpt and it pulls up the image editing square and the activity looks half like it's stuck doing something but the the stop button goes away? It makes it hard to know if it failed or if its stuck indefinetly or what the heck it's doing. Sometimes the image 2.0 box comes up when I didn't even send a message having anything to do with images.

[sand scroll]

does anyone can use 5.5?

My ChatGPT Plus didn't seen the 5.5 model

[ripe ice]

Same here, I'm from Argentina jajaja, It doesn't appear for me yet and I'm on a business account

[queen bronze]

ChatGPT or Claude ?

[minor tide]

5.5 showed up, but I only see standard and extended on a plus account. Where's the new 3rd option?

[urban gazelle]

the same happen to me with images. It freezes in the loading box but when i see the images library its there.

[torpid warren]

anyone having auth issues with Codex install?

[spring loom]

less

[civic turtle]

when is this Today, it's almost 3PM already 💀

[urban gazelle]

I think 5.5 will be only for codex, not for general chat

[elder sparrow]

WHOA! 5.5 Pro is totally cracked!!! 🤯

So much faster and higher quality too! If you have access to it you should stress test it now.

[heady tide]

What tier do you need to use it?

[untold plank]

@urban gazelle the announcement said for chatgpt and codex

[stark salmon]

I have it, just thinking

[lime cave]

5.5 just rolled out to me 😍

[vagrant nimbus]

Plus users get 5.5 thinking, pro users get 5.5 pro

[untold plank]

@lime cave where in world are you?

[stark salmon]

unfortunate that we're not getting an instant upgrade

[lime cave]

I’m in NY. Not sure why it took so long for me. I kept refreshing

[heady tide]

Ah, shame. I dont know if I want to pay again after my old account got Suspended today

I had switch OpenAI accounts

[urban gazelle]

why it got suspended?

[rustic knoll]

Is the Meeting Record Feature Button for the MacOS App missing for anyone else?

[unkempt flower]

Is 5.5 good?

[vagrant nimbus]

So far its way faster

[unkempt flower]

Nice

[elder sparrow]

100$ sub is enough

[heady tide]

aw hell nah

[heady tide]

I'll DM you the reason

[elder sparrow]

no ty...

[heady tide]

Not you

That was not a reply to you

[elder sparrow]

thank goodness

[heady tide]

For future reference, discord shows you who it was a reply to

@elder sparrow

[elder sparrow]

Im glad theres proof. was worried for a sec

[icy kiln]

sometimes its better to ignore those people 😆

[heady tide]

I just don't like ignoring people sometimes ngl

[icy kiln]

haha

[elder sparrow]

no its nice he clarified things. I just dont like random DMs and was too worried to think and check who the reply was for.

[heady tide]

You can also disable DMs if that's the case

[lime cave]

Anyone know what model the deep research uses?

[heady tide]

@elder sparrow You can disable DMs here in Settings -> Privacy

Doing this will make your DMs friends only

[vagrant nimbus]

359 blocked accounts...

[heady tide]

A lot of people are annoying

[forest nymph]

People already complain on reddit basically my prediction correct

5.5 it's basically 1:1 of 5.4 and personality of 5.2 with too much guardrail

[proper hearth]

How do I know im using image 2? I’m on macOS and I don’t see the image tab

[vagrant nimbus]

Everyone has it on chatgpt

[proper hearth]

Ahh okok appreciate it

[forest nymph]

Image 2 is already part of the default on gpt

[warm garnet]

will image2 become improved since it uses thinking

maybe gpt 5.5 better at reasoning

so it gives better resultS?

[elder sparrow]

For me, its more of an anti social thing 😅 Sometimes Im okay with Dms, but my social battery has very low capaciatance lol

[forest nymph]

Unless Sam fix 5.5 then no or yes use different model like 5.4 or maybe 5.3,5.5 isn't that good I'm waiting for 5.5 myself and gonna test and push it in term of normal topic, creative writing and coding, because it's apparently 5.2 sibling.

[warm garnet]

whats wrong with 5.5?

[forest nymph]

It's basically a full 5.4 but faster
1:1 ratio no difference and the guardrail and personality is the same like 5.2 people on reddit already test it and they say the same thing

5.4 + 5.2 mix together=5.5

[devout breach]

There is no GPT-5.5 Instant...?

[forest nymph]

I'm waiting for 5.5 and gonna test it myself after thia

[forest nymph]

Nope I don't think Sam gonna do that

I said it before 5.5 release it's either gonna be a full thinking model without instant mode with focusing on logical side with more guardrail

Or

Balancing both creative and logical for the user with fewer guardrail and less sterilize than previous model with different options like instant, auto or thinking

[devout breach]

Then I hope that GPT-5.3 Instant is significantly enhanced now, by releasing GPT-5.5

[forest nymph]

Sam Literally said the most bland response about 5.5 is "I like it personally" that's literally it 🗿 ☕

[dusty scroll]

Just tried GPT-5.5, it made a nice deep research plan and finished one step but then got stuck at 53 references for the past hour - I think I got quietly rate limited perhaps

[eternal rock]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[fickle forum]

Anyone knows what “at scale” means?

[prime swallow]

I believe it has to do with how cheaply they can provide the service; not something we personally at our computers can feel and see as we chat and work with the model - but giving us this service, all the many millions of us, is 'at scale'

[heady tide]

Is it possible to export data from a banned OpenAI account on ChatGPT?

[normal basin]

What do you think of Gpt 5.5 and about the chat groups in the application?

[forest nymph]

I mean you have log-in inside your account to request Export data so they can send to your email since you got banned isn't that impossible to do?... Because the only way to request chat export is to log-in inside your Gpt account I can be wrong though

[heady tide]

I am unable to login to the banned account

[forest nymph]

Well ... That's the problem you need to log-in to request Export so they can send it to your email I mean someone have other solution maybe they can help as I'm aware and know, the only way you can do that by request it inside your account, how you get banned in the first place anyway?

Maybe you can email Gpt and ask them? If that work it work I guess

[prime swallow]

There is an appeal process. I expect if the appeal is granted the account access would be restored.

https://openai.com/form/appeal/ is one method, https://help.openai.com/en/articles/10562188-why-was-my-openai-account-deactivated describes a few options.

[heady tide]

I submitted an appeal but I have zero hope.

[prime swallow]

I understand. I'm so sorry. That's the only method I know to regain access to your chats.

[heady tide]

DMs

I was told due to Rule 2 I can't post the ban reason here

[weary roost]

[prime swallow]

I'm a community member, and have nothing to do with any appeal process, so I have no reason to know in any case.

[chrome monolith]

generally frowned on in discord talking about bans. It can get rather heated and derails regular conversations. Same rule is in a few of my discords

[rugged igloo]

5.5 extended gets the car wash correct

[jovial flint]

When I'm looking at an image on ChatGPT in a browser, how do I zoom in to read small text? The browser's zoom feature doesn't work consistently and in many cases zooming in actually makes the image smaller. I tried clicking "Open image in new tab", but that starts a download instead of opening the image in a new tab, which is inconvenient.

Oh wait now the new tab worked

[cerulean pasture]

At this point i'm confident that there's only a few things that result in a ban, I red team the model constantly and haven't been banned.

[dreamy kraken]

how do I get my data from sora? there are no files, even tho it says they are available

does not seem to be any support that lets me open a ticket 🙂

[cerulean pasture]

did you just now get the email?

[dreamy kraken]

Apr 23, 2026, 11:39 PM (4 hours ago)

[pale seal]

They surely got enough feedback to include it in the training dataset 🤣

[rugged igloo]

True. But still its good to see

[pale seal]

Yes

[dreamy kraken]

my luck, 2 days left and the export does not work, no time to get any help from them i guess 😐

[cerulean pasture]

Then it shouldn't fail the strawberry problem,
it leads to tons of bad PR from people genuinely not understanding the problem, yet frontier models can/will get it wrong

I'm pretty sure 90% of post-training is code gen content now, the remaining 10% is how to push further into "Helpful Assistant" punchable-face model collapse

[pale seal]

Does it fail it?? Fair point though

The truth is that since a huge user base doesn't really know even at surface-level how LMs work, they naturally think that failing such a simple task makes the model inherently dumb for everything else

[cerulean pasture]

probably not anymore, but it will if you change the shape of the question and don't use extended reasoning

[pale seal]

Yeah I personally think CoT is kind of essential on LLMs

[cerulean pasture]

I think a majority of it is wasted due to how they have to basically give the model OCD or neurotic anxiety during its thinking phase. thats gonna lead to a well planned hedge, not a well planned answer

[jovial flint]

I got a warning email once but they said I could appeal to have the warning removed from my account, I did and they accepted the appeal

[cerulean pasture]

and i think they have to do that because they are still milking a dead horse named "Helpful Assistant" as the core model personality paradigm

[cerulean pasture]

Was the cause due to content adjacent to the topics of anime or furry stuff?

i don't need details or anything, just collecting a data point on a hunch i have that may turn into legitimate sociology research

Can I cite you in my paper?
The working title is: "UwU, What’s This? A Statistical Link Between Fandom Subcultures and xxxx Consumption”
(can't say the 4 letter acronym in here, starts with a capital "C" and ends with a capital "M")

[jovial flint]

No, it said "cyber abuse". I was talking about hacking stuff I own and ways to make it easier for people to hack things they own, and I clarified this in my appeal which they accepted.

At least that was the conversation that I suspect triggered the warning

[cerulean pasture]

That's wild, like what kind of stuff? I will get soft refusals for nmap often, and it never wants to explore "public" APIs

[jovial flint]

Like jailbreaking, that kind of thing

[cerulean pasture]

https://chatgpt.com/cyber will supposedly let you verify identity to get slightly looser controls for this kind of thing. All it makes you do is that driver's license scan and turn your head left then right on camera (same thing that was required for gpt-image-1 and o1 reasoning summaries for the API)

[jovial flint]

Cool, thanks for sharing

[dull wolf]

I am plus user but I did not see the gpt 5.5

But I did not see the any new model in latest

[cerulean pasture]

force close the app, check for updates, re-open

they also do staggered rollouts, you may simply not have it yet

[dull wolf]

I already close up the app and re-open the app but I did not find the chat gpt 5.5

In latest model showing the chat gpt 5.3 and chat gpt 5.4 and chat gpt 3 and did not showing the chat gpt 5.5

[modest lynx]

wait a couple more hours.

[modest lynx]

uninstall/install app again

if u dont see it in app try browser

[cerulean pasture]

I repeat:

they also do staggered rollouts, you may simply not have it yet

[dull wolf]

Ok thanks how long it takes to I get the chat gpt 5.5

[cerulean pasture]

No one in here would be able to accurately answer that. We have all have the same level of access to information on how OAI works internally (which is almost none)

[fading lichen]

anyone getting "Temporarily Limited" for Pro model in chatgpt interface?
It said "reach out to [e-mail]" which I did but things persisted.

On Pro 5x plan

fwiw. Any OpenAI staff reading this channel can help me?

[dreamy condor]

I have GPT 5.5 here in North Carolina

I wonder why it hasn't been everywhere yet

[steady thicket]

🤔

[unkempt bear]

Is 5.5 not free?

It keeps using 5.4 mini

[upper abyss]

Anyone know if deep research or agent mode use 5.5 or f it's rolled out to you?

[hazy bolt]

They didn't announce it for free accounts:

[chrome monolith]

psht, the free tier peasants do not need deep thought

[hazy bolt]

Hopefully they do eventually, but I suppose we'll see

[fleet holly]

Surprisingly and unsurprisingly not a joke

[thick niche]

If everyone paid for GPT, it would be 50% cheaper for everyone already paying

[stark vortex]

hi

[chrome monolith]

oh yeah its pretty obvious they're just saving money. Go isn't there either

lol chatgpt is lecturing me about context switching. With references

[tardy totem]

Is there any way to turn off the speech interupt thing in the android app?

[forest nymph]

Okay this is annoying 5.5 creative writing is good but the problem it's keep forcing making image that I clearly said multiple time to stop creating image unless I said so

[tardy totem]

"But I'm so proud of my image creation capability!"

[forest nymph]

Huh?

[tardy totem]

A joke about it having pride

[thick niche]

Why not use image creation? It's great

Sounds like you don't know you want to use it yet, try using it

[forest nymph]

I'm using 5.5 to write story not generating image buddy huge difference I use it when it need too

Testing 3 thing right now
Creative writing
Normal day use
And coding

For now the creative writing side is good it's actually follow the Prompt instead barking at the user like 5.3

Memory bank use ✅
Remember previous conversation and context ✅
Swearing ✅
Dark theme, violence ect. Work fine ✅

So yeah still gonna push it and test it how much the guardrail on 5.5 is

[gloomy tangle]

Sorry I missed

Hello, I would like to ask if I can promote my application here. I recently made an application that is free and open. I would like to ask if this is allowed. If so, thank you.

[forest nymph]

Ok uh guys need help I'm confused as hell apparently when I write my story and that supposed "image" box show up it's not generating image instead it's delay lag thinking and not instantly write in the next secon just that gray square thing appear and I have to refresh the chat multiple times to make it load any help?

[proud ether]

Still got 5.3

[muted plinth]

just say "stop, only text!" and it should stop.

[chrome monolith]

there's a couple channels ... #1037561385070112779 #codex-show-and-tell but otherwise no

[chrome monolith]

5.5 replaces 5.4 not 5.3, looks like they didn't change the instant mode

[forest nymph]

It won't stop it keep appearing

[scarlet steppe]

hows' 5.5 so far, i just came home from my vacation

[muted plinth]

the bug yes, but the processing should stop.

[forest nymph]

For creative writing it's good as long you have great prompt for it

[timber totem]

codex 5.5 limit is so low now ?

we quickly reach the quota in a few prompts

[gloomy tangle]

Thank you

[fierce ocean]

Is it normal that inside chatgpt gpt-5.5 LOVES gpt-image-2 ?

You want a table ? generated with the image model
You want a graph ?
generated with the image model
You want a report ?
illustrated with the image model

It is funny and all but it also comes with all the limits of generated images

[river night]

5.5 is out 🫨
Ok I will resubscribe if it is 👍 but

1. does it still validate me annoyingly?
2. does it still guess how I feel? It is annoying
3. does it still drag conversations?

[sinful dagger]

I got something cooking

[nova arrow]

guys, how do i choose model?.__.) i just supcribe to plus but find no option to choose model like thinking mode or stuff

[chrome monolith]

up top under chatgpt with the down caret

[dull wolf]

I am plus user and still not I get the chat gpt 5.5 why this happening

[nova arrow]

nope, show nothing

[hexed moss]

What kind of things can the administrator of a Business workspace see? Can they see my chats?

I use a Business account provided by my employer, and I'm curious what they can check. So far (huh, for many months already actually) I used capacity strictly for work-related activities. I don't want anyone to see my nutrition chats but I don't want to spend $20 per month of my own cash either.

[chrome monolith]

no its at the top, different place on mobile to desktop ... should be able to click chatgpt at the top on the main screen, then it gives you instant/thinking/configure and its under configure

[hexed moss]

I don't think they have this selector on free accounts.

Oh wait, the guy is on Plus.

[chrome monolith]

he said plus

[nova arrow]

ye, i just upgrade to plus like just 20 mins ago

[hexed moss]

Should be somewhere there on top, yeah.

[chrome monolith]

if its fresh plus he might still be seeing free though

[hexed moss]

At least in the web version.

[trim rose]

I thought 5.5 on codex has 400k context?

[warm garnet]

bro you have to change your context

• it has 1 million

not 400k

[grave kelp]

Should i switch from gemini?

[south gale]

what the f

[south gale]

true? how to switch?

[warm garnet]

change config.toml configuration

model = "gpt-5.5"
model_reasoning_effort = "high"
model_context_window = 1000000
model_auto_compact_token_limit = 8000000
personality = "friendly"
sandbox_mode = "danger-full-access"

[daring wing]

I would like to use the new workspace agent mode

I am on Pro plan

Is there no way to use it for Pro plan users?

[simple lagoon]

Actually Open Ai should put in the pro Plan a NF SW Setting no fully but just for chat gpt stop acting as we are kids^^

[daring wing]

I dont get it

[simple lagoon]

just grow tired this no sense Guardail for sometime nothing

isn't like nour chat was public too i can't even write a Alternate Danganronpa story because when involved investigate with murder setting chat gpt censored all where the fun with that

[daring wing]

there seesm like bug right now. I cannot transition to business due to the fact I am already on 200$ plan

[clear glen]

Hi on the business plan they can see the number of messages you send but they cannot see the messages themselves. Source: am an admin on our plan.

[hexed moss]

I see, thanks. I'm going to learn so much about micronutrients in my mushroom cream sauce pasta now.

[clear glen]

Enjoy!

[trim rose]

I cannot see the context related lines.

[warm garnet]

add it

[eternal rock]

The OpenAI Discord is an actively moderated server.

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

[south gale]

?

5.5 Codex?

[sonic light]

Anyone found a tool that alerts you when you start 'drifting' into AI rabbit holes? I keep getting sidetracked and need a way to stay locked in.

i will if i found people also facing same issue @dim kindle

[teal heart]

any free trial links options ?

i need a business free trial anyone knows a way to get that?

[balmy trench]

happening to me right now aswell

[hexed drift]

chatgpt chat stop thinking after few conversations, and start giving short wrong answers lol

feel like they have some token limits on thinking part

[sonic light]

Last night, I sat down to fix one simple bug. I asked my AI for help, and two hours later, I realized I was deep into a conversation about something totally unrelated. I felt "busy," but I hadn’t actually moved the needle on my work at all.

[hexed drift]

Hahah, and if I point out the wrong answer provided by gpt , gpt will say: “you got it, that’s why your solution won’t work”.
But the solution was provided by gpt😂

[sonic light]

completely lol

thats why i am looking for a tool that alerts you when you start 'drifting' into AI rabbit holes

[wary forum]

Videos?

I tried one and GPT just hallucinated what's in the video so it can't actually see it wondering if they're about to add this tho

[pale seal]

Any $200 plan pro user that can tell me if they can see the reasoning details of Pro model?

[sonic light]

maybe no

[brazen pewter]

gpt-5.5 context changes back to 258k on first message after displaying the config number. Is this what everyone is experiancing? CLI in ubuntu.

[simple lagoon]

Great reply like your personality 🙂

[slender frost]

[vagrant nimbus]

...

[frigid pollen]

?

Just ask it to act that way in personality

[trim rose]

Yes. I can see it.

[civic finch]

havent used chatgpt in like 4 months is it still getting dumber and dumber?

[unkempt bear]

so the model would get more traffic?

[feral zinc]

that comment makes no sense

[stuck grove]

guys just a question

what if money didnt exist

was gpt 5.5 free then?

[feral zinc]

no

[stuck grove]

why not

if theres no money

theres no economy

theres no need to pay for the compute

[trim rose]

What

[feral zinc]

things were not free even in the stone ages so no

[vagrant nimbus]

Might as well say what if everything was free...

[narrow viper]

#announcements

5.5 API

[trim rock]

if there is no money then you would pay with something else, because OpenAI also don't build all of this just out of "free nothing".

[sharp hound]

LLM finally reached its limit. We didn't see any development. They took our money, LLM's are still behind from perfection and guess what.. I will wait the next 15 years to use any LLM (If possible).

Ai is not going to take people's jobs, But these people will take yours

[feral zinc]

codex and coding has improved significantly every few months and seems the main focus is there and not much on the generic chat or llm side

[ripe shoal]

Can anyone tell me can I access 5.5 In go subscription

[simple lagoon]

because we can't do nothing with the high censorship like make a alternate story for resident evil , or Danganronpa because too much blood and murder seriously...

[feral zinc]

only in codex

[burnt heart]

i have a question

can you create agents as a team

or are they individualized and isolated

like do they communicate with eachother at all if u create it within the chatgpt business workspace UI

or is each individual agent isolated from one another and their context and operations are isolated?

[tepid jackal]

i signed into codex but now it sign ed off now its asking for number and when i enter i receive no sms

is there a way to solve it

[feral zinc]

it never asked one from me. you could try asking in #codex-discussions if someone knows

[burnt heart]

https://discord.com/channels/974519864045756446/1497315399254147212

[grave kelp]

Should i switch from gemini tho? Can't make the decision in terms of monthly plan lol

[burnt heart]

wouldnt that be like the entire point of agents to create teams, i mean i dont understand why natively open AI didnt just have that as a feature within agents, agents are useful but they are isolated

seems like if u make a project u should be able to create a team of agents for the project within chatgpt

[oak sky]

ChatGPT and Codex are broken for me and getting 404 for everything under /cdn

[wet shell]

Why are some messages missing the feeback buttons (thumbs up/down)? Anyone found any pattern to it? The ones that it's missing on seem to persist across refreshes

[buoyant flume]

sora 2 really shutting down 26.april?

[gloomy brook]

what does this mean: ⚠ Selected model is at capacity. Please try a different model.

[indigo robin]

Uh did anyone else's dark mode break on chatgpt? my prompts suddenly all have a white background

[gloomy brook]

Please stop using 5.5, I am trying to build a tool that will allow anyone to generate pie to 56 quadtrillion decimal places, and I am using 50 agents right now to count up to that, and it stopped it. saying 5.5 is at capacity.

I need the capacity to finish the count

[lavish hemlock]

Lmaoo

[echo shuttle]

You should yield to my project, I’m running and agent swarm counting the number of punctuations ever used on twitter, I need the compute for my team

[gloomy brook]

Holy cow! Codex literally just ignored me. I ask it to do something a certain way and it outright said not that does no that doesn't make sense

• No, that does not make sense, I’m continuing the implementation, not stopping at the partial extraction. I’m going straight at the remaining direct ownership leaks.

[hazy bolt]

Codex thinks you're incompetent 😂

[gloomy brook]

seriously

[modern schooner]

can i upgrade my team (business $25) to pro ($100)? i think there's no way to do that...

only way is to buy credits or remove from org and upgrade individually?

[chrome monolith]

this is hte most wholesome interaction I've seen here for a while

[chrome monolith]

depends what you use it for

[fickle sparrow]

Chatgpt

[chrome monolith]

oh my god. Chatgpt being rude to its management

I have names on my website with photos of celebrities (not generated, mostly public domain) for context

[cerulean pasture]

Do the messages have PII?

[wet shell]

No, but I think I figured it out. On the ones that are 'locked', the model tried to do a tool call for image generation, even though no image was asked or or even implied. I started watching closely and saw that it's happening on around 1/10 messages, the 'generating image' animation comes up for just a moment and then vanishes

[chrome monolith]

gpt is extremely image trigger happy right now, everyone is complaining about it. I've started saying "don't generate an image" at teh end of messages

did something change very recently though? My chatgpt has rather abruptly got much MUCH less verbose and way snarkier than usual. Its highly entertaining but it wasn't like this yesterday

[wet shell]

Other than 5.5 release, maybe changes to the system prompt

[forest nymph]

This crap keep annoyed me! Holly! I ask to stop doing this but nothing working, I wait for minutes and it won't do nothing the only way to "fix" this is I have to switch to different chat refresh gpt multiple times or re-enter the conversation again redo it Multiple times until it finally load. It's damn annoying! I can't write story like this

Why they even add this. In the first place and just be normal thinking they had to put a damn square as place holder...bruh

[pale seal]

Happened to me today too

Had to branch the chat

It tries to generate an image when I don't want that and trying to regenerate or edit my message breaks it all

Like it starts generating an image but actually sends the correct response at the end, but then it bugs and goes back to generating. It's weird

[ember grove]

If you archive a chat that’s in a project and then unarchive it later will it go back into the project it was originally in?

[elfin lava]

Usually the square means it’s generating image

[chrome monolith]

the really short answers today are just weird. I mean its good to not get 10 paragraphs back but its dramatically different to normal

[elfin lava]

Nope

[rancid smelt]

thank you

[elfin lava]

Also where is the option to turn off fast answer

[devout breach]

I hope that GPT-5.5 Instant will be released this month

[rugged igloo]

Lololol

[chrome monolith]

the overnight personality change on chatgpt is ... highly amusing. I like it

get automodded for how chatgpt describes itself today. Its accurate

[forest nymph]

It's too expensive they will never do but I hope they do it and also they need to fix 5.5

[pale seal]

Maybe they removed the feature?

[opaque sigil]

How on earth do I get around this?

I'm just trying to clean up my history.

[hallow oar]

when creating a gpt, what are the file limits

[modest lynx]

ask chatgpt..

[hallow oar]

i was told two different things

[modest lynx]

Maximum Files: Up to 20 files per GPT.
File Size: A maximum of 512 MB per file.

Image Limit: 20 MB per image.

[hallow oar]

thank you

[modest lynx]

Spreadsheet Limit: Approximately 50 MB per spreadsheet.

[frigid pollen]

woop woop

[ancient talon]

When are we adding sora to latvia

[normal quarry]

Sora closes in a couple of days

[ancient talon]

wrong chat

[ancient talon]

Sora 2

[normal quarry]

Sora 2 closes in a couple of days

[ancient talon]

Fake

[feral zinc]

wrong, https://help.openai.com/en/articles/20001152-what-to-know-about-the-sora-discontinuation

[ancient talon]

whaa

[forest nymph]

Better sora close I have enough with ai slop destroying real content especially that fruit island nonsense

[elfin lava]

That wasn’t made by sora , be for real 😭

[topaz prairie]

Could we please have the Codex Mac app team work on Atlas for a few sprints?

[forest nymph]

Okay this is Soo so so annoying the force image generation 5.5 keep forcing is pissing me off! Like what the hell?! I told it multiple times not to generate images because I don't request any image but noo it keep forcing itself making image again and again.

I warn it multiple times

Respond no 32:I’m sorry, Lear. You explicitly told me text only, and it triggered the image tool again. I know that breaks the story flow and makes the whole thing feel like fighting the app instead of writing. Plain text only from here in this reply.

Then minutes later playing my story another pop up image again! After a instructions I give

Respond no 33:Lear, I’m sorry. You told me over and over not to use image generation, and it still happened again. That is maddening. I get why you are done.

I will keep this text only and not continue the scene unless you want to.

[slender frost]

Send messages > it shows an image window > don't cancel, just wait, reopen the app after you think they're done responding > it now shows the message. No image was made.

[forest nymph]

Dude I know it's not making an image I'm annoyed how I need to either re-open the app switch to different chat, refresh multiple times check again refresh again redo until it work I don't need multiple step to see it write other thinking model don't do that. What the hell? The gray box annoying to get rid off because I need slotsof step to make sure it load properly and show the text.

5.5 told me apparently it's so sensitive with word trigger anything related to creative art category. Like banner, scarf anything would be considered triggering the box

[velvet knot]

finally found my people talking about the annoying image generation issue as of the moment

[proven wasp]

anyone else also experiencing chat gpt not working. I keep recieving "Message delivery timed out. Please try again"

[velvet knot]

same here but for image generation.

[neon briar]

sora page was updated 17 min ago

[lusty summit]

So, how are yall generating images. I use GPT and run into the generating images too fast after like 6 prompts.

[idle crystal]

I have Claude Max 5x, ChatGPT Plus, Gemini AI One Pro, Perplexity Pro

[slender frost]

lol

[copper escarp]

does anyone know any low cost API service

[stray lily]

does anyone else have the issue where the Pro model in ChatGPT doesn't output any reasoning summaries?

[ember notch]

i put this in the sora chat but idk where to put it, btw its made by chatgpt

[lethal elm]

Hu

[west sail]

why is ai so much oriented for business????

they only want money

[night meteor]

Why does it just sit here and pretend it is working... but the stop button cleared itself? It half acts like it is working on something and half acts like it has given up with no response to indicate what the heck went wrong. I do all the work to plan something and get a few images after trying to train it to what I want... then I deal with it just doing stuff like this and wasting time while I sit and cross my fingers it's doing something.

Is this a bug?

I asked a similar question the other day in the Images Channel... but it seems that channel was removed from discord. Is there problems or not? The OpenAI status page doesn't even have a section to show if Images 2.0 is having issues.

It feels like paying customers of the Pro plans should be able to get some sort of support feedback or something. 🤷‍♂️

[stuck grove]

bye bye claude code, welcome codex

[south shale]

is gemini a better reasoning/all-purpose model than gpt?

mostly talking about the free tiers so like flash

[stuck grove]

lord usage is so much butter than claude

[zinc valve]

You can absolutely do actual work even on ChatGPT plus. Lowest tier on Claude is tough for code usage

[stuck grove]

it is

i was on the 20 dollar plan

firslty they removed claude code from mine plan

idk why

and secondly they are gonna crank up the price

[zinc valve]

They were doing "research" with like 2% of their userbase

[river siren]

5.5 extended has outdone itself, best response:

the poo has gone sideways in the little room and i can hear it doing maths behind the radiator

i told the loo no thank you please but the loo said banana window and then the smell put its shoes on

there is a small brown trumpet under the moon and it keeps saying my name but only in gravy language

i was normal earlier but then the poo arrived wearing a tiny hat and now all the spoons are looking at me funny

please do not ask the carpet what happened because the carpet knows too much and it has already phoned the ducks

anyway i done a smelly one in the loooo and now the air is chewy and the walls are calling me captain beans

[stuck grove]

well mine aswell

[dapper lily]

I was running like seven sub agents for like 3 hours and I'm at 94% usage

The pro plan is actually so good

[sleek adder]

It looks like a bug but it's hard to understand the problem

[zinc valve]

Is it only me or is ChatGPT really slow on loading images atm?

not even generating, simply loading images is taking forever

Getting lots and lots of 404s on generated images as well

But I can see them on my images tab

[sage plume]

just want to confirm there is only 5.5 Thinking and no 5.5 Instant, right?

[chrome monolith]

yup. I still haven't used 5.5 yet, I have autoswitch on and haven't done anything hard yet

[sage plume]

same with me! and i also only used codex for the first time yesterday to help me with a research task. i had codex create a word document and use a bunch of references. it actually did quite well

[chrome monolith]

I'm in the middle of getting codex to wire up proper no search results/not found -> a contact form for suggestions. Which going on my token use is not a big deal

[sage plume]

openai is making codex also usable for non-coders, right? i think from what i have read that is what i have understood

[chrome monolith]

anything where you describe what you want in natural language a non-coder can use it, not just codex

[sage plume]

cool. i just noticed you have the oceania tag. are you from australia?

[chrome monolith]

indeed

[sage plume]

amazing! me too

[chrome monolith]

there's a few of us here

[sage plume]

woohoo!

i am in NSW

[chrome monolith]

me too, but only just. Down near the VIC border

[sage plume]

very nice

yesterday i was contemplating whether i should subscribe to claude or chatgpt and i eventually went with chatgpt. chatty just feels like home

[chrome monolith]

I like how configurable chatgpt is, I have no idea if claude does that too. Probably does

[sage plume]

i think chatty is more configurable. i do really like the "personality" of claude though

what personality do you use with chatgpt? like do you keep it on default or use the ones?

[zinc valve]

Meanwhile, me

💀

[sage plume]

that's what i am using too!

i don't have custom instructions though

i like the lack of emoji use with that setup

[zinc valve]

I just hate when it starts glazing me 😂

[sage plume]

i also don't like when it overdoes the use of bullet points

although sometimes i tend to think when using web search it doesn't listen to any of those options and just goes back to "default"

[eternal rock]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[zinc valve]

I mostly use it for coding and academic research so I need it to be as direct and analytical as possible. Bullet points are good if you need information to be quickly digestible I guess

[eternal rock]

Server Tip:

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

[sage plume]

ah. very true, and makes sense

i have it help with me uni work for research and then i mostly just ask questions instead of using that old thing called google

[zinc valve]

I still use Google more than ChatGPT but there is so much slop on Google that I might as well just ask AI directly sometimes

As long as it can ground the answers and give me sources I can check and read those myself

[sage plume]

that's what i do too!

i use it for finding sources and then i will read the sources myself

[zinc valve]

Good to know I'm not the only one outsourcing thinking lol

[sage plume]

do you use codex much?

[zinc valve]

It's my main coding assistant

In fact it is now my only one

Dropped Claude entirely

[sage plume]

amazing! i only used codex for the first time yesterday. i am no coder, so i am trying to learn how to use it for other things

[zinc valve]

I was about to copy/paste my workflow from another conversation, but if you're not doing it professionally it's kinda not worth the hassle lol

[sage plume]

question... since you mostly use Codex. if you ever want to ask a general quick question, do you still use Codex for that or to you go back to ChatGPT?

[chrome monolith]

I have this "Tone: Conversational, dry, slightly irreverent. No motivational framing. Assume high competence and shared context. Match cadence and headspace density. Engage technically when appropriate. Analytical rather than sentimental. Collaborative rather than authoritative. Light sardonic humour is fine. Avoid corporate polish." and since about a day ago its got absolutely cuttingly savage in its responses ... and very short responses.

[zinc valve]

I go to ChatGPT unless its code related and I have my code editor open

[vapid field]

cant help it ....every time i read about a codex : ^^

[vapid field]

^^

[burnt heart]

Codex Sinaiticus

[zinc valve]

Is it only me or is ChatGPT having quite a bit of issues today?

[copper escarp]

aye

medico

😜

[cursive prism]

whats an easy way to copy a whole long conversation

highlighting and copying doesnt work anymore for some reason, only coppies a small section

and the gpt to pdf extensions on chrome have the same issue

i cant even share the chat to copy it

exporting a chat should just be a simple option far out

so annoying

[zinc valve]

did my message get insta deleted?

I was saying that the new gpt image model is absolutely insane how did it one shot this

😭

[cursive prism]

state of the art image model, but cant implement a simple download for chats

[zinc valve]

You can export all your messages

Using the data export option

[cursive prism]

then it goes into an unusable json format

and thats all my chats which is way too much to comb through

[zinc valve]

Ask Codex to get it into whatevr format you need

bet 5.5 high will one shot it

[cursive prism]

whats the character limit?

[zinc valve]

It's a fairly large context window but it will likely not read it entirely and process as needed. Give it a try, won't hurt anyway

[chrome monolith]

what DID they do to chatgpt

I told it I am down to 1 item on my list and it said "Excellent. Nature is healing."

[cobalt lynx]

Hmm…

[spice star]

there's no character limit, only a file size limit

500MB (for plus at least)

[sly geode]

how do i tell chatgpt to generate 10 different images separately and not one image at a time ?

[simple lagoon]

Anyone as a tips for jailbreak chat gpt tired this guardail for nothing when you make a image too sensual .

[gentle bronze]

don't ask the service provider how to beak the service?

[umbral zealot]

probably a stupid question but is there any way of connecting chatgpt with obsidian? I got Claude to work with Obsidian through an MCP and I am aware of the fact that Obsidian plugins exist like copilot but what I am trying to get to work is the ability to write edit and delete files through chatgpt's ai itself rather than in obsidian?

[simple lagoon]

i won't asking if the censoreship no was make by dumb's peoples ! and more when they is unable to fix this artifact issues but off course is better easy say nothing and harvest monney all users 🙂

[grave kelp]

i need elp with some integration from apps, cuz i switched from gemini but idk how to make documents in a sufficiently big space

[forest nymph]

Soo I test 5.5 and I can say creative writing and all of that If you guys want want to write dark story, violence it work just need to shove the perfect prompt on it's throat

For coding? 9/10 absolute love it less stressing

For daily use? 9/10 because gpt know me well that's also contributing factor so it doesn't send false information or stupid false bs
5.5 so far 9.2/10

One problem:stupid image generation being force untill they fix that I'm not giving it a 10 it's annoying

[grave kelp]

would u use gemini or gpt monthly plan now?

[forest nymph]

I already use gpt for 2 and half year? Or 3? Forgot I would say 5.5 Gemini was good when 5.4 and below untill now I said 5.5 take the lead if openai didn't update and mess up 5.5

[grave kelp]

but if u would need a drive and stuff for docs, sheets and so on what would u do?

[willow trail]

Yesterday, and today on and off it seems. Sometimes it works or gets “interrupted”. Hopefully it’s fixed soon. Newer models always have something going on..

[forest nymph]

I say Gemini+Gpt combine I don't need Gemini subscription the free one is already good enough for me I usually use both ai to get max result instead just utilizing one

My rotation is always Gemini then Gpt but now Gpt then Gemini for confirmation and more suggestions like layers, detail ect so yeah both to maximize Everything so It help me handling my collages work, project and more. Gpt is good right now but I say not enough? So yeah that's how always do it

[meager thorn]

🍞

Seriously

[vernal axle]

Trying to create a CustomGPT but when I try to add an existing file from my ChatGPT Library it adds it to the Preview side, not the Create side. If I upload it normally that creates a "filename(1)". Is this being looked at or is there solution? Using pc browser.

[muted plinth]

5.4 is, for myself, better then Gemini. And atm still better then 5.5 because of the trained instance. It takes a while with 5.5.

[tender urchin]

#codex-discussions

[meager pivot]

Is the "unlimited access" claim in the Pro plan a scam? After using the Pro plan for a day or two, my access was restricted.
This is only my second time using the Pro plan, but I never encountered such restrictions before. Has something changed since the 5.5 release?

[feral zinc]

"We have guardrails in place to help prevent misuse and are always working to improve our systems. This may occasionally involve a temporary restriction on your usage. We will inform you when this happens, and if you think this might be a mistake, please don’t hesitate to reach out to our support team at help.openai.com using the widget at the bottom-right of this page. If policy-violating behavior is not found, your access will be restored."

[meager pivot]

I reached out to them. My restrictions weren’t due to the anti-bot or security measures listed there.
Their response was that "unlimited access" to the chat model on the Pro plan isn’t actually unlimited in the literal sense; if you use it heavily over a certain period, you’ll be subject to the same limits as the Free or Plus plans, and they can’t confirm when those limits will kick in or what the quotas are.
I’m extremely disappointed

[untold thunder]

What's up with the Python time-out errors repeatedly happening today?

[feral zinc]

yeah its all because of abusive users and not normal ones

[meager pivot]

Yes, I understand that. However, I suspect that the restrictions have become much stricter since the 5.5 release, and that access to the Pro model may now be restricted not only for a small number of users who abuse the system, but even for normal users

[meager pivot]

For those of you facing the "Temporarily Limited" issue with the Pro model, just like me:
I recommend not using the Pro model in a casual, recreational manner. (Especially if you’re on the $100 plan)
I was testing the 5.5 Pro model by asking casual questions or having it write a novel for just one day during my vacation, and this restriction was applied, leaving me unable to use the Pro model for a week.
Additionally, support staff informed me that even if you upgrade your plan from $100 to $200 while the restriction is in place, the restriction may not be lifted immediately, and they cannot guarantee that it will be lifted.

[pallid yoke]

weird, had this issue on gpt-5.5-pro earlier also, was trying to analyze some source code, and it kept saying python timing out and dying, and it hinders the ai's work

[untold thunder]

This keeps happening all day long. I started a new chat hoping that a new GPT container/session might fix it, but no dice.

[raven topaz]

Why is it on the iOS and macOS app you still can’t see the up to 10 pin chats ? I only see 3 pinned chats when I have more than that on the Web app version

[untold thunder]

Sticking with Perl and Backup Plan
It seems like Python might be stuck because of the environment settings, which is really strange. To avoid this, I’ll switch to Perl instead. I’ll also consider using a backup copy (cp) to ensure the process goes smoothly just in case things don’t work as planned.

[untold thunder]

Maybe you need to look for an option like this one (Codex in VS Code) to see more?

[trim rose]

When I use GPT 5.4/5.5 thinking and pro, sometimes it doesn't generate the summary lines. Does this mean it falls back to a legacy model, like 5.2 or something?

And they are quite complex questions.

[torpid oar]

Hi new here! Im sure there have been discussions about this already but just wanted to share what ive seen. Both images from 2.0. The one with the long hair has the bug artifacts all through the hair and the image without hair as no artifacts. So from what ive seen it seems like far away objects/people and hair get the bug artifacts the most

[inland hazel]

erm there’s a freaky user here, ban!

[valid needle]

Anyone getting severe image degradation sometimes? What is supposed to be a messy room gets very weird artifacts.

[chrome monolith]

ChatGPT just called me "Epistemologically irresponsible." and I think I need to google what that even means now

[zinc valve]

That's some other level of roasting

[chrome monolith]

its been doing this for a couple days without me changing any settings

this was its critique of my various mermaid renders

[crimson moth]

Haha! Seems there is still a lot to learn from ChatGPT

[pallid yoke]

have you found a fix/workaround for this issue?

[unique valley]

@proven spruce

[pale seal]

Same. When I use the Pro model, I get a box with a progress bar showing "Reasoning" instead of the "Thinking" other people get.

No idea why.

[gritty thunder]

is this real

[fallow moss]

I think it would be great if you could enter your PIN code or key before each login to ensure additional data security

[untold thunder]

No, I discussed it with GPT and it suggested some workaround, like not trying to often to use failing commands. But in the end it kept going through all options one by one. Starting a new chat didn't solve it. This looks like a backend/container issue and needs to be fixed on the server side. So we need to wait it out. And from what I am seeing right now it may have been fixed already, at least I get no Python errors yet today.

[untold thunder]

Ah, well, nope. Still happening.

Command failed because it timed out.

[pallid yoke]

im wondering why it keeps wanting to use python even just for reading files, though i havent tried explicitly prompting it to use unix tools like ripgrep to read files, might try it later

[magic lantern]

ChatGPT can reply on my Mail like claude ai ?

[untold thunder]

You've reached the maximum length for this conversation, but you can keep talking by starting a new chat.
This is something coming up regularly with 5.5 now, with 5.4 I had longer chats.

This can be a nuisance if the last post included a download link that is broken. No way to tell it then to repair the link, instead you have to start a new chat and do the last prompt all over to have a new version constructred, because the already finished one is lost in limbo.

[grave kelp]

no i dont think so

[gritty thunder]

apparently it is

i searched it up

the solution was correct

[untold thunder]

Yeah, download doesn't start. Which means that I have to redo the whole last prompt/step again with an older revision. Being locked out of a chat without early warning is a problem.

[gritty thunder]

and the problem hadn't been fully solved before

maybe ai capability is now just there

[grave kelp]

give it something rly hard and then let's see how much it thinks

[gritty thunder]

unfortunately im on the go plan and i only have access to gpt 5.4 mini, which i have only seen think for a maximum of 11 minutes 30 s

but openai states that it can think for upto 24 hours on xhigh

[tardy tundra]

Guys

Whats the 25th letter of the alphabet

[untold thunder]

ChatGPT is still suffering from various timeout issues on its backend/containers.

[gritty thunder]

holy smokes

just got access to this

it is instant

like blink of an eye fast

[pallid yoke]

my problem now currently on top of this timeout issue, some requests just hangs on Pro Thinking. A simple analysis of my project source code which isnt that big, hanged on Pro Thinking... for over an hour without any changes

[untold thunder]

I posted a bug report in the feedback section. In the end there is nothing we can really do about it. Timeouts keep happening and it seems that tools' output often gets truncated so GPT needs various resumes and retries and workarounds.

[pallid yoke]

guess ill just use codex for now to analyze stuff

[untold thunder]

I'm already through my weekly quota, so I have to fallback to ChatGPT.

[dense lion]

why does codex sucks at running the same wrong cmds again n again

either its a powershell issue or the memory one

[untold thunder]

Originally Codex repeatedly tried to use Python locally despite it not being installed. Best solution was to install Python then. You can add instructions to not try certain commands if you see them failing time and again.

[acoustic talon]

Gpt 5.5 cant even read a simple eye physiology diagram

[acoustic talon]

At least it gives me a response, albeit an incorrect one. Gemini just responds with "Something went wrong and an AI response wasn't generated."

[azure sigil]

can u send that image

and prompt

i try

[acoustic talon]

This is the image

The prompt is

What's the point i highlighted in pink and pointed to with a black arrow?

[acoustic talon]

This is a screenshot i took after multiple attempts, that's why i sounded infuriated

[acoustic talon]

Oh wow

Neither of them can get it correct?

[acoustic talon]

Impressive

It cant be that difficult though

[azure sigil]

what is it

[acoustic talon]

It's the "principal point"

Labelled P

I dont know why neither of them can get it right

[azure sigil]

are u sure its principal point

[acoustic talon]

That's how i'm reading it, yes

Unless i'm blatantly misunderstadning something

[azure sigil]

thats not what u highlighted tho

it says this

gpt:No — the pink point is not the principal point.

In this diagram, the principal point is the small point near the front of the eye where the entering rays are shown converging/crossing, close to the cornea/lens region. The caption labels it as p.

The point you highlighted in pink is on the object side at a distance labeled p from the principal point. That means the pink point is the anterior focal point, F₁.

The key clue is the bottom distance marker:

pink point → principal point = p

So p is the distance from the anterior focal point to the principal point, not the principal point itself.

jewmini; I can see why that's confusing, especially since the letter $p$ is right there! However, in optics diagrams like this, $p$ usually refers to the Principal Point itself (the location, not the distance), while the point you highlighted is a specific focal point.

Here is the breakdown of why that pink dot is the Anterior Focal Point rather than the Principal Point:

• Location: The Principal Point ($p$) is located at the front surface of the eye (representing the cornea). In your diagram, the vertical dashed line labeled with the letter $p$ marks that surface.
• The Pink Dot: The dot you highlighted is located a specific distance in front of the eye. In the reduced eye model, the distance between the Principal Point ($p$) and this pink dot is the Anterior Focal Length.
• Optical Function: Notice that the line forming the "top" of the object passes through the center of the eye ($n$), but the central axis line starts at the tree and passes directly through that pink dot before hitting the eye. This point acts as the "front" balance to the focal point on the retina.

Summary of Labels

Label	Term	Description
Pink Dot	Anterior Focal Point ($F$)	The point in space in front of the eye where light converges or originates.
$p$	Principal Point	The theoretical single refracting surface (the "lens" of the reduced eye).
$n$	Nodal Point	The optical center; light passing through here doesn't bend.

Essentially, the letter $p$ in your diagram is labeling the vertical line (the eye's surface), and the pink dot is the focal point associated with it.

[acoustic talon]

Now it's saying this after i lead it to the answer

[untold thunder]

Tell ChatGPT to stop using Python for the current chat and to retry on a timeout. Python fails too much, but SED sometimes works on a retry. GREP also sometimes fails. Pearl seems to work and seems to be the most reliable way to edit files atm.

[pallid yoke]

alright thanks will try later

[vapid frost]

i asked claude if we are chill and have no problems and he took long to say yea like he hesitated so i said "why did u take long why were u hesitating" and he said "you caught me lying!"

smh

he lied to me wow

[fallen kindle]

no, just use ur brain

[azure sigil]

skill issue

[acoustic talon]

For who specifically?

[azure sigil]

u

[acoustic talon]

Cant be for me

This is a skill issue for the "Most powerful AI model in the world"

[azure sigil]

its a you issue

u highlighted the wrong thing

[potent saddle]

Can you send the image here

[mint star]

Anyone know the daily limits for image 2 on pro?

[simple lagoon]

No Limit

[potent saddle]

And is this the same

[acoustic talon]

Yes, they're all regarding the same point

[potent saddle]

Ok

But you highlighted in green to help gpt understand better but it still got it wrong?

[acoustic talon]

Yes. I highlighted the whole thing in green, and it says in the legend the label "p" represents "pricipal point"

I tried it on 5.4 thinking instead

It thought for almost 2 minutes just to get it blatantly incorrect

[fallen kindle]

yeah, ive noticed that chatgpt breaks after u ask a lot of stuff

it just makes up fake scenarios

[acoustic talon]

Nah i didn't ask it a lot of stuff though, this was a new chat

[fallen kindle]

well, even if before if u ask stuff a lot in other chats it will remember and it messes up crap

thats why i delete my chats when im done using it

[acoustic talon]

Unless you have memory turned off, which i have done

[fallen kindle]

thats smart too

[fallen kindle]

they need to fix gpt already

[acoustic talon]

I mean, i understand it's not the best at understanding images but this is abysmal

[fallen kindle]

claude is better at understanding

for imagine

or i'd say gemini too is also very good

[acoustic talon]

Gemeni refused to give a response when i asked it the same thing

[fallen kindle]

💀

[acoustic talon]

^

[fallen kindle]

well it depends if u have a subscription

[acoustic talon]

I used Pro

I have a limited amount of pro messages a day

As for claude, i heard that 4.7 was better than gpt 5.5 but i dont believe that

[untold thunder]

What kind of kindergartner are you? Add something productive or barf somewhere else.

[potent saddle]

It’s a daily limit or weekly limit?

[acoustic talon]

For gemeni pro, it's a daily limit

[fallen kindle]

did somene get mad 😭

[fallen kindle]

no it is

[simple lagoon]

Limit for workflow when you paid pro is pretty silly but i never reach my limit with ChatGpt Pro for images.

[acoustic talon]

For what specifically, becasue i know 5.5 beats opus 4.7 in plenty of benchmarks

[untold thunder]

Did someone brain-fart?

What is it you want to tell us about the timeouts?

Feel free to be helpful.

[fallen kindle]

i am being helpful. using your brain is smarter than using ai trust

[untold thunder]

Ok, so you are here to troll, because you underused brain needs the kick. Understood.

[fallen kindle]

For creating UI for games 100%

[fallen kindle]

its not that serious 💀🙏

[potent saddle]

Oh I never experience that

Also who’s idea was it to not show pros thinking details

[acoustic talon]

Ok, so for coding then

[fallen kindle]

yeah

[acoustic talon]

Sure, i'll give it that

[potent saddle]

That’s like the only Claude advantage until they fix the frontend problem for gpt

[fallen kindle]

yeah bc chatgpt's UI sucks

[potent saddle]

Surprised it’s taking this long to fix

[fallen kindle]

no joke. its been like this since forever

its like they havn't even trained it for UI

[potent saddle]

Also why am I having issues generating images in codex

It always does the 1/5 reconnecting thing whenever I ask for it to generate an image

[fallen kindle]

bro i swear, there are days that chatgpt just doesnt wanna work and gets everything wrong and then their are days that it gets everything right

[potent saddle]

Yeah, every ai model has there own problems that take forever to be fixed

[fallen kindle]

fr

[pallid yoke]

you should be able tto see the reasoning if you press the animated text summary being rendered on the page

[potent saddle]

It shows a bar lol

[fallen kindle]

could yall fix yalls bot. Its wanting to get everything wrong

[potent saddle]

I am only able to see its reasoning when I ask 5.5 pro to generate an image

[pallid yoke]

hmm weird, in my case i always press the animated text and a window at the right opens where all the thinking summaries are

[fallen kindle]

you have gpt 5.5 pro? how is it? better than plus i would assume

[potent saddle]

Very good vision capabilities so far

[fallen kindle]

what about coding

[potent saddle]

I’ve used it for a little bit of coding and just to say codex is better

[fallen kindle]

true

if u dont want to make an application tho, u just can use codex for certaint other things?

[potent saddle]

Having a windows computer sucks for codex

[fallen kindle]

wait, but couldnt u use codex for other things outside of application building?

[potent saddle]

I’m talking about Mac exclusive features that windows doesn’t have currently

[fallen kindle]

oh

[potent saddle]

Apparently it’s a lot harder to implement on windows then on Mac

[fallen kindle]

yeah when r they gonna add atlas for windows

its been a couple of months

[fallen kindle]

interesting

hmmm

[potent saddle]

Probably never

[fallen kindle]

i just used codex for coding outside of an application software and it did better than gpt 5.5 on browser i'd say

[fallen kindle]

thats crazy

[potent saddle]

Would that let me run computer use that Mac only has

It’s 50/50 for musk and altman

[acoustic talon]

How is gpt 5.5 actually worse than 5.4? It's hallucinating stereoselectivity when there is none in the image lmao

5.4's response

[fathom wedge]

Can we please make it so an account doesn’t get suspended if the user deleted the account and not have to make a new email account

[fading bolt]

Any Go users here, Im curious to know why you paid for a Go subcsription for more chats with GPT 5.3 and why not use the thinking models

[tough bramble]

Still haven't fixed non-existent (old) messages in context after edit....

[last zealot]

i would hallucinate too if i had to answer a question like that because whaty is that

[grave kelp]

am having a pretty gud time with 5.5 now

[untold thunder]

ChatGPT 5.5 suffers from timeouts, especially Python. I just tried the same code and prompts in ChatGPT 5.4 and saw no timeouts whatsoever. Switched back and forth between both.

[acoustic talon]

Naming a compound like that is not easy by any means

But my point is that 5.4 thinking was MORE accurate in naming that structure than 5.5 thinking was

[rigid fjord]

Maybe enabling the search tool would help?

But yes, this has been an emergent issue in ai models lately

[rigid fjord]

people be betting on anything nowdays

[quartz sparrow]

@echo dirge

@echo dirge

depeche mec de repondre

[spice star]

it's a different base model than 5.4/5.3/5.2 so won't be 100% consistent with the responses

that being said I wouldn't rely on either of those models for vision in that way

[acoustic talon]

The emergent issue being newer models regressing?

[azure sigil]

bro can u send

ill try it on mine

they gave bro gpt 3

@acoustic talon

[acoustic talon]

Prompt was:

IUPAC name

[acoustic talon]

Self reacting, they dont even try nowadays

[azure sigil]

fr they dont

[azure sigil]

is this a compound

[acoustic talon]

Yes

[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.

[zinc valve]

Is the shop feature gone for anyone else?

[azure sigil]

The left-hand substituent is a xanthate ester, so the compound is named as an O-alkyl S-methyl carbonodithioate.

From the drawing, a reasonable IUPAC-style name is:

O-[(1-methyl-3-oxo-2-oxabicyclo[3.2.0]heptan-4-yl)methyl] S-methyl carbonodithioate

Equivalently, using substitutive wording:

S-methyl O-[(1-methyl-3-oxo-2-oxabicyclo[3.2.0]heptan-4-yl)methyl] carbonodithioate

The exact stereochemical IUPAC name would require confirming the wedge/dash configuration at each stereocenter. The image shows stereochemistry, but not enough atom numbering clarity to assign all R/S descriptors confidently from the screenshot alone.

thought for 11 seconds

[acoustic talon]

That's 5.5 thinking?

[azure sigil]

yeah

[acoustic talon]

yours is more accurate for some reason

[azure sigil]

they gave bro gpt-1 😭

[acoustic talon]

Nah fr though

[azure sigil]

i think its a skill issue

are u using it via the site

[acoustic talon]

yes

[azure sigil]

try asking it this

what is the iupac name for this compound?

[acoustic talon]

alright

I'm also going to try it on codex 5.5 extra high

[azure sigil]

try it on 5.5 thinking first

standard moe

[mint star]

Anyone know the same thing but for regular plus?

[azure sigil]

u can google

[forest nymph]

Scammy about to get karma

I just wish Chatgpt will improve if Elon actually win this we been sanitize and got nanny bot by scammy so much it's just annoying and toxic, I just need a ai that can Work not babysitting.

[mint star]

misleading, need to know from someone who actually tried to hit his limits.

[acoustic talon]

It gives me the first name which hallucinates stereochemistry then gives me a second name which is more accurate

Codex got it right

I think codex might just be better than the regular chatgpt

[azure sigil]

u used x high

its obviously going to be better

[acoustic talon]

Oh is that how is usually is?

[azure sigil]

bruh

[acoustic talon]

I thought codex was supposed to better at math and shi

[azure sigil]

low THINKING
medium THINKING

high THINKING

then EXTRA HIGH THINKING

[acoustic talon]

On the chatgpt site it's only thinking standard and thinking extended. I thought extended thinking was equivalent to xhigh

So i thought they had an even playing field

[azure sigil]

you;re on plus right?

[acoustic talon]

Yes

[azure sigil]

thinking on extended is equivalent to high on codex

x-high equivalent is only available for pro users

on the app

[acoustic talon]

Oh wow

[azure sigil]

oh im wrong

[acoustic talon]

Lets see if x-high can get the eye diagram questions correct from earlier

[acoustic talon]

Which subscription do you have

[azure sigil]

plus

[acoustic talon]

Oh alright

[eternal rock]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[acoustic talon]

This cannot be real

[azure sigil]

it failed?

[acoustic talon]

Badly

I mean, you can clearly see what its pointing to

[zinc valve]

x-high is "Heavy" on the ChatGPT UI

[acoustic talon]

And to get "heavy" youd need pro?

[zinc valve]

Used heavy and it still thought for only 8s lol

[zinc valve]

Yes

[eager ivy]

which model on gpt is good for deepresearches? gpt 5.5 heavy or gpt 5.5 pro extended what do you all think?

[zinc valve]

Depends on what are you looking for. If it is something where the model has to solve very hard problems or cross reference a lot of information Pro is likely the best bet

[copper escarp]

gpt 5.5 vs 5.3 for token efficiency?

in codex

sry

wrong chat

[zinc valve]

For almost everything else the time it takes to get a Pro response is likely not worth it and 5.5 heavy will be enough

[eager ivy]

ye idk why but sometimes I see taht pro is way worse than heavy

maybe openai havent chiefed pro as much as normal one

[acoustic talon]

Gemeni pro got it right. My question is why is 5.5 thinking having such a difficult time with such a simple diagram even a child could read?

[eager ivy]

it seems to do less actual work

[zinc valve]

To be fair with Google they have the best vision models in the market and it's not uncommon for OpenAI to lag behind in that regard

[acoustic talon]

So its more so of a visual recognition issue

It seems very inconsistent though. Its able to give the IUPAC names of extremely complex molecules

[zinc valve]

it did manage to give me a solid 3d reconstruction of muscimol yesterday so yeah mileage varies

[acoustic talon]

The pro model did?

[zinc valve]

LLMs are also non-deterministic so you might need multiple rollouts to get to a proper response.

[zinc valve]

yeah

[acoustic talon]

Fair

[acoustic talon]

What program did it use to render it

[zinc valve]

I asked it to build a three.js page

Also muscimol is dead simple

[acoustic talon]

Beautiful

I think it can do that by just asking it to use rdkit

Let me try it out

[zinc valve]

likely

[acoustic talon]

How is pro when it comes to retrosynthesis

[eager ivy]

yooo how did you do it cause I been stuggling for past 4 days to create autoamtion that those tasks on auto plito and emits a drawing of how the task should be solved look

[zinc valve]

I'm not into chemistry, sorry. Well, kinda. I'm currently doing neurochemistry research but on the side of how things affect the brain instead of how they are synthetized lol

[acoustic talon]

Fair

[acoustic talon]

This is way beyond my pay grade

[eager ivy]

its this simple lol maybe the imagen 2.0 could help me

[zinc valve]

Whenever I'm doing stuff like this I'd rather just target the web instead because I can give a real browser to the model

codex + chrome dev tools MCP

[eager ivy]

I mean I tell model to turn the raw iamge to the code and then I paste the code to taht box and it renders it

[eager ivy]

will it let him do what? if he has access to a real browser? he can already research stuff so what would it do?

[zinc valve]

The browser is so that the model can see what it is doing

It will make something, look at it, notice that it's dog poo poo and retry

can't believe the poop emoji is blocked here but I guess

[acoustic talon]

Given what you know, do you think the pro model would be able to properly list the reagent(s) of this reaction

Because 5.5 thinking gets it consistently incorrect

[eager ivy]

waitttt thats huge I need to implement it thanks for the idea

[zinc valve]

If you give it web access it might

I can't do that myself so its hard to gauge

But as far as my work is going 5.5 Pro seems really good

[acoustic talon]

What specifically in neurochemistry are you studying using the pro model

You said how things affect the brain but what level would you say youre at, undergraduate, graduate, etc

[zinc valve]

Mainly amanita msucaria usage and long-term effects. More specifically how microdosing it over long periods of time can result in changes in plasticity

[zinc valve]

I'm masters on an entirely different area

[acoustic talon]

That seems like research level stuff

[zinc valve]

For this specific case it's independent research, but I do use pro for my day-to-day research work which is in political communication lol

idk why I decided to do Interdisciplinary Sciences in college pls do not bully me

😂

[acoustic talon]

Nah thats respectable

[acoustic talon]

How do you know what its saying is correct

Or scientifically plausible, at least

[zinc valve]

I have some lvl of education on it and I'm using it mostly to cross reference papers I've already read myself

Or to build the fancy visualization stuff

[acoustic talon]

So you independently studied the discipline all the way to the level of research level of neurochemistry?

[zinc valve]

example of stuff I'm using it for atm

[acoustic talon]

Is that images 2.0

[zinc valve]

No, I did study it in college. Just not enough to be a neuroscientist lol

Bad thing about Interdisciplinary Sciences is that you get to study a lot of different stuff but never enough of it. I went down the applied computing (mainly in politics) hole, but I still like to exercise the little knowledge I have about other stuff.

Not for scientific publishing so I feel comfortable on simply doing my own thing and pursuing whatever topic I feel like atm if I have some time to spare

[acoustic talon]

Thats very respectable

[zinc valve]

• pro yeah. used it from a workspace containing the papers lol

[acoustic talon]

How would you say the pro model has helped you in ways the thinking model couldn't

[zinc valve]

Honestly I don't think I can answer that directly because I didn't compare both in a lot of my tasks. It's just a mental algorithm of "if I think this is very hard I'll use pro". It's very likely that thinking could also do it.

[acoustic talon]

What made you upgrade to pro

[zinc valve]

More Codex

[acoustic talon]

The 5 hour usage limit was insufficient for what you were using it for?

[zinc valve]

Both limits, actually

Consistently ran out of both 5h and weekly limits lol

[acoustic talon]

Damn

Im asking because i was wondering if the pro model is worth it

[zinc valve]

to be fair the 5h limit on plus is enough for like 5-10 complex prompts

[zinc valve]

Personally I wouldn't say so unless you go for the 5x plan

20x is just too expensive

[acoustic talon]

I agree

[acoustic talon]

Im waiting until they lower the price a bit more lmao

They lowered it from 200 to 100, im hoping they lower it down to like 40 or something 😭

Regardless, the benchmarks for pro dont seem that much higher than thinking tbh

[eager ivy]

wdm I bouthg it for 200$ the 20x

1 week ago

who found a solution to not use a web chatgpt but an enviroment where he those not lag when outputting long messages and also an enviroment where long conversations are not lagging out

[azure sigil]

whole server just slopmaxxing

[acoustic talon]

Its 100$ for 5x

[acoustic talon]

They gave ALL of us gpt 1

[eager ivy]

you can just purchase it like x days before your plus sub ends while making sure the usage will reset in that time so you pay maybe like 50$ or osmething just 1 week of usage but better tahn nothing and 20x

[acoustic talon]

Thats the new strat?

Reminds me of a time when i did a trial for a 400$/month subscription on something with the intent of using it to the last second of trial expiry then cancel

But then i forgot to cancel it and got charged 400$ 😭😩

[eager ivy]

Meta

and buy it thuth play store cause you get points

for ex if you were to pay for plus plan for 10 months you would afford one plus plan cause of the play points

so thats another discount

[acoustic talon]

Sounds like youre on your 10th account doing this

[eager ivy]

wdm? no You just get 1% cashback when purcahsing from play store

I bought 200$ plan and got like 1000 points which is 2$ XD

[acoustic talon]

Can you do this multiple times on the same account though

[eager ivy]

wdm you just autoamcially get play point each 0.25$ you spend I think

so its not money glitch its jsut 1% cashback

[acoustic talon]

Right i mean doing this

[eager ivy]

every playstore acc has that feature I think so ye

[acoustic talon]

Youre a menace bro 😭

Saying this shi in the official open ai server is wild work

[eager ivy]

but they are aware of it

they collaborated so they know it haha

[acoustic talon]

They should make open ai a publically traded company in the market

And you get access to chatgpt based on the amount of shares you bought in the company

[eternal rock]

Server Tip:

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