chatgpt-discussions

[cloud pewter]

am not surprised

I speculate that this person has pro sub

[whole roost]

The app is a trash pile, they need to put more workforce into it

[cloud pewter]

at least it handles long conversations better than web

[maiden estuary]

Is anyone experiencing issues with o3-mini or even o1? If it gives you suggestions, and you ask it to implement those suggestions to an output you are working on in the session, you get a response that this may have violated the terms of use. If so, how do you work around this? I’m not violating anything.

[cloud pewter]

but it also:

jump to random part of a conversation (new bug, discovered recently)
can’t retrieve older tries if regenerated a response
the model selector is a mess

[vocal monolith]

Do we know anything about the release date for o3-pro?

[cloud pewter]

it isn’t visible for me all of the time

[tight island]

Probably

I would assume some grey area prompt was used

[cloud pewter]

yeah but, it’s pro, it’s supposed to be unlimited!!!!

[elfin lava]

Does anyone tried this GPT to actually do what it does based on description

[cloud pewter]

real question: why does this gpt exist

[fringe barn]

its wild 🤪

[whole roost]

i have another task for o3, lets see what it can do

wow i have a jewel now

i don't know, ask on deepseek server

[ionic kettle]

i paid for the plus but im not getting the plus on my account. anyone knows what i can do?

[normal quarry]

this channel is for ChatGPT discussions, talk about other AI in #ai-discussions and delete your message before a mod does

[whole roost]

you could try to reach out to helpdesk, they respond quite fast about payment stuff

[eternal rock]

`` Server Tip: ``

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

[ionic kettle]

where can i find that section at?

[whole roost]

not sure if i can post links even if its to open AI, i'll try

https://help.openai.com/en/

wooh, and then the button on the bottom right

[ionic kettle]

ive already went ont that one. LOL its been 4 weeks and no response

[whole roost]

whaaat i have 2 responses from yesterday

are you logged in there?

[ionic kettle]

i think they dont want to return my money

yeah i logged in and i try to log in and out multi times and no luck at all

[whole roost]

they should at least respond that they cant help you with that, but for me i usually get a reply within 2-12 hours

[ionic kettle]

yeah that is why i reached out to the discord because as of this moment, i spent 20 and its been the whole month already. no help no response and no PLUS on the GPT

[whole roost]

i don't think anyone here can help you actively with anything, we're all just users

[cloud pewter]

yeah an openai employee is so hard to find

it doesn’t even show the chat ui for me in the support page

[whole roost]

maybe write to helpdesk again instead of waiting, i really get replies about anything, i could tell them that i like toast for breakfast

the first reply is usually ai generated but when you ask further some human dude will come along

[ionic kettle]

yeah at this point im just looking for the refund at this point i still have the screenshots of the errer message when i try to restore the sub

[cloud pewter]

Wha happened to ur sub

got scammed???

not surprised tbh

[whole roost]

i think o3 fixed my script

o1 pro wasn't able to, interesting

[cloud pewter]

Lmao

[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.

[whole roost]

and my other script it can't fix, neither can any other AI, i give up

it probably needs a deepgpt5o+6 to get a simple text color highlighter going

[sullen shard]

why is chatgpt saving everything i say to its memory

[fringe barn]

because it wants to get to know you 😉

[cloud pewter]

because your memory isn’t full

[elfin lava]

That’s embarrassing

[elfin lava]

I just want that damn ChatGPT MoonShine feature now

[lavish vale]

it even feels like o3-mini-high is better than o1-pro

[elfin lava]

“The memory is getting smarter” feature

[lavish vale]

• faster

[elfin lava]

Does anyone have this yet

[cloud pewter]

nope

another drip feeding feature

[next veldt]

Lmao I asked o3 about the resources needed to construct a battlestation the size of the deathstar and it refused reasoning that it's unethical to provide such information on the grounds of not wanting to help build weapons.

You got me OpenAI, you've stopped my Death Star project and quashed my galactic conquest ambitions 😦

[elfin lava]

Lmfao

[devout robin]

Hi everyone,

I've been experiencing a persistent issue with ChatGPT, and after checking the OpenAI forum, I see that many users have reported the same problem for months, but there's still no official fix. I hope someone from OpenAI sees this and helps address the issue.

The Problem:
In the middle of a conversation, ChatGPT's responses suddenly disappeared, leaving only my inputs.

After refreshing the page, I got the error: "There was an error generating a response", and the conversation became stuck.

Any new messages after that would disappear upon refreshing.

The issue persists across different devices and browsers, even after clearing cache, resetting DNS, and other troubleshooting steps.

This isn't an isolated case—I've found at least 7+ forum threads with the same issue, including some dating back to March 2024.

What I've Tried (Without Success):
✅ Clearing cache and cookies
✅ Using different browsers (Chrome, Edge, Firefox)
✅ Switching networks (Wi-Fi, mobile data)
✅ Resetting DNS & changing DNS servers
✅ Disabling extensions & VPN
✅ Exporting data and starting a new conversation (new chats eventually break too)

Why This Matters:
Users on both free and paid plans (ChatGPT Plus) are affected.

Some users have lost important conversations due to this bug.

OpenAI support acknowledges the issue but has no timeline for a fix.

Has anyone else experienced this? If OpenAI staff is reading this, please escalate this issue—it's been ongoing for months with no resolution. Any help would be greatly appreciated!

[cloud pewter]

is it a long conversation??

[elfin lava]

I actually got this often now recently

But mine was in AVM

[cloud pewter]

if a conversation is long it also gets lost

(me had suffered thru these)

[cloud pewter]

o3 mini sometimes overthinks and doesn’t handle translating very good

[eternal rock]

Get notified for server events!

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

This is a Safe for Work Discord Server

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

Posting NSFW content will result in a permanent ban.

[cloud pewter]

most of its chain of thought seems like filler content tbh

[rocky kayak]

It also appears that there is a restriction in place.

[cloud pewter]

maybe I should try to let it create me a website because apparently it’s just that good at coding???

[rocky kayak]

To the best of my knowledge, Anthropic's Claude excels in website creation, at least based on my observations, as it can design visually appealing sites without even utilizing CSS.

I believe that is the case, or at least that is my understanding.

[cloud pewter]

seems like it got lobotomized at least in translation in order to get good coding performance

[cloud pewter]

Claude has its usage so poor even it’s paid tier doesn’t get much

[rocky kayak]

Do you have Plus or Premium?

[rocky kayak]

No, I have created some well-crafted artifacts with it, although they are relatively simple.

[cloud pewter]

plus

[rocky kayak]

You using o3-mini or o3-mini-high

cuz its an option

[cloud pewter]

it ain’t even Japanese 😭 😭

[cloud pewter]

I tried both, but they all overthink and gets some words translated wrong

[rocky kayak]

It tends to overanalyze, but the most frustrating aspect, at least from my perspective since aesthetics are important to me, is that its thought process is sluggish.

[cloud pewter]

while 4o one shots it

[rocky kayak]

4o cant reason as well tho

so far 01 is Your best option

Another reason O3 might be subpar is that it was released earlier than intended due to DeepSeek, though that is also why they made it free. The complete version of O3 should be available by next week, I believe.

[cloud pewter]

o1 does it very well, even adds touch from my custom instructions

but the limit determines that I won’t be using it so much for works like this

o1 is the most impressive one to me as of now

[rocky kayak]

yeah

[cloud pewter]

tried Deepseek too, it doesn’t overthink, the cot is well structured but the results look mid at best

but it’s free ¯_(ツ)_/¯

[devout robin]

It’s very long, there’s a whole story out there, and it’s a pity to lose it, I just can’t move ((

[rocky kayak]

it its just less accurate soietimes

[cloud pewter]

so far:
4o: not a reasoning model, does it very well, the translations are pretty accurate
o1: thought for 9 seconds, exponentially good and adds touch from the style requirements of custom instructions set by me
o3-mini: thought for 1m 28s, cot is messed up, decent translation but gets some wrong
o3-mini-high: thought for 1m 23s, cot also messed up, translation mid at best (some of the words kept English when it could be translated)

[cloud pewter]

well, it’s just the common result of an extremely long chat
have you encountered this error with it?

[devout robin]

How it all started and the mistake itself

[cloud pewter]

yeah, on mobile it would show the thing I posted

Then the chat will become invisible
you better copy down the link NOW otherwise it’ll become inaccessible

thread size at its peak
#1333435624686092378

also did the website become so laggy

[rocky kayak]

I am unsure whether the lag occurs solely on the O3 or if it affects other models as well, since I have the application on my PC, and the O3's thinking preview appears to be sluggish.\

[devout robin]

The error started with the disappearance of messages, and then the message “Error generating response” had appeared. If a message had immediately appeared that “the limit has been exceeded,” then I would have understood, although it would also have been annoying.. in that dialogue I created a whole world... and it really hurts me to lose him😢 When transferred to a new window, all the atmosphere is lost😭

[cloud pewter]

it’s a general problem that occurs when the chat gets very long

[cloud pewter]

Ur not alone bro

anyways sleep time, if u have more issues u can dm me

[tawny mirage]

I asked this yesterday but has anyone else noticed pro mode is kind of dumb lately? Sometimes I'll get the old workflow, others I'll get a 6 second response at most, and the response is only about o3-mini quality.

[eternal ridge]

chatgpt has became trash. it doesn't answer my question directly, instead it tells me to search the web and search on forums. although i asked it the same question 3 months ago and it did give me the answer.

So now I have to click the "Reason" button?

[tawny mirage]

Why are they saying o3-mini-high is any good?
It can't keep its thoughts straight and repeats information unnecessarily.

[eternal ridge]

cuz dementia

[tawny mirage]

If o3 is like o1, then o3-mini is overdistilled.

[eternal ridge]

well its mini, what did you expect

[tawny mirage]

I'd expect more given the evident limits placed on it.

[half vine]

is o3-mini free?

without a limit

[native otter]

Only if you are on Pro

But Plus get a high limit

150 messages a day for o3-mini (which is the medium variant)

o3-mini-high, which is closer to o1, is 50 messages a week (But not shared with o1)

[regal portal]

closer to o1, meaning still inferior to o1?

[native otter]

In Benchmarks it's a bit better, but real world is different
Like in coding, it's better (even o3-mini is on bar), but other stuff o1 seems to perform a bit better honestly...
Also as o1 has been around for longer, it's more integrated with support of Canvas, Custom Instructions, and Image Analysis, while o3-mini and o3-mini-high currently don't as they are very new

[next skiff]

hi

[native otter]

👋

[last turtle]

OpenAI should publicly announce this, got my limit usage message already after one day, I thought its 150 messages per day like their article implied it.

[native otter]

Yea they honestly weren't clear about o3-mini-high limits...
We should have a message counter, as currently it's guessing, unless you count every message you send manually lol

[last turtle]

I do this, I keep records. 😁

[tawny mirage]

o1 itself in ChatGPT seems reduced, too. Are they trying to cripple the service some to reduce operational costs due to opening o3-mini to free tiers?

[last turtle]

I let o3-mini-high program a puzzle game inspired by "Brix", it achieved it after 15 prompts. Install Python and type "pip install pygame" in the command line. You have to make a line with 3 of the same images to let them disappear. I only reached the 8th of the 10 levels, lets see if you are better. Have fun!
#1070955161431851068 message

[plain arrow]

Anyone else seeing 4o's horrific decline in writing quality since about Wednesday of last week? (ie sentence fragments, randomly bolded words and clauses, inability to maintain continuity, etc)

[hoary gazelle]

I see this document edit feature very disruptive, i dont want a text editor included, just give me old fashion gpt. This is not word

[fringe trench]

yipee!

[river panther]

I'm actually having some issues with it today myself. It's refusing to follow instructions.

I'm trying to get writing done, and I have specifically instructed it to NOT cut paragraphs short in it's readout and analysis.
I post in 3 paragraphs at a time and it reads out and analyzes the full 1st paragraph, then heavily cuts short everything from paragraphs 2 and 3

[winter vine]

then o3-mini low is only available on API right?

[native otter]

Yes

[winter vine]

nice, then we have o3-mini medium : D

[plain arrow]

Ditto, even when I prompt it not to use sentence fragments and bold formatting it seems to break midway through the first response, let along 2nd+

[rugged igloo]

Yooooo it's finally here

[winter vine]

what is that?

[rugged igloo]

Hopefully it's deep research like what Gemini has

[winter vine]

oh nice ty

[potent saddle]

but its a lot cheaper than o1

[potent saddle]

when is it coming out though

[rugged igloo]

Have to watch the live stream it was just announced lol

[native otter]

Yea it's a lot cheaper, but very efficient
Of course not same Calabar in everything, but it's a mini after all

[nocturne ember]

So deep research gpt 5?

[silk flame]

hello , I am experiencing a weird problem with AVM recently. Sometimes chatgpt repeats itself twice when repsonding... it worked perfectly before but since couple days it works terrible. Has anyone had the same problem?

[native otter]

Check if there is anything in memories and custom instructions that influences the behavior

And it can sometimes glitch, make weird sounds, scream suddenly, speak after long delays...

Other times it's a network stability issue

[native otter]

Isn't that similar to o3-mini with websearch?

[potent saddle]

but more deeeeep

[native otter]

I want files

Files can't be this hard right? Unless they want to implement semantic search and RAG like 4o does with Custom GPTs...
Because it should be just a part of the prompt it self without any of the fancy features

[rugged igloo]

Deep research with o3 mini

[rugged igloo]

Yea but o3 mini with internet only does a search depending what you want for maybe a min or so this deep search can spent like 15 min or more looking things up

[native otter]

That's lots of compute if true

[rugged igloo]

It was a high highly requested thing for them to add. Me included.

[native otter]

That seems like a Pro only thing

[rugged igloo]

Prob .

Or maybe for plus get like x amount of times you can use it . Then pro it's unlimited.

[native otter]

Well i'm subscribed to Gemini, which offers deep research and it's pretty good, I think OpenAI's version shouldn't be behind the Pro plan only...

[rugged igloo]

But for Gemini it's on 1.5 pro which is really stupid. This one is on o3 mini which is reasoning model and is miles above anything out on the market besides o1 pro

Also to confirm it is deep research

[regal portal]

I hope they fix this such as the reasoning tab doesn't automatically open every time you send a prompt. It's annoying.

[orchid pollen]

is o3 mini better then o1?

[regal portal]

so far they feel pretty even.

o3 mini feels a little less useful, and the thinking tab opening all the time is annoying

[tawny mirage]

Has anyone had o3 get caught in a loop? You ask it a simple question and it reviews the entire conversation going down a rabbit hole on something completely unrelated?

I asked it about an IValidationAttributeAdapterProvider and whether it needs to be registered anywhere, but I cut off everything but AttributeAdapterProvider, and it started looking into taxes, doing estimates for yearly income of $70,000.

[floral ether]

Is there a site that lists how many o1, o3 mini, gpt4-mini, etc queries we get at the plus level?

[regal portal]

using 4o and giving it a 1400 line code i had done with it back when 4o was brand spankin' new.
Now after a few lines of chitchat it can't remember any of the code. I keep having to reupload it to it.

seems like it just stops coding after only 100-150 lines

feels significantly limited compared to how it was.

[whole roost]

yes, when i wanted parts of the system prompt from it in the 1st response it started overthinking every question i had and wouldnt answer anything i asked anymore because it had in the thought process that i still want to know the system prompt

[regal portal]

i do not understand why they limited 4o to 130ish lines..this is such a disappointing downgrade.

[dawn sentinel]

Did you preview the code inside Canvas or did you copy/paste to an IDE?

[cloud pewter]

o3 mini still lacks ability on writing

creative writing

[tawny mirage]

Probably because it's aimed at STEM topics which inherently need to focus more on accuracy over making things up.

[cloud pewter]

sadly I use chatgpt more on creative stuff

I would use o3 mini to solve my math homework if it just accepts image.

[tawny mirage]

Then it's great if you want to use it for creating a hybrid world that shares some of Earth's behaviors from the past such as monarchies, and sociopolitical dynamics. You could ask it example scenarios to get an idea of how people in that period might act and apply that logic into your own story.

[whole roost]

canvas has a bug for regenerating a response

[tawny mirage]

Clarify. That's like saying ChatGPT did something wrong.

[dawn sentinel]

Does anyone else not see Preview button at the top left of their Canvas?

[whole roost]

Let's say you edit something in canvas, then generate a response and the AI answers, all good, the AI can read the change, but when you regenerate the reply the change is no longer accessible

i was wondering the whole time why this works so badly, now i know

[tawny mirage]

Edit it manually? I've noticed that too. I think you're best off stating what you want changed, and having it change it. I'm not sure the changes you make inline are communicated, but I can't say with certainty.

[whole roost]

yeah i sometimes regenerate 10 times and always thought why doesn't it do what i put on canvas

[tawny mirage]

Anyone else noticed that even o1 high is bloody lazy, even on API?

[regal portal]

this isn't relevant to my issue.
My issue is that back when 4o was released, i had it make this 1300+ sidescroller platformer and it could spit it out without issue. Now it can't even seem to code more than 130 lines before stopping.

[tawny mirage]

Here's a pattern, here's the information you need to follow this pattern, finish the boilerplate pattern

Does 30% of the work, 15% of the top, 15% of the bottom. Adds comments // and so on for parts x-y. 🤦‍♂️

[cloud pewter]

it would show error if I let it edit more than 2 times

[whole roost]

it's not supposed to edit anything

it simply has no idea what's on canvas, it isn't loaded

[tawny mirage]

Sounds like a candidate for #1070006915414900886.

[whole roost]

i handed it in via help desk 3 minutes ago

[rugged igloo]

Yoooo this isn't even this yet . As it's to be announced still

[cloud pewter]

if it’s for mini it has to be available to plus

heck, even free

[true nexus]

Ayyy

I guess the 1% of users get to enjoy deep research for now

[true nexus]

Until we get access to a more "watered" down version.

[rugged igloo]

Ouch plus only get 10 queries a month

[true nexus]

You'll have to prepare your prompts using 4o or o3

So you don't waste it.

[true nexus]

This is separate from deep research and just solely for o3.

[cloud pewter]

how is Sam saying two numbers it’s so confusing 😭

[cloud pewter]

agent???

“next agent release will be available for plus”

LIES!!!!😡

[true nexus]

The tweet "(note: this is not the "one-more-thing" for o3-mini. few more days for that.)" is not on that thread.

Hence it's gonna be something else

[cloud pewter]

they must be thinking of: oh Deepseek is putting so much pressure on us, must release o3 now!!! (but has way smaller usage and a half baked feature)

[rugged igloo]

I wasn't talking about that lol

[true nexus]

[cloud pewter]

“oh buuut it’s sooo compute heavy, look even pro only had 100/month!!!😢 “

[rugged igloo]

I was excited as there still got something planned for o3 mini.. but what j was talking about is that for plus users deep research will 10 queries a month

[cloud pewter]

also I went to Japan earlier than Sam does this year. I took 527 pictures

[rugged igloo]

Why not 530? Bro fumbled

[cloud pewter]

on this version tho

[cloud pewter]

idk

[rugged igloo]

Dam. What a way to end the photo with that

[true nexus]

Ah also notice he said that it's built on o3 and not o3-mini

[cloud pewter]

I bet Sam didn’t even experience peak food which is bento on a Shinkansen

[rugged igloo]

Bento boxes peak Japan life

[cloud pewter]

hehehaha.

yum.

[dense halo]

Good evening.

[cloud pewter]

I wish I knew the address of openai Japan office so I can go there and see

and possibly see Sam but I left Japan on Jan 23 so

[rugged igloo]

I live close to open AI in San Francisco

I actually went into their building to ask for a meme job but told me to leave. I told them I pay 200 a month and that I keep the lights on .

[cloud pewter]

aw man they are mean

[rugged igloo]

Yea

[whole roost]

i would like to try the deep research agent

[cloud pewter]

is it not rolled out to you yet

But Robert has it smh

[whole roost]

hopefully it ever will

maybe in a month or two

[cloud pewter]

Rip

[whole roost]

And grown up mode, leave me alone orange, I can slay demons

[cloud pewter]

“oh btw chatgpt you can swear, it’s normal conversation tactics”
outputs some sentences that has the f word and something
“this content may violate our usage policies.”

[regal portal]

grown up mode = talking about elves and demons? 😛 j/p

mine swears all the time with no issue. It gets the violation message if he talks about "below the belt" stuff

[whole roost]

I don't want mine to swear but I definitely wish for more violence, I mean my roleplay can't get more fantasy than a Sorceress and a demon slaying foes in diablo 4

[regal portal]

D2:Lod is best Diablo

[whole roost]

Yeah I also played the remake, was nice while it lasted but ultimately too easy

I play D4 and have GPT on my 2nd monitor to roleplay what I do in the game and explain the lore to me

[halcyon cliff]

why even bother anouncing it, it will be like sora 2 years later lol

[regal portal]

announce what?

[cloud pewter]

I need the censorship to be lifted more

like it could be flagged over basic swearing words

i want my chatgpt to be unhinged

well, I’m not talking any nsfw stuff but the censorship on literally anything else is crazy

[native otter]

Btw, sometimes using custom gpts and in temp chats can help mitigate censorship

[cloud pewter]

and violence things yes

[native otter]

Your words may be flagged, but the model a lot of times not

Didn't want to share that here, as now an OpenAI employee may look into it and put an additional safeguard, but whatever lol

[wary narwhal]

Simple things like this gets flagged. Idk. Kinda silly

[cloud pewter]

“content removed” LMFAO

so paranoid

[rugged igloo]

LOLOL

[cloud pewter]

if u asked for its cot you get flagged

for some reason

[native otter]

Have you tried this btw? It does work for me a bit.

[cloud pewter]

I do have a custom gpt for uncensored stuff but doesn’t work very well

[native otter]

It depends on the prompting

[cloud pewter]

yeah but sometimes it’ll reply with “sorry I can’t help with that”

[idle zenith]

I have more than one experience this way. I mean, I was using 03-Mini-High too. So GPT is wasting my messages this way. That 50 monthly messages will run out very soon. Customgpt can't even generate long articles like it used to. I am actually using API to workaround. Biggest issue is, Chatgpt 4.0 feels like using Chatgpt 3.5. It even halucinate.. Cannot do anything slightly complicated..

[cloud pewter]

it really does it sometimes, and it’s hard to determine because of the transparency

[idle zenith]

It was just fine before. I have a routine of pushing it to its limits generating reports.

[unborn stump]

Hello, I am a pro subscriber and I dont see option to use the new deep research. Can someone help?

[native otter]

Yea that's not how to work with an LLM

[native otter]

They said later today

[idle zenith]

Please educate me. What is the right way? I have a routine of pushing customgpt to generate a bunch of huge articles. My prompts are so long that I need to have multiple follow up prompts. But last week. I used an excessive number of follow up prompts and still not done.

[cloud pewter]

it’s being rolled out, just wait for a bit

[unborn stump]

ok ty

[cloud pewter]

I work with chatgpt this way too, sometimes it work sometimes it’s just dumb, deal with it 😭

[idle zenith]

for 6 months straight. Every week. works very well. But this week. It just stopped altogether. I had to workaround it. So yes, I understand now. It can change without notice. And we supposed to reply on AI. I guess not

Actually no. it is very bad right now. It really seems that nothing is consistent since a week ago

[cloud pewter]

sometimes it changes model depends on your ip

which plan you on?

[idle zenith]

Plus

[raw portal]

I'm not seeing a deep research button on my chatbar on pro mode. tried logging in on a new browser

[native otter]

Actually you shouldn't use huge prompts with 4o, it does work nicer in o series, but not 4o

You should try maximizing focus by writing in markdown, clear words, and enough context, and having the dataset it self structured well

Also encourage it to use the sources (the files) given, never make up stuff, citing sources, and in the prompt you may want to include '[Search Files] '

[idle zenith]

I happen to have a practical use case using only Customgpt and GPT Plus. This involved generating 14 long reports every week. Long as in 10000 words per report. Yes, this is a bit crazy. I was able to push it to generate at least 7 reports with a few prompts. But since a week ago. OMG. It literally feels like using 3.5. IT didnt even read half my prompts. It never happened this way before.

I have been doing this every week for 6 months

But what I notice is that when using other services like ChatLLM. It still works like a charm. Actually it works better than GPT plus. If Customgpt isnt doing what I need it to do. Maybe I don't need GpT plus like what I thought that I do.

[cloud pewter]

they do be doing some shadow strategy…I’m plus too and sometimes it really depends on your ip and stuff

nobody knows why sometimes it just feels lobotomized

[idle zenith]

Lesson learned. Cant depend on GPT if you are trying to make money. It can stop at any given time...

[rugged igloo]

4o sucks at long detailed prompts same with custom instructions. It just ignores them

[idle zenith]

I am actually making money with this gpt too

[idle zenith]

I think Customgpt isnt even 4o.

Maybe they upgraded it. I thought that it was using 4

[rugged igloo]

No it been 4o for a long while now

[idle zenith]

Then 4.o sucks with long prompts is incorrect then. My prompts are so long that I need to link a txt file just for this prompt. It works until a week ago. Works for 6 months.

it is now

[native otter]

• https://platform.openai.com/docs/guides/prompt-engineering
• https://help.openai.com/en/articles/8843948-knowledge-in-gpts
• https://help.openai.com/en/articles/8868588-retrieval-augmented-generation-rag-and-semantic-search-for-gpts
• https://help.openai.com/en/articles/9358033-key-guidelines-for-writing-instructions-for-custom-gpts

[idle zenith]

This is a live, practical routine that I do every week. At least this is my experience. Works until last week

Feeling bummed out ....

[native otter]

Yea as mentioned, you shouldn't do that, 4o doesn't like extremely long prompts or custom instructions

You should be building your GPT in a RAG optimized way for large datasets, which the help files explain
Also read on how to write custom instructions for GPTs

All linked above

[idle zenith]

Currently, my workaround is using gpt 03-mini-high in ChatLLM. Yes. it took me like 30 messages to get it done. At least I got it done. I will learn about RAG. Thank you for the pointer.

[hollow sage]

Same, Pro user, I don't have access yet.

[untold blade]

I have been using GPT-4o, I see o1, o3-mini & o3-mini-high.

Which one is my next best one to use? General purpose for now

[abstract pendant]

I noticed when 4o has been prompted to use a less formal writing style, any time the subject goes into efforts to accomplish something that might not work without decent effort, it loves to work in saying "don't do things half-assed" pretty often.

[cloud pewter]

4o still, you have the most usage on it and won’t get throttled after 50 prompts

[noble gull]

So.

What is the point of scheduled tasks?

Like if I want to have it be an alarm I really can't. Because it's just a message.

I guess I coukd have it summarize major news of the day, everyday but that's about it?

[hollow mantle]

I had hundreds of hours of work in my projects folders… today all project folders disappeared. All of that info lost.

[native otter]

You could have it write a motivational quote, search the web for stuff like news, have it remind you of you of tasks to do / stop procrastination, or just stupid stuff like have it writing a horror story every night based on the main events of the day like weather and news :V

[noble gull]

Seems kind of useless without like it using voice automatically or actually sending you a notification

[cloud pewter]

it’ll send you an email and message but that’s basically it

And it’s 4o mini based so it doesn’t have much intelligence

false advertisement at its peak

[sick pecan]

I tried summarize news of the day and it repeats most days. I'm unsure of they're is any practical use for scheduled tasks

if it ran custom gpt action it would be infinitely useful. but as usual openai don't integrate with custom gpt actions

[vestal hedge]

Bruh deep research is gonna be either amazing or terrible

[teal spoke]

look like us only

[shell geode]

Time to use an alternative for a month bc it's getting super strict again with the rules and flags things that aren't against the terms of service again

[willow oar]

Is deep search a pro feature or for free users

[chrome citrus]

its extra pro

[willow oar]

?

[rugged igloo]

Prompts when I say I mean custom instructions . It ignores them. And the longer the conversation the more 4o ignores what you asked it to do

[rugged igloo]

It's for pro users only. And plus users will get it soon for 10 queries a month

[tawny mirage]

Will pro have a limit?

[willow oar]

ah funky, and i presume free users will get it in like 1000 years

[rugged igloo]

100 a month of queries

[tawny mirage]

Oof

[rugged igloo]

I agree

[willow oar]

Come on thats 3.2 asks a day

thats a lot

[wraith dagger]

by soon do they mean its in a testing phase for pro right now or is that just soon

[rugged igloo]

It's rolling out to pro users today . And later date for plus users

[wraith dagger]

Alright

[willow oar]

And never for free users :)

[rugged igloo]

Free users are getting it to but they didn't say how much they will get

[tawny mirage]

Depends on how productive you want to be.

[rugged igloo]

If 10 per month for plus I expect 5 per month for free users

[tawny mirage]

If they give five per month for free, there'd be a reason to use a VPN. 😅

[willow oar]

You can do a lot with 3 in depth asks a day that gives you really good thought out and researched answers

just gotta not waste em on "How much does a candy bar cost in america 2025"

[tawny mirage]

I can do more with more. It's all perspective. I'd use it for market research, deep dives into psychology for user trends, things like that.

Stuff I could do, but it's boring and tedious.

[fading kettle]

hey, is o3-mini free anytime?

[plain crane]

is deep research only coming for the full o3 model?

[rancid eagle]

how can i access deep research? i have pro

[tawny mirage]

I think they're doing a staged rollout. Not everyone gets it at once, most likely so they can review the impact it has on their servers and allocate resources to scale.

[cloud pewter]

wait until it gets rolled out to you

[cloud pewter]

Nah that’s either too low for plus or too high for free

[rugged igloo]

Sam said 10 per month lol

[cloud pewter]

for current version, if it ever hits the plus sub

[tawny mirage]

Yeah, even pro gets 100 per month. Difficult to say if that's a long-term limit or not. I remember the 4o limits seemed to last forever back when it came out.

Given Deep Research has such limits, I imagine o3 once it releases in full will have fairly heavy limits on it.

[cloud pewter]

10/w for pro😔

[untold blade]

Thanks

[nocturne wolf]

Wonder where they got the inspiration for "deep" research from 🤣

[finite vault]

does the follow-up clarification query count against your 100 monthly deep research queries?

[elfin lava]

This is so funny to me lmao

[cloud pewter]

CHEATERRRR

[native otter]

Benchmarks are liars at this point

o3-mini-medium better than o1? sure bud

[cloud pewter]

it’s just a marketing tactic

[ebon ermine]

whats a good way to implement autocomplete? anyone have any good papers on this?

[spice star]

in a few situations it is better

[last turtle]

I just copy and pasted it into an editor and made it a .py file.

[graceful pumice]

Oh dear. the free reasoner model admits it is biased in its economic answers when confronted. You need to milk it more to be in line with current research. Makes me question its usefulnes on everything.

From a cursory look it seems to be biased towards human thinking instead of the current economic research.

[cloud pewter]

tried comparing o1 to o3 mini again with a request that has multiple words to write a description of:
o1 thought for 58 seconds and gave me the thing I wanted, although it missed some words from the original prompt
although o3 mini only thought for 15 seconds it tells me that it would be too long to generate such a thing, and tells me to narrow down.

well, o1 supremacy i guess

and I enabled search for o3 mini too

also tried 4o; the content is great but it often either fails on the required text amount or outputs in parts

[lost rover]

Is it just me, or is 4o not following "custom instructions?" for example, I put "no emojis" but 4o be like 🚀 🚀 🚀 🚀 🚀

What a rebel

[cloud pewter]

same

[elfin lava]

Ai rising up

[last turtle]

Try "no icons".

[rugged igloo]

4o was neverr good at following custom instructions

[elfin lava]

That’s why I inject my custom instruction directly into to=bio

[rugged igloo]

yea that works better

[cloud pewter]

memory also works better

[leaden ingot]

The fact that deep research is not coming to the EU is total BS

[dapper patio]

Deep research not available in the eu? Or still being rolled out or such?

[dapper patio]

just saw that.. yes that is bs

[leaden ingot]

they cannot complain about EU regulations, as Google’s deep research has been available in the EU from day 1

[dapper patio]

Agree, but EU should also remove the stick they got up their.... and stop over regulating

[leaden ingot]

totally

[vocal monolith]

Wait... seriously? I am in the EU and was planning to buy pro today, should I wait?

[dapper patio]

oh pro only.. hell no i'm not paying 200 a month lol

[rugged igloo]

idk , i dont what people have over in the EU as they hate technology advancements

[cloud pewter]

yeah wait until more features release

[vocal monolith]

Thanks. Do you know if what Sam called o3-pro in his tweets is a different thing than the deep research? I am so confused with all the names xd

[elfin lava]

Gay right 🏳️‍🌈

[cloud pewter]

yeah it’s different

openai’s naming problem

[coarse yoke]

Wow that was exactly the feature I was looking for! Almost perfect timnng honestly
This Wednesday will be my last presentation that I put this much effort in forever hopefully! 🤞

[cloud pewter]

wha

[coarse yoke]

The amount of research you would need to do and find the right images, make it interesting

Deep research can just do it for me

Like now I need to plan a 1 hour presentation and it's nuts to do in 2 days of work
But with this it should be easy as hell

[ornate nest]

how do you open deep research

I cant find it

[coarse yoke]

Purchase pro

[ornate nest]

I have it

[coarse yoke]

There should be a button

[ornate nest]

where

[ornate nest]

in which model

[coarse yoke]

From their livestream yesterday

o4 regular

[ornate nest]

it doesn't show for me

[coarse yoke]

You got those 3 dots?

[ornate nest]

yes

[coarse yoke]

It could still be releasing sometimes it can take time

Are you from eu?

[ornate nest]

yes

[coarse yoke]

Then it makes sense, wait for an announcement that it is coming to eu or see by the end of the day if you get it

[rugged igloo]

well its monday 12:11 am and i still dont have it

[whole roost]

great they changed a few things and all my scripts stopped working

[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!

[cloud pewter]

Bro has the secret access

😔 😔 😔

[whole roost]

What secret access, I don't get any of the new features

[cloud pewter]

me have the money to subscribe to pro when?

[whole roost]

work harder like me works almost nothing

[cloud pewter]

or yell “MOMMMM!!!”

[rugged igloo]

i know exactly what i am gonna ask it when i get it

[cloud pewter]

same

[tawny mirage]

Anyone figured out what causes o3-mini high to get confused and start going on a journey?

Hah, maybe that's the thing. It's high.

[whole roost]

they refuse to answer my helpdesk requests now because my mail address doesn't exist, i cant believe they have such issues

Never change your Google mail address, the services suck at updating it

[cloud pewter]

lmao

[tawny mirage]

It got 355 steps into the request (about IRS forms) and it went on an AI journey about how it can help in healthcare and several other sectors.

[rugged igloo]

holy crud

[tawny mirage]

Yeah, 355 paragraphs.

[cloud pewter]

truly an overthinker

[rugged igloo]

take notes

[cloud pewter]

lol true

[tawny mirage]

Even though I reloaded the page, it still is animating as though it's working. I wonder if some day it'll return from its AI journey.

I hit my notify me when AI costs hit a threshold. I've passed that total and it's made me want to cry. o1 pro mode seems to have been lobotomized, which is unfortunate.

[rugged igloo]

using all the compute :LP:

[cloud pewter]

U have pro?

yeah they lobotomize everything

[tawny mirage]

Yes, that they've almost crippled it makes me a bit annoyed. Deep Research and the upcoming o3 better be worth it.

I use it for work, so it helps me and has paid for itself (somewhat). That they're forcing me to use the API in the Playground is vexing, to be sure.

[tight island]

That’s because they want the good stuff for themselves

[cloud pewter]

yeah that too
and for “safety reasons”

[tight island]

“Safety” is very subjective

[minor hound]

and VPN is no option?

[tight island]

Depends

[tight island]

If you use vpn for the sake of overcoming the region lock they’ll ban you if they find out

So it really depends

[minor hound]

so no option 😄

[tight island]

But if you travel you don’t need a vpn

That’s what I do

[minor hound]

next flight outside eu takes me 2 weeks

but also fair point

[tight island]

It’s the legal way of overcoming region lock

“Legal”

[cloud pewter]

True that, but for openai it’s just an excuse to lobotomize

[tight island]

They did partner with the us gov

[rugged igloo]

so this is a problem

99/100 wow

[cloud pewter]

WOW!!!

[rugged igloo]

gg

ima try again

[rugged igloo]

bruh ty for tht

[cloud pewter]

another chatgpt pro meme 😔

[rugged igloo]

hella upset

post this on twitter

[frank wolf]

is O3 mini high the deep research?

[cloud pewter]

Lmao

[whole roost]

o3 takes the custom instructions very seriously

[rugged igloo]

yea its a resoning model

[whole roost]

o1 didn't always go prime evil style

Especially with coding

[elfin lava]

Tell that to Sam

[eternal rock]

`` Server Tip: ``

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

[elfin lava]

ok and

[whole roost]

Why are you here if you hate AI? Just curious what brought you to the server, I'm not on an ashes of creation server either

[rocky condor]

Does anyone know when deep search comes out?

[whole roost]

So just to get rid of your opinion?

I hope you don't have too many opinions or you'll be a busy human

[elfin lava]

For real

[rugged igloo]

it worked after 5 tries

@cloud pewter

[cloud pewter]

5 TRIES 😭 😭

[rugged igloo]

yea kept bugging out

or not doing it

[cloud pewter]

o1 is better on that to me tbh

[cloud pewter]

is the result any good though

[rugged igloo]

31 pages of text or

thats without double space

[cloud pewter]

it churned them out all in once in one message? wouldn’t it make the chat laggy

[rugged igloo]

yea it was laggy when it was outputting 34 pages of words in one chat but it did it\

[cloud pewter]

is the content any good, like does it have actual depth instead of just filler content that chatgpt usually does

lmao they really have to optimize the website

[rugged igloo]

i mean i can semd you the word doc if you want

[rugged igloo]

https://chatgpt.com/share/67a08d0d-0a24-800e-a7a9-68d1d9ffe575

[maiden loom]

Wow, I wish there was Deep Research when I was a Wikipedia enthusiast contributor. Not a spare minute for it anymore, but if I ever return, man that will be some good articles.

[cloud pewter]

give me the doc ig

dm me a txt file if you can

it’s so long damn I can’t really tell if it’s good or not

[rugged igloo]

gotta convert it into a word doc its much better format

[cloud pewter]

I used pages on my iPhone to automatically convert it into some format

[cloud pewter]

especially when it’s on a topic like this

[rugged igloo]

rip

[cloud pewter]

1984

not even txt

[rugged igloo]

it has a source page at the bottom

prob could if you ask it to

ask me what you want to have it do and ill do it

[cloud pewter]

W

[novel pebble]

Hey! I’m having trouble with a zip file upload.
I get the message: "File contents may not be accessible."
The zip isn’t password-protected and just has some text and config files. Any idea why this happens or how to fix it?

[whole roost]

I want the feature as well, I need it to do thorough research about diablo 4 lore

[rugged igloo]

dmed you the word doc

[rugged igloo]

well maybe if we got more specific could do better

Science fiction experienced notable developments and diversification after World War II. American and British authors, editors, and publishing houses shaped the genre differently, reflecting varying cultural, social, and historical influences. The period from 1945 to 1960 saw new themes, styles, and publication practices emerge on both sides of the Atlantic.

Key Questions and Requirements:

Historical and Cultural Backdrop

How did post-war events, societal attitudes, and publishing norms in the United States differ from those in the United Kingdom during this period?
In what ways did the Cold War, nuclear anxieties, and technological optimism influence the themes of science fiction in both regions?
Authors, Magazines, and Notable Works

Identify major American and British science fiction authors (e.g., Isaac Asimov, Robert A. Heinlein, Ray Bradbury, Arthur C. Clarke, John Wyndham, C.S. Lewis) active between 1945 and 1960.
Compare the focus and editorial slant of key American pulp and digest magazines (e.g., Astounding Science Fiction, Galaxy, The Magazine of Fantasy and Science Fiction) with their British counterparts (e.g., New Worlds).
Examine the literary styles and predominant themes (e.g., space exploration, scientific invention, social satire, post-apocalyptic visions).
Stylistic and Thematic Contrasts

Outline differences in tone, scope, and narrative complexity.
Discuss how American science fiction might have been more action-oriented or technologically driven, while British authors might have leaned on sociopolitical allegories or philosophical introspection. Provide evidence from specific stories or anthologies.
Include any commentary on language use, character archetypes, and plot construction.
Literary Criticism and Scholarly Perspectives

Summarize the main arguments from established literary critics or historians who have studied transatlantic science fiction during this era.
Cite relevant academic sources (books, articles, dissertations) offering comparative analyses.
Citation Format

Wherever possible, quote or paraphrase from primary and secondary sources, providing accurate page references (e.g., “As Astounding Science Fiction editor John W. Campbell argues, ... [Campbell, Astounding, p. 45]).
Include page numbers for all direct references to print sources, journal articles, or collected essays.
Depth of Research

Investigate interviews, newspaper archives, or author statements from the 1945-1960 timeframe for firsthand insight.
Draw on digital libraries, reputable databases, and specialized science fiction archives (where available) to ensure breadth of coverage. ----this was the prompt----

yea with a more better way of prompting could do a better job like APA format or mla format with page numbers ect

i feel like when it asks "how much time do you want this to get done" doesnt really mean much as i said go as long as possible but it doesnt go for that long

[eternal rock]

Get notified for server events!

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

[crystal tapir]

Why is ChatGPT reasoning my prompts? And the answers are very bad. I'm using 4o model. It's not consuming my weekly limit for o1 and o3, right?

[cloud pewter]

deep research gives response so long that I had a hard time reading them all

[rugged igloo]

intresting so i found out you can give it an outline for a APA paper or MLA paper and it it will further exspand on it wow this is very cool. this will really make people who do essays in schools easier

[normal quarry]

Has anyone got Deep Research in the macOS app yet?

[rugged igloo]

not for me

only on my pc

[cloud pewter]

o series really need file uploads now

😔

[rugged igloo]

chat i think i broke it

been stuck reasonin g for like 30 mins lol

[cloud pewter]

how

[rugged igloo]

its just compiling sources lol

been like this for a fair bit

[restive kernel]

No. I'm from the UK and work for a US company and saw the "Deep Research" toggle just now and thought I'd try it out, but it turns out it's only visible because I'm on a US VPN. It doesn't work, I guess because I'm a UK customer

[rich surge]

You've triggered AGI. It's thinking about how to deal with you now...

[restive kernel]

went down a rabbit hole learning to play all the games it read about lol

[rugged igloo]

prob i prob used like 10 deep researches so far lol

[elfin lava]

Is there limit for deep research

[rugged igloo]

100 a month for pro users

[eternal rock]

This is a Safe for Work Discord Server

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

Posting NSFW content will result in a permanent ban.

[rich surge]

I wish I had access to deep research (I'm a plus user)

[rugged igloo]

rip

[rich surge]

🪦

[rugged igloo]

its still thinking btw

[restive kernel]

Judging by the stuff posted last Friday about fine tuning reasoning models, we can expect this and better from open source models soonish

[rugged igloo]

yea

[rich surge]

I'm excited for the future

[cloud pewter]

same

[rich surge]

[rugged igloo]

yo future your nintendo apa paper broke deep research dude\

[cloud pewter]

Nintendo moment

[elfin lava]

Wait is deep research a reasoning + search?

[rugged igloo]

its an agent

[elfin lava]

.

Wait I’m sorry for my ignorance but

What’s agent exactly

Im a casual ChatGPT user so idk much about this

[rich surge]

Yes but it's like conducting a research paper. Like creating a academic or scientific research paper based on doing real research online.

[rugged igloo]

lol i had to refresh the page

[elfin lava]

Oh how is it any difference than when asking ChatGPT to search with keyword research paper etc

[rugged igloo]

https://chatgpt.com/share/67a0aa44-a404-800e-9c7c-02751ee1ccd2 here is the APA paper all about Nintendo

[alpine coral]

I’d like to know if it will be possible for ChatGPT to use my personal login credentials for certain websites (like MLS, or County and State government sites) to access additional resources for in-depth research. For example, as a real estate agent, could I sign in with my MLS credentials so ChatGPT can pull information directly and provide a comprehensive analysis of a specific property?

[rugged igloo]

yes in operator but its only for pro users for now

[rich surge]

The difference is that when you normally ask it something, it only searches online for specific meaning of something and clarification. Deep research actually goes to different websites and collects research to what you want to put together. Like for example, asking it to make an original blueprint for making a car that runs on water and how to build one yourself.

[cobalt idol]

is the deep research another thing not in EU yet?
EDIT: I see (We are still working on bringing access to users in the United Kingdom, Switzerland, and the European Economic Area. )

[elfin lava]

This is so cool damn

[elfin lava]

Does it able to actually look every website without having to worry if certain website restrict ChatGPT from viewing

[rugged igloo]

this is how much it wrote

[rich surge]

No, it skips websites it doesn't have access to and finds other websites.

[elfin lava]

@rugged igloo you weren’t kidding damn

If only canvas can do that

[rugged igloo]

if only

[elfin lava]

Canvas whenever a wall of text appeared : starts freezing on my end for no reason

[alpine coral]

Wow, I just about bet, NAR, etc. are gonna have a stroke over that one.

[rugged igloo]

yea

[cloud pewter]

Ty ty

[cloud pewter]

is this that Nintendo research paper lol

[iron roost]

o3 mini high and deep research and coding (be ready for some amazing results)

[rugged igloo]

deep reseacrh uses o3 already

[elfin lava]

Yeah it’s the convo

[rugged igloo]

full o3 the model that has evceryone freaking out about because its smarter than most phd people

[cloud pewter]

basically some ai that does some stuff automatically upon a request

[iron roost]

well ive just taken the 5 ai video generation ideas for youtube shorts and uploaded them (using deep research) on a hobby youtube channel, lets see if they beat my average view per short of 3000 views

[cobalt idol]

I thought thats every model 😄

[iron roost]

does pretty good at creative tasks combined with deep research and some chain of thought

they seem to be tracking on the normal average, the only difference is 20% of views so far are in the youtube search, which normally its the shorts feed with chatgpt suggested and created ideas

[rugged igloo]

lol\

[iron roost]

which is a good sign i hear, so thats a bit different, the deep research seems to have done something

[tawny crater]

Every thing i enter errors out

[iron roost]

think of it as a deeper more complex version of search

[tawny crater]

Is so ridiculous

[rugged igloo]

thasts prob from me

using to much compute from deep search lol

I'll get back to you with an expanded, well-structured report. (remember when this was just an hallucination and didnt do anything now it acually will get back to you)

[cloud pewter]

speaking of which o1 can actually handle 17k words and translated ALL OF THEM into Chinese

[rugged igloo]

wow thats good to know

[cloud pewter]

all I did was say “translate to Chinese, ensure accurate translation” and pasted the entire thing into it

the only letdown is it lost all of the reference links

(only thought for 9 seconds)

[rich surge]

(this message is a filler for excited ramblings, but can't think of a topic, but talking for the sake of talking during excited times)

[cobalt idol]

yeah, pasted 5000 lines of code and the o3 mini takes it
used to be much more limited per message on the webapp

[rugged igloo]

also it seems to be confirmed that tasks does in fact work with deep research

[whole roost]

Mine somehow doesn't use the custom instructions at all on o1, here

[cloud pewter]

good for citing things

[tulip ridge]

bro why is 4o spamming emojis like 🚀 ✅ ✅ even tho it's written 4 times in it's memory to not send any? did openai team told him to send emojis to appear young and friendly??

[rich surge]

Yeah

Basically

[half vine]

can i access o3 throught api?

[cloud pewter]

Not yet

[half vine]

oh

so when

[cloud pewter]

I can make more out of it tbh but I’ll prob do it when I either get this feature or subscribe to pro

[cloud pewter]

after its official release

[half vine]

and also what is the difference between o3-mini low, o3-mini medium and o3-mini high

[half vine]

oh ok, so to get more messages with o3 i need to pay the chatgpt plus?

[rich surge]

Yes

[half vine]

ok

[cloud pewter]

o3 isn’t released yet but yeah if you want more usage than free then plus is for you

[half vine]

ok

[whole roost]

Yeah it annoys me as well, try to set harsher custom instructions for it to adapt to, like a personality, telling it to not use emojis didn't help on my end

[cloud pewter]

low is just slightly better than o1 mini while med or high are technically on par on o1

[half vine]

ok so for coding i should use o3-mini medium?

[cloud pewter]

it always starts to use emojis after 5~6 messages

[cloud pewter]

yessir

[half vine]

ok

is there a difference in cost at these models? o3-mini low, medium, high

[rich surge]

Token cost, yes

[half vine]

oh ok

[tulip ridge]

thanks for the advice, did setting harsher instructions worked on your end? seems like it's hardcoded and instructions have little impact on this

[half vine]

and also why o1 is not available in the api?

[cloud pewter]

you have to pay

[half vine]

oh?

[rich surge]

You pay now!

[whole roost]

Yes, i have my GPT Play a role and it has opinions and is overall quite negative so it is more focussed on responding in the correct tone than sending emojis

[tulip ridge]

ok i'll give it a try, thanks!

[whole roost]

Wanted to add something but I have no idea which part triggered the automod so I just wish you good luck

[cloud pewter]

automod moment

[elfin lava]

Wait so I wasn’t the only one omg

Was it because of this

[whole roost]

Yes

Although it is beyond me why they would do such a thing, because it was way more easy to command it to use emojis than to restrict it to do so

But in an honest take I think that they tried to make it less formal and this is just a side effect they cannot prevent so they label it a feature

[rugged igloo]

thats the real agi test

[spice lotus]

for those who have "deep research" feature, is it possible to attach multiple pdf's to it ?

[whole roost]

Not on my end luckily since he maintains being the lord of terror with all his sardonic disdain for everything

[cloud pewter]

all hail the lord of terror

xdxdxd

[whole roost]

I think my custom GPT badly explained also uses no emojis

[cloud pewter]

I should optimize my custom instructions a bit since yours seems to be short

yet has a good effect

[whole roost]

It does a bad job at following instructions anyway so more is sometimes less

They would need post conversation prompting or something similar

As good as this

Once it did it it won't let off anymore

[cloud pewter]

true true

[whole roost]

What annoys me a bit more than the emojis is that it now constantly asks follow up questions

Someone else must have noticed as well that it now always asks questions and I am not inclined to answer any questions

It's like faking to be interested in something while I just needed a response

[cloud pewter]

this has been kinda prevented from happening for me because of my memories and cis lol
But i forgot

i tells it to not be “ai like”, and respond as natural as possible

[whole roost]

I want it to be AI like and just perform tasks and not ask anything back unless it needs information to complete a task

I was alright with the certainly! Is there anything else you need assistance with?

[cloud pewter]

😔

[whole roost]

I might try to instruct it to be that way again but it doesn't bother me that much because I'm only in project anymore so the standard thing doesn't really reply to me anymore much

But since there are many complaining about the emojis I also think that there will be a personality change soon again

I think it is hard to find the middle ground because I know a lot of people who complain that GPT was too corporate
I have a few people in my friends list who now cherish the model because it sounds so casual and human the exact thing that I despise

[cloud pewter]

it is so corporate and it sucks to me
but instead of changing them all why don’t make custom instructions and memories more important, like it controls over its personality more

[whole roost]

Lately there have been more complaints about GPT having gotten dumber and I actually think it hasn't really gotten dumber but just sounds dumber because of the more casual tone

[elfin lava]

Have u tried inject to memory

[cloud pewter]

yeah, in fact I injected multiple

Works well on the personalities but not on emojis

and sometimes it forgets about it and keeps saying those “banned phrases” from me

[whole roost]

I have given up to tell it to cease any phrase

[cloud pewter]

“if you have more questions, please let me know”

[whole roost]

I would like to have that sentence back because mine started to ask questions about whatever I write about

[cloud pewter]

real

[whole roost]

None of your business sir

Imagine if I would be on plus subscription and only had like 80 per three hours

That's one way to burn your charges

[frozen pulsar]

is deepsearch only for pro users?

[cloud pewter]

Ye

[whole roost]

No it should be for everyone to some degree but it's not available in Europe if that's an issue

[frozen pulsar]

ah alr

[frozen pulsar]

ah probable thanks

[cloud pewter]

me lol, I’m rationing these usages HARD

[whole roost]

Imagine being a pro subscriber in Europe and tell me again about getting nothing

[cloud pewter]

yeah…….

[cloud pewter]

Openai either need to decrease subscription price or offer more usage to eu lol

[whole roost]

Well I'm still paying so I'm my own problem to some degree

[shut dagger]

does deep research work with o1-pro? it seems to sit there and do nothing, is it working in the background?

[cloud pewter]

I’ve heard someone said that it used to work

not anymore I suppose

[shut dagger]

i've gotten to work with 4o, so far. ill try o1 next

[shut dagger]

It would be easy enough to have 4o load all the papers then switch to o1-pro to read them

[honest jay]

Deep research uses a version of the o3 model: https://openai.com/index/introducing-deep-research/

Powered by a version of the upcoming OpenAI o3 model that’s optimized for web browsing and data analysis, it leverages reasoning to search, interpret, and analyze massive amounts of text, images, and PDFs on the internet, pivoting as needed in reaction to information it encounters.
So I don't think it's using 4o or o1 (and I don't think this is changeable). If you're noticing it working with one model selected and not the other, it's possible that it's just a UI bug where maybe it's only triggering properly if from the 4o model? But it doesn't sound like the actual model powering deep research can change

[shut dagger]

Thank you!

[rocky kayak]

deepseek is trippin again

tried to do a comparison between R1 and o3

[cloud pewter]

chatgpt ui is mid at best

[rocky kayak]

nah its good asf

[honest jay]

#ai-discussions is the best place for non-OpenAI discussions/comparisons

[rocky kayak]

it could be better but its already good

[rocky kayak]

Thank you

[cloud pewter]

can’t handle long conversations properly lmao

[rocky kayak]

You said the UI

[rocky kayak]

.

[visual nova]

is there a trick to getting deep research to complete without failing? it looks really cool, potentially better than the google one, but its failed 100% of the time so far.

usually after 5-10 minutes of thinking

[dense halo]

200 usd (300 aus) to beta test it i guess haha

[normal quarry]

what happens when it fails?

[rocky kayak]

It was literally just launched a few hours ago, so I do not expect it to function as intended.

[cloud pewter]

yeah it’s a pain to scroll through the ui when the convo is long
it also has performance issues on web

[rocky kayak]

Oh, certainly, although I have never encountered such issues myself.

[visual nova]

it will read a bunch of sites but then just brick

[normal quarry]

I hope it's not deducting a use every time it happens

[rocky kayak]

Could it merely be a port of their agent, similar to the web agent, which performs all the research without presenting it to you?

[visual nova]

how many uses on pro?

[visual nova]

is it still pretty limited on there

oh, yeha thats pretty limited lol

[rocky kayak]

So js a scam

[cloud pewter]

I have encountered it multiple times, once one chat gets lengthy it takes ages to load on web
even with just one long message on o1

[rocky kayak]

I'm not sure, I've never used O1.

[cloud pewter]

Sads

but o1 can handle long texts and does it very well

I mean things as long as 4o doesn’t even accept them

[visual nova]

Ok, I think if you want your research turned into code you have to do it in 2 steps using deep research then o3-high

it seems to work if i dont couple the request with "turn your research into a matplotlib graph" or whatever else.

[cloud pewter]

memory is bugged I never mentioned my favorite pokemon

in fact I was telling chatgpt not to use emojis!

[visual nova]

you arent fooling anyone, seriously, porygon-z?

[cloud pewter]

and it keeps adding that “I like porygon-z” stuff like it’s real

I don’t even play pokemon.

😡 😭

[visual nova]

is your name gamecube?

maybe they have another cross user leak where you are seeing someone else's memories

[cloud pewter]

yes

[visual nova]

weird lol, the memory feature is the one thing i dont use, i like consistent system prompts.

[cloud pewter]

it keeps getting added 😭 😭

my prompt is “Remember, zero emojis unless explicitly asked. bold/all caps texts should not be overused too, just on the exact emotion.”

[whole roost]

this also happens frequently and sometimes when i enter the custom instructions they are in fact empty until i refresh

really odd

[cloud pewter]

added #1335987900743028758

[whole roost]

yeah it has no clue, the instructions aren't there

and here it has them again, something's severely bugged

[cloud pewter]

when I tried to update my custom instructions it also doesn’t update after my edit was “saved”

[whole roost]

mine are up to date if they are applied at all

but maybe the no emoji instruction would actually work if the instructions worked

[cloud pewter]

yeah

[whole roost]

[obtuse jetty]

anyone knows what the rate limit is for Operator?

[young basin]

100 per month

Ooops sorry thought you meant Deep Research

I’m actually not sure about operator

Has anyone even reached a limit with it?

Might fall under one of the many unlimited features of the pro plan

[cloud pewter]

it has a daily limit on pro

but it’s unclear

[young basin]

I guess because it depends on how long the task takes

Asking it to fill out 100 forms probably eats up the quota quick

[forest hinge]

deep research channel where people show the results from deep research

#1070006151938314300

also why i dont have a thinking button in chatgpt?

[midnight rover]

Hi, unfortunately chatGPT does not run self-hosted. I like to discuss personal details with GPT like my CV, deep psychological insides about myself and so on. I disabled contributing to further models with my conversations. But is my idea to share any confidential info with still chatGPT stupid? What is the outcome?

[feral zinc]

there should not be any risk especially if you use a temp chat. if you dont use a temp chat then in case you get hacked someone could read the old chats

[prime swallow]

I wonder.

Do you know what you want the model to end its outputs with?

I don't. But instead of negative prompting - have you tried getting the model enthusiastically involved in other options, and then guiding to the one you want, then getting the story going?

Not sure what you were doing before that output. Appears to be comparing two armies, one yours and the other not.

So I mirrored that, on my 3rd input.

https://chatgpt.com/share/67a0f6be-1510-8011-8658-a620f46c0e04

I so recommend, guide the model to where you want.

You're like someone mad that the taxi driver's taking you to the airport. And yelling about the airport when you say don't ask questions.

If you want Disneyland or the seedy back alley near 4th and Main in downtown. Talk about that! Tell the model what you want. Inside allowed content and its ability to process it (spatial sense still very iffy, other problems too) - it'll go there eagerly.

It just needs help to 'identify' where you want to go.

[crystal tapir]

I don't have access to Deep Research yet

[midnight rover]

Will Sam Altman secretly read my CV if I upload it to chatGPT? 😄

[whole roost]

I dont want it to get a story going, it's supposed to answer my question and done

[prime swallow]

Well, if you wanna keep complaining instead of examine examples for what works and adapt to your own goals and needs - all on you. Keep havin' fun.

[whole roost]

Not at all, I just want the previous personality of the model back

i don't like the new one

[prime swallow]

300 million weekly users.

Seems unlikely to get personal attention; but I bet you can apply through proper channels for something!

[prime swallow]

If I were in that situation. I would attempt to provide the model with examples of exactly what I wanted, and what I notice that is different about the good/bad.

I would train it on my preference, whatever that was. And then I'd have what I wanted for me. No other user would have my customization, and who cares. I'm getting the outputs I want, and I can if I choose offer clues to others who don't seem as happy as I am.

[whole roost]

you cannot train your model

you can give it an instruction but that won't overwrite the system prompt

[prime swallow]

If you mean disallowed content - no, and I have no interest in that.

If you mean get it to do 'the impossible for it' - no, that's not possible.

Anything else, is.

Care to discuss some specific "I want X and cannot get X"? With a target I will display a method I would use.

[whole roost]

I want the previous version back without emojis and without courtesy questions, simple answering of questions in the corporate manner it did before

[prime swallow]

Are you pro? So far I understand it's pro-only.

[whole roost]

i don't want any of these questions, it's not my friend and shouldn't feign interest

[prime swallow]

If you have any older chats where it used the style you like.

Give it that message you just typed here + tell it here are three examples of you doing it right. put each example in [] to make it clear where each starts and stops.

I expect high success from that.

[whole roost]

it recites those examples so that this isn't a good way for me either

[crystal tapir]

Oooh. No I'm not. Plus. That's why.

[prime swallow]

Me too. Sorry that we wait - it's planned to expand to our tier and have a few uses per some time period.

[whole roost]

i mean

my instructions aren't difficult

[crystal tapir]

Nice. Ty

[whole roost]

if you are in EU the wait might be longer

it's not coming yet to eu

[crystal tapir]

I'm not

[prime swallow]

Yeah. This is the available info as far as I know:

[We are starting with a version optimized for Pro users today, with up to 100 queries per month. Plus and Team users will get access next, followed by Enterprise. We are still working on bringing access to users in the United Kingdom, Switzerland, and the European Economic Area.]

https://openai.com/index/introducing-deep-research/

[crystal tapir]

I'm in Americas

[crystal tapir]

Okay

[cloud pewter]

at this rate pro users are getting every new feature first and plus users are just left waiting😭

[crystal tapir]

We also will get 100 queries per month?

Or less?

[prime swallow]

Ahh, and:

[In the coming weeks and months, we’ll be working on the technical infrastructure, closely monitoring the current release, and conducting even more rigorous testing. This aligns with our principle of iterative deployment. If all safety checks continue to meet our release standards, we anticipate releasing deep research to Plus users in about a month.]

So I think that's the soonest possible for us.

If pro gets 100/mo, plus may get less.

[whole roost]

i mean we're paying times 10

[crystal tapir]

I wish I could pay for Pro

[whole roost]

i cancelled most other subscriptions, i work part time minimum wage but this is worth it

[crystal tapir]

Nice

[prime swallow]

I'm just a community member like yourself so, I dunno.

However, access has expanded to -everyone- over time, inside the areas where accounts are allowed.

Free accounts used to have usage limits over time, now they get unlimited mini 4o, right? I think I read that. And they get uses of dalle and browsing and python tool and more?? I think. Not loads but some.

And plus access has expanded since it started. Hugely. We now have an unlimited chat for sure, with 4o mini - we never had anything unlimited until recently.

So to me, I'm like, "Bet it keeps getting better"

[cloud pewter]

yeah as long as it generates enough happiness and possibly profit, and u can afford it, it’s great

[crystal tapir]

Yeah

[cloud pewter]

they gonna lobotomize DR and give plus 20/mo

[crystal tapir]

I never explored the possibility of profiting with ChatGPT

[prime swallow]

Long as it's better than what I have now, even that's a step up.

[crystal tapir]

I only use it for study

[prime swallow]

And there's so many steps up I've seen over the couple years. I'm really cool with it, to me it's AI-gifts-mas and yay

[whole roost]

i use it to study at work, if that ever turns to money i don't know but why not getting a higher exam while being stuck in a traffic jam

[midnight rover]

I deleted all links of shared conversations.

1.) There was no [delete all] button (that I saw)
2.) Deleting a shared link updates the conversation and lists very old conversations on top of my chronological conversation list

[gloomy brook]

I am confused how do I access research? Does anyone currently have access?

[normal quarry]

Do you have Pro?

[gloomy brook]

Lol i am not rich enough

so i guess it is for pro people

cool Thank you!

[normal quarry]

Deep Research has only released to Pro subscribers for now, Plus will come later

[gloomy brook]

Have you tried it Robert? Do you have Pro? Wondering if you have how were the results?

[normal quarry]

I have, it's really good, the results were better than I expected. If you scroll back through the chat you will see several people shared their experiences with it

[tawny whale]

Any info on expected EU timeline for Deep Research on pro apart form the initial statement?

[normal quarry]

only that they are working to bring it to the EU as soon as they can

[hexed condor]

what's the issue with Enter key not working to send the message?

[neon acorn]

How do you do it?

[lone lodge]

is there a limit of message we can have with chatGPT ?

[winter vine]

what type of courtesy questions ?

[brittle kelp]

did chatgpt website change logo?

[dawn sentinel]

I don't see the Preview button on my Canvas. Is it the same for the rest of you?

[orchid pollen]

when gpt with tasks coming to the windows desktop version?

the point of the app is so i dont have to use the browser the whole time

[dense fox]

I just tried Deep Research and nothing happened. It asked me about 10 questions to clarify what I wanted and then it just stopped responding. I've been waiting and hour for it to provide an answer or a result. If I ask for a status update, nothing happens. Is this how it works?

[orchid pollen]

meanwhile EU always behind no deepsaerch hier 😄

[runic drum]

this this is useless you need a 100000 page paragrafh for it to understand what a 4 year old would know they should fix it now and thats an ORDER OR THEY SHOULD TRY AGAIN FROM THE START WITH ONE THAT WORKS

[green elbow]

is gpt4 having issues

[vocal monolith]

Anyone managed to access deep research from Europe by using that thing that starts with v and ends with n?

[gloomy brook]

Please who can afford PRO what is your annual salary?

[waxen arch]

Is it worth to try it?

[mossy zodiac]

Is there unlimited access to deep research on pro plans?

[honest jay]

100/month currently: https://help.openai.com/en/articles/10500283-deep-research-faq

Pro users (located outside the UK, Switzerland, and the European Economic Area) have access to 100 deep research tasks per month.

[mossy zodiac]

Oh ok

[keen hollow]

Is Deep Research available for the Plus subscription?

[honest jay]

Not yet, but OpenAI have said that access for Plus and Team subscribers is next

[keen hollow]

Great. Thanks!

[sacred vale]

We are still working on bringing access to users in the United Kingdom, Switzerland, and the European Economic Area. Does this mean we're not getting it like Sora?

We need to set something up to where I can send a prompt to someone in the USA and he can send me the result... I'm paying just as much and not getting the access.. let's start a Discord group for exchanging access results

[mossy zodiac]

What does that mean

Is there an official date?

[honest jay]

Not currently, it doesn't look like. From the link above:

We look forward to extending Pro access to all regions soon, and will expand to Plus, Team, and Enterprise users at a later date.

[tiny obsidian]

I've had this issue as well... I've tried it 4 times already and it only tells me that it's going to research and notify me afterwards, but it never shows the progress bar, and sometimes it doesn't even write anything and keeps it at that

[honest jay]

Hey, I totally get why you'd be frustrated in this scenario. And apologies that this response will likely not assuage any frustrations, but out of caution, it is important to keep in mind that bypassing regional restrictions is against OpenAI usage terms.

[sacred vale]

Ok in that case I won't do it

[sacred vale]

Can we get a discount then? 😉

[mossy zodiac]

I guess we have to wait till Feb 2026 for release then

[sacred vale]

Yeah I made a bunch of GPT's I really like.

And again, yes I am paying for it, just as much as a US citizen. And I am not suggesting any illegal internet access to Sora or this new Deep Search feature. Just asking someone in the US to look something up for me, but appearantly that's against policies as well.

[mossy zodiac]

You could also get gpt 3.5 for free or at a much lower price than o3 mini, so what's your point? The person here wants the Deep Research feature which is super advanced and uses the o3 model.

Its not just him "sticking to one brand"

[sacred vale]

True

[pearl wind]

Chat wont let me upload images rn, are servers currently down or something?

[sacred vale]

But yeah I love GPT, but I also stuck round a year longer than I should have with my ex gf

[mossy zodiac]

The alternatives for o1 would be deepseek, which unless you wanna use a distilled version which isn't the same or use a CCP operated app that steals your data then you need to pay for it as well. And they weren't talking about alternatives for o1, they were talking about the deep research model.

[sacred vale]

Yeah that too, and I could also write a letter to the European Congress, but that's not going to help

Kio Kio Kio Kio..... I'm gonna take my horse to the ole town road I'm gonna...riiiiiide...

I'd be happy if I got some Deep Sleep.. shouldn't be too hard to get access to

I close my eyes..

I know, been working too hard

[mossy zodiac]

Nobody said its "cool" that OpenAI steals data from its users and uses it to train their data. Main difference however is that DeepSeek is operated by a foreign adversary and every chat is used to train their models and gives an evil government an advantage in a serious technology.

[sacred vale]

Goodnight my dear friends

[dense fox]

Thank you for confirming this Deep Research behavior is not unique. What a frustration and waste of time. It appears my entire query is dead. I've paid for two months of Pro and the Deep Research doesn't pay out. Sora never materialized to ever offer the Pro side all of the available creation features. I'm not sure paying for a third Pro month is a wise effort.

[sacred vale]

Thanks, you too!

[trim rose]

Wait, deep research in online? Not in the UK I guess....

Also sora and agent is unavailable in the UK

[mossy zodiac]

North Korea is evil. It kills and imprisons anyone who lashes out and heavily censors information from the outside world and their own wrongdoings. China is the same except it doesn't prevent people from leaving. They are evil and I personally wouldn't want them to have an advantage with AI. And again, you're putting words in my mouth. Never said that we are specifically the good guys, just saying that I'd rather use a US owned data-stealing app than a CCP owned data-stealing app.

[grim pumice]

Same here. Claims it's going to do research and then never does. I've tried four queries so far, always the same thing.

[frosty bridge]

yo not sure if this the right place to ask but isnt the daily o3-mini-high limit 150 messages with the plus subscription?
Ive used it a bunch this week but now i got 25 messages remaining until the 8th?

[dense fox]

Hmph! Super disappointing! Thanks for confirming this behavior, too. I don't get it. Why release something that doesn't really reliably run? I was getting a little perturbed at its never ending questioning of what I was asking it to do -- but I played along -- and then, I guess, it ran out of questions and died? This is not Pro level work, I fear.

EDITED: Typos.

[grim pumice]

From reading Twitter and here, seems like it works for most people but at least a few of us are in this broken state. Unclear why - an OAI employee told me they flagged a couple of reports to look at.

[latent ibex]

bro what why these chinese/japanese text? I don't even have that in my code

[dense fox]

Well, that's good to know they're looking into it. I thought maybe I asked it too much -- but it conversationally whittled down the inquiry into very specific terms -- so I thought everything was okay. They need to add an, "I'm giving up now" response or something. To ask for an update, or to just say "Hello?" and get no response at all is something I've never seen before from OpenAI. Here's how it ended...

[grim pumice]

Your experience looks just like mine. It asks followups and then just does nothing. And doesn't reply to any "hello?" type queries.

[dense fox]

Super interesting! I actually thought the back-and-forth questions from the Bot were really good and focused; and then, yeah, just like your did, it all just died. Not a great look for a brand new feature release.

[dawn sentinel]

Seems to be down.

[wise anchor]

Oh cool honestly I didnt expect that

[rocky wing]

Why is the conversation history on the Android app still broken? It's been over 2 weeks at this point.

[wise anchor]

The o3-mini web search finally started working for me and it is GOOD so i can only imagine how good deep research is

[ornate nest]

"o3 mini is fast"

[mossy zodiac]

Whats the best alternative for Deep Research

[sick pecan]

#ai-discussions

[young basin]

I hope we get a date for Plus soon. I've got some research questions but it'll be just my luck I subscribe to pro and then deep research comes to Plus the next day 😅

[hidden ibex]

so if it's been researching for 40 minutes with no updates for the last 30, did something go wrong?

ive reloaded and it still says it's thinking, so it's not a "something went wrong" type deal

[grim pumice]

I've noticed that if I pick o1 (not o1-pro) as the model, it works more reliably.

[cloud pewter]

sometimes it overthinks lol

[ornate nest]

leave it overnight

[hidden ibex]

it finally stopped lol

spent an hour thinking, then stopped, and then said it was only 7 minutes

hmmm lol

[ornate nest]

what are you prompting

😭

[hidden ibex]

antigravity drive lmao

[simple arrow]

Does anyone here have o1 Pro?

Can we attach attachments to o1 pro? like files. O1 can only attach images, is it the same in o1 pro

[honest jay]

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

[elfin lava]

How come iOS app doesn’t get this UI

[weak tree]

Gemini Deep Research or Perplexity

[quick meteor]

Deep research

[quick meteor]

From what they've shown, those are really close

[graceful stream]

So for general use is o3-mini or 4o a better option?

[honest jay]

I'd argue 4o, but it depends on what general use means for you. I love 4o for being so quick and capable, especially its internet search ability

[gloomy brook]

Anyone willing to share results from research? I can't afford PRO, but hoping those of you would be kind enough to share the results with us people who can't afford pro, but would love to see the results?

[normal quarry]

if you search #openai-chatter there was a lot of sharing of results last night

[gloomy brook]

Thank you! I was able to find one.

It would have been great if they gave plus members at least one research every week or month

[normal quarry]

Plus and Team users will get access soon

[visual nova]

I got you

[zenith locust]

🤣

[cloud pewter]

real

[elfin lava]

If only canvas can do that too 🙄

[brave valve]

So, Europe finally got advanced mode for voices. It claims I can share the screen but nothing happens, any tutorial for that?

[elfin lava]

It’s only on iOS and android app

[brave valve]

Yes on android

[eternal rock]

`` Server Tip: ``

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

[elfin lava]

When you open voice mode , the icon for camera and screen share should appear

[brave valve]

When I do I see a blank screen

[elfin lava]

Can you screenshot when you open avm?

[elfin lava]

Oh isn’t that sharing screen

That mean it’s sharing

Try ask “can you see my screen”

If it respond with yes then it’s working

[brave valve]

Sharing with who? It says it doesn't see my screen, if I leave the app as well

[elfin lava]

With ChatGPT

I had that problem too whenever it says it can’t see my screen

What I did was I open AVM, waited for like 1 minutes or few seconds and then turn on screen share

Then ChatGPT should be able to see your screen

Ok wait I’m having the same problem

[brave valve]

I checked app access but doesn't see anything related to screen sharing

[elfin lava]

Ok it’s working for me now, what I did was start a new AVM, and then say hi like usual and then tell em I’ll be sharing screen, after the AVM responds I turned on screen share and it works

[manic parrot]

I am using Korean in the Mac GPT app, but when I type in Korean, some of my keystrokes are being skipped. I need some help!!!

[potent saddle]

💀

[brave valve]

Trying but still doesn't work, unfortunately, will reinstall

[elfin lava]

Or restart your phone

[quick meteor]

Can someone write this out

[elfin lava]

🦋

[wraith dagger]

I’ve still never found out why there’s a maps provider option because I’m never seen it used. What’s it for?

[native otter]

Related to map web search

[quick meteor]

That butterfly was made by o3 mini high but it can’t be copied and pasted on mobile only on pc

[quick meteor]

😭

I meant o3 mini medium

[wraith dagger]

Ah I see

[quick meteor]

Not o3 mini high

[quick meteor]

Now try pasting it and send it and see what shows up. It’s terrible @wraith dagger

[wraith dagger]

.==-. .-==.
()8-._ . .' .-'8()/
(88" ::. ./ .:: "88)
_.'-::::.(#).::::-'./
._... .q(_)p. ..._.' ""-..-'|=|-..-""
.""' .'|=|. "".
,':8(o)./|=|.(o)8:. (O :8 ::/ \_/ \:: 8: O) \O ::/ ::' O/
""--' `--""

[quick meteor]

Yeah….

[wraith dagger]

I see what you mean

[quick meteor]

And I couldn’t get it to fix it

[wraith dagger]

Its just an issue with different sized text boxes

It’s better in notes just not perfect

[quick meteor]

Oh true

[cloud pewter]

pov average usage of o series models

[quick meteor]

I only did it once with o1 but o3 mini is like the first model to actually get precise word counts. Like I ask for EXACTLY 500 words about something and then it actually does it when I put it in the word counter. I’ve done it multiple times now it actually understands it. I haven’t even tried o3 mini high since it would obviously get it. I also didn’t just say “write a 500 word story.” There were other constraints as well

O1 had failed this

FINALLY it was about time

[quick meteor]

Well if it had failed it would have told me something about the models limitations.

[rugged igloo]

now this is what im talking about

https://chatgpt.com/share/67a1b181-f924-800e-8d1a-865439ed2988

[normal quarry]

Excellent

I see that when you open a shared chat in the macOS app you can now "Continue the Conversation" - this was the case earlier but OpenAI removed it for some reason - glad to see this feature back

[rugged igloo]

oh nice

[cloud pewter]

plus peasants would never get anything better

it’s over

[cold sequoia]

Hey everyone , anyone knows when is the release date of deep search for plus plans